Monitoring variation of water turbidity and related ... · Nature Reserve, China Abstract Poyang Lake is the largest fresh water lake in China. There are pronounced spatial-temporal

Monitoring variation of water turbidity and related environmental factors in Poyang Lake

National Nature Reserve, China

Liu Wei March, 2007

Monitoring variation of water turbidity and related environmental factors in Poyang Lake National

Nature Reserve, China

by

Liu Wei Thesis submitted to the International Institute for Geo-information Science and Earth Observation in partial fulfilment of the requirements for the degree of Master of Science in Geo-information Science and Earth Observation, Specialisation: Environmental Systems Analysis and Management Thesis Assessment Board Chairman: Prof. Dr. Ir. Alfred de Gier, NRS Department, ITC External examiner: Primary supervisor: Associate Prof. Dr. Chris Mannaerts, WRS Department, ITC Secondary supervisor: Prof. Dr. Liu Yanfang, SRES, Wuhan University Internal examiner: Dr. Michael Weir, NRS Department, ITC Internal examiner: Prof. Dr. Liu Yaolin, SRES, Wuhan University

INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATION

ENSCHEDE, THE NETHERLANDS

Disclaimer This document describes work undertaken as part of a programme of study at the International Institute for Geo-information Science and Earth Observation. All views and opinions expressed therein remain the sole responsibility of the author, and do not necessarily represent those of the institute.

Monitoring variation of water turbidity and related environment factors in Poyang Lake National Nature Reserve, China

Abstract Poyang Lake is the largest fresh water lake in China. There are pronounced spatial-temporal patterns in Water turbidity in Poyang Lake National Nature Reserve (NNR), China. This study aims to understand the temporal and spatial dynamics of water turbidity and related environmental factors which influence the water turbidity situation in Poyang Lake NNR. The Moderate Resolution Imaging Spectroradiometer (MODIS) of medium-resolution data were used in this research for mapping the tempo-spatial dynamics of Secchi Disk Depth (SDD) in Poyang Lake National Nature Reserve. Remote sensing images have successfully been applied to map SDD which is selected as the indicator of water turbidity of inland water bodies. A most suitable empirical model between MODIS reflectance and SDD which is validated by the field data is used to mapping the water turbidity situation. There are pronounced temporal dynamics, which are high water transparency values or SDD are observed during the summer season while the most turbid situations always occur in winter. In different years, the trend is similar while the occurrence of detailed peaks (high or low SDD values) is a little different in the same lakes. Comparing the situation in different seasons, the most turbid places show in different directions. The turbidity difference in the low-water season is less than the varying in the other seasons. According to the seasonal and temporal dynamics of the water turbidity situation in Poyang Lake NNR, the roles of related environmental factors are analyzed. Statistical methods were used to quantify the influence of some factors such as water level, wind speed, temperature and rainfall. A multivariate regression model is built which permitted to relate lake turbidity and SDD to water level and mean wind speed variations in Poyang Lake NNR area. Further statistic analysis is used to judge the accuracy of the model. Some ancillary environmental factors which also can play a role such as fishing, dredging and bird’s influence are analyzed by theoretical deduction, supported by field investigations. Also historical data of the Poyang Lake NNR were used to assist in the analysis. Key Words: MODIS, water turbidity, Poyang Lake National Natural Reserve, Secchi Disk Depth, spatio-temporal dynamics

