Application of calibrated Swiss catchment model parameters ...

Technical University of Munich

Department of Civil, Geo and

Environmental Engineering

Chair of Hydrology and

River Basin Management

Prof. Dr.-Ing. Markus Disse

Swiss Federal Institute for Forest,

Snow and Landscape Research (WSL)

Mountain Hydrology and Mass Movements

Hydrological Forecasts

Dr. Massimiliano Zappa

Application of calibrated Swiss catchment model parameters to hydrologically assess a

microscale catchment in the Peruvian Andes

Carina Anne Pfeuffer

Master Thesis

Matriculation number: 03618614

Degree Course: Environmental Engineering

Field of Study: Environmental Hazards and Resources Management

Supervisors: TUM Prof. Dr. Markus Disse

WSL Dr. Massimiliano Zappa

WSL Norina Andres

April 6th, 2017

III

Declaration of Authorship

I hereby declare that the thesis submitted is my own unaided work. All direct or indirect sources

used are acknowledged as references.

I am aware that the thesis in digital form can be examined for the use of unauthorized aid and in

order to determine whether the thesis as a whole or parts incorporated in it may be deemed as

plagiarism. For the comparison of my work with existing sources I agree that it shall be entered

in a database where it shall also remain after examination, to enable comparison with future

theses submitted. Further rights of reproduction and usage, however, are not granted here.

This paper was not previously presented to another examination board and has not been pub-

lished.

__________________________ ___________________________

Place and Date Signature

IV

Acknowledgments

First and foremost, I would like to express my sincere and appreciative gratitude to my supervi-

sors Dr. Massimilano Zappa and Norina Andres for the continuous support of my master thesis,

for all the patience, motivation, useful remarks and constant source of knowledge. Your guidance

essentially helped me in the writing of this thesis.

I would like to acknowledge and thank Prof. Dr. Markus Disse for supporting this thesis from the

side of the university, for the encouraging talks and helpful advice. Thank you also for joining my

presentation in Zurich – this really meant a lot to me.

Another big thank you goes out to the entire WSL family for an amazing working experience, that

made me feel like home. Thank you for teaching me some Swiss German and the patience of

having to repeat multiple times. Especially, I would like to thank Käthi Liechti for being such a

great office mate and always helping me out with R Studio. A big thank you also goes to Daniel

Farinotti for making my scholarship possible, I really appreciate the help.

Additionally, this thesis would not have been possible without the field work and data provided

by Jan R. Baiker. Thank you for the cooperation.

I would like to acknowledge helpful advice and comments for the completion of the thesis from

my Dad, Regina and Anna.

Getting through my degree and this master thesis required more than academic support, and I

have many people to thank for listening to and, at times, having to tolerate me over the past

years. It‘s hard to express my gratitude and appreciation for your friendship – you know who you

are. Thank you for being unwavering in your personal support always there for a kind word and

a great talk. I cannot wait to move back to my beloved Munich and you are one of the biggest

reasons. I would also like to thank the Wolfswinkel-WG including Benni, Chris, Tiziana, Daniel

and Lisa who opened their home to me when I first arrived in the city.

Finally, I would like to express my very profound gratitude to my loving parents and to my best

friend and partner in crime Chris, for providing me with unfailing support and continuous encour-

agement throughout my years of study and through the process of writing this thesis. This ac-

complishment would not have been possible without you. I will be grateful forever for your love.

V

Content

Declaration of Authorship III

Acknowledgments IV

Content V

List of figures VIII

List of tables X

List of abbreviations XI

Abstract 13

1 Introduction 14

1.1 Integration of the thesis into the frame of the current research project .......................... 14

1.2 Motivation and objectives ............................................................................................. 15

1.3 Structure of the thesis ................................................................................................... 15

2 Introduction to the study site 17

2.1 Ampay National Sanctuary ........................................................................................... 17

2.1.1 Geology and geomorphology .................................................................................. 17

2.1.2 Climate and hydrology ............................................................................................ 17

2.2 Study catchment ........................................................................................................... 19

3 Status of scientific research 21

3.1 Hydrological processes of the rainfall-runoff system ..................................................... 21

3.1.1 Precipitation ............................................................................................................ 21

3.1.2 Evaporation ............................................................................................................ 21

3.1.3 Evapotranspiration .................................................................................................. 22

3.1.4 Empirical parameter relationships ........................................................................... 24

3.2 Manual calibration versus parameter donation ............................................................. 24

3.2.1 Manual calibration .................................................................................................. 24

3.2.2 Parameter donation ................................................................................................ 26

3.2.3 Evaluation of the two approaches ........................................................................... 27

4 Experimental basis 29

4.1 Hydrological model system PREVAH ........................................................................... 29

4.1.1 General information ................................................................................................ 29

4.1.2 Runoff generation tuneable parameters .................................................................. 30

VI

4.1.3 Parameter estimation ............................................................................................. 30

4.2 Model input data ........................................................................................................... 32

4.2.1 Physiographical information ................................................................................... 32

4.2.2 Climatology ............................................................................................................ 32

4.2.3 In-situ measurements ............................................................................................. 34

5 Methodology 37

5.1 Catchment subdivision and HRU generation ................................................................ 37

5.2 Application of the Swiss catchment tuneable parameter sets ....................................... 37

5.2.1 General idea of the methodological approach ........................................................ 37

5.2.2 Generation of comparison situations ...................................................................... 39

5.2.3 Data processing and quantile plot generation ......................................................... 39

5.3 Parameter set optimization ........................................................................................... 40

5.3.1 Visual quantile comparison ..................................................................................... 40

5.3.2 Numerical quantile comparison .............................................................................. 41

5.3.3 Comparison of tuneable parameters of well performing donor sets ........................ 42

5.3.4 Decision on set for further analysis ......................................................................... 43

5.4 Sensitivity analysis ....................................................................................................... 44

5.4.1 Visual sensitivity analysis ....................................................................................... 44

5.4.2 Numerical sensitivity analysis ................................................................................. 44

5.5 Additional in-depth analysis of meteorology and hydrology .......................................... 44

5.5.1 Temperature extreme value analysis ...................................................................... 44

5.5.2 Time series curves ................................................................................................. 44

5.5.3 Additional analyzing plots ....................................................................................... 46

5.6 Donor parameter set tuning .......................................................................................... 47

5.7 Water balance .............................................................................................................. 48

6 Results 49

6.1 Sensitivity analysis ....................................................................................................... 49

6.1.1 Surface runoff ......................................................................................................... 49

6.1.2 Interflow ................................................................................................................. 49

6.1.3 Deep percolation .................................................................................................... 50

6.1.4 Baseflow ................................................................................................................ 51

6.1.5 Comparison between visual and numeric sensitivity analysis ................................. 52

6.2 Additional in-depth analysis of meteorology and hydrology .......................................... 53

VII

6.2.1 Temperature extreme value analysis ...................................................................... 53

6.2.2 Time series curves ................................................................................................. 53

6.2.3 Additional analyzing plots ....................................................................................... 55

6.3 Donor parameter set tuning .......................................................................................... 58

6.4 Water balance results ................................................................................................... 58

6.4.1 Boxplots ................................................................................................................. 59

6.4.2 Barplots .................................................................................................................. 63

7 Discussion 66

7.1 Explanation of the hydrology in the catchment .............................................................. 66

7.1.1 Temperature ........................................................................................................... 66

7.1.2 Precipitation ............................................................................................................ 66

7.1.3 Runoff ..................................................................................................................... 67

7.1.4 Evapotranspiration .................................................................................................. 70

7.1.5 Effect of geology and El Niño ................................................................................. 72

7.2 Uncertainties and limitations ......................................................................................... 75

7.3 Evaluation of the donor parameter approach ................................................................ 76

8 Conclusion 77

References 79

Appendix 88

Appendix A: Tables ............................................................................................................... 88

Appendix B: Figures .............................................................................................................. 88

VIII

List of figures

Figure 2-1: Overview of the country of Peru with a detailed map showing the Ampay National Sanctuary`s core area (blue), the study catchment (red) and the bofedal area (green) Additionally visible: Ampay glacier in the NW of the ANS and the location of the closest two SENAHMI weather stations and the closest ERA Interim location. ................................................................................................................. 18

Figure 2-2: Catchment of all five subareas from a northern looking perspective including location and abbreviations of the v-notch weirs and acronyms for the subareas (source: Google Earth). ......................................................................................... 19

Figure 2-3: The bofedal from a southern looking perspective during different months of the year indicating the variability in greenness and soil moisture in the bofedal (source: Jan R: Baiker). ......................................................................................... 20

Figure 3-1: Water cycle with important spatial subunits and hydrological processes for the hydrological simulation in PREVAH (diagram changed after [17]). ........................ 22

Figure 4-1: Schematic diagram of the PREVAH structure including tuneable parameters, storage modules and hydrological fluxes (source: Viviroli et al. 2009 [90]). ........... 31

Figure 4-2: Map of the distribution of measuring devices and botanical plots in the bofedal area (source: Jan R. Baiker, map prepared by Dina Farfán Flores) ....................... 35

Figure 4-3: Currently performed bucket measurement at a v-notch weir (source: Jan R. Baiker). .................................................................................................................. 36

Figure 5-1: Diagram summarizing the methodology used in the thesis subdivided by simulation, data input and plotting results. The lines indicated the source of input and the red labels the chapters of additional information. ...................................... 38

Figure 5-2: Diagram with exemplary output for one tuneable parameter set. .......................... 40

Figure 5-3: The linear and corresponding logarithmic plot of a poor approximation on the left (AlpEin) and a particularly good donor set on the right (Tic_34). ...................... 41

Figure 5-4: Explanation of quantiles and intervals used in the plots and calculations. ............. 42

Figure 6-1: Diagrams for the threshold storage for surface runoff for a high (left), medium (middle) and small (right) value in direct comparison ............................................. 49

Figure 6-2: Diagrams for storage coefficient for surface runoff for a high (left), medium (middle) and small (right) value in direct comparison. ............................................ 50

Figure 6-3: Diagrams for storage coefficient for interflow for a high (left), medium (middle) and small (right) value in direct comparison........................................................... 50

Figure 6-4: Diagrams for percolation for a high (left), medium (middle) and small (right) value in direct ompareson. .............................................................................................. 50

Figure 6-5: Diagrams for the storage coefficient for fast baseflow for a high (left), medium (middle) and small (right) value in direct comparison. ............................................ 51

Figure 6-6: Diagrams for the storage coefficient for fast baseflow for a high (left), medium (middle) and small (right) value in direct comparison. ............................................ 51

Figure 6-7: Diagrams for the storage coefficient for delayed baseflow for a high (left), medium (middle) and small (right) value in direct comparison. .............................. 52

Figure 6-8: Diagram comparing daily in-situ temperature minimum (green), mean (blue) and maximum (red) values (as mean over HOBO 1-3). ................................................ 53

IX

Figure 6-9: Hydrograph curve simulating the years 1995-2009 using the HOBO1-3 daily data (Area 4 compared to V-notch weir 1). ............................................................54

Figure 6-10: Mean simulated evapotranspiration compared to scaled in-situ daily data. .........55

Figure 6-11: Scatterplot water table (in-situ) versus simulated SLZ. ........................................56

Figure 6-12: Hydrograph curve comparing in-situ precipitation and corresponding water table. .....................................................................................................................57

Figure 6-13: Comparison of the simulated potential evapotranspiration aggregated to the sequences of in-situ evapotranspiration measurement to the mean over all five evaporation pans. ..................................................................................................57

Figure 6-14: Scatterplot of in-situ potential evapotranspiration versus precipitation both summed over the evapotranspiration measurement sequences. ...........................58

Figure 6-15: Hydrograph curve simulating the years 1995-2009 using the HOBO1-3 daily precipitation and temperature data for Tic_34_mod (Area 4 compared to V-notch weir 1). ..................................................................................................................59

Figure 6-16: Diagram indicating the waterbalance as comparison between simulated and in-situ data as an example for area4. .....................................................................61

Figure 6-17: Simulated compared to in-situ data subdivided by evapotranspiration, precipitation and runoff. .........................................................................................62

Figure 6-18: Simulated precipitation contrary to the sum of actual evapotranspiration and total runoff (area4). ................................................................................................64

Figure 6-19: In-situ water balance with precipitation contrary to scaled reference evapotranspiration and total runoff (area4). ...........................................................65

Figure 7-1: NASA's IMERG data collected for the time period February 23-29, 2016 indicating the surplus of total rainfall over South America which is partly provoked by El Niño. For the study region a total precipitation of about 200 mm is estimated (source: www.nasa.gov [60])..............................................................74

X

List of tables Table 3-1: Advantages and disadvantages of manual calibration and parameter donation. .... 28

Table 4-2: Elevation [m] and distance [km] of the SENAHMI and ERA-Interim locations to the bofedal. ........................................................................................................... 33

Table 4-3: Outline of the available meteorological data including additional information. P: precipitation, T: temperature, RH: relative humidity, S: sunshine duration, W: wind speed. IDW: inverse distance weighting, DIDW: detrended inverse distance weighting, LPR: lapse rate (source: Andres et al. (2014)). ..................................... 34

Table 5-1: Ranking based on a school grade system. The maximum of 50 is marked by the number of measurements performed to date by Jan Baiker. ................................. 42

Table 5-2: Table comparing the well approximating donor sets in the visual and numeric analysis. V: visual; N: numeric. X marks the positive approximation. ..................... 42

Table 5-3: Overview over parameters with best performance range, tested range and mean. It sums up the datasets of best performance. ...................................................... 43

Table 5-4: Summary of important PREVAH location parameters of the Tic_34 region and the study catchment in the Peruvian Andes [98]. ................................................... 43

Table 6-2: Delta storage values [mm/ month] for in-situ compared to simulated data. The last column compares the sum over the 12 months. .............................................. 65

XI

List of abbreviations

Δs Delta storage value as P-(ETR+RGS)

ANS Ampay National Sanctuary

a.s.l. Above sea level

CG1H Storage time for quick baseflow

CTD Electrical conductivity, temperature, water depth

DEM Digital elevation model

EbA Ecosystem-based Adaptation

EC Electrical conductivity

ECMWF European Centre for Medium-Range Weather Forecasts

ENSO El Niño Southern Oscillation (El Niño)

ERA-Interim Global atmospheric reanalysis produced by the European Centre for

Medium-Range Weather Forecasts (ECMWF)

ETP Potential evapotranspiration

ETR Actual (real) evapotranspiration

FAO Food and Agriculture Organization of the United Nations

GWN Percolation to saturated zone

HBV Model concept by Bergström (1976) and Lindström et al. (1997)

HOBO Brand name for certain data logger (providing precipitation and tem-

perature data)

HRU Hydrological response unit

ING National Geographical Institute (of Peru)

K0H Storage time for surface runoff

K1H Storage time for interflow

K2H Storage time for slow baseflow

MOD12Q1 MODIS Land Cover product

MODIS Moderate Resolution Imaging Spectroradiometer

N# Number of subarea (1-5)

NA Not available

NGO non-governmental organization

O In-situ observation

P Precipitation

PACC Programa de Adaptación al Cambio Climático

PERC Percolation rate

PREVAH Precipitation Runoff Evapotranspiration Hydrotope

PUB Prediction in Ungauged Basins initiative

R Runoff or discharge

R0 Surface runoff

R1 Interflow

http://www.fao.org/

XII

R2 Total baseflow

RGS Total runoff

SDC Swiss Agency for Development and Cooperation

SENAHMI Peruvian Meteorological and Hydrological Service

SLZ Lower zone runoff storage

SLZ1MAX Storage limit for fast baseflow storage

SRTM Shuttle Radar Topography Mission

SUZ Upper storage reservoir

V# Number of V-notch weir (1-7)

VWC Volumetric water content

WINMET Meteorological data pre-processing tool in PREVAH

WMO World Meteorologial Organization

WSL Swiss Federal Institute for Forest, Snow and Landscape Research

13

Abstract

Hydrological modelling of ungauged catchments is as challenging as it is relevant. To adequately

transfer tuneable model parameter sets from gauged and calibrated catchments to the location

of interest, research introduced diverse approaches. For the study, parameter donation and rel-

atively short-term runoff measurements are combined. More precisely, tuneable model parame-

ters from various calibrated alpine catchments are transferred to a microscale catchment in the

Peruvian Andes to allow an integrated assessment of the catchment hydrology. The applicability

of the donation across continents and climate zones to successfully approximate in-situ meas-

ured runoff is evaluated. The results are validated with the sparse local information.

The experimental basis for the study are 44 representative catchments in Switzerland and North-

ern Italy, that have been successfully calibrated with the hydrological modelling system PREVAH.

The investigation period of the evapotranspiration, precipitation and runoff in-situ measurements

used for the analysis in the master thesis is March 16th, 2015 to December 29th, 2016, however,

the measurements in the field are still ongoing.

PREVAH requires values of six meteorological variables. Due to the remote location of the study

catchment, only ground station and ERA Interim constitutes to this climatology data. The hydro-

logical similarity between the simulated, climatology based parameters and the in-situ measure-

ments is tested and the sensitivity of the entire hydrological system to the single tuneable param-

eters analyzed. The best approximating donor set is used for a water balance generation.

A comparison of the parameters runoff, evapotranspiration and precipitation functions as an in-

dicator of the storage filling and emptying and the corresponding timing of groundwater regener-

ation or reduction. The main research questions are how the water balance of the catchment can

be described and whether the catchment hydrology can be successfully characterized by an in-

vestigation period of solely 1.75 years.

Results show, that high precipitation occurring in the two austral summer months of December

and February as well as uniform evapotranspiration and total runoff throughout the year, domi-

nate the in-situ water balance. The findings result in a discrepancy from the sinusoidal pattern

visible in the simulated data. The in-situ delta storage values do not nearly achieve the volitional

result of adding up to zero by the end of the year.

In general, it can be concluded that the study catchment is more buffered and reacts more slowly

and inert than indicated by the simulation. The information gained in the region and the progress

in understanding the interaction of its hydrological processes allow future analysis of ecological

and botanical aspects in the catchment which in turn enable a comparison of potential future

climate change adaptation strategies.

A comparison of the in-situ to simulated runoffs and additional measurements in the field support

the assumption of severe sampling issues. As both evapotranspiration and runoff are subject to

high measurement uncertainties, the quality of the overall water balance is also questionable.

Furthermore, the investigation period coincides with the 2015/ 2016 El Niño Southern Oscillation,

which could be one of the reasons for the peak precipitation significantly influencing the balance.

The findings indicate that the parameter donation is an approach that works to a certain extent.

A clear evaluation of its usability is limited by low quality in-situ data especially regarding the

measured runoff. The findings highlight that at least another year of measurement and data anal-

ysis needs to be performed and measurement improvements made.

1 Introduction

14

1 Introduction High data scarcity and low quality constitutes one of the major challenges in hydrology [43]. Mul-

tiscale spatial and temporal heterogeneity of processes across different landscapes are to the

date not fully understood, among others because only a limited number of small and remote

catchments are gauged [43]. The Peruvian Andes comprise an area of the world, which despite

being acknowledged as highly vulnerable to climate change and its resulting impact on humans

and the environment, has only restricted data available [6].

Hydrological data series going back a longer period in time are the starting point for water re-

sources management and water-induced natural hazards restriction [95]. The benefit of even

brief runoff measurements has lately been recognized [95]. Runoff measurements allow the hy-

drological modelling of river basins and catchments. They are an indispensable tool to estimate

the elements of the water cycle in the area of interest [8] and facilitate the assessment of future

climate and land use change effects [65]. Hydrological modelling systems adequately illustrate

the heterogeneity of hydrological processes at various spatial and temporal scales [43]. Their

application is of high relevance albeit being a challenging task [8,95]. In order to target ungauged

catchments, the scientific community developed multiple approaches for transferring tunable

model parameters from a gauged and calibrated donor to an unknown and ungauged receiver

catchment [95].

1.1 Integration of the thesis into the frame of the current research project

The project “Programa de Adaptación al Cambio Climático (PACC Peru)” [89] was initiated by

the Peruvian Ministry of Environment in collaboration with the Swiss Agency for Development

and Cooperation (SDC) as an answer to severe vulnerability of the Peruvian Andes to climate

change, especially in the regions of Cusco and Apurimac [6,89]. PACC unifies a great variety of

different institutions in Peru and Switzerland, including governments on national, regional and

local levels, NGOs, universities and research institutes [73].

The program is subdivided into three main features: water management, disaster prevention and

food security [6,89]. Even though the data basis in the region is scarce [73], an analysis is inevi-

table to examine the imaginable consequences of climate change and quantify the present water

resources [6] in order to initialize resource management decisions that ensure development and

availability of water resources [73]. PACC has a great potential to develop guidelines, tools and

methods for a better handling and adjustment to climate change effects, regardless of the defi-

cient data base in the Peruvian Andes [73].

In-situ data for this master thesis was collected during a research project which is guided by Jan

R. Baiker and which has been implemented and funded by the SDC in the frame of the program

PACC Peru. The fieldwork is currently executed in the Ampay National Sanctuary (ANS), a small

area protected by the Peruvian state, located in the Abancay province, in the north of the Apuri-

mac region [7]. In his research, Jan R. Baiker’s attention is focused on high Andean wetland

ecosystems and the adaption measures taken to confront the negative impacts of global change,

including both climate and anthropogenic effects. One of the research questions is encouraged

by the impact of climate change on the water balance and the vegetation structure of a high-

1 Introduction

15

mountain wetland ecosystem (bofedal) at ANS [100]. The research plan includes a continuous

monitoring phase with data acquisition and analysis, followed by hydrological and ecological/

botanical modelling. In a third phase, a discussion of the results, obtained with steps one and

two in the context of the adaptation strategy called “Ecosystem-based Adaptation (EbA)” [7], is

performed. The input of this master thesis is directed to the analysis of the acquired field data

with primary focus on the water balance factor.

1.2 Motivation and objectives

Before the realization of the PACC program, an analysis of the study site in Peru was limited by

data scarcity. Now, with an extensive measurement system in the bofedal area, the investigation

of various elements of the hydrological cycle is possible.

The hydrological modelling system PREVAH (Precipitation Runoff Evapotranspiration Hydrotope)

[93] is applied to the study catchment, requiring a set of associated runoff generation tuneable

parameters. The optimal set can either be provided by manual calibration – which is featured by

limitations and restrictions caused by the short measurement period – or tuneable parameter

donation from various regions in the Swiss and North Italian Alps to the study catchment in the

Peruvian Andes. The original intent of parameter donation across continents, climate and vege-

tation zones is classified as more promising and hence followed. The first research question of

the thesis is therefore:

Is it possible to successfully approximate in-situ measured runoff using

parameter donation across continents, climate and vegetation zones?

