Assessment of erosion hotspots in a watershed: Integrating the WEPP model and GIS in a case study in the Peruvian Andes Guillermo A. Baigorria a, * , Consuelo C. Romero b a Agricultural and Biological Engineering Department, University of Florida, Gainesville, FL 32611, USA b Departamento de Suelos, Universidad Nacional Agraria La Molina, Lima 12, Peru Received 8 April 2005; received in revised form 3 January 2006; accepted 9 June 2006 Available online 22 August 2006 Abstract This paper presents a case study in assessment of erosion hotspots in an Andean watershed. To do this, we made use of an interface called Geospatial Modelling of Soil Erosion (GEMSE): a tool that integrates Geographical Information Systems (GIS) with the Water Erosion Predic- tion Project (WEPP) model. Its advantages are: (i) it is independent of any special GIS software used to create maps and to visualize the results; (ii) the results can be used to produce response surfaces relating outputs (e.g. soil loss, runoff) with simple inputs (e.g. climate, soils, topogra- phy); (iii) the scale, resolution and area covered by the different layers can be different among them, which facilitates the use of different sources of information. The objective of this paper is to show GEMSE’s performance in a specific case study of soil erosion in La Encan ˜ada watershed (Peru) where the hillslope version of WEPP has been previously validated. Resulting runoff and soil loss maps show the spatial distribution of these processes. Though these maps do not give the total runoff and soil loss at the watershed level, they can be used to identify hotspots that will aid decision makers to make recommendations and plan actions for soil and water conservation. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: Geospatial modeling; WEPP; GIS; Soil loss; Runoff; Andes Software availability Name of software: Geospatial Modelling of Soil Erosion (GEMSE) Developer and contact address: G.A. Baigorria, Frazier Rogers Hall, University of Florida, Gainesville, FL 32611, USA Coding language: Delphi 7 Software requirements: Any GIS software only for visualiza- tion purposes Hardware requirements: PCs with Windows 98, Windows 2000 or Windows XP. Program size: 1.1 Mb Available since: 2004 1. Introduction Modeling has formed the core of a great deal of research focus- ing on inherently geographic aspects of our environment, and has led to the understanding of distributions and spatial relationships in everything from astronomy to microbiology and chemistry (Parks, 1993). In the case of soil erosion, simulation models have become important tools for the analysis of hillslope and wa- tershed processes and their interactions, and for the development and assessment of watershed management scenarios (Santhi et al., 2006; Miller et al., 2007; Lu et al., 2005; Metternicht and Gonzales, 2005; He, 2003). Since erosion can adversely affect ecosystems on-site as well as off-site, the estimation of runoff and soil loss in catchments is becoming more important as con- cerns about surface water quality increase (Cochrane and Flana- gan, 1999). For this, the ‘‘hotspots’’ (source areas of sediments) within a watershed need to be identified. However, many of the predictive models do not examine the problem in a geographic context (Pullar and Springer, 2000). * Corresponding author: Fax: þ352 392 4092. E-mail address: [email protected]fl.edu (G.A. Baigorria). 1364-8152/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.envsoft.2006.06.012 Environmental Modelling & Software 22 (2007) 1175e1183 www.elsevier.com/locate/envsoft
9
Embed
Assessment of erosion hotspots in a watershed - Guillermo
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Assessment of erosion hotspots in a watershed: Integrating theWEPP model and GIS in a case study in the Peruvian Andes
Guillermo A. Baigorria a,*, Consuelo C. Romero b
a Agricultural and Biological Engineering Department, University of Florida, Gainesville, FL 32611, USAb Departamento de Suelos, Universidad Nacional Agraria La Molina, Lima 12, Peru
Received 8 April 2005; received in revised form 3 January 2006; accepted 9 June 2006
Available online 22 August 2006
Abstract
This paper presents a case study in assessment of erosion hotspots in an Andean watershed. To do this, we made use of an interface calledGeospatial Modelling of Soil Erosion (GEMSE): a tool that integrates Geographical Information Systems (GIS) with the Water Erosion Predic-tion Project (WEPP) model. Its advantages are: (i) it is independent of any special GIS software used to create maps and to visualize the results;(ii) the results can be used to produce response surfaces relating outputs (e.g. soil loss, runoff) with simple inputs (e.g. climate, soils, topogra-phy); (iii) the scale, resolution and area covered by the different layers can be different among them, which facilitates the use of different sourcesof information. The objective of this paper is to show GEMSE’s performance in a specific case study of soil erosion in La Encanada watershed(Peru) where the hillslope version of WEPP has been previously validated. Resulting runoff and soil loss maps show the spatial distribution ofthese processes. Though these maps do not give the total runoff and soil loss at the watershed level, they can be used to identify hotspots that willaid decision makers to make recommendations and plan actions for soil and water conservation.� 2006 Elsevier Ltd. All rights reserved.