i


Acknowledgements When I finished my postgraduate thesis in the end, I would like to record my hearty thanks and deep appreciation to all of organizations and persons that gave me generous support during this time. I would like to acknowledge the International Institute for Geo-Information Science and Earth Observation (ITC) and Wuhan University for giving me this precious chance to study in the Netherlands. It broadened my view and expanded my knowledge. Thanks to all the NRM and other ITC staff, my classmates. The precious time together with you will have a special place in my heart! My deepest thanks were addressed to my supervisors Associate Prof. Dr. Chris Mannaerts for his invaluable guidance and support. I feel so fortunate to have the chance to carry out my work under his supervision. Thanks to my Wuhan supervisor Prof. Dr. Liu Yanfang. She provided many suggestions and critical reviews to my thesis. To Dr. Jan. D. Leeuw and Dr. Michael Weir, I give my heartfelt thanks to you for your continuous help from the time even before I went to ITC. I want to express my appreciation to Prof. Dr. Liu yaolin, Prof. Zhang wanshun, Prof. Du Qingyun, Prof. Li Lin, Ph.D candidate Huyong who kindly offered me precious advices on the research methods. Thanks to Dr Hu Chaozhen, Dr Xiao Mei, Dr Liu Huilong and other staff in Chemistry Research Laboratory in School of Resource and Environmental Science in Wuhan University for their help and guidance of the laboratory work. I am grateful to all the staffs of Poyang Lake National Nature Reserve who have provided me great help during my fieldwork, especially Mr. Zhou Feilong, Mr.Yi Wusheng and Mr. Zeng Nanjing. My special thanks go to Ph. D candidate Wu Guofeng for his great support and encouragement. Through the crossing-continent discussions by E-mail and telephone with him, many doubts and obstacles in the research had been cleared. Thanks to all my dearest Chinese friends there in ITC. I will remember your friendship forever, especially your encouragement and taking care when I was ill in the Netherlands. My genuine thanks also go to all classmates and colleagues who have helped me in my study and living. Thanks to Nilam, Orgil, Paco, Boris, Emma, and Dr. Rossiter for your friendship and courage all the time. The library staffs are also thanked for their enthusiastic help. My volleyball friends also created me an important memory. Thanks for all the persons that made me grown up. Finally, I’d like to say to my parents: I love you! Thank you so much for creating me a warm and fragrant family atmosphere, and giving me so much support! Without you, I can not go this far. I love you forever!

ii


Table of contents 1. Introduction ......................................................................................................................................1

1.1. Background .......................................................................................................................1 1.2. Research Problems ............................................................................................................2 1.3. Objective ...........................................................................................................................3

1.3.1. General objective .......................................................................................................3 1.3.2. Specific objectives .....................................................................................................3

1.4. Specific research questions................................................................................................3 1.5. Research Approach............................................................................................................4

2. Study Area and Materials .................................................................................................................5 2.1. Study area ..........................................................................................................................5

2.1.1. Location .....................................................................................................................5 2.1.2. Hydrology and Climate..............................................................................................5

2.2. Research materials.............................................................................................................7 2.2.1. Satellite images ..........................................................................................................7 2.2.2. Field data....................................................................................................................7 2.2.3. Ancillary data.............................................................................................................7

3. Research methods .............................................................................................................................8 3.1. Field data collection and analysis......................................................................................8

3.1.1. Field survey................................................................................................................8 3.1.2. Laboratory measurement and analysis.......................................................................9

3.2. Relation between SDD and water turbidity.....................................................................10 3.2.1. Some related terminologies......................................................................................10 3.2.2. The relation between SDD and TSS, SDD and Turbidity .......................................11

3.3. Mapping temporal and spatial dynamics of water turbidity by MODIS images .............12 3.3.1. Remote sensed data acquisition and processing ......................................................12 3.3.2. Empirical approach ..................................................................................................14 3.3.3. Empirical models related and validation..................................................................15 3.3.4. Detecting the pattern of water turbidity by MODIS images....................................16

3.4. Analysing the temporal dynamics of water turbidity using a statistical multivariate model 17

3.4.1. Analyzing the related environmental factors ...........................................................17 3.4.2. Statistic methods related ..........................................................................................18 3.4.3. A statistical multivariate water turbidity model.......................................................19

4. Results ............................................................................................................................................20 4.1. SDD distribution maps in Poyang Lake NNR using satellite data ..................................20 4.2. Spatial dynamics of water turbidity in Dahuchi Lake of Poyang Lake NNR .................27 4.3. Temporal dynamics of water turbidity in Poyang Lake NNR.........................................28

4.3.1. Temporal dynamics observed from MODIS in Dahuchi Lake ................................28 4.3.2. Temporal dynamics observed from MODIS in Shahu and Banghu Lake ...............30

4.4. Analysing the effects of environmental factors ...............................................................30 4.4.1. Related environmental factors .................................................................................30 4.4.2. The correlation between SDD and environmental factors .......................................31 4.4.3. Error analysis ...........................................................................................................32

iii


4.4.4. The temporal variation of SDD predicted by the multivariate regression model ....33 4.4.5. Accuracy validation between temporal pattern predicted by MODIS images and by the multivariate regression model ..........................................................................................34