The approach of parameter donation is critically evaluated and its benefits and deficiencies in

the particular application in the subtropical, mountainous and high-elevated study catchment are

lined out.

One aim is to understand the hydrology in the catchment and to analyze its water balance. The

target is to generate a better comprehension of the interconnections between hydrological func-

tions such as the response of the catchment to the input and to the local physical properties. The

work turns from initial sole parameter fitting in order to approximate the in-situ with simulated

runoff in an optimal way, towards an extensive understanding of the processes. Further resulting

research questions are:

How can the water balance in the study catchment be described?

Is it possible to characterize the hydrology in the catchment based on

an investigation period of solely 1.75 years?

The answering of these research questions is of high value, since it constitutes a basis for climate

change estimation and its corresponding impact on the hydrology and ecology in the bofedal.

1.3 Structure of the thesis

To build up the basis to answer the questions above, initially the region including the study catch-

ment is introduced regarding its geology, geomorphology, climate and hydrology. The status of

1 Introduction

16

the scientific research is lined out, evaluating the important hydrological processes of the rainfall-

runoff system and their corresponding empirical relationships. Furthermore, the two approximat-

ing approaches – manual calibration and parameter donation – are presented and evaluated. The

experimental basis with the hydrological modelling system PREVAH and the input data, subdi-

vided by physiographical information, climatology and in-situ data is described in chapter 4. The

complex methodology is introduced in chapter 5. In chapter 6 the results of the investigations are

shown and discussed in chapter 7. The discussion focuses on the following three aspects: ex-

planation of the hydrology in the catchment with the uncertainties and limitations and evaluation

of the performed donor parameter approach. The conclusion in chapter 8 sums up the findings.

2 Introduction to the study site

17

2 Introduction to the study site Peru is located in the West of South America, bordering several countries to the North (Ecuador,

Colombia), East (Brazil, Bolivia) and South (Chile), and limited to the West by the Pacific Ocean.

The Andean Mountains (Andes) are rising parallel to the Pacific Ocean and subdivide the country

into three regions: the costa (coast), which is the narrow and predominantly arid plain in the West,

the sierra (highlands) located directly in the Andes and the selva (jungle) to the East covered by

the Amazon rainforest [45]. Peru has a complex geography and various climatic conditions, which

create an enormous heterogeneity of ecosystems and manifold biodiversity [72].

2.1 Ampay National Sanctuary

Apurimac is a region located in the highlands of southern-central Peru bordered by the Cusco region

in the East and subdivided into seven provinces including Abancay [104] (compare Figure 2-1).

The area around the Nevado or snow-capped Ampay was declared National Sanctuary (ANS) in

1987 and covers an area of 36.36 km2. The altitude ranges between 2900 m and 5235 m above

sea level (a.s.l.). To date the glaciated peak acts as a groundwater supplier and as a regulator

of water quantity.

2.1.1 Geology and geomorphology

The Ampay massif was formed with the rising of the Andes, which began in the Tertiary and

continues during the Quaternary [56]. Structurally it constitutes the Paleozoic sedimentary rocks

and Quaternary deposits. The Paleozoic corresponds to the groups Copacabana and Mito [103].

The Copacabana group forms the main constituent of the mountains while being composed of

gray limestones with intercalations of shales. The texture of the limestones in combination with

weathering and deterioration created the characteristic Karst landscape. The Mito group depos-

its, that appear in the foothill of the Ampay, overlap the Copacabana group. Its red continental

deposits, resulting from erosion of the emerged zones, are composed of coarse-grained sand-

stones, red sandy shales and conglomerates. The Quaternary is represented by glacial moraine

clusters in the middle and lower levels of the mountain. In the deeper subsurface, the rocks of

the Mito group are serving as stagnation layer for the described quaternary materials. A great

part of the plain is covered by fluvial, glacial, alluvial and eluvial quaternary deposits especially

in depressions [103].

The geomorphology has mainly been generated by glacial and river erosive activities. Traces of

the glaciations can be observed above 3500 m a.s.l. by U-shaped valleys and morainic deposits

[56]. Furthermore, a large number of hanging valleys are visible.

2.1.2 Climate and hydrology

Due to the topography and existence of several elevation levels, climate characteristics are highly

variable in space and time in the Ampay National Sanctury (ANS). While the capital of the region,

Abancay, has a climate with average temperatures of 18°C, the region between 2300-3600 m

a.s.l. reaches temperatures between 11 and 16°C, the area between 4000-4800 m a.s.l. is char-

acterized by the cold climates of the Puna region with temperatures of 0° to 10°C and the region

above 5100 m a.s.l. has a very cold climate with snowfall and temperatures below 0°C [103].


18

Figure 2-1: Overview of the country of Peru with a detailed map showing the Ampay National Sanctuary`s core area (blue), the study catchment (red) and the bofedal area (green) Additionally visible: Ampay glacier in the NW of the ANS and the location of the closest two SENAHMI weather stations and the closest ERA Interim location.


19

In general, the Sanctuary shows a very distinct seasonality with the rainy season between No-

vember and April (austral summer) and the dry season between May and October (austral win-

ter). During June to September low temperatures occur – accompanied by infrequent frost events.

From September to December first precipitation occurs and moderate temperatures at an aver-

age of 14°C allow the greening of the vegetation. The peak of the rainy season is between Jan-

uary and March [103].

Around the Ampay massif the hydrological activity is the most important relief shaping aspect.

The water drains through the hydrographic ridges in form of streams, composing waterfalls and

lagoons. The runoff has a seasonal behavior with the highest observable volume of water be-

tween January and March (peak of the rainy season) [103].

2.2 Study catchment

For the hydrological analysis, a catchment expanding to an area of ca. 4.4 km2 is considered.

The area is enclosed by the mountain ridge to the North, West and East and by a topographic

step equipped by a waterfall to the South. The catchment is subdivided into five different subar-

eas primarily based on the topography. Figure 2-2 shows the relevant hydrological catchment in

a 3D perspective. The upper region (N4-yellow and N3-light green) feeding the bofedal (N5-red)

covers roughly 2.0 km2 between an elevation of 4000 and 4600 m a.s.l. [7]. The entire hydrolog-

ical area comprises the aforementioned region and two areas in the lower part (N2-dark green,

N1-dark blue). Figure 2-2 furthermore indicates the location of the v-notch weirs (marked as V#)

important for the analysis (for details on measurement devices please refer to chapter 4.2.3).

The area is minted by a strong seasonality, which is shown in Figure 2-3 from a southern looking

perspective. The comparison of the pictures, taken in the different months, clearly indicates a

difference in greenness and thereby of the soil moisture content [7].

Figure 2-2: Catchment of all five subareas from a northern looking perspective including location and abbreviations of the v-notch weirs and acronyms for the subareas (source: Google Earth).


20

The term “bofedal” describes areas of wetland vegetation in the Peruvian Andes. According to

the Ramsar Convention classification of wetland types, bofedales are defined as “peatlands with-

out forests” [14]. The Peruvian General Environmental Law identifies bofedales as fragile eco-

systems (Law No. 28611, Article 99) [28]. Bofedales are located in areas that receive water ad-

ditionally to precipitation from rivers, lakes, underground aquifers or glaciers and store it in the

upper basins of the cordillera. As the runoff from the wetlands is slow and filtered, this ecosystem

functions as a regulator of the downhill water flux. Even though bofedales are not able to hold as

much water as glaciers for instance, they yet take an important role in storage capacity [28]. The

wetlands form at constant year-round edaphic humidity in flat areas and contrast to the surround-

ing drier land with their greener appearance throughout the year [29] (Figure 2-3).

Figure 2-3: The bofedal from a southern looking perspective during different months of the year indicating the varia-bility in greenness and soil moisture in the bofedal (source: Jan R. Baiker).

3 Status of scientific research

21

3 Status of scientific research Hydrological simulations require the classification of various hydrological processes such as run-

off, evapotranspiration or evaporation and the understanding of their relationship is important. In

the first part of the description, considering the research status, the fundamental relations of all

relevant processes are presented and discussed.

In the second part, the analytical approaches used in secondary literature for poorly gauged

catchments are lined out and the two main approaches namely manual calibration versus tune-

able parameter donation are evaluated.

3.1 Hydrological processes of the rainfall-runoff system

In a process-based model the combination of various components contribute to the overall dy-

namics of the hydrological system [27]. Figure 3-1 gives a supporting overview of the processes

and zones mentioned here in the following.

3.1.1 Precipitation

Precipitation is the condensation of atmospheric moisture that falls onto the earth in form of rain,

sleet or snow. Precipitation restocks water bodies on the surface, renews soil moisture and re-

plenishes aquifers [55,93].

In the ventilated soil zone the precipitation that infiltrates directly or with delay can follow a num-

ber of options. It may form quick interflow or be slowed down by the top soil layer to create

delayed interflow. The precipitation can also be stored for a shorter or longer time in the ventilated

soil zone. Following gravity the water may also percolate to the groundwater zone or be taken

out from the soil by evapotranspiration [93]. The share of the precipitation that actually reaches

the groundwater can form a baseflow by running off more or less delayed to the stream network.

The water can either stay in the groundwater zone for a long time, return to the aforementioned

ventilated zone through capillary rise or even percolate to deeper but inactive soil zones. Plants

and humans also have an impact on the hydrological cycle – humans by extracting the ground-

water for multiple purposes and plants by transpiration [93].

3.1.2 Evaporation

Evaporation, defined as the process of converting liquid water into a gaseous state, occurs from

bare soils or open water. Evaporation postulates the availability of water and atmospheric humid-

ity below the humidity of the atmospheric surface [93]. Unlike precipitation, runoff and soil mois-

ture, evaporation cannot be measured directly [82]. The parameter can be subdivided into po-

tential and actual evaporation.

Potential evaporation is defined as the amount of water evaporating from an extensive free and

ideal water surface at given atmospheric conditions such as radiation, temperature, humidity and

wind.


22

Actual evaporation is the evaporation generated from a surface under certain climatic conditions.

The actual evaporation can only be determined indirectly from water budget equations or as an

analysis result of the relation to other climatic factors [82].

The measurement of evaporation is potentially affected by the perturbation of the background

conditions. However, pan evaporation measurements are still regarded as an acceptable esti-

mator and as a reliable indicator of the evaporation variation [82].

3.1.3 Evapotranspiration

Definition

Just like evaporation, evapotranspiration is the transformation of water into atmospheric water

vapor. It is a combination of evaporation processes from plant and water surfaces as well as soil

and transpiration through plant canopies. This parameter is a significant factor for the energy and

water balance of a surface as it comprises more than half the energy flux from the Earth surface

to the atmosphere [92]. The link between evapotranspiration and the amount of runoff is guided

by the plant-available water content in the surface layers of the soil [93]. Evapotranspiration de-

creases constantly over dry periods [92].

Potential evapotranspiration (ETP) is the evapotranspiration originating from an idealized vege-

tated surface with satisfactory moisture availability at all times [46,87,93]. A reference potential

plant evapotranspiration value measured in [mm d-1] is defined “as the evapotranspiration that

occurs from an idealized grassy surface with a vegetation height of 12cm, an albedo of 0.23 and

a surface resistance of 69 sm-1” [92].

Real or actual evapotranspiration (ETR) is influenced by various factors subdividable into weather

parameters, vegetation factors, environmental conditions (salinity and fertility) and further site-

Figure 3-1: Water cycle with important spatial subunits and hydrological processes for the hydrological simulation in PREVAH (diagram changed after [18]).


23

specific parameters (e.g. geology, soil, vegetation type). The weather parameters include radia-

tion, humidity, air temperature and wind speed. Resistance to transpiration, vegetation height and

roughness, reflection, root characteristics and plant density all in combination with unique envi-

ronmental conditions lead to a certain amount of evapotranspiration. Cultivation and irrigation

affect the microclimate and vegetation characteristics as well as measures such as windbreaks.

The effect of the latter may significantly express itself under dry, warm and windy conditions [3].

Estimation of evapotranspiration

The estimation of evapotranspiration in distributed hydrological models relies on approved and

widely adopted equations introduced by e.g. Penman (1948) or Penman-Monteith (1975, 1981)

[62,66,67].

Potential evapotranspiration

The Penman (1948) equation for potential evapotranspiration is defined by

𝐸𝑇𝑃 = 𝑈𝐹𝐾

𝜌𝑤𝐿

∆×𝑅𝑁 + 𝛾 + 𝐸𝑎

∆+𝛾 using 𝐸𝑎 = 0.263 (0.5 + 0.537 𝑢) (𝑒𝑠 − 𝑒)

𝐿

86400

where

𝐸𝑇𝑃 Potential evapotranspiration [mm d-1] 𝑈𝐹𝐾 86 400 000 (conversion from ms-1 to mm d-1)

𝐸𝑎 Aerodynamic term [W m-2], ventilation 𝐿 Latent evaporation heat [Jkg-1 K-1]

𝜌𝑤 Density of water ≅ 1000 [kg m-3] 𝑢 Wind speed 2 m above ground [m s-1]

𝛾 Psychromatic constant [hPa K-1] 𝑒 Actual water vapor pressure [hPa]

∆ First derivative of the saturated vapor pressure curve (around temperature 𝑇) [hPa K-1]

𝑒𝑠 Saturated vapor pressure at actual temperature [hPa]

𝑅𝑁 Net radiation [W m-2]

A detailed observation indicates that each of the parameters is based on an equation, which is

dependent on the air temperature. In the PREVAH model the potential evapotranspiration is de-

termined as average daily values in [mm/ d] [93].

Actual evapotranspiration

Actual evapotranspiration is directly related to the available moisture content. During dry periods

evapotranspiration is reduced constantly and actual evapotranspiration possibly drops substan-

tially below the value of the potential evapotranspiration. According to Viviroli et al. (2007) the

process can be described best with the extraction or reduction function 𝑟(Θ):

𝐸𝑇𝑅 = 𝑟(Θ) ×𝐸𝑇𝑃

For simplification reasons, a stepwise linearization dependent on the dimension of the volumetric

soil moisture content Θ is used subdivided by covered (𝑟𝑏) and uncovered (𝑟𝑢) soils.

If the vegetation-covering grade 𝑉𝐵𝐺 (0 ≤ 𝑉𝐵𝐺 ≤ 1) defines the share of plant-covered soil, the

𝐸𝑇𝑅 (𝑏 + 𝑢) of partially covered soil surface can be approximated with the following equation:

𝐸𝑇𝑅(𝑏 + 𝑢) = 𝐸𝑇𝑃[(𝑟𝑏(Θ)×𝑉𝐵𝐺) + (𝑟𝑢(Θ)×(1 − 𝑉𝐵𝐺))] [93].


24

3.1.4 Empirical parameter relationships

In the beginning of the 20th century, the relation between precipitation, runoff and evaporation

was first acknowledged by hydrologists [17,74]. Several particular parameter relationships are

described to date even though large-scale observations of evapotranspiration are not available

and make a global portioning into transpiration, soil evaporation and canopy evaporation very

challenging [51]. While on uncovered soil evaporation appears exclusively, on covered soil the

share of transpiration increases with increasing vegetation density. Consequently the fraction of

evaporation is reducing to almost zero for a closed vegetation cover [92]. Because evaporation

and evapotranspiration appear simultaneously, differentiation between the two processes is par-

ticularly challenging [3]. The few available measurements suggest, that transpiration is the pre-

dominant element on global scale, followed by soil and canopy evaporation [51]. Observations

by Baumgartner and Reichel (1975), that are quoted to date stress, that actual evapotranspiration

returns about two third of long-term annual precipitation back to the atmosphere [9,57,86]. Actual

evapotranspiration varies between 90% in Australia and 60% in Europe [9,57]. However, evapo-

transpiration was often not adequately studied in catchment and water balance investigations

[57]. The low temperatures in high-elevation mountain regions and in tropical alpine regions also

the high frequency of fog, cloud cover and humidity, significantly reduce evapotranspiration rates

resulting in high water production [21]. According to Wang and Zlotnik (2012) with high available

moisture content, actual evapotranspiration can equal potential evapotranspiration [98].

3.2 Manual calibration versus parameter donation

Making predictions in ungauged basins is an indisputable challenge for the hydrologic community

[81]. Two different approaches are evaluated here regarding their potential of approximation:

manual calibration and parameter donation.

3.2.1 Manual calibration

All rainfall-runoff models require calibration to adjust the free parameters to the characteristics of

a study basin and imitate the particular hydrological behavior. Streamflow data is essential for

their calibration. However, few recommendations or guidelines are being made on the needed

duration or number of single streamflow records. The scientific community shares widely different

conception on the requirements [69]. There are further efficiency differences depending on the

season, regime type and size of the study catchment.

Sampling strategies

Various sampling strategies for manual calibration are used. Not all parts of the hydrograph and

all measurements are equally informative to estimate model parameters in demand [69]. Seibert

and Beven (2009) found that strategies including sampling of maximum flows performed better

than those involving minimum or mean flows [80]. In their paper issued in 2015 they furthermore

mention, that targeted gauging of high discharge events are more helpful for model development,

calibration and its use [81]. Moreover it was found that additional data to the plain streamflow

observations provide beneficial information for a calibration process especially when based on

limited discharge data [47,49,81]. Such soft data may include soil depth, bedrock permeability or

the catchments flow signal and flow source components [81,81].


25

Number of needed data points for calibration

A review on calibration period and point data requirements indicates that opinions strongly differ

regarding the minimum requirements or maximum number to achieve calibration improvements.

The generalization is limited due to the fact that most studies are solely carried out with a single

model on a single catchment [69]. Sorooshian (1983) suggests a minimum of a full hydrological

cycle corresponding to one entire year of data as a minimum requirement for calibration and most

existing studies seem to adopt without reconsideration [69,83].

While some papers illustrate limitation to few measurements as reliable, others support the opin-

ion of Sorooshian (1983). Viviroli and Seibert (2015) postulate that even a few short measure-

ments of mean runoff possibly allow models of higher efficiency than those based solely on hy-

drological similarity [95]. They even support the fact that the improvements in data-scarce regions

might even be higher [95]. According to Seibert and Beven (2009) a plateau is reached with 32

runoff observations so that additional observations will not help to improve the model perfor-

mance. About 6 to 16 streamflow measurements even helped to constrain model calibration [80].

Perrin (2007) confirms the results provided by Brath et al. (2004) that reliable estimates could even

be achieved with 50 calibration points of well-chosen measurements in a streamflow sequence

[16,69]. In some cases, results achieved with only 10 streamflow measurements for calibration

were still acceptable. He furthermore came to the conclusion, that informative streamflow data is

not limited to high flood events [69].

However, many measurements were necessary to exceed the efficiency of the extended region-

alization scheme introduced by Viviroli et al. (2009) [94]. According to Perrin (2007) 350 calibra-

tion days sampled out of a longer dataset holding wet and dry background conditions suffices to

acquire reliable estimates of model parameters [69]. Additionally, the number of necessary point

measurements is strongly dependent on the model used – the more complex the model, the more

measurement data is needed [69].

Duration of needed data for calibration

A study by Seibert and Beven (2015) indicates, that one high flow event is nearly as informative

as a three month continuous streamflow measurement [81]. The result is surprising since it is

normally assumed, that the longer the time series, the better the calibration efficiency. Still, it is

encouraging that a hydrologically intelligent choice towards a small number of observations com-

pared to a regularly or randomly choice achieves positive outcome [80]. Whenever subsets of

runoff data are used for calibration, good performances are achieved for entire events and par-

ticularly for larger events [81].

Especially rainfall-dominated catchments profit by longer measurements because of rapid

changes of their runoff [95]. Earlier studies for example by Brath et al. (2004) and Perrin (2007)

showed that, to efficiently calibrate a hydrological model, about one year of data is needed

[16,69]. Continuous data is especially valuable as it provides information on the recession which

is a powerful tool for successful calibration [81].

Although longer duration measurements are slightly more efficient, other factors such as the

season of data acquisition or the type of flow regime had a stronger influence than the duration [95].


26

Further influences on the performance of a manual calibration

Improvements mainly depend on the regime characteristics of the study catchment and season

of data acquisition. The most suitable season for measurements varies depending on the regime

type. In pluvial regimes, rainfall is the dominant factor throughout the entire year and the temporal

differences of rainfall may vary significantly even in the course of one year. Therefore, limiting

measurements to one season is less effective. For other catchments that are dominated by snow

and ice melt for instance, even two measurements during spring or summer significantly improve

the efficiency [95]. Furthermore, a tendency toward poorer calibration results for smaller catch-

ments were observed [80]. In the study by Seibert and Beven (2009) the smallest catchments

measured a size of 6.6 and 14 km2 [80]. It was additionally noted, that field measurements are

especially valuable in mountain areas [95]. The high value of short field campaigns is intensified

by the limited meteorological stations at high elevation [68,91].

3.2.2 Parameter donation

Predicting streamflow at ungauged catchments gives the opportunity to transfer the hydrological

information from a donating gauged to a receiving ungauged catchment – a process that may

also be called regionalization [15,48,65]. Regionalization in hydrology encountered outstanding

development through the Prediction in Ungauged Basins (PUB) initiative (2003-2012) of the In-

ternational Association of Hydrological Sciences [43].

The approach requires two steps: after identification of a donor catchment of high hydrological

similarity, the relevant information, either model parameters or streamflow values, is transferred [65].

Similarity may be defined by either physical (e.g. soil type, topography) or spatial (distance) prox-

imity measures [65]. Both similarity approaches have their advantages and disadvantages: spa-

tial proximity hinders high hydrologic similarity of catchments of high distance but is regarded as

one of the most reliable methods [61,64,114]. Furthermore, the limitation of spatial transfer over

various climatic and geographic regions is uncertain [65].

Attributes transformed during regionalization should represent the hydrological response of a

catchment including good water balances and sufficiently reproduce the variability of the daily

discharge [8,48]. As the hydrological behavior of the ungauged catchment may only be assumed,

the identification of hydrological similarity proves to be challenging [65]. Also model parameters

occur as complementary parameter values and changes to one parameter may be compensated

by another [8]. Different sets of model parameters lead to similar performance and make the

identification of one best unique dataset virtually impossible [12].

Sampling strategies

Mainly three different parameter regionalization approaches are developed that are applicable

whenever manual calibration is impossible: nearest neighbor, kriging and regression [94].