Keywords: Geospatial modeling; WEPP; GIS; Soil loss; Runoff; Andes
Software availability
Name of software: Geospatial Modelling of Soil Erosion(GEMSE)
Developer and contact address: G.A. Baigorria, Frazier RogersHall, University of Florida, Gainesville, FL 32611,USA
Coding language: Delphi 7Software requirements: Any GIS software only for visualiza-
tion purposesHardware requirements: PCs with Windows 98, Windows
2000 or Windows XP.Program size: 1.1 MbAvailable since: 2004
1364-8152/$ - see front matter � 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.envsoft.2006.06.012
1. Introduction
Modeling has formed the core of a great deal of research focus-ing on inherently geographic aspects of our environment, and hasled to the understanding of distributions and spatial relationshipsin everything from astronomy to microbiology and chemistry(Parks, 1993). In the case of soil erosion, simulation modelshave become important tools for the analysis of hillslope and wa-tershed processes and their interactions, and for the developmentand assessment of watershed management scenarios (Santhiet al., 2006; Miller et al., 2007; Lu et al., 2005; Metternicht andGonzales, 2005; He, 2003). Since erosion can adversely affectecosystems on-site as well as off-site, the estimation of runoffand soil loss in catchments is becoming more important as con-cerns about surface water quality increase (Cochrane and Flana-gan, 1999). For this, the ‘‘hotspots’’ (source areas of sediments)within a watershed need to be identified. However, many of thepredictive models do not examine the problem in a geographiccontext (Pullar and Springer, 2000).
Under these circumstances, a Geographical InformationSystem (GIS) becomes a valuable tool. A GIS is a powerfulset of tools for collecting, storing, retrieving at will, transform-ing and displaying spatial data from the real world (Burrough,1986). GIS has made a tremendous impact in many fields ofapplication, because it allows the manipulation and analysisof individual ‘‘layers’’ of spatial data, and it provides toolsfor analyzing and modeling the interrelationships betweenlayers (Bonham-Carter, 1996). Coupled to an environmentalmodel, a GIS can interpret simulation outputs in a spatial con-text (Pullar and Springer, 2000). It is presumed that better in-tegration of GIS and environmental modeling is possible byexploiting the opportunity to combine ever-increasing compu-tational power, more plentiful digital data, and more advancedmodels. GIS/modeling tools necessarily encourage the bestimplementation of new and better ‘‘hybrid’’ tools. Accordingto Parks (1993), there are three primary reasons for integra-tion: ‘‘(1) spatial representation is critical to environmentalproblem solving, but GIS currently lack the predictive and re-lated analytic capabilities necessary to examine complex prob-lems; (2) modelling tools typically lack sufficiently flexibleGIS-like spatial analytic components and are often inaccessi-ble to potential users less expert than their makers; and (3)modeling and GIS technology can both be made more robustby their linkage and co-evolution.’’ Both GIS and simulationmodels have been developed with their own conventions, pro-cedures and limitations. However, linking them at a technicallevel does not guarantee improved understanding or usefulprediction (Burrough, 1986). More quantitative quality indica-tors, together with spatial statistics and error analysis, areneeded to improve the value of GIS/modeling interfaces(Hartkamp et al., 1999).