5. Conclusions and discussion ............................................................................................................36 5.1. Conclusions on the spatial and temporal variation of water turbidity .............................37 5.2. Conclusions on the effects of related environmental factors...........................................37 5.3. Recommendations and future research on monitoring water turbidity............................38

References ..............................................................................................................................................39 6. Appendix ........................................................................................................................................43

iv


List of figure Figure 1-1 Framework and structure of the study ....................................................................................4 Figure 2-1 The geographical locations of Poyang Lake NNR and its inclusive lakes .............................5 Figure 2-2 The meteorological and hydrology situation in year 2004 (left) and 2005 (right) .................6 Figure 3-1 Field work samples.................................................................................................................9 Figure 3-2 Distribution, Turbidity, TSS and SDD ...................................................................................9 Figure 3-3 Locations and COD numbers of the samples........................................................................10 Figure 3-4 Relation between SDD and TSS, SDD and Turbidity..........................................................12 Figure 3-5 Flowchart of images processing ...........................................................................................14 Figure 3-6 Scatter plots revealing the agreement of SDD predicted from MODIS images with SDD by field survey.............................................................................................................................................16 Figure 3-7 Distribution of the residual ...................................................................................................16 Figure 3-8 Flowchart of detecting seasonal pattern of water turbidity ..................................................17 Figure 3-9 Analysis the influence of wind on water turbidity ...............................................................19 Figure 4-1 The SDD images in year 2004..............................................................................................22 Figure 4-2 The SDD images in year 2005..............................................................................................23 Figure 4-3 The SDD images in year 2006..............................................................................................25 Figure 4-4 SDD images in Dahuchi Lake in 2004 .................................................................................27 Figure 4-5 SDD images in Dahuchi Lake in 2005(left) and 2006(right) ...............................................28 Figure 4-6 Temporal dynamics of SDD derived from MODIS images in 2004 ....................................29 Figure 4-7 Temporal dynamics of SDD derived from MODIS images in 2005 ....................................29 Figure 4-8 Temporal dynamics of SDD from MODIS in Shahu Lake and Banghu Lake in 2005 and 2006........................................................................................................................................................30 Figure 4-9 Scatter plots revealing the agreement of SDD predicted using the multivariate model with field observed SDD ................................................................................................................................32 Figure 4-10 Distribution of the regression residuals between modelled and observed data ..................33 Figure 4-11 Temporal dynamics of SDD derived from the multivariate regression model in 2004......33 Figure 4-12 Temporal dynamics of SDD derived from the multivariate regression model in 2005......33 Figure 4-13 Scatter plots revealing the agreement of SDD predicted using the SDD from MODIS.....34 Figure 4-14 Environmental factors related with water turbidity and their effect ...................................35 Figure 4-15 The number of birds in 2004 (left) and 2005 (right) ..........................................................36 Figure 4-16 Temperature in 2004 and 2005...........................................................................................36

v


List of tables Table 2-1 Dates of the MOD09 images used in the study........................................................................7 Table 3-1 MODIS land products ............................................................................................................12 Table 3-2 MODIS visible, thermal bands and potential applications.....................................................13 Table 3-3 MODIS/Terra Surface Reflectance Daily L2G Global 500m SIN Grid V004 products........13 Table 4-1 Correlation matrix between SDD and environmental factors ................................................31 Table 4-2 Summary of the regression analysis between average SDD and environment factors ..........32

vi


Abbreviations MODIS Moderate Resolution Imaging Spectroradiometer NNR National Nature Reserve TSS Total Suspended Solids SDD Secchi Disk Depth SSC Suspended Sediment Concentration TM Thematic Mapper DEM Digital Elevation Model EOS Earth Observation Satellite COD Chemical Oxygen Demand NTU Nephelometric Turbidity Units EOS Earth Observation Satellite ANOVA Analysis of Variance s.e. Standard Error