Regression, being the predominantly used transferring technique, is the parameter estimation

from relations to the catchment attributes [48,94]. Regression was performed on multiple occa-

sions [1,48] even though the transfer is hampered as summed up in Bárdossy (2007), because:

▪ optimized parameter sets are dependent on the particular model used [35,54]

▪ the uncertainty of the parameters themselves [50]

▪ the equifinality of possible parameter sets reaching similar model performances [12].


27

Kriging is the parameter interpolation in physical space for each calibrated model parameter in-

dependently from each other [94].

The nearest neighbor approach is defined as the transfer of the parameters from catchments

with similar attribute space [94]. After finding a calibrated donor catchment with high similarity

(distance or behavior) to the ungauged receiver catchment, the tuneable model parameters are

transferred as a complete set [94].

Efficiency

Several factors tend to lead to good approximation when performing tuneable parameter dona-

tion. Smaller distances or closer proximity are likely to achieve good approximation [65]. How-

ever, contrary to expectations, high gauging density in the vicinity of an ungauged catchment

does not automatically guarantee positive predictability [65]. Furthermore, the majority of high

predictability catchments are characterized by low values of evaporation and aridity indices,

higher channel slopes exceeding 1%, high permeability, forest density and distinct topography

[65]. This corresponds to humid mountainous regions. Many well predicted catchments have high

values of baseflow, but the trend is not especially significant [65]. Accordingly, spatial proximity

alone fails to explain good predictability and needs the contribution of regional climate variability

and geology [65]. Limited quality of donation may be explained by one of the following three

aspects: too high distance between donor and receiver catchment, too high spatial climatic vari-

ability around the receiver catchment or idiosyncrasy of the ungauged catchment due to deep

groundwater sources, karst or deficit of water due to regional aquifers [65].

3.2.3 Evaluation of the two approaches

Several advantages and disadvantages of manual calibration and parameter donation follow di-

rectly from the literature review in chapter 3.2.1 and 3.2.2. The most prominent ones are sum-

marized in Table 3-1.

The improvements of manual calibration in comparison to plain hydrological similarity analysis in

data-scarce regions, based on some single streamflow measurements, are expected to be higher

than in data-rich regions. Soft data such as soil depth or bedrock permeability may provide addi-

tional beneficial information and are easily collected in the field. Several papers mention a low

number of needed measurements (6, 10, 16, 32, 50) for acceptable efficiency.

Quantity and quality of runoff data significantly influence manual calibration and the robustness

of the model. The number of necessary point measurements is strongly dependent on the model

used. Simpler models need fewer measurements for calibration. PREVAH for instance is a rather

complex model. Furthermore, maximum flows performed well and one high flow event is almost

as informative as three months of continuous measurements. This indicates however, that a clear

evaluation of the possible highest magnitudes is postulated, which is not necessarily available

for remote catchments with discontinuous measurements. As in pluvial systems the temporal

difference of rainfall varies significantly and rapidly, calibration of limited data is questionable and

longer measurements preferable. High uncertainty and possible errors of the parameters them-

selves as well as poorer calibration results for smaller catchments especially hinder promising

manual calibration.


28

Parameter donation can be based on the two factors proximity and behavior. In case that prox-

imity is prohibited by the intention of transferring parameters across continents, hydrological sim-

ilarity has to be achieved. This hydrological similarity and information on the water balance and

variability of daily discharge is however hard to evaluate for to the date ungauged catchments.

The “trial and error” approach of testing best approximation in parameter donation is time-con-

suming. Different sets may achieve similar performance and this equifinality makes the approach

so challenging. Parameter donation is dependent on the particular model used and on the un-

certainty of the parameters themselves. The transferability over various climatic and geographic

regions is questionable.

The dismissal of donor catchments with strong differences is a strong advantage of the approach.

Also, good predictability was achieved for catchments equipped with low values of evapotranspi-

ration, high channel slopes, high permeability, distinct topography and high values of baseflow.

The original intent of the thesis was to try parameter donation of tuneable parameter sets from

Switzerland to Peru. After testing and performing a first sensitivity analysis, also manual calibra-

tion came into the focus as potential alternative. However, it was abandoned due to all afore-

mentioned disadvantages of manual calibration. The main disadvantages are that this different

approach does not follow the original intent of the thesis and using 50 point measurements for

calibration leaves no data for validation purposes.

Table 3-1: Advantages and disadvantages of manual calibration and parameter donation.

Advantage Disadvantage

Manual

calibration

▪ improvements with a few short meas-

urements especially in data-scarce

regions [95]

▪ acceptable efficiency with low num-

ber of measurements (postulated in

several papers)

▪ even more significant improvements

in case of year-round available

measurements [95]

▪ strongly dependent on quantity and quality of the data

▪ longer measurements needed for higher efficiency; es-

pecially for more complex models [69]

▪ questionable applicability for remote catchments with

uncertain highest peak magnitudes

▪ longer measurement periods needed for high variable

pluvial regimes [95]

▪ contrary to original intent of master thesis

▪ additional requirements of measurement data for vali-

dation purposes

▪ dependence on the size of the catchment: poorer cali-

bration results for smaller catchments

Parameter

donation

▪ Rejecting completely different hydro-

logical systems

▪ high performance for humid, moun-

tainous regions (postulated in [65])

▪ hydrological similarity only understandable after testing

▪ time-consuming trial-and-error approach in case of sev-

eral different donor sets

▪ equifinality of donor sets [12]

▪ transferability over climatic and geographic regions

questionable

4 Experimental basis

29

4 Experimental basis In the first part of the chapter, the hydrological simulation system PREVAH is described and its

runoff formation module including the single tuneable parameters. The chosen PREVAH options,

regarding evapotranspiration and other modification opportunities, are subsequently shortly lined

out.

The second part of the chapter gives an overview of the available input data subdivided into

physiographical information, climatology and in-situ data.

4.1 Hydrological model system PREVAH

Hydrological models based on meteorological data are mandatory for a substantial simulation of

interacting hydrological processes at catchment scale [93] and they proved to provide valuable

information allowing an estimation of water resources [6,77]. Spatially distributed modelling

turned into an acknowledged device for investigating the elements of and potential modifications

to the hydrological system [93]. All simulations for this study have been performed with the spa-

tially distributed hydrological simulation model PREVAH (Precipitation Runoff Evapotranspiration

Hydrotope) [36,95]. PREVAH is predicated on the concept of hydrological response units (HRUs)

[37]. A detailed outline of the model structure is out of the scope of this master thesis and only a

brief introduction is given here. For more details of the model’s physics, structure and parame-

terization please refer to Viviroli et al. (2007).

4.1.1 General information

PREVAH is a semi-distributed hydrological catchment modelling system based on a conceptual

approach with a modular set-up. The original intent lies in an improved understanding of hydro-

logical processes in catchments with complex mountainous topographies and immense spatial and

temporal variability [93]. PREVAH has an expansive record of successful utilization therein [96].

The efficient and dynamic spatial discretization of PREVAH is based on the subdivision of gridded

spatial information into clusters of similar hydrological response, the hydrological response units

(HRU) [93]. The generation of hydrological response units (HRUs) is based on the aggregation

of hydrologically equal surface units, which includes the analysis of the topography of the catch-

ment based on the spatial or physiographical information [93]. The similar or homogenous hy-

drological response is primarily directed to the factors with impact on important hydrological pro-

cesses such as evapotranspiration and runoff-generation (compare chapter 3.1). The size of el-

evation zones is selected with respect to the intended spatial model resolution and quality of the

meteorological data available – in this thesis, 50 m elevation zones are chosen. As a result of the

process, each HRU is attributed with a set of parameters containing all the information. It is stored

in a table and assimilated by PREVAH during the model initialization. HRUs may be featured with

an irregular shape and small size in areas with high spatial variability of soil, land surface and

topography [93].

Three types of input data are required to run PREVAH: physiographical information containing

the physiographical properties of each hydrological response unit (HRU), meteorological input

for the altitude zones and a control file containing site-specific information needed for modelling,


30

such as the tuneable model parameter values. The latter, also referred to as calibration param-

eters, are used to adjust the model to the prevailing conditions in the explicit catchment.

4.1.2 Runoff generation tuneable parameters

Some parameters are defined beforehand through the characteristics of the physiogeographical

basin and values extracted from literature. However a number of tuneable parameters need to be

adjusted in PREVAH to fit the particular modelling site [96].

They are subdivided into the following six groups: water balance adjustment, differentiation of

precipitation into liquid and solid, snowmelt module, glacier module, soil moisture module and

runoff generation module [93] (compare with model structure in Figure 4-1). The latter is intro-

duced in detail below.

The tuneable parameters of the runoff generation module are used in the sensitivity analysis

(chapter 5.4 and 6.1) with the primary goal to find the optimal set of parameters that achieves

maximum correspondence between observed and simulated hydrographs. Runoff (R or Q) is

defined as water volume per time unit ([m3 s-1], [l s-1]) which leaves the catchment through surface

and subsurface ways [93].

The runoff generation module relies on the HBV model concept by Bergström (1976) and

Lindström et al. 1997 [10,53]. In contrast to the HBV model, the runoff generation in PREVAH

has a spatially distributed representation (see [36,38]). The runoff generation in the soil’s unsatu-

rated zone is affected by storage times for surface runoff (K0H [h]) and interflow (K1H [h]) [96].

The baseflow is generated by the combination of two linear groundwater reservoirs [78] with a

fast and a delayed component, described by two explicit storage times (CG1H [h] and K2H [h])

[96]. The two groundwater reservoir approaches adopted by PREVAH allows high flexibility and

good simulation performance during dry periods [38].

The storage threshold SGR [mm] describes the generation of surface runoff, while the percolation

rate (PERC [mm ΔT-1]) and the storage limit for the fast baseflow storage (SLZ1MAX [mm]) con-

trols the flux from the unsaturated to the saturated soil zone [96].

4.1.3 Parameter estimation

Calculation of evapotranspiration in PREVAH

Particular attention is granted to the evaluation of evapotranspiration in PREVAH. In the PREVAH

model one may choose between several methods for the computing of evapotranspiration: po-

tential evapotranspiration with the Penman (1948), Penman-Monteith (1975, 1981), Hamon

(1961), Turc (1961) and Wendling (1975) method [39,62,66,67,88,99].

Depending on the selection of the implemented evapotranspiration method, different meteoro-

logical input is obligatory. Table 4-1 indicates that Penman-Monteith requires most meteorological

parameters including precipitation, air temperature, global radiation, wind speed, water vapor

pressure, relative humidity and sunshine duration. By default, the analysis in PREVAH is run with

the regular Penman approach.


31

Table 4-1: Meteorological requirements for different evapotranspiration approaches implemented in PREVAH.

Variables Penman-Monteith Hamon Turc Wendling

Precipitation required required required required

Air temperature required required required required

Global radiation required required required

Wind speed required

Water vapour pressure/ relative humidity required

Sunshine duration required

Further parameter settings

Some additional parameter settings are chosen in WINPREVAH before performing the simula-

tion. For a detailed explanation please refer to Viviroli et al. (2007) [93].

For the simulation in PREVAH, as typically, a “single full run” configuration is chosen which

means a run of the simulation is performed for the selected catchment between the given start

and end dates. For the snowmelt module the version according to Hock (1998) with constant

melt factor is used [40]. For a detailed analysis please refer to Zappa et al. 2003 [111]. The daily

Penman method is selected for computing snow evaporation.

Figure 4-1: Schematic diagram of the PREVAH structure including tuneable parameters, storage modules and hy-drological fluxes (source: Viviroli et al. 2009 [96]).


32

Furthermore with Penman (1954) a method for estimating the short-wave and long-wave radiation

budget is chosen [22,75]. Penman 1954 is used for the estimation of the aerodynamic term as well.

4.2 Model input data

The available input data is subdivided into physiographical information, climatology background

data and in-situ measurement data. The meteorological and physiographic data are both directly

pre-processed in PREVAH [93,96].

4.2.1 Physiographical information

The basic parametrization of PREVAH is based on the topographical analysis of a digital eleva-

tion model (DEM), on land cover and use characteristics and on soil type maps [95]. While the

DEMs and land use data offer relatively high resolution and accuracy, the soil types and their

corresponding soil properties are only estimates [93].

Digital elevation model

The digital elevation model (DEM) constitutes one of the most substantial data sets for hydrolog-

ical modelling systems, especially for the application to mountainous catchments. The DEM ac-

quired by the Shuttle Radar Topography Mission (SRTM) provides a sight of view resolution of

three arc seconds corresponding to approximately 30 m. DEMs help to determine terrain slope

and aspect but also allow hydrological estimates for e.g. flow directions, flow accumulations or

the river network [93].

Land use and land cover map

Land use or land cover can be identified and determined through digitalization, analysis and sub-

sequent interpretation of remote sensing images. Most land use information is provided by the

MODIS Land Cover product (MOD12Q1) [30]. The land use is applied for the parametrization of

several vegetation specific variables that are strongly linked to important processes within the

hydrological cycle such as surface roughness, vegetation density, soil heat flux, available water

content and maximum interception storage [93].

Soil map

Soil information supports the modelling of the interaction between atmosphere, soil and vegeta-

tion and therefore provides important input to the models in terms of crop growth simulation,

water balance and environmental impact estimation. Digital soil maps allow an a priori parametri-

zation of crucial soil specific variables such as plant available field capacity, soil depth and hy-

draulic conductivity [93]. Soil information is extracted from the FAO Soil Map of the World [25,60].

Additionally, in-field experiences by Jan R. Baiker are integrated in the soil determination of the

microscale catchment.

4.2.2 Climatology

The hydrological model PREVAH requires hourly values of six meteorological variables namely

precipitation, air temperature, wind speed, global radiation, relative humidity and relative sun-

shine duration (see Table 4-1). The climatology data collected for Andres et al. (2014) is used


33

here as well as input data for the PREVAH simulations of this master thesis. Due to the remote

location of the study catchment, for the master thesis solely the ground station and ERA Interim

data is extracted from the data pool and used as input to the model [6].

Ground station data

The PACC program incorporates the creation of a data portal with historic climate data gathered

from more than 100 SENAHMI (Peruvian Meteorological and Hydrological Service) stations

around Cuzco (see also [77]). Out of the existing data base, thirty-six stations were selected

containing homogenized daily data on precipitation, temperature and relative humidity for the

time span 1960 to 2009 [6]. Interpolation results in gridded meteorological data with a 540 m

spatial resolution based on the available set of ground station data and a daily time resolution

[96]. The elevation and distance to the study catchment of the closest stations can be extracted

from Table 4-2.

ERA-Interim

ERA-Interim represents the reanalysis of global surface and atmosphere conditions during the

period 1979 to the present [11] performed by the European Centre for Medium-Range Weather

Forecasts (ECMWF) [6]. The dataset offers a multivariate, spatially complete and coherent record

of the global atmospheric circulation. For the period between 1998 and 2009, daily resolution

data for the gridded sunshine duration and wind speed were received from ECMWF [6]. The eleva-

tion and distance to the study catchment of the closest stations can be extracted from Table 4-2.

Table 4-2: Elevation [m] and distance [km] of the closest SENAHMI (S) and ERA-Interim (E) locations to the study area.

S1 ERA1 S2 ERA2 S3 S4

Distance 14.3 15.5 18.1 40.2 54.7 70.5

Elevation 2868 3246 2954 2305 2578 3220

Spatial and temporal interpolation of meteorological input data

PREVAH operates with data sets of hourly resolution. In the case that only daily resolution is

available as input data, intermediate values are generated through interpolation. For the param-

eters air temperature, precipitation, wind speed, water vapor pressure and relative humidity, 24

identical values are assumed [93]. The approach for the interpolation of the global radiation re-

quires a division between sunrise and sunset, depending on the calculated potential clear-sky

direct radiation [41,76].

As the available data does not cover the entire domain of interest but exists as point measure-

ments, spatial interpolation becomes necessary. The PREVAH tool WINMET has been specially

developed to easily interpolate meteorological data [93]. This procedural method is motivated by

the observation, that normally spatial points closer to one another are more likely to show the

same magnitudes of the parameter than points further apart [19]. Spatial interpolation transforms

data, derived from a set of sample points such as various rainfall stations, to a continuous and

discretized surface [93]. Exact interpolators as one kind of interpolation methods predict an at-

tribute value at a sample point that is identical to the measured one. The exact interpolators used

on the ground station meteorological data are inverse distance weighting and detrended inverse

distance weighting [93]. ERA-Interim data is interpolated with inverse distance weighting and

lapse rate (Table 4-3).


34

Table 4-3: Outline of the used meteorological data including additional information. P: precipitation, T: temperature, RH: relative humidity, S: sunshine duration, W: wind speed. IDW: inverse distance weighting, DIDW: detrended inverse distance weighting, LPR: lapse rate (source: Andres et al. (2014)).

Ground station data ERA-Interim

Available data P, T, RH P, T, RH, S, W

Time resolution daily daily

Interpolation to grid IDW, DIWD IDW; LPR

Further processing averaged to meteorological subunits averaged to meteorological subunits

Available time period 1960-2009 1998-2009

4.2.3 In-situ measurements

Various instruments were installed over the catchment in order to specify changes. With these

instruments, different important hydrological parameters are determined either through continu-

ous or periodic measurements. Figure 4-2 gives an overview of most installed measurement

devices in the catchment [7].

Continuous measurements

Continuous measurements are performed every five minutes or at least every time a precipitation

event occurs. They comprise precipitation, air and soil temperature, groundwater table, volumet-

ric water content (VWC) and electrical conductivity (EC) data. The precipitation (liquid and solid)

is determined with data logging rain gauges called HOBOs hereafter (HOBO® is a registered

trademark of the manufacturer). The HOBOs are also used to determine the air temperature at

1 m above the ground. The HOBO temperature and precipitation data is transformed to hourly

and daily mean, maximum and minimum values. The VWC is acquired by automatic sensors.

Perforated piezometers are used to manually quantify the groundwater table. Furthermore a CTD

(Electrical Conductivity, Temperature, Water Depth) automatic sensor is fixed approximately 1 m

below the soil surface in one of the botanical plots [7].

Periodic measurements

Periodic in-situ measurements, performed every two to four weeks, include the runoff, the evap-

oration at ground level, the soil moisture, the water table (manually measured with a well whistle)

and precipitation with manual accumulative rain gauges (totalizators). The runoff is measured

with triangular or v-shaped Thomson v-notch weirs using buckets measurement techniques in

four locations (V1,3,5,6). Parshall weirs are calibrated in the laboratory and the runoff can be

directly read out. They are used in flat terrain where no overflowing allows bucket measurements

and are well applicable to small runoffs (V2,4,7) (compare Figure 4-3). In the following for simpli-

fication reasons, all weirs are referred to as v-notch weirs.

The seven implemented v-notch weirs can be subdivided according to their location in the bofedal

into surface inflow weirs (No. 1,2,3,4 and 7), surface outflow weirs (No. 5) and “groundwater”

outflow weirs (No. 6) (compare Figure 4-3). Evaporation pans are used as manual mean to quan-

tify the evaporation at ground level. A mobile soil moisture measurement kit, as well as qualitative

field methods (see [71]), are used for the qualitative and quantitative determination of the soil

moisture, that are however not used in the course of this master thesis. The soil moisture is

computed in 96 vegetation/ botanical plots of 1 m2 with monthly manually performed measure-

ments resulting in nine points in each plot [7].


35

Figure 4-2: Map of the distribution of measurement devices and botanical plots in the bofedal area (source: Jan R. Baiker, map prepared by Dina Farfán Flores, satellite image from Google Earth)


36

Figure 4-3: Parshall weir (top left) and currently performed bucket measurement at a v-notch weir (bottom left and right) (source: Jan R. Baiker).

5 Methodology

37

5 Methodology The methodology forms the key aspect of this master thesis since the approach is not straight-

forward and uses the in-situ measurement to evaluate the hydrological similarity of several donor

catchments. It is an approach of selecting complete parameter sets from a collection of calibrated

catchments in the Swiss and North Italian Alps and applying them as an unaltered set to a poorly

gauged area in the Peruvian Andes.

A short supporting overview of the used methodology is presented in Figure 5-1. Input data (shown

on the left side) is processed with simulation steps (at the top) resulting in diagrams and plotting

results as depicted at the bottom right. The type of lines indicate what kind of input data is used.

While climatology input data is plotted with dashed lines, in-situ data with solid ones. The diagram

additionally contains indicators to chapter numbers that lead to detailed information on the data

processing.

At first, physiographical information is used for HRU generation and catchment subdivision, then,

different tuneable parameters of donor catchments are applied and fed into PREVAH. One pa-

rameter set (Tic_34) is used for further analysis resulting in a multitude of plots describing the

processes and catchment hydrology. The donor set is subsequently modified by adopting good

performing parameters identified during the sensitivity analysis and the same analyses and plots

are iterated.

5.1 Catchment subdivision and HRU generation

WINHRU serves as the processing tool to produce the required raster-based grids and manages

the generation of the hydrological response units (HRUs) on user-specific criteria (see chapter

4.1.1). The created grids include information such as the aspect, slope, flow direction, flow accu-

mulation and basin area. The tool classifies different elevation zones that are subsequently used

to aggregate the meteorological data and to define the HRUs. In this analysis elevation zones

are used with a resolution of 50 m.

Results from field evaluations indicate that the study catchment can be subdivided into five sub-

units (compare chapter 2.2). In WINHRU only one pixel in each subarea is selected – in total five

pixels. Based on this definition and the physiographical information stored in the control file,

WINHRU automatically generates the corresponding subareas.

5.2 Application of the Swiss catchment tuneable parameter sets

5.2.1 General idea of the methodological approach

The core idea of the methodological approach is the identification of calibrated donor Swiss/

North Italian (from now on called Swiss) alpine catchments and adaptation of corresponding

tuneable parameters.

Experimental basis for the study are 44 representative donor sets, which in the course of numer-

ous published papers have been successfully calibrated and regionalized with the use of PREVAH

[2,38,52,70,90,110,112,113]. The sets include analyses on “Sihl” and “Rittelsbach” rivers, how-

ever most datasets are extracted from a study performed by Andres et al. (2016) [5] with focus

5 Methodology

38

Figure 5-1: Diagram summarizing the methodology used in the thesis subdivided by simulation, data input and plotting results. The lines indicated the source of input and the red labels the chapters with additional information.

5 Methodology

39

on sub-basins in the Swiss Ticino and bordering regions (refer to the mentioned studies for de-

tailed information on the catchments).