A comprehensive description of some of the most popularmodels of watershed hydrology in the world can be found inSingh (1995). As an example, we can mention some ofthem. The TOPMODEL (Beven et al., 1984) was developedas a distributed hydrologic model that uses digital elevationdata and spatial information on soil, vegetation and precipita-tion to estimate the soil moisture distribution at catchmentlevel, thereby taking account of the spatial heterogeneity ofboth topography and soils. One of the most promising of thephysically based models currently used to model erosion isthe Water Erosion Prediction Project (WEPP) model (Flanaganand Nearing, 1995). But it was not developed with a flexiblegraphical user interface for spatial and temporal scales appli-cations (Renschler, 2003). The first application of WEPPwith a raster-based GIS was by Savabi et al. (1995). Anothereffort to integrate WEPP and GIS was by Cochrane andFlanagan (1999) for watershed erosion modeling, using aninterface between Arc View and WEPP. In both cases, theintegration of WEPP with a GIS was done to facilitate andimprove the application of the model. Another computerinterface called Erosion Database Interface (EDI) processesthe surface hydrology output of the WEPP model resultingin a georeferenced estimation of erosion and runoff. The re-sults were erosion (Ranieri et al., 2002) and runoff (de Jongvan Lier et al., 2005) of a sugarcane growing area at
southeastern Brazil. The Geo-Spatial Interface for WEPP(GeoWEPP) (Renschler, 2003) is another example of a toolthat combines GIS and WEPP. It utilizes readily available dig-ital geo-referenced information from accessible Internet sour-ces like topographic maps, digital elevation models, land useand soil maps (Renschler et al., 2002), with the aim of evalu-ating various land-use scenarios to assist with soil and waterconservation planning. For those users of WEPP with no expe-rience with commercial GIS packages there is a new web-based WEPP-GIS system that only requires a user to havea network connection and web browser (Flanagan et al.,2004). The digital elevation data are processed on the serverside to delineate watershed, channels and hillslopes that,once located, WEPP simulations are conducted. Results ingraphical format are sent as images to the client computer.These two last examples’ applicability, however, can failwhere the availability of digital data is restricted, which oftenoccurs in developing countries.
This paper presents a new tool capable of integrating pro-cess-based models with Geographic Information Systems(GIS) for improving the analysis of point-estimated resultson larger scales. This interface, called Geospatial Modellingof Soil Erosion (GEMSE), makes use of the Water ErosionPrediction Project (WEPP) model, producing different mapsin GIS format as a result of this integration. Analysis of thesemaps gives insights useful for the evaluation of land resourcesand agricultural sustainability and for estimating risks in a spe-cific area.
2. Materials and methods
2.1. The study area
Field data for running the model were obtained in the northern Andean
Highlands of Peru, in La Encanada watershed. The study area is approximately
6000 ha and it is located at 7�4 0 S latitude and 78�16 0 W longitude, ranging
between 2950 and 4000 m above sea level (a.s.l.) (Fig. 1a).
Two main climate regimes can be identified during the year in this area: the
rainy season and the dry season. Three automatic weather stations were set up
in the study area to record the climate data on a daily basis. A summary of
climate conditions is shown in Table 1. A detailed description about rainfall
characteristics in the study area is given in Romero (2005) and Romero
et al. (in press).
According to the Soil Taxonomy classification (USDA and NRCS, 1998)
the main soil orders in the watershed are Entisols, Inceptisols and Mollisols
(INRENA, 1998). The spatial distribution of the main soil groups is shown
in Fig. 1b. In the highest part of the watershed there are deep soils with
a high content of organic matter. Shallow soils are also found; their low or-
ganic matter content is mainly because the topsoil has been removed by ero-
sion. Approximately 65% of the area has a slope gradient less than 15%. Very
steep slopes (up to 65%) are also present, increasing the risk of erosion in this
mountainous area. As steep slopes often occur adjacent to the river, water ero-
sion will contribute directly to the river sediment load.
The land use in La Encanada watershed is divided into croplands (55%),
cultivated pasture (13%), natural pasture (20%) and scrub (12%) (INRENA,
1998). Deep soils with the largest amount of organic matter are used as crop-
lands, with cereals, potato, maize and legumes the most important crops. How-
ever, crop yields are variable, depending on soil fertility and also on climatic
conditions. Poorly fertile shallow soils that show soil erosion characteristics
are also cropped, even though most of these areas are only appropriate for nat-
ural pasture (Proyecto PIDAE, 1995). The planting date for the main crop
varies temporally and spatially. For instance, a survey of the planting dates
In the present case study in La Encanada watershed, potato, barley and fal-
low land uses were simulated in different areas according to the land use map
of the study area (INRENA, 1998). In the case of crops, planting dates were
determined according to the field survey performed by Baigorria et al., (sub-
mitted for publication). These planting dates were established as the ones
used most frequently by the farmers in the study area.