vii


1. Introduction

1.1. Background

With the launch of Moderate Resolution Imaging Spectrometer (MODIS) on board the Terra and Aqua satellite, water monitoring applications have increased continuously (Duane Nellis et al. 1998; Houborg and Soegaard 2004; Miller and McKee 2004). Multiple channels of MODIS Terra/Aqua satellites provide data for land, including inland water studies at 500 m spatial resolution. Besides, two bands of 250 m spatial resolution, 620–670 nanometer (nm) which is band 1 and 841–876 nm which is band 2 can be used to monitor the optical properties of inland waters (Houborg et al. 2007). The advantage of MODIS instrument is that near-real time images are provided frequently at 2–4 images per day and the spatial resolution is appropriate for monitoring suspended particulate matter distribution and water turbidity. A common disadvantage of using MODIS remote sensing data for monitoring water turbidity in high latitudes is when frequent cloud coverage happens, to obtain satellite images information on water turbidity distribution are not available (Tan et al. 2006). In this case study, application of MODIS images for monitoring the seasonal and spatial situation and change of suspended matter and water turbidity are present. The study area is located Poyang Lake National Nature Reserve (NNR) in Poyang Lake Wetland, one of the world’s six top wetlands designated for the List of Wetlands of International Importance (http://www.poyanglake.org/). The optical properties (i.e. reflectance) of water depend on the concentration and character of suspended sediments, phytoplankton and dissolved organic matter (Scheffer 1998). Sensors aboard satellites can measure the amount of solar radiation at various wavelengths reflected by surface water, which can be correlated to water quality parameters, for examples, Total Suspended Solids (TSS). It is more practical to monitoring water quality with remote sensing (Macia et al. 2003). This constitutes one of the means to estimate water quality, which offers three significant advantages over ground sampling. First, the near-continuous spatial coverage of satellite imagery allows for synoptic estimates over large areas. Second, the global coverage of satellites allows for the estimation of water quality and variation in remote and inaccessible areas. Third, the relatively long record of archived imagery (e.g. Landsat since the early 1970s) allows estimation of historical water quality, when no ground measurements can possibly be performed (Sawaya et al. 2003). However, there are also significant disadvantages of satellite estimates. First, the ability to distinguish among the various constituents of the water is limited. Second, the sampling depth is limited to the surface, varies with water clarity and is not controllable. Third, the spatial and temporal resolution can be inadequate and is not controllable (Hellweger et al. 2004). Whereas either of these approaches can be used alone, the combination of ground and satellite estimates is often the most effective approach. As an example, satellite imagery can be used to extrapolate ground measurements to areas and times with little or no coverage. This reduces the number of ground samples and increases the spatial and temporal coverage of the estimates. Satellite estimates of water monitoring have found widespread application. Many parameters including TSS, Secchi Disk Depth (SDD), Suspended Sediment Concentration (SSC) and Turbidity are

1


estimated using the MODIS sensor onboard the Terra and Aqua satellites (Hellweger et al. 2004; Hu et al. 2004). The turbidity of lakes is highly dynamic. Temporal variations of the water turbidity are among the most important factors that influence observed patterns of species abundance, composition and biological mechanisms. Water turbidity dynamic have been attributed to factors including phytoplankton (such as algal blooms), the concentration and character of suspended sediments, detritus and nutrient factors (dissolved organic matter) etc. (Marten Scheffer 1998; Scheffer 1998; Gomez et al. 2004). Water turbidity also reveals spatial variation. Such patterns are not easily revealed by in situ based measurement which is time-consuming, but are very easily recognized in remote sensing imagery. Satellite systems for remote sensing provide an ability to measure the optical characteristics of suspended sediment (Duane Nellis et al. 1998). Landsat Thematic Mapper (TM) was used to estimate the magnitude and spatial variability of water quality in a major Kansas reservoir, Tuttle Creek, as it progressed through the flood event of 1993 (Duane Nellis et al. 1998). The above indicates that water turbidity reveals spatial and temporal patterns. Satellite estimates of water turbidity have found widespread application to assess such spatial and temporal dynamics. Satellite imagery such as MODIS and TM can be used to monitor the distribution of the water quality and turbidity, such as mentioned in the research in New York harbour (Hellweger et al. 2004). TM imagery is used to detect the temporal and spatial sediment re-suspension in the Changjiang Estuary (Yang et al. 2001). Multiple channels of MODIS Terra/Aqua satellites is also applied to monitor the transport of suspended matter (Sipelgas et al. 2006). Poyang Lake has remarkable spatial and temporal patterns in water turbidity (Wu and Ji 2002; Liu 2006). It appears more turbid near the mouths of inland rivers, the connected area between Poyang Lake and the Yangtze River. From the north to south, the turbidity decreases. The SSC is extremely high in the middle of Poyang Lake at the high-water season (Liu Qian 2006). They suggested a number of possible causes for this. In the main part of Poyang Lake, the main reasons are the phenomenon that the sediment comes from the Yangtze River during the flood season, the human activity of exploiting sand (Liu Qian 2006). There are also pronounced spatial-temporal patterns in water turbidity in Poyang Lake NNR (Wu and Ji 2002). Wu et al. (2002) reported that lakes at the periphery of Poyang Lake NNR were clear in summer and become turbid in winter. The characteristic of the dynamics was different for different lakes. The water clarity fluctuations of Dahuchi Lake and Dachahu Lake are larger, while those of Banghu Lake and Shahu Lake are smaller. Some possible causes were given for this. The water level is the main reason to the spatial dynamics (Liu Qian 2006). The after effect of the foraging activity of over-wintering of waterfowls affects the spatial pattern in April, May or October. And also unexpected phenomena, such as storms and water backflow, as well as some potential possible reasons such as fishing and dredging affect. The bottom-feeding swans maybe cause high SSC in winter (Liu Qian 2006).