All tuneable model parameters are transferred from the donor to the target catchment as an

unaltered and complete set. Since the physiographical information, which defines the geological

differences, is stored in the HRU control file, the influence of these factors while in the meantime

adopting the Swiss catchment parameters is reduced to a minimum. However, the donor catch-

ments widely differ with respect to their elevation, size and location, which influence the hydro-

logical components and as a result this is reflected in the runoff generation tuneable parameters.

5.2.2 Generation of comparison situations

The comparison of simulation output results with in-situ runoff measurements and the corre-

sponding plots are created for four cases, each in linear and natural logarithmic scale quantile

plots. The cases compare a specific simulation area or the sum of several areas with the meas-

urements taken at one v-notch weir or the sum of several weirs (compare with map in Figure 2-2

and Figure 4-2) and are defined as follows:

Case : comparison of the surface inflow from the northwest (N4 vs. V1)

Case : comparison of the surface inflow to the bofedal (N34 vs. V12347)

Case : comparison of the simulated inflow to the bofedal and the measured groundwater

flow (N345 vs. V6)

Case : comparison of the total simulated surface inflow and the measured outflow (N345

vs V56).

5.2.3 Data processing and quantile plot generation

The simulation is separately performed for each of the tuneable parameter datasets. The data

sets are then, one after the other, supplied to PREVAH. The catchment-specific tuneable param-

eters are presented in the Appendix (Table A- 1, p. 89). This first step of the analysis is solely

based on the climatology background data provided for the years 1995-2009 (chapter 4.2.2). No

climatological information obtained by the in-situ measurements are used in the simulation itself.

Both, precipitation and temperature are imported as daily information and PREVAH runs at an

hourly time-step.

In-situ runoff measurement data is subsequently introduced as single observational points to the

plots and compared with the simulated runoff.

For each of the donor tuneable parameter datasets, each case is calculated and presented in a

natural logarithmic and linear scale plot. This sums up to a total number of 44×4×2 = 352 quan-

tile plots created in R Studio (44 ≙ number of donor datasets; 4 ≙ number of cases; 2 ≙ natural

logarithmic and linear scale). Figure 5-2 shows this relation as exemplary output of one simulation.

The diagrams in Figure 5-3 give a first impression of the quantile plots` appearance. The y-axis

is assigned to the quantiles of the (natural logarithmic) runoff or discharge given in [l/s] versus

the x-axis as the time scale over one calendar year. Additionally, the in-situ runoff point meas-

urements are plotted as black dots and grey triangles against the simulation data. The colored

ribbons represent the quantiles or percentage values for the location based on daily climatology

data for all 15 years between 1995 and 2009 in comparison. The blue bars, corresponding to

5 Methodology

40

high values, indicate the low quantiles, that are only reached by a small percentage of the daily

values. The red bars are the small values that are exceeded by most (light red) to all (darkest

red) values for the one day of the year, as variance over the 15 years. The thick black line sepa-

rating the red and blue ribbons, indicates the 50% quantile, that represents the daily value that

is exceeded and undercut by the same amount of years.

The data and thereby the resulting plots are smoothed by an odd-numbered climatology moving

window of 31 days duration. All available data within the window region is statistically summa-

rized: the total number and average of these points is evaluated, their minimum and maximum

values formed and the standard deviation generated. The results are again processed as point

values at the center of the moving windows and the statistical indicators are the attributes of the

windows [102].

Be aware that not every case allows the same quality of comparison and natural logarithmic scale

plots have proved to provide a better comparability among the datasets and therefore focus is

put on them. Especially by cutting the high runoff peaks in the austral summer months, the loga-

rithmic plots show a more smoothed look and pretend an inexistent uniformity.

5.3 Parameter set optimization

The quantile natural logarithmic plots of climatology data are visually (subjectively) and numeri-

cally (objectively) evaluated according to their quality of approximation compared to the in-situ

point measurement data. The better the 50% quantile line fits the in-situ measurement data

points, the better the underlying tuneable parameter set approximates the point measurements.

The aim is to detect the optimal parameter set defined by the best fit.

5.3.1 Visual quantile comparison

On a visual basis, the logarithmic quantile plots for all donor tuneable parameter sets are evalu-

ated regarding their quality of approximation. Particular focus lies on the timing and magnitude

of maxima and minima and the ability of the simulation to capture the steepness of the flanks

between highs and lows correctly. Figure 5-3 shows in comparison the quantile plot approxima-

tions of two case plots of different quality: the left one based on AlpEin data of weak quality

and the right one on Tic_34 with a good approximation. The first approximations of all tuneable

parameter sets can be found in the Appendix (Figure A- 1 to Figure A- 15, p. 95-109).

Figure 5-2: Diagram with exemplary output for one tuneable parameter set.

5 Methodology

41

The visual comparison results in the following donor tuneable parameter sets for particularly good

approximations: Tic_01, Tic_05, Tic_13, Tic_23, Tic_34, Tic_37 together with the gas100 da-

taset. The geographical background in the northern part (not strongly glaciated) makes all sets

except Tic_23 highly interesting for a donation.

Figure 5-3: The linear and corresponding natural logarithmic plot of a poor approximation on the left (AlpEin) and a particularly good donor set on the right (Tic_34).

5.3.2 Numerical quantile comparison

Introduction to the ranking system

A numerical analysis is performed to validate this visual impression and to determine the optimum

donor parameter set to be used to achieve best approximation of the manual measurements.

At first for each dataset the number of in-situ measurements are identified, that are located within

the following intervals of the simulation: +/- 25%, +/- 47.5% (excluding the ones in +/- 25%), out-

liers and missing values (NA). Additionally, the number of measurements lying in the sum of both

5 Methodology

42

intervals +/- 25% and +/- 47.5% is calculated. Figure 5-4 explains the intervals observed in com-

bination with the quantiles.

Figure 5-4: Explanation of quantiles and intervals used in the plots and calculations.

For each donor set and simulated case, school grades are assigned. Hereby the number of in-

situ measurements that lie within a quantile-interval are counted and the grade is allocated ac-

cording to Table 5-1. The grades are summed up over each dataset. The result allows a com-

parison of the approximation quality to other datasets. The detailed ranking table is given in the

Appendix (Table A- 2/3, p. 90 f.). It can be observed, that the grades of “outlier” are reverse with

the lowest number corresponding to the best grades.

Table 5-1: Ranking based on a school grade system. The maximum of 50 is marked by the number of measurements

performed to date.

Grade 1 2 3 4 5

+/- 25%

+/- 47.5%

+/- 25% + 47.5%

41-50 31-40 21-30 11-20 0-10

Outlier 0-10 11-20 21-30 31-40 41-50

Evaluation of the ranking table

A number of conclusions can be drawn with the performance of this school type grade ranking.

The total sum of grades over all datasets averages roughly at a value of 51 and ranges between

47 for the best suiting and 64 for the worst approximation. The average grade is 3.2. Be aware

that the smaller the value, the better the fitting approximation is.

The best approximating donor datasets identified by this numerical analysis are Tic_04, Tic_13,

Tic_29, Tic_30, Tic_33, Tic_34 and gas100, which all achieve an overall grade of 47. With these

results the aforementioned visually identified best datasets, could be confirmed for gas100,

Tic_13 and Tic_34 (compare Table 5-2).

Table 5-2: Table comparing the well approximating donor sets in the visual and numeric analysis. V: visual; N: numeric. X marks the positive approximation.

5.3.3 Comparison of tuneable parameters of well performing donor sets

In a next step, the tuneable parameters of the donor sets are compared in order to get an indicator

of successful variable ranges and thus sensitive parameters. Table 5-3 compares the total tested

range (44 donor sets) and its mean to the range of the 12 best performing sets and the calibration

Tic_

01

Tic_

04

Tic_

05

Tic_

13

Tic_

23

Tic_

25

Tic_

29

Tic_

30

Tic_

33

Tic_

34

Tic_

37

Gas

100

V X X X X X X X

N X X X X X X X X

5 Methodology

43

range proposed by Vegas et al. (2012). A comparison shows, that the positive tested spread over

a wide range within the tested range which is an indicator for the challenge of equifinality (intro-

duced in chapter 3.2) [89]. The Tic_34 parameters are added for the further analysis.

The visual analysis was a good first indicator defining the quality of the datasets. However, only

the numeric analysis allows a fast and extensive comparison of all datasets at the same time.

Therefore, the performance of a numeric analysis is inevitable to find the optimal donor set of

parameters.

Table 5-3: Overview over parameters with best performance range, tested range and its mean, calibration range re-ported by Vegas et al. (2012) and the Tic_34 donor set.

5.3.4 Decision on set for further analysis

Donor set Tic_34 was selected for further evaluation of this kind of approach and for the analysis

of the hydrology at catchment scale. Tic_34 is located in Northern Italy, in the Piemonte region

and encompasses Valle Anzasca. The sub-region is called Anza in the analysis performed by

Andres et al. (2016). Table 5-4 summarizes and compares the PREVAH catchment information

of Tic_34 in Northern Italy and the study catchment in the Peruvian Andes. While the area of the

Anza location is almost 60 times larger and the elevation range much higher, the number of

HRUs and meteorological zones of the study catchment is even higher. Nevertheless, the per-

formed analysis to this point is showing a promising solution for the very small study catchment.

Table 5-4: Summary of important PREVAH location parameters of the Tic_34 region and the study catchment in the Peruvian Andes [108].

Location Area [km2] No. of HRUs Mean area per HRU

[km2] No. of meteo

zones Elevation

range

Anza 257.25 328 0.78 40 248-4665 m

Study catchment 4.5 472 0.01 53 3825-4588 m

total range tested mean

tested

range best

performance

calibration range

(Vegas et al. 2012) Tic_34

from to from to from to

Exponent for soil

moisture recharge 1.8 4.83 3.79 1.80 4.83 3.83

Threshold moisture

saturation for ETR 0.7 0.7 0.70 0.70 0.7 0.70

Threshold storage

for surface runoff 10 70 55.08 33 70 10 50 50.0

Storage coefficient

for surface runoff 5 40 31.65 10 40 10 30 35.0

Storage coefficient

for interflow 30 200 162.9 30 200 50 150 154.0

Percolation 0.05 0.5 0.28 0.14 0.5 0.02 0.20 0.19

Storage coefficient

for fast baseflow 200 1000 812.48 554 1000 200 1000 699.0

Maximum storage

for fast baseflow 25 300 203.60 108.3 300 225 250 129.2

Storage coefficient

for delayed

baseflow

1000 4000 256474 1000 3802 1000 4000 2951

5 Methodology

44

5.4 Sensitivity analysis

The Tic_34 donor parameter set is subject to a detailed sensitivity analysis. The overall set of

parameters is fixed throughout the analysis while only one parameter is fine tuned in a linear

fashion and the effect on the result controlled. By this method, the system response to the single

tuneable parameter is determined, while simultaneously the influence of the others is sup-

pressed. The test values are located in or slightly outside the calibration range reported by Vegas

et al. (2012) as shown in Table 5-3. The natural logarithmic quantile plots are subject to a visual

and numeric sensitivity comparison.

5.4.1 Visual sensitivity analysis

In terms of visual sensitivity comparison, the quantile plots related to the variation of a single

parameter are compared. The following aspects are considered:

▪ How sensitive is the system to the parameter? This is observable in the tendency of

the diagrams to remain or alter with the tuning of the parameter.

▪ How does the extent of the low quantiles change? This is observable with the dark

blue and dark red bands` tendency to extent.

▪ How does the variance change? This is observable by the tendency of the flanks to

increase or decrease in steepness.

▪ How does the approximation quality change? This is observable by the location of a

single point in-situ measurement compared to the mean value of the simulation.

▪ How does the mean value change? For this it is especially interesting whether the mean

value changes at all and whether it has the tendency to reach zero in any of the plots.

5.4.2 Numerical sensitivity analysis

The variance of the overall system performance to the change of a certain parameter of Tic_34

is also analyzed numerically. The numerical analysis is performed with the same approach ex-

plained in chapter 5.3.2 using a school grade ranking system on the quantile plots.

5.5 Additional in-depth analysis of meteorology and hydrology

5.5.1 Temperature extreme value analysis

The temperature data measured with HOBO 1, 2 and 3 every few minutes is transformed to total

hourly minimum, mean and maximum values over all three HOBOs. Each hour lasts from xx:45

to (xx+1):44. Not all HOBOs provide data for the same time span and missing data due to empty

batteries and full memory cards is given in August 2015. Artifact values generated by taking the

HOBO out of the wind and sun shield during data read-out, are deleted as well as the subsequent

10 min of data acquisition.

5.5.2 Time series curves

The donor tuneable parameter set Tic_34 is used for extensive analysis of the hydrology in the

catchment. In a first step, time series curves of the parameters runoff and potential/ actual evapo-

5 Methodology

45

transpiration are generated over the time period. To conduct a simulation in PREVAH with the

in-situ measurements of precipitation and temperature from the HOBOs, data for wind, solar ra-

diation and evapotranspiration is needed. This data is taken from the climatology data (1995-

2009) and subdivided in 1.75-year time slots. Each of the 1.75-year time slots of the climatology

data starts on March 15th and last until December 29th of the following year. The last of the 15

years is shorter and covers only the period from March, 2009 until the end of December, 2009.

The initial state conditions for the simulation is also taken from the climatology data set and is

given by the starting date of the time slot. In the end, the 1.75-year time slot is simulated 15

times, which results in 15 different time series curves. These curves are referred to as simulated

time series curves.

The influence of the parameters wind, solar radiation and evapotranspiration as well as the initial

conditions can be determined.

Four different scenarios of the time series curves are generated:

▪ Scenario 1A: daily precipitation and temperature data from HOBO in-situ measure-

ments (Figure A- 21, p. 116)

▪ Scenario 2A: hourly precipitation and temperature data from HOBO in-situ measure-

ments (Figure A- 21, p. 116)

▪ Scenario 1B: daily precipitation data from HOBO in-situ measurements and daily tem-

perature from the climatology data (Figure A- 20, p. 115)

▪ Scenario 2B: hourly precipitation data from HOBO in-situ measurements and hourly

temperature from the climatology data (Figure A- 20, p. 115)

The difference between the defined scenarios is expressed in: the source of the temperature

data and the resolution of the input data of either daily or hourly values. Daily and hourly values

are generated for the temperature as a mean over hour or day and for the precipitation as a sum

over hour or day. These time series curves even though simulated based on in-situ precipitation

(and temperature) measurements, are referred to as simulated data.

When using the in-situ temperature data for the simulation in scenario 1A and 2A, several extra

preparation steps have to be performed. The hourly temperatures have to be created and after-

wards interpolated over the mean height of the three HOBOs (4107 m). As the in-situ data shows

one time period in August, 2015 (August 3rd - August 22nd) where none of the three HOBOs

provides data and PREVAH needs continuous data, this time span has to be filled with the cor-

responding data of the same sequence in 2016.

Runoff hydrograph

The runoff hydrograph gives an impression of the simulated versus the in-situ measured runoff

over the period of 1.75 years. By simultaneously plotting the associated point measurements

(black dots), the quality of approximation to the in-situ data can be observed.

Evapotranspiration time series curves

Potential evapotranspiration (ETP) and actual evapotranspiration (ETR) are available in the

PREVAH output file. They are generated based on the four different scenarios. As the evapo-

transpiration pan measurements are falsified by the collected precipitation in the device over the

5 Methodology

46

time sequence, the data needs to be transformed to a precipitation adjusted reference evapo-

transpiration. The measured ETP is first subtracted from the precipitation in the same period and

then multiplied with an evaporation coefficient of 0.5 for the dry season (months May through

September) and 0.6 for the rainy season (October through April). Moderate wind speeds of 2-

5 m/s, a windward side distance of green crop of 1 m, pan placement on short green cropped

area and a low relative humidity >40 during the dry season and medium between 40-70 for the

wet season are assumed in order to obtain the coefficients based on the “FAO Guidelines for

computing crop water requirements” [26].

The in-situ potential evapotranspiration measurements are not continuous but sums over variable

time sequences. Therefore, the measured “evapotranspiration” data in [mm], available for se-

quences of varying length, are scaled compared to the simulation output according to the follow-

ing formula that is then solved to the daily observed evapotranspiration 𝐸𝑇𝑃𝑂𝑖 :

𝐸𝑇𝑃𝑃𝑖

𝐸𝑇𝑃𝑃̅̅ ̅̅ ̅̅ ̅

= 𝐸𝑇𝑃𝑂

𝑖

𝐸𝑇𝑃𝑂̅̅ ̅̅ ̅̅ ̅

with

𝐸𝑇𝑃𝑃𝑖 Evapotranspiration (ETP) simulated by PREVAH (P) with the climatology data for a

particular day i

𝐸𝑇𝑃𝑃̅̅ ̅̅ ̅̅ ̅ Mean over the evapotranspiration (ETP) simulated by PREVAH (P) with the clima-

tology data over the time sequence defined by the in-situ measurement

𝐸𝑇𝑃𝑂𝑖 Evapotranspiration (ETP) in-situ observation (O) for a particular day i

𝐸𝑇𝑃𝑂̅̅ ̅̅ ̅̅ ̅ Mean over the evapotranspiration (ETP) in-situ observation (O) over the time se-

quence defined by the in-situ measurement

5.5.3 Additional analyzing plots

As introduced in chapter 4.2.3, several additional parameters are measured in the field especially

by the botanical plots. Not all of these factors are mutually informative and important for the

understanding of the hydrology in the catchment; however, some of them are valuable for further

analysis. All data referred to as simulated is based on in-situ precipitation and temperature meas-

urements.

Water table depth analysis

The in-situ water table depth is only available on an hourly basis starting on March 17th, 2016

and lasting to January 12th, 2017. Therefore, the observable time span is limited to this period.

Be aware, that 380 mm is the maximum level of the water table depth that can be measured and

this critical value is well visible in diagrams. The water table depth is compared to the SLZ (Lower

zone runoff storage), which is simulated based on in-situ HOBO precipitation and temperature

data, and the in-situ HOBO precipitation.

The SLZ is an automatic given result of the PREVAH simulation. The water table depth daily

means are plotted against the SLZ daily values. A scatterplot with the SLZ on the x-axis and the

corresponding water table on the y-axis is created.

The water table is compared to the precipitation mean of HOBO 1-3 in-situ measurements. A

time series curve with the time span of 1.75 years forming the x-axis and the precipitation and

water table depth forming the y-axis with two different scales is formed.

5 Methodology

47

Potential evapotranspiration analysis

As described before, time series curves are performed based on the simulated (with in-situ P/ T)

and scaled in-situ potential evapotranspiration data. Further evapotranspiration analysis is per-

formed with sequence sums instead of scaled daily values.

A boxplot is generated comparing the potential evapotranspiration in-situ to the simulated data,

where the x-axis corresponds to the time sequences and the y-axis represents the potential

evapotranspiration. While the simulation ETP data is illustrated by a box over the variance within

the 15 years of simulation based on the climatology data, the in-situ measurements correspond

to points indicating the position in relation to the simulation.

A comparison of the in-situ potential evapotranspiration to the in-situ soil moisture fails to indicate

a clear trend and is therefore not further described.

Potential in-situ evapotranspiration is furthermore compared to the in-situ precipitation. The pre-

cipitation data is summarized over all three HOBOs and to the sequences specified by the in-situ

evapotranspiration data. A scatterplot is then created comparing the two parameters for each

evaporation pan separately indicated in different colors and additionally a bisecting line is in-

serted.

5.6 Donor parameter set tuning

Tic_34 is used as a donor tuneable parameter set and analyzed in detail. The resulting plots

indicate, that unlike supposed before, the parameter set is not optimal for the approximation of

the observed study catchment hydrology and does not satisfactorily represent the in-situ runoff

measurements.

There are two options persistent for further analysis and improvement of the data alignment:

manual calibration or selection of a new donor parameter set. Manual calibration is subject to

limitations, questionable applicability to the extremely small study catchment and represents a

completely different approach to the rest of this thesis (compare chapter 3.2). For this reason,

the manual calibration is not performed but a different donor parameter set chosen as input.

The primary goal is to use a set with more slowly reacting parameters. An increase of the in-

ertance can be achieved by the maximization of the following parameters: percolation, threshold

storage for surface runoff, storage coefficient for surface runoff and storage coefficient for inter-

flow. None of the other tuneable parameter sets with positive performance in the quantile plots

(Table 5-2) shows striking differences to Tic_34 regarding the parameters strongly influencing

the inertance (see Appendix Table A- 6, p. 94 for a comparison of the tuneable parameters).

The donor set ultimately used implies a modification of Tic_34 according to successful fine-tuning

parameters during the sensitivity analysis. The storage coefficient for interflow, percolation, stor-

age coefficient for fast baseflow and maximum storage for fast baseflow, that proved to achieve

the best approximation in a visual and numerically performed sensitivity analysis, are plotted

again in all possible variations of one to four modified parameters filled up with the original Tic_34

tuneable parameters. A comparison of the quantile plots identifies Tic_34 with a modified storage

coefficient for interflow (slightly higher) and a storage coefficient for fast baseflow (significantly

higher) as the most promising new set (modification “mod3” in Appendix Table A- 6, p. 94). It is

5 Methodology

48

now referenced as Tic_34_mod. Tic_34_mod is subject to the same runoff hydrograph curve

analysis as Tic_34 before (Figure A- 22/23, p. 117 f.).

5.7 Water balance

The water balance evaluation summarizes the aforementioned data and findings. It is the key to

understanding the hydrology in the catchment and the basis for further ecological analysis. Main

question behind the water balancing is the timing of water storage and release in the catchment

and its magnitude. The timing of groundwater and runoff generation comes along with it.

Fundament for the balancing is a grid generation possible with PREVAH, for the eight variables:

the potential evapotranspiration (ETP), the actual evapotranspiration (ETR), the adjusted inter-

polated precipitation (P), the percolation into the saturated zone (GWN), surface runoff (R0),

interflow (R1), total baseflow (R2) and total runoff (RGS). For each parameter, monthly mean

values are given out over the 15 years for each of the five areas. Since the surface runoff R0 is

zero for almost every month, it is not further described in the following.

A boxplot is generated over the 12 months of one year (x-axis) for the simulated variables (ex-

cluding R0) and the in-situ measurements for runoff, evapotranspiration and precipitation (y-axis).