2.7. Output generation
After the simulations, runoff and soil loss maps under different land uses
were aggregated. Note that the term soil loss represents the sediment yield out-
put from WEPP.
3. Results and discussion
3.1. Runoff
The runoff map for La Encanada watershed is shown inFig. 5. The estimated runoff values are the annual averageof a 4-year continuous simulation on simulated hillslopes of50 � 50 m (pixel size), expressed as mm year�1. We can ob-serve the runoff distribution on the map at pixel level or in ap-parently homogeneous areas presenting the same value. Theestimated runoff values ranged from <5 mm year�1 to40 mm year�1. Only a few pixels showed values over40 mm year�1. Two important areas are clearly visible onthe map: the northern area, presenting low values of runoff,and the central/southern area with the highest estimate of run-off. The northern part corresponds to the highest part of the
Outputs
Storing analyzed outputs
*.Sol *.Slp *.Man
ClimateDatabase
SoilDatabase
TopographicDatabase Management
GIS
&
dbf
*.Cli
Statistical data analysis
WEPP
Dbase file withpoint coordinates
Outputmaps
All pointcoordinates No
Yes
Fig. 4. Flowchart of GEMSE.
watershed, where deep soils are present and La Toma climateprevailed. The 4-year rainfall analysis in this watershed re-ported that around 90% of rainfall events had an intensityvalue <7.5 mm h�1 (Romero, 2005; Romero et al., in press).A higher number of rainfall events with intensity values>7.5 mm h�1 were observed in Manzanas (16 events, witha maximum intensity of 147 mm h�1) than La Toma (7 events,with a maximum intensity of 130 mm h�1), which indicatedthat the former area could be prone to suffer more runoff orerosion effects.
The combined effect of the low erosive events plus the deepsoils found in the La Toma area promoted the infiltration ofwater and resulted in a low runoff production, shown on themap as the white area. Eighty percent of the surface areahad estimated values of runoff <5 mm, as we can see in thehistogram (Fig. 6a). Therefore, this area can be considereda stable zone or the buffer zone protecting the bottom of thewatershed. The main land use of this zone is natural pasture,which acts as a protective cover for the soil.
The central and southern part of the watershed, where Man-zanas is located, is the area where most crops are cultivatedand had more number of rainfall events with >7.5 mm h�1 in-tensities. This area is also prone to get flooded easily due tothe bad drainage characteristics of its soils. Greater amountsof estimated runoff can be identified on the map: almost
Fig. 5. Runoff map of La Encanada using the GEMSE interface and the WEPP
15% of the area of the map has estimated values from 5 to20 mm, and 5% has estimates exceeding 20 mm (Fig. 6a).The variability of climate, soils, slope and management iswell represented by the model.
3.2. Soil loss
The estimated soil loss map of La Encanada is shown inFig. 7. The results of running the model for 4-year continuoussimulation on each pixel of the DEM (representing hillslopesof 50 by 50 m) are expressed in Mg ha�1 year�1. Each pixelrepresents a single slope profile where the WEPP model wasapplied. GEMSE does not consider flow from cell to cell inthe DEM. The map shows areas susceptible to erosion. Asin the runoff map, we can observe two regions within the wa-tershed. The northern area, with low soil loss rates (<10 Mgha�1 year�1) corresponds to the area with the lowest estimatedrunoff in Fig. 5. This area is usually under natural pasture, alsopreferred by farmers for growing cereals, which has the char-acteristic to protect the soil surface against the erosivity ofrainfall. In the simulation, we established barley since it isthe crop that most resembles the natural pasture that normallygrows in this area. In addition, farmers do not disturb the soilwhen sowing barley. This is why most of the area does notshow a great amount of soil loss. However, there are someplots where higher values of soil loss can be observed thatwould correspond to those unprotected areas that normallyare located on the steepest slopes facing the river.
Fig. 6. Histograms showing the percentage of the area under different esti-
mated values of runoff (a) and soil loss (b).