1.2. Research Problems

Poyang Lake is the largest freshwater lake in China. Although the water quality is still good, erosion, degradation, negative influences of toxic pollution from heavy metals and human activities (exploiting

2


sands, fisheries, transport and industry) all threaten its water quality. At the same time, the spatial and seasonal change of the water turbidity in the lakes in Poyang Lake NNR is obvious. Present operative monitoring of lakes waters in Poyang Lake NNR includes point samples taken and measured weekly and annually. Some lakes have good water clarity in summer but become turbid in winter. Even in the same lake the turbidity distributes differently. This phenomenon can also be seen from satellite imagery but related research is not too much. Remote sensing water area studies have been carried out before in Poyang Lake (Liu 2006), but so far the use of satellite imagery in water turbidity pattern studies, especially the related influencing environment factors has not been reported for Poyang Lake NNR. More awareness should be put on this. How does the turbidity change seasonally and spatially in different lakes? How do environmental factors influence the turbidity and sediment? Using scientific and technical methods to do monitor and analysis can help a lot on the protection and management of the Poyang Lake NNR and the decision-making for the government. In particular, satellite remote sensing for assessment and monitoring of lake conditions can supply timely and reliable information to the decision-maker.

1.3. Objective

1.3.1. General objective

The objective is to study the seasonal and spatial variation of water turbidity in lakes in Poyang Lake NNR and to investigate the relation with environmental factors known to influence water turbidity.

1.3.2. Specific objectives

To detect the seasonal patterns in the water turbidity mainly in Dahuchi Lake of the Poyang Lake NNR

To detect eventual differences in water turbidity patterns among the lakes in Poyang lake NNR To analyse the causes of the seasonal patterns and the reasons of different seasonal pattern

between lakes To investigate whether wind induced re-suspension can explain the observed patterns in turbidity To investigate the possible role of other factors (e.g. birds, lake water fauna) in explaining the

turbidity during winter

1.4. Specific research questions

What is the seasonal pattern of water turbidity in lakes of Poyang Lake NNR? Do lakes of Poyang Lake NNR differ in temporal or seasonal turbidity pattern? What factors cause this seasonal turbidity pattern in Dahuchi Lake? Lake level variations, wind

driven bottom sediment re-suspension, presence of birds and feeding, local dredging? What factors cause the different seasonal turbidity pattern between lakes? Why do other lakes not

become turbid in winter? Water levels? Sediment texture? Other factors? Is there a relation between the turbidity and arrival or the number of birds? How do other factors

influence the water turbidity

3


1.5. Research Approach

Original MODIS09

images

Projection transformation

Clip image to Poyang Lake

NNR

Export to Erdas format

Resampling of Bands3-7 from 500m to 250m

Layer Stacking

Corrected MODIS09

images (Poyang Lake NNR)

Wind speed & wind direction

Water level

precipitation

Historical data

Analytical model

SDD

Experimental model

Field data

Temporal pattern of

SDD

Statistic analysis Simulation

Temporal pattern of

SDD

Step1: SDD dynamics from MODIS images

Step2: SDD dynamics got from environment factors

Step3: Accuracy evaluation

Temporal pattern of

SDD (step1)

Temporal pattern of

SDD (step2)