The boxes show the variability of the monthly simulated values over the 15 years of the climatol-

ogy data – the scaled in-situ measurements are indicated as point measurements (for 2015 as

colored circles, for 2016 as colored triangles). Additionally, to the seven variables an eighth box

is generated summing R1 and R2. All variables are given in [mm/month]. The figure is used as

an overview over all variables and their relation in magnitude. Another boxplot is subdivided by

category into evapotranspiration (top), precipitation (middle), and runoff (bottom) and is therefore

used for a detailed description and interpretation.

The evapotranspiration in-situ data is used as the scaled values gained by applying the approach

described in chapter 5.5.2. The precipitation monthly values are generated by simple addition of

all daily values within a month. In order to receive monthly runoff values the following equation

has to be applied:

𝑟𝑢𝑛𝑜𝑓𝑓 [𝑚𝑚

𝑚𝑜𝑛𝑡ℎ] = 𝑟𝑢𝑛𝑜𝑓𝑓 [

𝑙

𝑠] ×

3.6

𝑠𝑢𝑏𝑎𝑟𝑒𝑎 [𝑘𝑚2]𝑎𝑟𝑒𝑎[𝑘𝑚2]

×4.4 ×1000×24×30

The equation scales the daily point measurements in [l/s] to monthly values and transforms it to

[mm/month]. Afterwards, the mean over all scaled runoff measurements occurring in one month

is generated. The formula is applied to the v-notch V1 data. When using the V6 (baseflow) or

V5+6 (total runoff), it is scaled to the total catchment area (subarea/ area = 1). Additional barplots

show the water balance with the incoming water through precipitation (P) opposed to the sum of

the outgoing water (RGS, ETR). As there is no ETR available for the in-situ to the date of sub-

mission, the in-situ barplot is created with the sum of RGS+ETP instead. For all three processes

the median of the 15 year data is used.

6 Results

49

6 Results

6.1 Sensitivity analysis

This sensitivity analysis, performed on the Tic_34 donor parameter set, is subdivided by system

component into surface runoff, interflow, percolation and baseflow. A sequence of diagrams com-

posed of one each for a high (left), medium (middle) and small (right) parameter value is provided.

The corresponding parameter modification to the original set can be extracted from the Appendix

(Table A- 4/ 5, p. 92 f.). The findings are discussed and interpreted in a hydrological consensus

in chapter 7.

6.1.1 Surface runoff

Threshold storage for surface runoff

The parameter “threshold storage for surface runoff” [mm] that defines the threshold content of

the upper storage reservoir (SUZ) must be exceeded to initiate surface runoff generation [93].

Irrespective of the parameter value, the mean value of the simulation does not change (Figure 6-1).

Figure 6-1: Diagrams for the threshold storage for surface runoff for a high (left), medium (middle) and small (right) value in direct comparison

Storage coefficient for surface runoff

The “storage coefficient for surface runoff” [h] governs the generation of surface runoff. In the

model by Viviroli et al. (2007) it is referred to as K0 [h] [93]. Irrespective of the parameter value,

the mean value of the simulation does not change (Figure 6-2).

6.1.2 Interflow

The “storage coefficient for interflow” [h] governs the generation of interflow. In the model by

Viviroli et al. (2007) it is referred to as K1 [h] [93]. Irrespective of the parameter value, the mean

value of the simulation does not change. However, the smaller the parameter value is chosen,

the more the dark blue part extends (Figure 6-3).

6 Results

50

Figure 6-3: Diagrams for storage coefficient for interflow for a high (left), medium (middle) and small (right) value in direct comparison.

6.1.3 Deep percolation

The upper storage reservoir is emptied by the “deep percolation PERC” into the reservoirs of the

saturated zones measured in mm/h [93]. The smaller the parameter values are chosen, the more

the steepness of the flanks increases and therefore the approximation for the in-situ point meas-

urements degrades (Figure 6-4).

Figure 6-4: Diagrams for percolation for a high (left), medium (middle) and small (right) value in comparison.

Figure 6-2: Diagrams for storage coefficient for surface runoff for a high (left), medium (middle) and small (right) value in direct comparison.

6 Results

51

6.1.4 Baseflow

Storage coefficient for fast baseflow

The “storage coefficient for fast baseflow” indicates how long water can be stored before forming

a fast groundwater runoff [h] [93]. The smaller the values are chosen, the more the steepness of

the flanks increases and therefore leads to an increasingly bad approximation for the point meas-

urement data (Figure 6-5).

Figure 6-5: Diagrams for the storage coefficient for fast baseflow for a high (left), medium (middle) and small (right) value in direct comparison.

Maximum storage for fast baseflow

The “Storage coefficient for fast baseflow” is in direct correspondence with the “maximum storage

for fast baseflow” and calculated in mm [93]. For small chosen parameter values, the flanks are

flattening out which causes an increasingly bad approximation for the point measurement data

(Figure 6-6).

Figure 6-6: Diagrams for the storage coefficient for fast baseflow for a high (left), medium (middle) and small (right) value in direct comparison.

Storage coefficient for delayed baseflow

Delayed baseflow is the belated streamflow during periods without rain [85]. The “storage coeffi-

cient for delayed baseflow” defines the time in hours until the delayed baseflow forms. The visual

comparison shows that with decreasing parameter value the steepness of the flanks increases,

which leads to a worsening approximation for the point measurement data during the austral

winter and an improved approximation during the spring months (Figure 6-7).

6 Results

52

Figure 6-7: Diagrams for the storage coefficient for delayed baseflow for a high (left), medium (middle) and small (right) value in direct comparison.

6.1.5 Comparison between visual and numeric sensitivity analysis

The school grade ranking table of the Tic_34 sensitivity analysis can be found in the Appendix

(Table A- 4/5, p.92 f.). Overall grades between 48 and 63 are achieved, representing the high

variability of approximation quality. The average grade is 51.

More meaningful than a strict numeric sensitivity comparison of the parameters is a direct com-

parison to the visual sensitivity analysis.

An observation of Table 6-1 indicates that there is a controversy between visual and numeric

approximation for three parameters: percolation and storage coefficient for fast and delayed

baseflow. This controversy is due to the weak case N345 versus V6. In this case with low values

the approximation seems to improve by an increasing steepening of the flanks because then the

flanks are steep enough to approximate the data points in the far right side of the diagram while

failing to represent the points on the left side that include the global maxima and minima. Unlike

the choice of the best tuneable parameter set of the 44 possible ones, in this part of the analysis

the numeric sensitivity analysis is not promising. Further modifications are based on the visual

evaluation. The best approximating parameter values are lined out in Table 6-1.

Table 6-1: Comparison for Tic_34 for visual and numeric approximation quality with rotation of single parameters.

Visually Numerically Best values

Threshold storage for surface runoff all equal all equal independent

Storage coefficient for surface runoff all equal all equal independent

Storage coefficient for interflow high best high best 175

Percolation high best medium best 0.4 – 0.45

Storage coefficient for fast baseflow high best medium best 900

Maximum storage for fast baseflow medium best medium-high best 50-100

Storage coefficient for delayed baseflow high best all equal 3000-4000

6 Results

53

6.2 Additional in-depth analysis of meteorology and hydrology

6.2.1 Temperature extreme value analysis

The temperature data measured with HOBO 1, 2 and 3 every few minutes is transformed to total

hourly minimum, mean and maximum values over all three HOBOs. The extreme value distribu-

tion for the 1.75 year time span is plotted in one diagram.

The time series scatterplot on the temperature extreme value distribution (Figure 6-8) shows on

the x-axis the time span between March 15th, 2015 and December 29th, 2016 and on the y-axis

the temperature in [°C]. Each color represents a different extreme value as daily mean over all

three HOBOs: red representing the maxima, blue the mean and green the minima. While the

immense dispersion of the maxima data points disproves a distinct periodicity, it clearly becomes

observable for mean and minimum values. The highest temperatures occur between January

and April – the lowest between June and August. The spreading and therefore variability of the

daily maxima is the highest with roughly 5°C at any given time, while for the green it is about 3-

5°C and for the blue ones only about 2.5°C. The periodicity of the data was assumed. However,

the daily maxima were expected with higher magnitude and stronger amplitudes as they are more

or less equal throughout the year, always between 7.5 and 15°C.

Figure 6-8: Diagram comparing daily in-situ temperature minimum (green), mean (blue) and maximum (red) values (as mean over HOBO 1-3).

6.2.2 Time series curves

Time series plots of the parameters runoff and potential/ actual evapotranspiration are generated

over the same time period as above. The 1.75-year time slot is simulated 15 times, which results

in 15 different time series curves. Influencing parameters, such as wind, solar radiation and evap-

otranspiration can be determined by fixing precipitation and temperature inputs to the HOBO

values.

6 Results

54

Runoff hydrograph

By simultaneously plotting the associated point measurements (black dots), the quality of ap-

proximation to the in-situ data can be observed. Figure 6-9 shows the hydrograph for area4 with

the corresponding v-notch weir V1 runoff in-situ measurements. As area4 is expected to be the

most promising and reliable case, it is the one focused on. The area4 hydrographs are performed

based on daily and hourly, precipitation (and temperature) in-situ data are provided in the Ap-

pendix (Figure A- 20/ 21, p. 114 f.). Since they only show slight differences in approximation, the

daily graph simulated with in-situ precipitation and temperature is solely used in the discussion.

The small simulated runoffs occur between June and November of each year, high runoffs re-

spectively between December and May. In the second year, the peaks reach up to 120 l/s (daily)

Figure 6-9: Runoff hydrograph curve simulating the years 1995-2009 using the HOBO1-3 daily data (Area 4 compared to V-notch weir 1).

6 Results

55

and 350 l/s (hourly). The December 2016 values do not meet the high values achieved in De-

cember of the previous year. In the first few months until the end of October 2015 the in-situ data

is higher than the simulation. For the remaining period the in-situ is lower than the simulated

data.

The runoff hydrograph curve allows the drawing of two preliminary assumptions. As the HOBO

precipitation and temperature data was used as an input to the simulation, the simulated runoff

should be equivalent to the in-situ measurements. The presumption emerges, that the in-situ

measurements are not able to adequately represent the actual discharge. Furthermore, one can

observe an almost negligible influence of the meteorological parameters wind speed, global ra-

diation, relative humidity and relative sunshine duration as the difference between the single

simulated 1.75 time slots is almost negligible. More details may be found in the discussion in

chapter 7.1.3.

Evapotranspiration time series curve

For a smoothing effect in the plot (Figure 6-10), the mean over the 15 years based on HOBO

temperature and precipitation simulated data, is created and these daily values compared to the

scaled in-situ daily values. For low potential evapotranspiration, the scaled in-situ data is higher

than the simulated and neither does the in-situ data meet the extremes of higher simulated values

in the austral summer months.

Figure 6-10: Time series curve of mean simulated evapotranspiration compared to scaled in-situ daily data.

6.2.3 Additional analyzing plots

Several additional parameters are measured in the field but not all of these factors are mutually

informative and important for the understanding of the hydrology in the catchment. Albeit some

of them are valuable for further analysis. The focus here lies on the water table and evapotran-

spiration.

6 Results

56

Water table analysis

Water table data is only available between March 17th, 2016 and January 12th, 2017 and trans-

formed to daily values. It is important to stress, that 380 mm is the maximum level of the water

table depth that can be measured and this critical value is well visible in diagrams. The water

table is compared to the simulated (based on in-situ P/ T data) SLZ (Lower zone runoff storage)

and the in-situ precipitation.

The scatterplot in Figure 6-11 describes the direct correlation between SLZ and the water table

depth. Both parameters are mutually dependent on the availability of water. The up-lining points

indicate an exponential distribution. The lower the water table as distance to the soil surface, the

lower the corresponding SLZ is. If the water table depth is high, the water even sinks 38 cm below

the soil surface, then little available water percolates to the saturated zone.

Figure 6-11: Scatterplot water table (in-situ) versus simulated SLZ. Water table depth is understood as the distance to the soil surface.

A significant trend of the water table compared to the in-situ precipitation data is restricted due

to limited data. Figure 6-12 only shows a slight tendency of the water table to sink immediately

after the incidence of low daily precipitation in the beginning of the dry season.

Potential evapotranspiration analysis

The following potential evapotranspiration analysis is performed with sequence sums instead of

scaled daily values.

Figure 6-13 shows a clear trend of the potential evapotranspiration simulated (based on in-situ

P/T data) compared to the in-situ data. While in the first half of the investigation period, the in-

situ measurements tend to be higher than the background simulation, in the second half they

clearly drop below. Furthermore, the higher the simulated ETP are, the higher the variance is

therein.

6 Results

57

Figure 6-12: Time series curve comparing in-situ precipitation and corresponding water table depth (distance to the soil surface).

Figure 6-13: Comparison of the simulated potential evapotranspiration aggregated to the sequences of in-situ evap-otranspiration measurement to the mean over all five evaporation pans.

6 Results

58

A comparison of the potential in-situ evapotranspiration versus the in-situ precipitation (Figure

6-14) indicates, that for low ETP and precipitation, ETP seems to almost equal the precipitation

visible by the clustering of points in the bottom left corner around the 1:1 line. The higher the

precipitation gets; the higher evapotranspiration gets as well but a slower pace as the points then

collect in the upper half of the diagram. No significant difference between the five evaporation

pans becomes observable.

6.3 Donor parameter set tuning

Figure 6-15 is the runoff hydrograph curve image for area4 compared to v-notch weir #1 for the

Tic_34_mod tuneable parameter set simulated with both precipitation and temperature in-situ

data. A comparison to the same image for Tic_34 (Figure 6-9) shows a slight improvement of the

approximation for the high runoff but by contrary a deterioration for the low runoffs in the second

half of the investigation period. Due to the hence missing enhancement to the Tic_34 dataset, all

other additional plots are not repeated.

6.4 Water balance results

In this chapter the diagrams for area 4 are described and analyzed. As the total runoff for area 4

is measured directly with V1 it allows a direct comparison of simulated and in-situ data in the

barplots. No other area besides provides this opportunity. While Figure 6-16 gives an overview

of all relevant parameters, Figure 6-17 is subdivided into precipitation, evapotranspiration and

runoff. In the water balance analysis, no in-situ precipitation and temperature are used for the

simulation but solely climatology data. As all measurements are in mm though, also the entire

catchment area total runoff may be exemplified by a comparison to V5+6 and the baseflow with

V6 (Figure A- 26/ 27, p 120 f.). Additionally, Figure A- 24/ 25 (p. 118 f.) compares the area4 total

runoff of V1 to V5+6.

Figure 6-14: Scatterplot of in-situ potential evapotranspiration versus precipitation both summed over the evapotranspi-ration measurement sequences.

6 Results

59

6.4.1 Boxplots

Evapotranspiration

The top part in Figure 6-17 sums up the findings of monthly simulated ETP and ETR data com-

pared to monthly in-situ reference ETP data. The higher values of ETP occur between August

and April, the lower values respectively between May and July. For ETR the higher values occur

between October and April, lower values between May and August. The averages are above

zero in all months for both parameters and the variation within a month is higher between Sep-

tember and November and during the end of the raining season (April and May).

In every month ETP is higher than ETR. ETP starts earlier to increase in the dry season in July,

while ETR only increases after the dry season in October. Thus, the difference between the two

Figure 6-15: Runoff hydrograph curve simulating the years 1995-2009 using the HOBO1-3 daily precipitation and tem-perature data for Tic_34_mod (Area 4 compared to V-notch weir 1).

6 Results

60

values increases continuously between June and October. During the raining season the differ-

ence decreases and reaches its minimum with the peak of the raining season in February.

The variation throughout the year of the in-situ data is not as prominent as for the simulated data.

The in-situ data also decreases at the end of the rainy season around April/ May and then slightly

increases again by June/ July. Except at the end of the rainy season, in the fall months, there is

a tendency for the in-situ to be lower than the simulated ETP. A comparison of the 2015 and

2016 data shows that the difference between the values is always roughly up to 10 mm/ month.

Solely the first two months of measurement in April and May 2015 completely stick out with a

difference to the 2016 data and simulation of up to 25 mm/ month.

Precipitation

The middle diagram in Figure 6-17 sums up the findings of monthly precipitation of the climatology

data and in-situ P. The precipitation during the rainy season (October-March) slowly increases

and reaches its peak in February. Low precipitation values occur during the dry season of the

winter months, when they even reach zero. Both trends are visible for the climatology and in-situ

data. The higher the values (summer), the higher the variance of the climatology therein.

The point measurements are comparable to the corresponding climatology ones during the dry

season. However, during the fall months, especially in April, the in-situ measurements react more

inert than the simulation and for both 2015 and 2016, the precipitation does not reduce as fast

as implemented by the climatology. In October, by the start of the raining season, the in-situ data

however reacts faster than the climatology (2015 and 2016). The most remarkable point meas-

urements are the February, 2016 and December, 2015 values.

Runoff

The bottom diagram in Figure 6-17 sums up the findings of monthly simulated GWN, R1, R2,

R1+R2 and RGS. In all areas, R2 can be compared to scaled V6 in-situ measurements – only in

Area 4 additionally RGS can be directly compared to V1. After using the runoff scaling formula

presented in chapter 5.7 and transforming the runoffs to mm measurements, V5+6 may be used

as total runoff from both the catchment area and the subareas (boxplots with V5+6 are in the

Appendix Figure A- 24-27, p. 118 ff.).

R1 and R2 are summed up to R1+R2 to demonstrate by its equality to RGS, that the surface

runoff (R0) is negligible. R1 shows the highest values in January and February, reduces to zero

in the winter months and slowly increases in November. R2 is highest in February and March,

continuously reduces from April to July and slowly increases again between October and No-

vember. R2 is always higher than R1 and while R2 never reaches zero, R1 does in April until

October. Since RGS is mostly influenced by R2, it follows the same pattern described for R2.

GWN has the highest variation reaching median values higher than any other parameter in Jan-

uary and February, but reducing significantly during March to April and reaches zero for all winter

months. In October, it slowly and in November strongly increases.

For the high runoffs in February and March, the V1 in-situ data is below the simulation but almost

reaches the simulated medians of the boxes. The runoff starts decreasing in April throughout the

entire winter – the in-situ data is changing at a slower rate than the simulated. This becomes

especially visible in April and May. A comparison of the 2015 and 2016 data shows almost no

difference.

6 Results

61

Figure 6-16: Diagram indicating the waterbalance as comparison between simulated and in-situ data as an example for area4.

6 Results

62

Figure 6-17: Simulated compared to in-situ data subdivided by evapotranspiration, precipitation and runoff.

6 Results

63

Figure A- 24/ 25 (p. 118 f.) show an additionally plotted total runoff based on V5+6 data for area4.

One can see, that it is significantly lower that both the simulation and the V1 scaled in-situ meas-

urements in all months even though both values should be roughly equivalent. The same dis-

crepancy becomes visible in A- 26/ 27 showing the V5+6 total runoff compared to the entire

catchment area.

With increasing simulated runoff in November and December, the in-situ is clearly delayed. It is

also visible in January through March, but for January and March the evaluation can be based

on only one available measurement in the 1.75 year time span and hence the findings need to

be used with care. The in-situ measurement of V6 shows no big variance throughout the year.

Therefore, it only matches the magnitude of the simulated values for low runoff in the winter

months. The timing of high values matches well.

Comparison of the three diagrams

Comparing ETP and P indicates that the increasing potential evapotranspiration begins before

the end of the dry season – before the precipitation increases (simulated and in-situ). Even if the

precipitation is low or reaching zero (May - September), both actual and potential evapotranspi-

ration occur and never reach zero.

In both the simulated and in-situ data it is observable, that R1 and R2 react slower to the obvious

start of the raining season in October with one or even two months’ delay.

The percolation into the saturated zone is the predominant constituent to the reformation of

groundwater. It is generally high whenever precipitation is high and low when precipitation is low.

During the dry season percolation reaches zero. GWN shows a fast response to reduced P in

April. The reformation drops to a minimum and in the following months the baseflow originates

from water in remaining storage and no longer from newly built groundwater. As soon as precip-

itation increases in October, the percolation increases likewise and storages slowly fill up.

6.4.2 Barplots

The barplots make the water balance even more obvious than the aforementioned boxplots. P is

opposed to the sum of RGS and ETR (ETP). In comparison to the barplot based on simulated

climatology data, for area4 also an in-situ data based water balance barplot is created. Because

all measurements are in mm, the barplots for the total catchment area with in-situ V5+6 as the

total runoff corresponds to the area4 plot with V5+6. Both are in the Appendix (Figure A- 28,

p.122)

Simulated data

In Figure 6-18 the simulated climatology data comparison is shown. P displays a sinusoidal pat-

tern with the highest values during the austral summer months December to February. The am-

plitude is about 130 mm/ month. Simulated RGS + ETR shows the same sinusoidal pattern with

the same timing of peaks and lows, though with a different amplitude of 90 mm/ month. Also

visible is a slight shift of the curve to the right as the reduction of the precipitation in the fall

months April and May is a lot faster and also seems earlier than the corresponding RGS+ETR.

Furthermore, the increase in spring is a lot faster than the corresponding RGS+ETR. Due to the

different amplitudes, uniform sinusoidal appearance and about 1-2 months shift of the curve, P

6 Results

64

exceeds RGS+ETR during the six summer months and falls below in the six winter months. Pre-

cipitation exceeding RGS+ETR corresponds to a filling up of the storage or a “positive” storage

allowing the regeneration of groundwater – the corresponding “negative” storage to a deflation.

In March the emptying of the storages starts slowly and increases strongly in April; the depletion

lasts throughout the entire austral winter months until the end of September.

Figure 6-18: Precipitation of the climatology data contrary to the sum of simulated actual evapotranspiration and total runoff (area4).

In-situ data

The well observable pattern of the simulated case is not as obvious in the in-situ data in Figure

6-19. P exceeds RGS+ETP in seven months and deceeds in the corresponding five dry winter

month. Neither one of the sets shows a sinusoidal appearance – the P shows peaks in February

and December that are too high and the RGS+ETP barely changes in the course of the year and

only shows slight peaks in April and May. An investigation period of only one complete raining

season is used in comparison to the 15 year mean and therefore smoothed climatology. An elon-

gated investigation period may help to counterbalance high precipitation peaks. The timing of the

peaks is shifted by two months (February (P) compared to April (RGS+ETP)). RGS+ETP does

not react to increasing precipitation in spring but stays equally low. In October the storages begin

to fill up. When using V5+6 for the total runoff from the total catchment area and area4 (Figure

A- 28, p. 122) the magnitude of the RGS+ETP does not significantly change compared to the V1

data. However the peaks then appear in February and March. No increase of the RGS+ETP at

the beginning of the raining season is observable – it is also more or less uniform throughout the

year.