The central part of the watershed, where most of the farm-ing occurs, has pixels with different estimated soil loss values,representing the variability of soils, land use (crop or fallow),slope and climate. The lowest part of the watershed presentslow values of soil loss, since this area corresponds to the flat-test part of the watershed (valley); due to the availability ofwater it is cropped year-round with improved pastures. Forthese two areas, the estimated soil loss values ranged from<10 Mg ha�1 year�1 to >150 Mg ha�1 year�1.
Although it seems that the model predicts high rates of soilloss in the area, a different picture emerges when a histogramof the quantification of pixels is made: on almost 58% of thetotal area the estimates of soil loss are low (<10 Mg ha�1
year�1), nearly 10% of the area has estimates 25e50 Mgha�1 year�1, 12% has estimates from 50e100 Mg ha�1 year�1,10% has estimates from 100 to 150 Mg ha�1 year�1 and only10% has estimates >150 Mg ha�1 year�1 (hotspots) (Fig. 6b).The model estimates high values of soil loss (>100 Mg ha�1
year�1) specifically in those areas where slope angle exceeds40� (78% gradient).
It seems unlikely that, for example, 30 mm year�1 of runoffis able to carry 125 Mg ha�1 year�1 in this watershed. Thiswould mean 417 g of sediment per liter of runoff. However,a maximum value of 395 g of sediment per liter of runoffwas recorded at a runoff plot at the bottom of the watershedin a sandy clay loam soil at 10% slope inclination, during
Fig. 7. Soil erosion map of La Encanada watershed using the GEMSE interface
the previous validation study of the hillslope version of theWEPP model (Romero et al., submitted for publication).Note that the climate map shown in Fig. 1c had much influenceon the resulting runoff and soil loss maps, giving two well-defined areas in the maps concerned. This would be improvedif the interface could use high-resolution climate maps. Afterthe study was completed a better climate map for this specificarea became available (Baigorria et al., 2004; Baigorria, 2005);it is intended to test the interface with this new input.
4. Conclusions
GEMSE is operational software that integrates GIS prop-erties with the Water Erosion Prediction Project (WEPP)model in order to analyze the spatial variation of runoffand soil loss. In the present study, the objective was to testthe performance of GEMSE in generating soil loss and run-off maps from the WEPP model outputs in La Encanada wa-tershed (northern Peru). The generation of these maps madeeasier the visualization of the erosion process at spatial andtemporal scales according to the actual land use of thewatershed.
Areas at risk of runoff and soil loss were identified fromthe maps. For runoff, the risk areas were associated with theflattest part of the watershed. For soil loss, the susceptibleareas were related to the steepest slopes within the water-shed. Although the map does not give the total soil lossat the watershed level, it can be used to identify the mostsusceptible areas to be eroded in the area (what we called‘‘hotspots’’), thus helping not only farmers but decisionmakers to formulate recommendations for soil and waterconservation strategies. GEMSE can be used in either smallor large watersheds. This demonstrates that GEMSE is anoption that can be used for strategic applications of theWEPP model.
Acknowledgements
The International Foundation for Science (IFS), Stock-holm, Sweden, supported this research through a grant toC.C. Romero and USAID & Soil Management e CRSP pro-ject funded grant No. 291488. Joy Burrough advised on theEnglish. The author conducted part of this research whilein association with the Department of Production Systemsand Natural Resources Management of the International Po-tato Center. Thanks are due to Prof. L. Stroosnijder for hisuseful comments in the manuscript. The authors wish tothank three anonymous reviewers for their observations andadvice which helped us significantly improve the presentpaper.
References
Baigorria, G.A., 2005. Climate interpolation for land resource and land use
studies in mountainous regions. Dissertation thesis, Wageningen Univer-
sity and Research Centre, Wageningen, The Netherlands.
Baigorria, G.A., Stoorvogel, J.J., Romero, C.C., Quiroz, R., Weather and sea-
sonal-climate forecast to support agricultural decision-making at different
spatial and temporal scales. Computer and Electronics in Agriculture, sub-
mitted for publication.
Baigorria, G.A., Villegas, E.B., Trebejo, I., Carlos, J.F., Quiroz, R., 2004. At-
mospheric transmissivity: distribution and empirical estimation around the
central Andes. International Journal of Climatology 24, 1121e1136.
Beven, K.J., Kirby, M.J., Schoffield, N., Tagg, A., 1984. Testing a physically-
based flood forecasting model TOPMODEL for three UK catchments.