Spatial pattern of

SDD

Accuracy evaluation

How did the environment factors influence the

water turbidity pattern

Figure 1-1 Framework and structure of the study

4


2. Study Area and Materials

2.1. Study area

2.1.1. Location

Poyang Lake NNR located between 115º55′ - 116º03′E, 29º05′ - 29º 15′N is in the south part of middle Yangtze River. It’s in the north-east part of Poyang Lake in JiangXi Province which is shown in red in the sub-map of the left map in Figure 2-1, southern China (Jiang and Piperno 1999). The reserve is dominated by 9 lakes: Dachahu (8500 ha), Banghu (7300ha), Dahuchi (3000ha), Shahu (1400ha), Changhuchi (700ha), Zhonghuchi (600 ha), Xianghu (400 ha), Meixihu (300 ha) and Zhushihu (200 ha). The total area is about 22,400 hectares (Wu and Ji 2002). Figure 2-1 shows the geographical locations of Poyang Lake NNR and its inclusive lakes. The left map of China shows the location of Poyang Lake in China. The right map of Poyang Lake NNR describes the locations of inclusive lakes.

Figure 2-1 The geographical locations of Poyang Lake NNR and its inclusive lakes

1: Changhuchi; 2: Zhonghuchi; 3: Xianghu; 4: Meixihu; 5: Zhushihu.

2.1.2. Hydrology and Climate

Poyang Lake NNR is one of the most important bird sanctuaries in the world. Hundreds of thousands of birds over-winter here every year because of the specific climate and hydrology situation (Wu and Ji 2002). From the meteorological and hydrology data collected from Poyang Lake NNR which is shown in Figure 2-2., we obtained some detail information about the wind speed, precipitation and water level in 2004 and 2005 in Poyang Lake NNR.

5


0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1 31 61 91 121 151 181 211 241 271 301 33Julian

Total precipitation

1 361 day

0

1

2

3

4

5

6

7

8

9

10

11

12

13

1 31 61 91 121 151 181 211 241 271 301 33

Julian day

Mean wind speed (m/s)

1 361

14

14.5

15

15.5

16

16.5

17

17.5

18

18.5

19

Jan-

05

Feb-

05

Mar-

05

Apr-

05

May-

05

Jun-

05

Jul-

05

Aug-

05

Sep-

05

Oct-

05

Nov-

05

Da

Water Level

Dec-

05

te

(m)

Year 2004 Year 2005

14

14.5

15

15.5

16

16.5

17

17.5

18

18.5

19

Jan-04

Feb-04

Mar-04

Apr-04

May-04

Jun-04

Jul-04

Aug-04

Sep-04

Oct-04

Nov-04

Dec-04

Date

Water Level (m)

0

1

2

3

4

5

6

7

8

9

10

11

12

13

1 31 61 91 121 151 181 211 241 271 301 331 361

Julian day

Mean wind speed (m/s)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1 31 61 91 121 151 181 211 241 271 301 331 361Julian day

Total Precipitation

Figure 2-2 The meteorological and hydrology situation in year 2004 (left) and 2005 (right)

Receiving water from five inland rivers and moderating floodwater from the Yangtze River lead to a complex hydrology condition in Poyang Lake. The water level fluctuates seasonally, with low water in winter and highest levels in July and August (Wu and Ji 2002). The maximum and minimum water levels in 2004 were17.7 m in July and 14.5 m March. While in 2005, 18.6 m in September and 14.8 m December respectively were the maximum and minimum water levels instead. The water level was low in winter and early spring; it rose sharply in the late spring and peaked in the summer; then it started to fall in autumn and retained a low level in the whole wintertime. It provides the suitable wetland environment for the birds living in winter when the water level is low. The large variation of water level between low-water season and high-water season together with the wind induced suspended sediment make a great difference of water turbidity situation.

6


The precipitation figures show the daily values during the years 2004 and 2005. From these total precipitation graphs in Figure 2-2, we can see that the relatively high precipitations in 2004 occurred in April and June. There was no precipitation in October. The mean daily wind speed is also fluctuant all year around while there are still two peaks during spring and autumn. The fastest wind speed is around 9.5 m/s in April in 2004 and 12.1 m/s in March in 2005. The wind speed is not very high in summer and winter which is around 4 m/s.