Comparison simulation/ climatology - in-situ

Both highest P values for in-situ and climatology data occur in February, though with different

magnitudes. The shift by 1-2 months of RGS+ETR (ETP) is present in both as well.

Table 6-2 shows the ∆𝑠 as the amount of water filling or emptying the storages per month for the

simulated climatology data compared to the in-situ data. In the course of one year the “positive”

6 Results

65

storages and “negative” storages should roughly equal zero. While for the simulated data it sums

up to +10 corresponding to slightly more water coming into the system than going out, according

to the in-situ data significantly more water is coming in, especially provoked by the peaks in

February and December. The table furthermore indicates that the timing of storage and depletion

is shifted. While for the simulation data the storages start draining in March, for the in-situ data

they do not empty until May.

Table 6-2: Delta storage values [mm/ month] for in-situ (I) compared to simulated (S) data. The last column compares the sum over the 12 months.

Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Sum

I 47.7 196.1 39.3 4.6 -58.8 -37.0 -32.6 -25.5 -23.4 39.2 46.6 123.5 319.7

S 22.44 20.8 -1.0 -39.4 -29.4 -17.5 -12.9 -9.0 -7.8 13.6 28.7 40.9 9.5

Figure 6-19: In-situ water balance with precipitation contrary to scaled reference evapotranspiration and total runoff (area4).

7 Discussion

66

7 Discussion

7.1 Explanation of the hydrology in the catchment

Comprehensive conclusion of the “Result” chapter 6 is a deviation of the in-situ water balance

compared to the sinusoidal pattern of the simulated (based on climatology data) water balance.

To understand how the different parameters interact and individually contribute to the visible de-

viation the three major processes, precipitation, runoff and evapotranspiration adding up to the

water balance, are discussed. Afterwards, some findings are substantiated by the effect of a

unique catchment geology and the El Niño Southern Oscillation extreme weather phenomena.

7.1.1 Temperature

As described in secondary literature on the ANS, during June and September low temperatures

appear accompanied by infrequent frost events, caused by clear sky conditions during the nights

between May and September (Figure 6-8). The moderate temperatures between September and

December at an average temperature of 14 °C are confirmed for the study catchment. In the

beginning, a periodic temperature extreme value distribution was assumed. While the immense

dispersion of the daily maxima data points disproves a distinct periodicity, it could be confirmed

for mean and minimum values. The extremes were expected with higher magnitude and stronger

amplitudes than the variability between 7.5 and 15 °C. Higher magnitudes and amplitudes are

restricted by high fog formation in the bofedal in the raining season. The daily minima variation

is due to clear sky conditions during the nights, with temperature drops even below zero between

May and September. The flanks of increasing temperatures in spring and decreasing in fall are

not equally steep, indicating that it takes longer to build up higher temperatures than to reduce.

Furthermore, higher maximum temperatures occur earlier than the highest mean and minimum

temperatures.

7.1.2 Precipitation

Precipitation is the sole constituent of incoming water to the hydrological system and its timing

and magnitude is the driving force for all other processes. As expected, the seasons are distinct

with low precipitation occurring in the dry and cold season of the winter months and high precip-

itation in the warmer summer months accordingly.

Following literature descriptions for the ANS, first precipitation occurs after the dry season in

September and almost reaches its maximum by December. With the onset of precipitation, the

storages start to fill up again and new groundwater is generated. By the end of the raining season

all storages are full resulting in reduced retention time in the upper catchment area.

The precipitation in-situ measurements are sometimes higher, sometimes lower than the precip-

itation of the climatology data. The boxplots indicate (Figure 6-16, Figure 6-17) that in both con-

secutive years the December 2015 and 2016 measurements are higher than the precipitation of

the climatology data and that the February 2016 value is about 125 mm/month higher than the

climatology. During the fall months, especially in April, the in-situ P measurements react more

inert than the climatology and both for 2015 and 2016 the precipitation does not reduce as fast

7 Discussion

67

as obtained by the climatology. Therefore, according to the in-situ precipitation measurements,

the dry season is not as prominent and short as indicated by the climatology.

High spatial and temporal variability of the meteorological and hydrological factors occurs from

one valley to the next, resulting in high deviation of the in-situ to the climatology precipitation and

make a precise hydrological modelling extremely challenging. The climatology data is based on

the interpolation of surrounding meteorological station observations. However, as described in

chapter 4.2.2, the stations are not close and located at different elevation. Variability of precipi-

tation and temperature for the bofedal region is not likely to be adequately represented and there-

fore the HOBO in-situ measurements are expected to provide higher accuracy.

According to Jan R. Baiker, the only imaginable source of error in the HOBO precipitation meas-

urement arises in case of sleet generation and following congestion and overflow of the devices.

The HOBOs are then unable to measure the entire solid precipitation. A comparison to the tem-

perature extreme value distribution (Figure 6-8) indicates that these events are unlikely to occur

between December and February. This allows the conclusion, that the effective precipitation may

be higher in the cold dry season, but sleet generation fails to explain extreme precipitation peaks

in the summer months (personal communication, March 12th, 2017).

The effect of precipitation on the vegetation in the bofedal is shown in Figure 2-3. While the

maximum greenness is reached in the end of March with the end of the warmer raining season,

the bofedal still appears rather green but is already surrounded by brownish grassland in June

and August. In October, with only slightly more precipitation, the vegetation reacts quite fast to

the new water availability and directly soaks up the incoming water for vegetation activity. A

comparison to the runoff indicates that no water percolates to the saturated zone yet to produce

significant increase of runoff.

7.1.3 Runoff

The understanding of the runoff and its magnitude, delay, sensitivity and quality of the measure-

ments is of major importance for a comprehension of the hydrology in the catchment and various

diagrams consult the process.

Magnitude

The runoff hydrograph plot (Figure 6-9) and the boxplots (Figure 6-16, Figure 6-17) indicate that

the runoff peaks in magnitude expected by the simulation are not met in the bucket in-situ meas-

urements, but the lows seem to correspond fairly well. Unlike the magnitude, the timing of the

highest runoff peaks is matching in February/ March (Figure 6-16, Figure 6-17). While the in-situ

barplots (Figure 6-19) imply that the sum of total runoff and evapotranspiration barely changes

over the course of the year, the composition indeed shows slight variability.

Delay

Delay becomes visible on two occasions: runoff tends to respond with delay to increasing pre-

cipitation and in-situ runoff is delayed to the simulated data.

Baseflow (R2 in boxplots) peaks in March, when precipitation already reduces, which results in

a two months’ delay of the runoff to precipitation (Figure 6-16, Figure 6-17). Accordingly, the

7 Discussion

68

reduction of the in-situ runoff in the beginning of the dry season in April and May is not as prom-

inent. More water than expected by the simulation is still in storage and forms distinct runoff in

the months of already reduced precipitation – corresponding to months of draining storage (Fig-

ure 6-19). The delay of runoff is controlled by the choice of the storage coefficients, that the

catchment is simulated with. In the field, delay may be explained by the fact, that while precipita-

tion is a direct process, runoff reacts inert due to the interaction with the ground surface.

In secondary literature for the region, runoff is described with a seasonal behavior and with the

highest volume of water between January and March. Simulated and in-situ runoff (Figure 6-9,

Figure 6-16, Figure 6-17) is high in January but only peaks in February and March. Well visible

in the runoff hydrograph curve (Figure 6-9) is an about 1-2 months’ delay of in-situ runoff to the

simulated data, which becomes observable in November 2015. Accordingly, an increase of the

runoff with the onset of the raining season 2016/ 2017 does not occur until the end of the inves-

tigation period. In this case, the investigation period limits the delay detection and definition.

However, in-situ precipitation increases faster than the corresponding climatology precipitation

in October (Figure 6-16, Figure 6-17). Therefore, the assumption of delay due to missing incom-

ing precipitation can be omitted. More precipitation is required than simulated to actually initiate

increasing runoff. This indicates that the storage thresholds are even higher than presumed.

Furthermore, one can see in a comparison of the magnitudes, that the catchment´s hydrology

responds more inert than simulated. The ecosystem of the bofedal and the unique geology of the

catchment function as regulators of the downhill water flux and decelerate the runoff from the

catchment more than simulated, according to the definition of a bofedal area.

One might expect a similar delay of the in-situ highest runoffs to the simulated (Figure 6-9), but

the second peak precipitation in February occurs at a time of filled storages and high saturation

in the saturated and unsaturated zone. This is resulting in a fast, un-buffered and unfiltered runoff

due to reduced retention time.

Sensitivity

Both, the sensitivity analysis (Figure 6-1, Figure 6-2) and boxplots (Figure 6-16, Figure 6-17)

indicate, that the sensitivity of the system to surface runoff is very limited. The short living com-

ponent is of insignificant importance to the hydrology in the catchment; R0 is zero over almost

the entire year. The bofedal as a wetland that absorbs the precipitation like a sponge, transforms

it into vegetation activity and releases excessive water as percolation to the saturated zone,

where it forms fast and delayed baseflow throughout the year.

The hydrological system is moreover nearly insensitive to the storage coefficient for the interflow,

representing the only influencing parameter within the unsaturated soil zone (Figure 6-3). Albeit,

for small storage coefficients, the high discharges are simulated even higher. A comparison to

the interflow, represented as R1 in the boxplots (Figure 6-16, Figure 6-17), shows that it never

exceeds about 15 mm/ month, while being zero for more than half of the year. It confirms that

the impact of interflow increases for high runoffs and precipitation, but also accentuates the sig-

nificance of the baseflow as the sole runoff constituent forming year-round discharge.

Percolation is a parameter of high significance in a wetland system. A small percolation value

indicates that more surface runoff generates and little water percolates through the saturated

zone, going into storage and forming groundwater. As a result, runoff is zero during the dry sea-

son as no water is available in the upper storage tanks (Figure 6-4). According to the boxplots

7 Discussion

69

(Figure 6-16, Figure 6-17) in the winter months (May to July) the percolation sinks to a value

lower than the baseflow. The saturated zone is continuously emptied, but throughout the year

the storage never completely drains visibly as baseflow occurs at all times of the year. As men-

tioned above, baseflow is the main constituent to the particular hydrological system, resulting

from high percolation, which is plausible for the encountered quaternary and limestone karst ge-

ology. The increasing precipitation in October is stored in the unsaturated zone storage, in the

soil surface and rootstock of the plants and percolation to the saturated zone only slowly starts

with the beginning of the raining season. GWN in the boxplots (Figure 6-16, Figure 6-17) peaks

with the climax of precipitation in February, which proves the strong interdependence of percola-

tion to precipitation. The mean values of the percolation are roughly half compared to the high

precipitation summer values and about zero in the low precipitation winter months. Therefore,

according to the simulation, about 50% of the incoming precipitation reaches deeper ground lay-

ers and is not directly transformed to surface runoff (which is negligible as mentioned before) and

evapotranspiration. Nevertheless, a comparison to the in-situ barplot (Figure 6-19) shows, that

March and April are the only months where evapotranspiration and runoff have a higher magni-

tude. During all other months, the share of runoff is far less than 50%. It is strongly evident, that

the runoff measurements are of minor quality.

Baseflow is represented by three different coefficients in the runoff generation module of the

PREVAH simulation system. Small storage coefficients for fast and delayed baseflow correspond

to fast response of runoff to incoming precipitation. Small coefficients limit the reservoir extent

capabilities and lead to completely empty storages by the end of the dry season, resulting in

mean values of zero for the corresponding quantile plots (Figure 6-5, Figure 6-7). During the

rainy season the storage continuously fills up and empties during the dry months, but as seen in

the boxplots (Figure 6-16, Figure 6-17), the baseflow never reaches zero even in the dry months

with simultaneous zero percolation. A better fit is achieved for high storage coefficients for fast

and delayed runoff and therefore longer water storage before forming baseflow. The best fitting

coefficients for the delayed baseflow correspond to half a year of storage until the delayed

baseflow occurs. In contrast, best approximation is achieved for middle values of “maximum stor-

age for fast baseflow” corresponding to the amount in [mm] of stored water before producing

baseflow (Figure 6-6). If the parameter is chosen too high, the amount of water needed to pro-

duce baseflow is also too high and in the dry winter months the available water in the storage

tank sinks below the threshold, resulting in a zero runoff for these months. Reducing the storage

capacity for fast baseflow increases the importance of the delayed baseflow to the system. Figure

6-11 describing the direct correlation between the simulated (based on in-situ P/ T) lower zone

runoff storage (SLZ) and the in-situ water table depth, shows a logarithmic distribution. The

higher the water table depth, the smaller the distance to the soil surface, the higher SLZ and the

more water is available in the ground and therefore in storage. However, when the water table

depth first increases, also SLZ increases slightly but later faster, leading to this logarithmic func-

tion distribution. Both, geology and the wetland characteristics make a longer storage and buff-

ering more plausible. Summarized it can be stated, that the baseflow is the most dominant dis-

charge component in the study catchment, especially in the dry winter months. As stated by

Buytaert et al. (2006) the Andean wetland is able to bridge fairly large periods of distinct dryness

while maintaining a pronounced baseflow [20].

As detected in the sensitivity analysis, thresholds for Tic_34 are already chosen high within the

calibration range, indicating that water is stored long and high before baseflow is produced. But

7 Discussion

70

yet, the Tic_34 tuneable parameter thresholds seem to be too low to adequately simulate the

storages of the system, featured with the bofedal and karst characteristics. The controversy be-

tween the overall system inertness opposing the delay of in-situ compared to simulated data,

provoked by the choice of tuneable parameter values, is the main challenge of parameter tuning

for the study catchment.

Quality

Several indicators question the quality of the in-situ runoff measurements, which helps to under-

stand the deviation from the simulated data in the barplots. The non-continuous measurements

are not able to reliably portray every daily or weekly peak and therefore the magnitude of the

highest values stays uncertain. The delta storage (Table 6-2) observation with the surplus of water

in the in-situ plot is a strong indicator, that not all the runoff is caught with the weirs. The simulated

runoff hydrograph curve, based on in-situ precipitation and temperature, shows a strong discrep-

ancy between in-situ and simulated observations. Also, the comparison between percolation

changes and a more or less uniform runoff is highly questionable. In November 2016, additional

discharge measurements were performed in the field to compare to the v-notch weir V6 discharge

data located at the southern end of the bofedal area. They showed that weir V6 does not get hold

of the entire considerable discharge. The water seems to flow into the ground (saturated zone)

to exit below the runoff measurement weir V6 and then leaves the catchment through the water-

fall at the very southern end of the study catchment. The water percolation and re-appearance

can be explained by the geological characteristics of the catchment (see more in chapter 7.1.5).

The comparison of V1 and V5+6 data as total runoff for area4 in Figure A- 24/ 25 (p. 118 f.) is an

additional indicator of incomplete measurements at the weirs at the southern end of the bofedal

area (V5 and 6). The quality however mainly influences the magnitude but not the timing of the

high discharges, therefore the delay in the response cannot be explained with this argumentation.

The only option to determine the amount of water that is lost by the measurements is an additional

weir measuring the amount of water that is available below weir V6 or leaving through the water-

fall to the south out of the catchment. The missing reference data from one of these weirs pre-

vents a re-calculation of the V6 measured discharge by the magnitude of deviation factor from

the new weir. An important improvement is furthermore the installation of measurement devices

for continuous runoff measurements to evaluate peak runoffs within the catchment. For the actual

rainy season 2016/ 2017, i-buttons were installed for additional continuous measurements of the

water table in the drains at the locations of the weirs 1-7. After appropriate analysis, they will help

to provide a continuous runoff series (personal communication with Jan R. Baiker, March 23rd,

2017).

7.1.4 Evapotranspiration Potential evapotranspiration is by default higher than actual evapotranspiration (Figure 6-16, Fig-

ure 6-17). It has to be distinguished between simulated and in-situ potential (ETP) and actual

(ETR) evapotranspiration. In-situ actual precipitation is to the date not available and in-situ ETP

is more referred to as a scaled reference evapotranspiration (see chapter 5.5.2). The simulated

ETR drops substantially below the simulated ETP in the winter and spring months (Figure 6-16,

Figure 6-17). As stated by Viviroli et al. (2007), evapotranspiration decreases constantly over dry

periods, a confirmed fact for the simulated ETR but not for simulated ETP, which already in-

creases again in July during the peak of the dry season (Figure 6-16, Figure 6-17). In-situ ETP

is featured by an almost equal value throughout the year with a slight peak in March but drop in

7 Discussion

71

May to a value that is hold over all dry winter months. This lack of variation becomes visible in

the barplot (Figure 6-19), boxplots (Figure 6-16, Figure 6-17) and time series curve (Figure 6-9)

and corresponds to a more inert reaction of the in-situ. While in the dry months the in-situ values

are expected to sink even further, in the wet months in-situ ETP is expected to increase more

significantly. The boxplot (Figure 6-13) compares the sequence means over all evaporation pans

and the corresponding sum over simulated ETP. It provides an intensified representation of the

same result, visible in the boxplots that hold monthly sums instead of sequence sums (Figure

6-16, Figure 6-17). Scaled in-situ evapotranspiration put into sequences is higher in the first half

and lower in the second half of the investigation period compared to the simulated. In the second

part of the investigation period, the in-situ seems to have the same trend as the simulated, even

though not quite the same magnitude. The exceptionally high measured evapotranspiration in

the first half may originate from measurement errors in the beginning of investigation potentially

due to increasing cow intervention during the dry season.

Both, simulated and in-situ potential and actual evapotranspiration increase with increasing tem-

perature (compare Figure 6-16 to Figure 6-8). The fluctuations of the in-situ data in the course of

2-3 months (e.g. October 2015 – January 2015) is a lot higher than for simulated (Figure 6-10).

The boxplots reveal that for the simulated data, the variability in a month is smaller for small ETP

than for high (Figure 6-16, Figure 6-17). A possible cause for the found variability is the retarda-

tion of evaporation by missing solar radiation during high precipitation phases causing high hourly

variability, which consequently leads to high variability in the wet season. In theory, ETP is never

restricted as it presumes sufficient moisture. For this reason, the value never reaches zero

throughout the year. Even though by default ETR is based on ETP, the timing of the peaks is

shifted: while simulated ETP peaks in October, simulated ETR reaches the highest value in De-

cember (Figure 6-16, Figure 6-17).

ETR is strongly dependent on weather, vegetation, environmental conditions and further site

specific parameters, which makes it very sensitive in the beginning of the raining season to the

current circumstances. Only after the onset of the raining season, when the surface layers carry

sufficient water to allow vegetation activity, ETR increases as well (Figure 6-17). ETR is depend-

ent on the capability of plants to extract water from the soil and their corresponding root depth

compared to the water table. The significant drop of the water table depth with the onset of the

dry season in May is clearly observable in Figure 6-12. Once the water table depth is lower than

they can reach, limited ETR is facilitated solely based on the moisture left in the upper layers.

The more water is available, the more ETR equals ETP, as then both parameters are defined by

satisfactory moisture availability at all times as also stated by Wang and Zlotnik (2012) [98].

Evapotranspiration’s dependence on precipitation

The boxplots and barplots (Figure 6-16, Figure 6-17, Figure 6-18, Figure 6-19) allow a direct

comparison of evapotranspiration and precipitation. During the dry season period, when over

months lower moisture content and less precipitation are provided, less water is available and

hence evapotranspiration drops. Even in the driest months with zero precipitation, still evapo-

transpiration occurs. The scatterplot in Figure 6-14 even shows that for low in-situ ETP it almost

seems to equal precipitation but the higher precipitation is, the higher ETP rises but at slower

pace. Vegetation is able to directly respond to low precipitation and during the dry months all

precipitation is stored in the root zone and transformed to evapotranspiration without even reach-

ing the depth to create baseflow. Therefore, ETP is as high as P for low values. Once precipitation

7 Discussion

72

increases, not all water can be directly used for vegetation activity. While evapotranspiration is

indeed higher, its percentage of direct use compared to precipitation reduces. There is a contro-

versy of the trend. High precipitation followed by high evapotranspiration (ETP and ETR) signifies

more available moisture but at the same time high precipitation is expected to limit the amount

of possible transpiration due to reduced solar radiation. After all, both, high temperature and

precipitation, act as limiting factors for high evapotranspiration. As ETP describes conditions of

unlimited water, it may increase in the dry season without significant increase of the correspond-

ing precipitation. ETR (simulated) in turn reacts slower to increasing precipitation in the beginning

of the raining season (Figure 6-16, Figure 6-17).

Quality

The quality of the measured evapotranspiration is questionable just like the runoff data. In the

beginning of the measurements only a simple fence was used around the evaporation pans. In

some cases, cow intervention was detected due to marks on the ground, in other cases it may

only be assumed due to discrepancies. In May 2015, the pans were reinforced with barbwire

(personal communication with Jan R. Baiker, March 12th, 2017).

The plateaus, visible in the ETP time series curve for in-situ daily data (Figure 6-10), are created

by the in-situ measurement sequences and the use of these sequences during the scaling pro-

cess. The plateaus indicate how irregular measurement in sequences of varying length is. The

use of a moving window in the scaling process or an adaptation of the scaling formula could help

to eliminate these shortcomings. The outstanding peak in the time series plot in May 2015 results

from an exceptionally high in-situ value. Technically measured with the evaporation pans is the

potential evaporation and not the evapotranspiration, since the evaporation pans represent an

open water surface. Furthermore, the measurement of evapotranspiration is extremely challeng-

ing, because it may potentially be affected by perturbation of the background conditions or the

existence of measures such as windbreaks also stated by Allen et al. (1998). However, in the

bofedal close to the evaporation pans, no real windbreaks can be found (personal communication

with Jan R. Baiker, March 12th, 2010).

The applied pan evaporation coefficient has a high influence on the scaling for particular daily

evapotranspiration generation. However, a comparison with the FAO guidelines indicates, that

higher differences between the coefficients to 0.4 and 0.7 for instance are not plausible (compare

chapter 5.5.2). The reference evapotranspiration values erroneously used to this point in the in-

situ barplot (Figure 6-19), still need to be transferred to a real ETR with the use of a water stress

coefficient and the single crop coefficient, which potentially changes the data towards a rather

sinusoidal pattern.

7.1.5 Effect of geology and El Niño

Geology

As described in the introduction to the study catchment (see chapter 2) and during the preceding

discussion, the main constituent of the Ampay massif is the limestone dominated and karst gen-

erating Copacabana group with intermediate Flysch layers. Both the Flysch and the sandy-clayey

Mito group appearing in the foothills of the massif underlying the Copacabana may function as

ground water stagnation layers for the quaternary deposits. The karst in the flat area of the bofe-

dal is an underground aquifer and allows year-round edaphic humidity.

7 Discussion

73

A comparison of the coarse grained geological map of Peru with a 1:1,000,000 scale, a 1: 100,000

map of the National Geographical Institute (ING) with contour lines and the satellite images of

the region indicate, that the entire study catchment is in the Copacabana group. The waterfall

and its corresponding terrain ridge at the southern end of the study catchment (compare Figure

4-3), potentially divides the competent karst and incompetent Flysch layers as terrain ridges of

about 20 m in height are no coincidences but tend to occur at significant geological steps, in this

case possibly formed during past glaciation periods.

As a result, it can be assumed, that at the head of the waterfall all water of the catchments can

be caught in a measurement and no water is lost because of being dammed by the underlying

layer. Consequently, runoff measurements during dry and wet season are of highest significance

at this location within the catchment. Maintenance on a weir installed in this location is likely to

be challenging especially during the raining season.

El Niño

The El Niño Southern Oscillation (ENSO) is one of the most significant weather-forming phenom-

ena on Earth and has a noticeable effect on the distribution of wind and heat across the Pacific

region, while altering the rainfall patterns [109]. Even though the built-up of El Niño is identified

in advance, it typically reaches its peak between November and January [107]. Atmospheric

connections provoke impacts all over the world, especially in tropical and subtropical regions

including the Peruvian Andes [109]. The Andes/ Amazon transition is one of the rainiest regions

of the world [23] with the Andes forming a natural blockade for dry and cold winds originating

from the subtropical Pacific Ocean [31,42] and wet and warm westerly winds from both the At-

lantic Ocean and the central Amazon Basin [31,32,34,44,59,97]. This interaction between large-

scale circulation and complex topography determines exceptional spatial variability of associated

rainfall in the Eastern tropical Andes [23]. El Niño plays a not negligible role in regulating rainfall

variability over the central Andes at interanual time scale [79].

Since the beginning of the official recording in 1950 the three strongest El Niños occurred in the

following austral summers: 1972-73, 1982-83 and 1997-98. However, the second and third

named occurrences showed different effect on precipitation and hydrology in the Central Andes.

For El Niño 1982–1983 an extreme deficit of precipitation was observed throughout the Central

Andes [33,84], but during the El Niño 1997–1998 the same regions did not show significant dry-

ness [58,84]. Also other summers not classified as El Niño years were followed by pronounced

dryness as well [79]. Depending on the region in Peru, El Niño shows its own temporal and spatial

characteristics. The Northwest as well as the coastal regions are associated with an extended

and intensified raining season [84]. Even though the effects of the weather phenomena are not

as consistent in the Southeast in the Altiplano plateau, ENSO here is related to reduced precipi-

tation to the point of drought conditions [84]. In the Central Andes in the Mantaro basin (about

150 km NW of the study catchment), ENSO has a negative impact on precipitation [23]. The

study catchment seems to be located in a transition zone with a tendency for drier conditions.

The World Meteorological Organization (WMO) regards the El Niño 2015/ 2016 as condescend

[105]. It developed in March 2015 and climaxed in November and December 2015. Afterwards

the phenomena steadily decreased to reach neutral conditions in mid-May 2016 [109]. An anal-

ysis performed by the NASA Earth Observatory reveals, that 12 mm more rain than average

occurred over the warmer eastern Pacific and the extraordinary precipitation reached to the

northwestern South American continent affecting Peru and Ecuador [107]. By the end of February

7 Discussion

74

2016 several Peruvian regions, including Apurimac, were affected by high precipitation events

resulting in flooding and landslides (Figure 7-1) [101]. According to meteorologists, El Niño is

partly a responsible factor for the February 23-29th exceptionally high precipitation [106]. Unlike

expected, the ENSO amplified precipitation event has not an area-wide affect in wide and long

bands of the western Andes but rather appears in a punctual fashion.

As described before, December 2015 and February 2016 in-situ precipitation values are signifi-

cantly higher than the climatology, but unlike runoff and evapotranspiration measurements,

HOBO precipitation measurements in the summer months are reliable. Both monthly values co-

incide with high El Niño affected precipitation described in the literature.

Evaluating the El Niño 2015/ 2016 and its effect on the study region is challenging due to limited

reference data. A quick comparison of the now slowly available data for January and February,

2017, shows that the El Niño seemed to have two different influences on the local climatology

during the raining season 2015/ 2016: very high precipitation in December 2015 and February

2016 and a significantly reduced precipitation in January 2016. Furthermore, compared to the

now ongoing raining season 2016/ 2017 a shift from the peak precipitation in January to the

ENSO peak in February was recognized (personal communication with Jan R. Baiker, March

12th, 2017).

The two months of precipitation particularly influence the overall in-situ water balance as de-

scribed in the barplot (Figure 6-19) and the delta storage evaluation (Table 6-2). Assuming nor-

mal raining season precipitation of 125 mm/ month on average for both months, results in a re-

duction of the delta storage surplus to about 140 mm/ year. The rest of the surplus is expected

to be a result of measurement errors and uncertainty. Even though the simulated runoff may be

Figure 7-1: NASA's IMERG data collected for the time period February 23-29, 2016 indicating the surplus of total rainfall over South America which is partly provoked by El Niño. For the study region a total precipitation of about 200 mm is estimated (source: www.nasa.gov [106]).

7 Discussion

75

influenced by being simulated with the El Niño affected precipitation, the corresponding meas-

ured runoff should be influenced by ENSO as well and therefore be higher than during regular

years. However, the assumption is not met in the runoff hydrograph curve (Figure 6-9).

In summary, at least another year of monitoring is necessary to verify the measurements and

discrepancies initiated by the strong weather event.

7.2 Uncertainties and limitations

Several uncertainties and limitations arising in hydrological simulation are addressed in second-

ary literature but only some of them are applicable to the study catchment.

Model performance and parameterization is strongly affected by random and systematic errors

[43]. The potential uncertainty resulting from assuming uniform air temperature and precipitation

intensity throughout the day is neglected as parameters are available and used as hourly varia-

bles [93].

Of particular interest is a finding by Andréassian et al. (2004) and Oudin et al. (2004). Contrary

to precipitation, models are rather insensitive to potential evapotranspiration errors caused by

the buffering ability of soil moisture components and corresponding filtering features of models

[4,63]. They therefore conclude, that a model is highly sensitive to precipitation, which in turn in

the thesis analysis is strongly affected in magnitude by the El Niño Southern Oscillation. On the

other side, they postulate, that the selection of evapotranspiration estimation method (here Pen-

man) is insignificant.

As results potentially differ from year to year, leading to misleading findings, several consecutive

years should be evaluated [80]. Especially the recession curve analysis is of major importance

for an extensive understanding of the catchment hydrology [95]. Various uncertainties arise in

every hydrological modelling: every place has unique and to some extent unknowable character-

istics, boundary and auxiliary conditions; their reflection in tuneable parameter values in a simu-

lation model is always challenging and never flawless [13]. Not negligible is also the uncertainty

by the hydrological model itself and the calibration period of previous calibration [24].

A variety of uncertainties arises from the unique location of the study catchment, based on:

▪ a beforehand completely ungauged catchment and hence missing reference

▪ stations providing reference climatology data are at least 15 km away and 1200 m lower

in elevation

▪ unique karst geology combined with bofedal wetland characteristics

▪ complex simulation system PREVAH, requiring precise information

▪ runoff measurements restricted to 50 data points and obviously not catching the entire

discharge

▪ erroneous evapotranspiration pan measurements, due to cow impact and interpretation

difficulties caused by sequences of changing length

▪ limitations of runoff and evapotranspiration scaling

▪ only complete raining season measurements corresponds to an extreme El Niño event

▪ high uncertainties in the in-situ measurements of two out of the three parameters forming

the water balance: runoff and evapotranspiration.

7 Discussion

76

7.3 Evaluation of the donor parameter approach

Initially, in the status of scientific research chapter 3.2, the two different approaches manual cal-

ibration and parameter donation are evaluated. After testing and performing a first sensitivity

analysis based on parameter donation, also manual calibration came into the focus as potential

alternative. However, due to various reasons, the thesis is only focusing on parameter donation.

The step by step understanding of the catchment hydrology, accompanied with different dia-

grams of analysis indicates, that the approach works to a certain extent. With Tic_34 a good

parameter set was detected, but no optimal solution for approximation is possible.

The combination of the various aforementioned factors of uncertainty in the catchment analysis

restricts a clear evaluation of the approach itself and whether parameter donation across conti-

nents, climate and vegetation zones is promising. All these factors, especially the low quality and

reliability of the runoff data reinforces the assumption, that parameter donation rather than manual

calibration is the most constructive approach to represent the ungauged study catchment in Peru.

8 Conclusion

77

8 Conclusion This master thesis investigated the applicability of model tuneable parameter donation from

gauged Swiss catchments for an integrated assessment of the hydrology of a microscale catch-

ment in the Peruvian Andes. The approach is followed by a validation with the sparse in-situ

information. The innovation in the approach lies in the donation across continents, climate and

vegetation zones for a catchment with limited runoff measurements and a short investigation

period of solely 1.75 years to the date of submission.

The main task was the use of 44 available runoff generation tuneable parameter sets from the

Swiss and North Italian Alps and their application as a complete set to the high-elevated, moun-

tainous but glacier-free Peruvian catchment. Each set was simulated using the hydrological mod-

elling system PREVAH. Based on the best approximating data set in the initial analysis, a region

in Northern Italy, a sensitivity analysis was performed. This allowed an evaluation of the influence

of each process subdividable into surface runoff, interflow, percolation and baseflow. An addi-

tional in-depth analysis of the meteorology and mainly hydrology was conducted allowing detailed

analysis of the processes evapotranspiration and runoff as well as a comparison of the in-situ

and simulated data therein. Based on the aforementioned sensitivity analysis, a donor parameter

set fine-tuning was performed. However, the main task of the set was not met, due to inadequate

approximation compared to the in-situ data especially regarding high runoffs in the raining sea-

son. The originally best donor set was used to generate a water balance evaluation to summarize

the aforementioned findings. The water balance is the key to an understanding of the hydrology

in the catchment and the basis for further ecological analysis. Precipitation is opposed to the sum

of evapotranspiration and total runoff – separately for in-situ and simulated/ climatology data.

High precipitation in the two summer months December and February as well as uniform evap-

otranspiration and total runoff throughout the year dominate the water balance. The findings re-

sult in a discrepancy from the sinusoidal pattern visible in the simulated data. While the simulated

delta storage values per month add up to almost zero by the end of the year, the in-situ do not

achieve this volitional result. A combination of these factors leads to a restricted utility of the

water balance.

This deviation, the missing quality of approximation in other diagrams and additional comparison

runoff measurements performed in the field, support the assumption of severe sampling issues.

As both evapotranspiration and runoff are subject to high measurement uncertainties the quality

of the overall water balance is also questionable. The investigation period coincides with the

2015/ 2016 El Niño Southern Oscillation which is blamed to be the reason for some of the peak

precipitations significantly influencing the balance. The author of this thesis is aware that studying

a remote catchment with only one raining season of data is not fully appropriate, especially while

being influenced by an extreme weather phenomena.

Therefore, it is suggested to repeat the statistical analysis after at least another raining season

and compare the findings. Furthermore, a manual calibration may be conducted based on the 50

in-situ data measurements and validated with future measurements. An eighth weir should be

installed further south than the existing ones to find the losing factor of the runoffs obtained further

up and measurement devices set up for direct continuous runoff measurements. The evapotran-

spiration measurement should be re-evaluated and possibly a more continuous approach chosen

rather than the sum over sequences of variable length. Also, the evapotranspiration scaling to

daily values leaves space for improvements.

8 Conclusion

78

The effect of the low quality and reliability runoff measurements is too high in the water balance

to allow a clear evaluation of the parameter donation approach across continents. Nevertheless,

the outcome of the manual calibration would have been excessively influenced by the low-quality

runoff point data, which supports the choice towards parameter donation that was made from the

beginning.

In general, it can be concluded that the study catchment is more buffered and reacts more slowly

and inert than indicated by the simulation. This is remarkable as the parameter set chosen for

simulation already has inert characteristics.

With the new information and the good and detailed understanding of the hydrological processes

in the bofedal and its entire catchment area, a basis is developed for future analysis and prog-

nosis of the ecological and botanical part of the project. According to Jan R. Baiker the interface

between the hydrological and ecological part is given by the soil moisture which is affected by

the analyzed parameters such as precipitation, evapotranspiration or the height of the ground-

water table. The idea is to evaluate the soil moisture and its changes over the course of one year,

based on the analyzed data and additional soil moisture measurements installed in the botanical

plots of the bofedal. The evaluation will help to achieve a soil moisture gradient over the area of

the bofedal, which in turn allows a prognosis of changes of the vegetation in reaction to different

climate change scenarios. The water balance analysis additionally allows an assessment of the

ecosystem services namely “water” and “plant biomass for pasturing”, which makes different cli-

mate change adaptation processes comparable. The analysis of the water balance and subse-

quent ecology and botanical investigation allows a comparison of the two main adaptation strat-

egies, damming versus ecosystem-based/ nature-based solutions. It may create fundamental

awareness of the unique environment of the bofedal and the impact of a damming structure,

which is already planned by the government (personal communication, March 12th, 2017).

References

79

References

[1] ABDULLA, F.A., LETTENMAIER, D.P. (1997) Development of regional parameter estimation

equations for a macroscale hydrologic model, Journal of Hydrology 197 (1-4), 230–257.

[2] ADDOR, N., JAUN, S., FUNDEL, F., ZAPPA, M. (2011) An operational hydrological ensemble

prediction system for the city of Zurich (Switzerland): Skill, case studies and scenarios, Hy-

drol. Earth Syst. Sci. 15 (7), 2327–2347.

[3] ALLEN, R.G., PEREIRA, L.S., RAES, D., SMITH, M. (1998) Crop evapotranspiration - Guide-

lines for computing crop water requirements - FAO Irrigation and drainage paper 56, FAO -

Food and Agriculture Organization of the United Nations, Rome.

[4] ANDRÉASSIAN, V., PERRIN, C., MICHEL, C. (2004) Impact of imperfect potential evapotran-

spiration knowledge on the efficiency and parameters of watershed models, Journal of Hy-

drology 286 (1/4), 19–35.

[5] ANDRES, N., LIEBERHERR, G., SIDERIS, I.V., JORDAN, F., ZAPPA, M. (2016) From calibration

to real-time operations: an assessment of three precipitation benchmarks for a Swiss river

system, Met. Apps. 23, 448–461.

[6] ANDRES, N., VEGAS GALDOS, F., LAVADO CASIMIRO, W.S., ZAPPA, M. (2014) Water re-

sources and climate change impact modelling on a daily time scale in the Peruvian Andes,

Hydrological Sciences Journal 59 (11), 2043–2059.

[7] BAIKER, J.R. Presentation on “Climate change and human impacts on high-mountain wet-

land ecosystems (bofedales) of the Central Andes – Implications for Ecosystem-based Ad-

aptation in the Ampay National Sanctuary (Apurimac, Peru)”.

[8] BÁRDOSSY, A. (2007) Calibration of hydrological model parameters for ungauged catch-

ments, Hydrol. Earth Syst. Sci. 11, 703–710.

[9] BAUMGARTNER, W.C., REICHEL, E. (1975) The world water balance. Mean annual global,

continental and marine precipitation, Elsevier, Amsterdam.

[10] BERGSTRÖM, S. (1976) Development and Application of a Conceptual Runoff Model for

Scandinavian Catchments, Bulletin Series A, No. 52. University of Lund, SE.

[11] BERRISFORD, P. (2009) The ERA-Interim Archive, ERA Report Series, No. 1, ECMWF:

Reading, UK.

[12] BEVEN, K., FREER, J. (2001) A dynamic TOPMODEL, Hydrol. Process. 15 (10), 1993–

2011.

[13] BEVEN, K.J. (2002) Towards a coherent philosophy for modelling environment, The Royal

Society 458, 2465–2484.

[14] BLANCO, D.E., BALZE, V.M. (2004) Los Turbales de la Patagonia: Bases para su Inventario

y la Conservación de su Biodiversidad (Peatlands of Patagonia: Basis for Inventory and

References

80

Biodiversity Conservation). Publication No. 19, Wetlands International, Buenos Aires, Ar-

gentina.

[15] BLÖSCHL, G., SIVAPALAN, M. (1995) Scale issues in hydrological modelling: A review, Hy-

drol. Process. 9 (3-4), 251–290.

[16] BRATH, A., MONTANARI, A., TOTH, E. (2004) Analysis of the effects of different scenarios of

historical data availability on the calibration of a spatially-distributed hydrological model,

Journal of Hydrology 291 (3-4), 232–253.

[17] BUDYKO, M. (Ed.) (1948) Evaporation Under Natural Conditions: (English translation, Isr.

Program for Sci. Transl., Jerusalem, 1963), Gidromeoizdat, St. Petersburg, Russia.

[18] BURLANDO, P., FATICHI, S. (2016) Hydrology 2: Modelling of Hydrological Systems, ETH

Zürich, Chair of Hydrolog and Water Resources Management.

[19] BURROUGH, P., MCDONNELL, R.A. (2000) Principles of Geographical Information Systems,

Oxford University Press, New York.

[20] BUYTAERT, W., CELLERI, R., BIEVRE, B.D., CISNEROS, F., WYSEURE, G., DECKERS, J.,

HOSTEDE, R. (2006) Human impact on the hydrology of the Andean páramos, Earth Sci-

ence Reviews.

[21] BUYTAERT, W., CUESTA-CAMACHO, F., TOBÓN, C. (2011) Potential impacts of climate

change on the environmental services of humid tropical alpine regions, Global Ecology and

Biogeography 20 (1), 19–33.

[22] DEUTSCHER VERBAND FÜR WASSERWIRTSCHAFT UND KULTURBAU E.V. - DVWK (Ed.) (1996)

Ermittlung der Verdunstung von Land- und Wasserflächen, 240th ed., Gesellschaft zur

Förderung der Abwassertechnik e.V. - GFA, Bonn.

[23] ESPINOZA, J.C., CHAVEZ, S., RONCHAIL, J., JUNQUAS, C., TAKAHASHI, K., LAVADO, W. (2015)

Rainfall hotspots over the southern tropical Andes: Spatial distribution, rainfall intensity,

and relations with large-scale atmospheric circulation, Water Resour. Res. 51 (5), 3459–

3475.

[24] EXBRAYAT, J.-F., BUYTAERT, W., TIMBE, E., WINDHORST, D., BREUER, L. (2014) Addressing

sources of uncertainty in runoff projections for a data scare catchment in the Ecuadorian

Andes, Climatic Change 125, 221–235.

[25] FAO, o. (Ed.) (1988) FAO/Unesco Soil Map of the World, Revised Legend, with corrections

and updates, World Soil Resources Reports 60, Rome.

[26] FAO, o. (Ed.). Allen, R. G., Pereira, L. S., Raes, D., Smith, M. (1998) Crop evapotranspira-

tion - Guidelines for computing crop water requirements - FAO Irrigation and drainage pa-

per 56, Rome.

[27] FATICHI, S., VIVONI, E.R., OGDEN, F.L., IVANOV, V.Y., MIRUS, B., GOCHIS, D., DOWNER,

C.W., CAMPORESE, M., DAVISON, J.H., EBEL, B., JONES, N., KIM, J., MASCARO, G., NIS-

References

81

WONGER, R., RESTREPO, P., RIGON, R., SHEN, C., SULIS, M., TARBOTON, D. (2016) An over-

view of current applications, challenges, and future trends in distributed process-based

models in hydrology, Journal of Hydrology 537, 45–60.

[28] FONKÉN, M. (2010) Comportamiento De La Vegetacion De Bofedales Influenciados Por

Actividades Antropicas (Bofedales Vegetation Influenced by Anthropogenic Activities).

Magister thesis, Pontificia Universidad Católica del Perú.

[29] FONKÉN, M. (2014) An introduction to the bofedales of the Peruvian High Andes, Mires and

Peat 15, 1–13.

[30] FRIEDL, M.A., MCIVER, D.K., Hodges, J. C. F, ZHANG, X.Y., MUCHONEY, D., STRAHLER,

A.H., WOODCOCK, C.E., GOPAL, S., SCHNEIDER, A., COOPER, A., BACCINI, A., GAO, F.,

SCHAAF, C. (2002) Global land cover mapping from MODIS: algorithms and early results,

The Moderate Resolution Imaging Spectroradiometer (MODIS): a new generation of Land

Surface Monitoring 83 (1–2), 287–302.

[31] GARREAUD, R.D. (1999) Multiscale Analysis of the Summertime Precipitation over the Cen-

tral Andes, Monthly Weather Review 127 (5), 901–921.

[32] GARREAUD, R.D. (2009) The Andes climate and weather, Advances in Geosciences 22, 3–

11.

[33] GARREAUD, R.D., ACEITUNO, P. (2001) Interannual Rainfall Variability over the South Amer-

ican Altiplano, Journal of Climate 14 (12), 2779–2789.

[34] GARREAUD, R.D., VUILLE, M., CLEMENT, A.C. (2003) The climate of the Altiplano: observed

current conditions and mechanisms of past changes, Palaeogeography, Palaeoclimatol-

ogy, Palaeoecology 194 (1–3), 5–22.

[35] GUPTA, H.V., SOROOSHIAN, S., YAPO, P.O. (1998) Toward improved calibration of hydro-

logic models: Multiple and noncommensurable measures of information, Water Resour.

Res. 34 (4), 751–763.

[36] GURTZ, J., BALTENSWEILER, A., LANG, H. (1999) Spatially distributed hydrotrope-based

modelling of evapotranspiration and runoff in mountainous basins, Hydrological Processes

13, 2751–2768.

[37] GURTZ, J., BALTENSWEILER, A., LANG, H., MENZEL, L., SCHULLA, J. (1997) Auswirkungen

von klimatischen Variationen von Wasserhaushalt und Abfluss im Flussgebiet des Rheins.

Schlussbericht NFP 31:“Klimaänderungen und Naturkatastrophen”, 1st ed., vdf Hochschul-

verlag AG an der ETH Zürich.

[38] GURTZ, J., ZAPPA, M., JASPER, K., LANG, H., VERBUNT, M., BADOUX, A., VITVAR, T. (2003) A

comparative study in modelling runoff and its components in two mountainous catchments,

Hydrol. Process. 17 (2), 297–311.

[39] HAMON, W.R. (1961) Estimating potential evapotranspiration, Journal of the Hydraulics Di-

vision. Proceedings of the American Society of Civil Engineers 87, 107–120.

References

82

[40] HOCK, R. (1998) Modelling of Glacier Melt and Discharge, Dissertation ETH, 12430; De-

partment of Geography, ETH Zürich.

[41] HOCK, R. (1999) Distributed temperature-index ice- and snowmelt model including poten-

tial direct solar radiation, Journal of Glaciology 45 (149), 101–111.

[42] HOUSTON, J., HARTLEY, A.J. (2003) The central Andean west-slope rainshadow and its po-

tential contribution to the origin of hyper-aridity in the Atacama Desert, International Jour-

nal of Climatology 23 (12), 1453–1464.

[43] HRACHOWITZ, M., SAVENIJE, H., BLÖSCHL, G., MCDONNELL, J.J., SIVAPALAN, M., POMEROY,

J.W., ARHEIMER, B., BLUME, T., CLARK, M.P., EHRET, U., FENICIA, F., FREER, J.E., GELFAN,

A., GUPTA, H.V., HUGHES, D.A., HUT, R.W., MONTANARI, A., PANDE, S., TETZLAFF, D.,

TROCH, P.A., UHLENBROOK, S., WAHENER, T., WINSEMIUS, H.C., WOODS, R.A., ZEHE, E.,

CUDENNEC, C. (2013) A decade of Predictions in Ungauged Basins (PUB) - a review, Hy-

drological Sciences Journal 58 (6), 1198–1255.

[44] INSEL, N., POULSEN, C.J., EHLERS, T.A. (2010) Influence of the Andes Mountains on South

American moisture transport, convection, and precipitation, Climate Dynamics 35 (7),

1477–1492.

[45] INSTITUTO DE ESTUDIOS HISTÓRICO–Marítimos del Perú (1994) El Perú y sus recursos: At-

las geográfico y económico, Auge S.A. Editores.

[46] JENSEN, M.E., ALLEN, R.G. (2016) Evaporation, Evapotranspiration, and Irrigation Water

Requirements. Evaporation, Evapotranspiration, and Irrigation Water Requirements, 2nd

ed., Tasl Committee on Revision of Manual 70.

[47] JUSTON, J., SEIBERT, J., JOHANSSON, P.-O. (2009) Temporal sampling strategies and un-

certainty in calibrating a conceptual hydrological model for a small boreal catchment, Hy-

drol. Process. 23 (21), 3093–3109.

[48] KOKKONEN, T.S., JAKEMAN, A.J., YOUNG, P.C., KOIVUSALO, H.J. (2003) Predicting daily

flows in ungauged catchments: model regionalization from catchment descriptors at the

Coweeta Hydrologic Laboratory, North Carolina, Hydrol. Process. 17 (11), 2219–2238.

[49] KONZ, M., SEIBERT, J. (2010) On the value of glacier mass balances for ydrological model

calibration, Journal of Hydrology 385 (1-4), 238–246.

[50] KUCZERA, G., MROCZKOWSKI, M. (1998) Assessment of hydrologic parameter uncertainty

and the worth of multiresponse data, Water Resour. Res. 34 (6), 1481–1489.

[51] LAWRENCE, D.M., THORNTON, P.E., OLESON, K.W., BONAN, G.B. (2007) The Partitioning of

Evapotranspiration into Transpiration, Soil Evaporation, and Canopy Evaporation in a

GCM: Impacts on Land–Atmosphere Interaction, Journal of Hydrometeorology 8 (4), 862–

880.

[52] LIECHTI, K., ZAPPA, M., FUNDEL, F., GERMANN, U. (2013) Probabilistic evaluation of ensem-

ble discharge nowcasts in two nested Alpine basins prone to flash floods, Hydrol. Process.

27 (1), 5–17.

References

83

[53] LINDSTRÖM, G., JOHANSSON, B., PERSSON, M., GARDELIN, M., BERGSTRÖM, S. (1997) De-

velopment and test of the distributed HBV-96 hydrological model, Journal of Hydrology

201, 272–288.

[54] MADSEN, H. (2003) Parameter estimation in distributed hydrological catchment modelling

using automatic calibration with multiple objectives, Advances in Water Resources 26 (2),

205–216.

[55] MAIDMENT, D.R., MCGRAW, H. (Eds.). Shuttleworth, W. J. (1993) Evaporation, Inc. New

York, New York.

[56] MAROCCO, R. (1975) Geologia de los cuadrangulos de Andahuaylas, Abancay y Cotabam-

bas, Lima, Peru.

[57] MCMAHON, T.A., PEEL, M.C., LOWE, L., SRIKANTHAN, R., MCVICAR, T.R. (2013) Estimating

actual, potential, reference crop and pan evaporation using standard meteorological data:

A pragmatic synthesis, Hydrol. Earth Syst. Sci. 17 (4), 1331–1363.

[58] MCPHADEN, M.J. (1999) Genesis and Evolution of the 1997-98 El Niño, Science 283

(5404), 950–954.

[59] MÉLICE, J.L., ROUCOU, P. (1998) Decadal time scale variability recorded in the Quelccaya

summit ice core delta18O isotopic ratio series and its relation with the sea surface temper-

ature, Climate Dynamics 14 (2), 117–132.

[60] MENZEL, L. (1996) Modelling canopy resistances and transpiration of grassland, Soil-Vege-

tation-Atmosphere Transfer at Different Scales 21 (3), 123–129.

[61] MERZ, R., BLÖSCHL, G. (2004) Regionalisation of Catchment Model Parameters, Journal of

Hydrology 287, 95–123.

[62] MONTEITH, J.L. (1981) Evaporation and surface temperature, Quarterly Journal of the

Royal Meteorological Society 107, 1–27.

[63] OUDIN, L., ANDRÉASSIAN, V., PERRIN, C., ANCTIL, F. (2004) Locating the sources of low-

pass behavior within rainfall-runoff models, Water Resour. Res. 40 (11).

[64] OUDIN, L., ANDRÉASSIAN, V., PERRIN, C., MICHEL, C., LE MOINE, N. (2008) Spatial proxim-

ity, physical similarity, regression and ungauged catchments: A comparison of regionaliza-

tion approaches based on 913 French catchments, Water Resour. Res. 44 (3).

[65] PATIL, S., STIEGLITZ, M. (2011) Hydrologic similarity among catchments under variable flow

conditions, Hydrol. Earth Syst. Sci. 15 (3), 989–997.

[66] PENMAN, H.L. (1948) Natural evaporation from open water, bare soil and grass, Proceed-

ings of the Royal Society of London. Series A. Mathematical and Physical Sciences 193

(1032), 120–145.

[67] PENMAN, H.L. (1975) Vegetation and the Atmosphere. Ed. MonteithJ. L. London: Academic

Press (1975) Vol. 1; principles pp. 298, £10: Vol. 2; Case Studies pp. 459, £15, Experi-

mental Agriculture 14 (2), 178.

References

84

[68] PERKS, A., WINKLER, T., STEWART, B. (Eds.) (1996) The adequacy of hydrological net-

works: a global assessment, World Meteorological Organisation (WMO), Geneva.

[69] PERRIN, C. (2007) Impact of limited streamflow data on the efficiency and the parameters

of rainfall-runoff models, Hydrological Science Journal 52 (1), 131–151.

[70] PESENDORFER, M., LOEW, S., ZAPPA, M. (2010) Environmental impacts of the Lötschberg

Base – and Crest Tunnels, Switzerland. Rock Engineering in Difficult Ground Conditions –

Soft Rocks and Karst – Vrkljan, 729th ed., Taylor & Francis Group, London, U. K.

[71] RINDERER, M., KOMAKECH, H., MÜLLER, D., WIESENBERG, G., SEIBERT, J. (2015) Qualitative

soil moisture assessment in semi-arid Africa – the role of experience and training on inter-

rater reliability, Hydrol. Earth Syst. Sci. 19 (8), 3505–3516.

[72] SAGÁSTEGUI, S., DILLON, M., SÁNCHEZ, I., LEIVA, S., LEZAMA, P. (1999) Diversidad Flo-

rística del Norte del Perú (Floristic Diversity of the North of Peru ), Editorial Graficart, Tru-

jillo, Peru.

[73] SALZMANN, N., HUGGEL, C., CALANCA, P., DÍAZ, A., JONAS, T., JURT, C., KONZELMANN, T.,

LAGOS, P., ROHRER, M., SILVERIO, W., ZAPPA, M. (2009) Integrated assessment and adap-

tation to climate change impacts in the Peruvian Andes, Adv. Geosci. 22, 35–39.

[74] SCHREIBER, P. (1904) Über die Beziehungen zwischen dem Niederschlag und der Wasser-

führung der Flüsse in Mitteleuropa, Z. Meteorol., 21 (10), 441–452.

[75] SCHRÖDTER, H. (1985) Verdunstung: Anwendungsorientierte Meßverfahren und Bestim-

mungsmethoden, Springer-Verlag, Berlin.

[76] SCHULLA, J. (1997) Hydrologische Modellierung von Flussgebieten zur Abschätzung der

Folgen von Klimaänderungen., 1st ed., Zürcher Geographische Schriften, Geographisches

Institut der ETH Zürich.

[77] SCHWARB, M., ACUÑA, D., KONZELMANN, T., ROHRER, M., SALZMANN, N., SERPA LOPEZ, B.,

SILVESTRE, E. (2011) A data portal for regional climatic trend analysis in a Peruvian High

Andes region, Adv. Sci. Res. 6 (1), 219–226.

[78] SCHWARZE, R., DROEGE, W., OPHERDEN, K. (Eds.). In: Diekkrüger, B., Kirkby, M.J.,

Schröder, U. (Eds.) (1999) Regional analysis and modelling of groundwater runoff compo-

nents from catchments in hard rock areas., IAHS Publication 254. IAHS Press, Walling-

ford, UK.

[79] SEGURA, H., ESPINOZA, J.C., JUNQUAS, C., TAKAHASHI, K. (2016) Evidencing decadal and

interdecadal hydroclimatic variability over the Central Andes, Environmental Research Let-

ters 11 (9), 94016.

[80] SEIBERT, J., BEVEN, K.J. (2009) Gauging the ungauged basin: how many discharge meas-

urements are needed? Hydrology and Earth System Sciences 13, 883–892.

[81] SEIBERT, J., BEVEN, K.J. (2015) Gauging the Ungauged Basin: Relative Value of Soft and

Hard Data, Journal of Hydrologic Engineering 20 (1).

References

85

[82] SOLOMON, S. (1967) Relationship between precipitation, evaporation, and runoff in tropi-

cal-equatorial regions, Water Resour. Res. 3 (1), 163–172.

[83] SOROOSHIAN, S., GUPTA, V.K., FULTON, J.L. (1983) Evaluation of Maximum Likelihood Pa-

rameter estimation techniques for conceptual rainfall-runoff models: Influence of calibration

data variability and length on model credibility, Water Resour. Res. 19 (1), 251–259.

[84] SPERLING, F., VALIDIVIA, C., QUIROZ, R., VALDIVIA, R., ANGULO, L., SEIMON, A., NOBLE, I.

(2008) Transitioning to climate resilient development perspectives from communities in

Peru., Environment department papers no. 115.

[85] STEWART, M.K. (2015) Promising new baseflow separation and recession analysis meth-

ods applied to streamflow at Glendhu Catchment, New Zealand, Hydrol. Earth Syst. Sci.

19 (6), 2587–2603.

[86] SUMNER, D.M., JACOBS, J.M. (2005) Utility of Penman-Monteith, Priestley-Taylor, reference

evapotranspiration, and pan evaporation methods to estimate pasture evapotranspiration,

Journal of Hydrology 308 (1-4), 81–104.

[87] THORNTHWAITE, C.W. (1948) An Approach toward a Rational Classification of Climate, Ge-

ographical Review 38 (1), 55–94.

[88] TURC, L. (1961) Evaluation des besoins en eau d'irrigation, évapotranspiration potentielle,

formulation simplifié et mise à jour, Annales Agronomiques 12, 13–49.

[89] VEGAS GALDOS, F., ANDRES, N., ZAPPA, M., LAVADO CASIMIRO, W.S., HILKER, N. (2012)

Current availability of water resources, evaluation of TRMM rainfall data and future impacts

of climate change on water resources in the southern Peruvian Andes - Cuzco and Apuri-

mac Region, Report for WSL, Zurich, Switzerland.

[90] VERBUNT, M., ZAPPA, M., GURTZ, J., KAUFMANN, P. (2006) Verification of a coupled hydro-

meteorological modelling approach for alpine tributaries in the Rhine basin, Journal of Hy-

drology 324 (1–4), 224–238.

[91] VIVIROLI, D., ARCHER, D.R., BUYTAERT, W., FOWLER, H.J., GREENWOOD, G.B., HAMLET,

A.F., HUANG, Y., KOBOLTSCHNIG, G., LITAOR, M.I., LÓPEZ-MORENO, J.I., LORENTZ, S.,

SCHÄDLER, B., SCHREIER, H., SCHWAIGER, K., VUILLE, M., WOODS, R. (2011) Climate

change and mountain water resources: overview and recommendations for research, man-

agement and policy, Hydrol. Earth Syst. Sci. 15 (2), 471–504.

[92] VIVIROLI, D., DÜRR, H.H., MESSERLI, B., MEYBECK, M., WEINGARTNER, R. (2007) Mountains

of the world, water towers for humanity: Typology, mapping, and global significance, Water

Resour. Res. 43 (7).

[93] VIVIROLI, D., GURTZ, J., ZAPPA, M. (Eds.) (2007) The hydrological modelling system

PREVAH. Model documentation and user manual., B.D.o.G. Geographica Bernensia P40,

University of Berne.

[94] VIVIROLI, D., MITTELBACH, H., GURTZ, J., WEINGARTNER, R. (2009) Continuous simulation

for flood estimation in ungauged mesoscale catchments of Switzerland – Part II: Parame-

ter regionalisation and flood estimation results, Journal of Hydrology 377 (1–2), 208–225.

References

86

[95] VIVIROLI, D., SEIBERT, J. (2015) Can a regionalized model parametrisation be improved

with a limited number of runoff measurements?, Journal of Hydrology 529, 49–61.

[96] VIVIROLI, D., ZAPPA, M., GURTZ, J., WEINGARTNER, R. (2009) An introduction to the hydro-

logical modelling system PREVAH and its pre- and post-processing-tools, Environmental

Modelling & Software 24 (10), 1209–1222.

[97] VUILLE, M., BRADLEY, R.S., HEALY, R., WERNER, M., HARDY, D.R., THOMPSON, L.G.,

KEIMIG, F. (2003) Modeling δ18O in precipitation over the tropical Americas: 2. Simulation

of the stable isotope signal in Andean ice cores, Journal of Geophysical Research: Atmos-

pheres 108 (D6).

[98] WANG, T., ZLOTNIK, V.A. (2012) A complementary relationship between actual and poten-

tial evapotranspiration and soil effects, Journal of Hydrology 456–457, 146–150.

[99] WENDLING, U. (1975) Zur Messung und Schätzung der potentiellen Verdunstung, Zeit-

schrift für Meteorologie 25, 103–111.

[100] www: http://eclim-research.ch/ (accessed on November 18, 2017).

[101] www: http://www.enso.info/anhang/El_Nino_2015_16.pdf (accessed on available March

6, 2017).

[102] www: http://www.gitta.info/ContiSpatVar/en/html/SpatDependen_learningObject2.xhtml

(accessed on March 6, 2017).

[103] www: http://www.idmaperu.org/ampay/tour30.htm (accessed on November 18, 2017).

[104] www: http://www.peru.travel/de/reiseziele/apurimac.aspx (accessed on November 18,

2017).

[105] www: http://www.wmo.int/pages/prog/wcp/wcasp/docu-

ments/WMO_ENSO_Nov15_Eng.pdf (accessed on March 6, 2017).

[106] www: https://www.nasa.gov/feature/goddard/2016/nasas-imerg-measures-flooding-rain-

fall-in-peru (accessed on March 6, 2017).

[107] www: http://earthobservatory.nasa.gov/Features/ElNino/ (accessed on March 6, 2017).

[108] www: http://www.cmvo.it/ValleAnzasca (accessed on March 6, 2017).

[109] www: https://www.munichre.com/en/reinsurance/magazine/topics-online/2016/04/el-

nino-la-nina/index.html (accessed on March 6, 2017).

[110] ZAPPA, M., ANDRES, N., KIENZLER, P., NÄF-HUBER, D., MARTI, C., OPLATKA, M. (2015)

Crash tests for forward-looking flood control in the city of Zürich (Switzerland), Proc. IAHS

370, 235–242.

[111] ZAPPA, M., GURTZ, J. (2003) Simulation of soil moisture and evapotranspiration in a soil

profile during the 1999 MAP-Riviera Campaign, Hydrol. Earth Syst. Sci. 7 (6), 903–919.

[112] ZAPPA, M., JAUN, S., GERMANN, U., WALSER, A., FUNDEL, F. (2011) Superposition of

three sources of uncertainties in operational flood forecasting chains, Atmospheric Re-

search 100 (2–3), 246–262.

References

87

[113] ZAPPA, M., KAN, C. (2007) Extreme heat and runoff extremes in the Swiss Alps., Natural

Hazards and Earth System Sciences 7 (3), 375–389.

[114] ZHANG, Y., CHIEW, F. (2009) Relative merits of different methods for runoff predictions in

ungauged catchments, Water Resour. Res. 45 (7).

Appendix

88

Appendix

Appendix A: Tables

Table A- 1: Original runoff generation tuneable parameters of 44 Swiss and North Italian catch-

ments

Table A- 2: Numerical sensitivity analysis of the original runoff generation tuneable parameters

divided by cases and intervals (PART 1)

Table A- 3: Numerical sensitivity analysis of the original runoff generation tuneable parameters

divided by cases and intervals (PART 2)

Table A- 4: Tic_34 sensitivity analysis overview of tuned parameter and the corresponding nu-

meric sensitivity ranking (PART 1)

Table A- 5: Tic_34 sensitivity analysis overview of tuned parameter and the corresponding nu-

meric sensitivity ranking (PART 2)

Table A- 6: Tic_34 fine tuning to find modified donor parameter set and numerical sensitivity

school grade ranking

Appendix B: Figures

Original 44 donor parameter sets quantile visual sensitivity analysis for case 1-4 (Figure A- 1 to

Figure A- 15)

Tic_34 modification quantile visual sensitivity analysis for case 1-4 (Figure A- 16 to Figure A- 19)

Runoff hydrograph curves based on simulated daily (top)/ hourly (bottom) P/T and P HOBO data

for area 4 compared to V1 in-situ data for Tic_34 (Figure A- 20 to Figure A- 21

Runoff hydrograph curves based on simulated daily (top)/ hourly (bottom) P/T and P HOBO data

for area 4 compared to V1 in-situ data for Tic_34_mod (Figure A- 22 to Figure A- 23)

Boxplots water balance as comparison between simulated and in-situ data as an example for

area4 with V5+6 compared to V1 total runoff (Figure A- 24 to Figure A- 25)

Boxplots water balance as comparison between simulated and in-situ data for total catchment

area with V5+6 as total runoff (Figure A- 26 to Figure A- 27)

Barplots of in-situ water balance with precipitation contrary to scaled reference evapotranspira-

tion and total runoff (V5+6) for area4 (top) and the total catchment area (bottom) (Figure A- 28)

Appendix A

89

Appendix A

90

Appendix A

91

Appendix A

92

Appendix A

93

Appendix A

94

Appendix B

95

Figure A- 1: Quantile visual sensitivity analysis for case 1-4 for AlpEin (left), BibBib (middle) and Gas100 (right).

Appendix B

96

Figure A- 2: Quantile visual sensitivity analysis for case 1-4 for MinEut (left), rhb50 (middle) and Rhine4 (right).

Appendix B

97

Figure A- 3: Quantile visual sensitivity analysis for case 1-4 for Tic_01 (left), Tic_02 (middle) and Tic_03 (right).

Appendix B

98


Appendix B

99


Appendix B

100


Appendix B

101


Appendix B

102


Appendix B

103


Appendix B

104


Appendix B

105


Appendix B

106


Appendix B

107


Appendix B

108


Appendix B

109

Figure A- 15: Quantile visual sensitivity analysis for case 1-4 for Tic_37 (left) and Ver500 (middle).

Appendix B

110

Figure A- 16: Tic_34 fine tuning to find modified donor parameter set with mod1 (left), mod2 (middle) and mod3 (right).

Appendix B

111


Appendix B

112


Appendix B

113

Figure A- 19: Tic_34 fine tuning to find modified donor parameter set with mod10 (left) and mod11 (middle).

Appendix B

114

Figure A- 20: Runoff hydrograph curve based on simulated daily (top)/ hourly (bottom) precipitation HOBO data for area 4 compared to V1 in-situ data for Tic_34.

Appendix B

115

Figure A- 21: Runoff hydrograph curve based on simulated daily (top)/ hourly (bottom) precipitation and temperature HOBO data for area 4 compared to V1 in-situ data for Tic_34.

Appendix B

116

Figure A- 22: Runoff hydrograph curve based on simulated daily (top)/ hourly (bottom) precipitation HOBO data for area 4 compared to V1 in-situ data for Tic_34_mod.

Appendix B

117

Figure A- 23: Runoff hydrograph curve based on simulated daily (top)/ hourly (bottom) precipitation HOBO data for area 4 compared to V1 in-situ data for Tic_34_mod.

Appendix B

118

Figure A- 24: Diagram indicating the water balance as comparison between simulated and in-situ data as an exam-ple for area4 with V5+6 compared to V1 total runoff.

Appendix B

119

Figure A- 25: Separated diagram indicating the water balance as comparison between simulated and in-situ data as an example for area4 with V5+6 compared to V1 total runoff.

Appendix B

120

Figure A- 26: Diagram indicating the water balance as comparison between simulated and in-situ data for total catchment area with V5+6 as in-situ total runoff and V6 as baseflow.

Appendix B

121

Figure A- 27: Separated diagram indicating the water balance as comparison between simulated and in-situ data for the total catchment area with V5+6 as in-situ total runoff and V6 as baseflow.

Appendix B

122

Figure A- 28: Barplots of in-situ water balance with precipitation contrary to scaled reference evapotranspiration and total runoff (V5+6) for area4 (top) and the total catchment area (bottom).

Application of calibrated Swiss catchment model parameters ...

Documents