2.2. Research materials

2.2.1. Satellite images

MOD09, a seven-band product computed from the MODIS level 1B bands 1 to 7, cantered at 645, 858, 470, 555, 1240, 1640 and 2130 nm respectively is used in this research. The data have been corrected for the effect of atmospheric gases, aerosols and thin cirrus clouds and thus provide an estimate of the spectral reflectance as it would be measured at ground level in the absence of atmospheric scattering or absorption (http://modis-sr.ltdri.org/html/surfref.htm accessed 16th February 2006). MOD09GQK contains MODIS bands 1 and 2 at 250 m spatial resolution, while MOD09GHK stores MODIS bands 1 to 7 at 500 m spatial resolution. The geo-location accuracy of the satellite is approximately 50 m at nadir following procedures to eliminate bias and other sources of error (Wolfe et al. 2002). MOD09 images of three years from 2004 to 2006 were selected and 34 cloud-free images were downloaded at the frequency of one image per month. Table 2-1 shows the date of the MOD09 images selected and finally used in the study.

Table 2-1 Dates of the MOD09 images used in the study Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

2004 21 14 10 9 30 22 28 16 4,28 18 12 2005 14 22 21 27 3 5 17 9 18,31 26 7 2006 9 7 3 17 20 11 12 14 26 24

2.2.2. Field data

During the field campaign, a number of 48 water samples were taken from the lakes. At every sample point, a bottle of lake water was taken as a sample for the later laboratory analysis. The longitude and latitude of every sample, SDD were recorded. TSS, Turbidity, SSC of each spot and COD on 3 different part of the Dahuchi Lake were determined using laboratory analysis. The collection and laboratory analysis methods are described specifically later in the next section.

2.2.3. Ancillary data

Three kinds of ancillary data were used in my research. 1) DEM: The digital elevation model (DEM) was made in 1998. The vertical accuracy is 0.1 m and the position positional accuracy is 20*20 m. The DEM covers the whole Poyang Lake NNR (includes part of the big Poyang Lake). It is used to derive the water depths and to asistant to analyse the wind-resuspension and the influence of the related environmental factors. 2) Hydrological and meteorological data: The daily wind speed, wind direction, rainfall and water level data from 2004 to 2006 were acquired from the Administration of Poyang Lake NNR. These data

7


are essential and used to provide background information and analyse the influence on the water turbidity. 3) Birds observation data: the data was acquired from Poyang Lake NNR. The observation activity was organized from Octobor to the next April every year since 2000. This data are used to analyse the relationship between the birds presence and activity and water turbidity.

3. Research methods

3.1. Field data collection and analysis

3.1.1. Field survey

The field data acquisition was conducted on 26th 27th and 28th September respectively in Dahuchi Lake, Poyang Lake and Shahu Lake. In order to improve the accuracy of the relationship between water turbidity and sun radiation reflectance of the MODIS satellite images, the sample range of the SDD was extended to include the whole range of observed SDD in the lakes. For this MSc research, and in order to have a preliminary sample and impression of the water turbidity variations and relationship to SDD in the study lakes, a transect sampling strategy (Younes et al. 2003) was applied. This information could serve eventually to design a full spatial and temporal sampling strategy by the Poyang lake NRR authorities or researcher in the future. A full sampling design was beyond the reach of this individual MSc level research. The SDD variation in different parts of the lake can be observed obviously from MODIS true colour composite combined by band 1 and band 2 (Liu 2006). According to the access policy of the Poyang Lake NNR authority, the economy and other feasibility factors, and using a preliminary visual interpretation of the MODIS images, the following sampling transect route in Figure 3-1 was used during the field work. The number of the samples was set to forty-eight, of which twelve originated from the Dahuchi Lake. The field samples were to related to the reflectance of MODIS images, so the sample spots were chosen to be the places more than 500 m from the lake banks. At each sampling point, surface water was pumped from approximately 0.2 m below water surface into 300 ml plastic bottle and the bottle was then capped, labelled, and placed in a box to be shipped for laboratory analysis (Wang and Yang 2004). The following parameters were measured at each location: longitude and latitude by portable GPS device (pinpoint locations within 15 metres), and SDD. The equipments required in the field survey are listed in the appendix. All the measurements were performed on a flat-water surface and in cloud-free condition.

8


Figure 3-1 Field work samples

3.1.2. Laboratory measurement and analysis

Water samples were examined in the chemistry laboratory in School of Resources and Environmental Science (SRES) in Wuhan University. Turbidity and TSS are measured by SGZ-1A Digital Turbidimeter and HACH DR2800 (810nm) separately. The distribution of Turbidity, TSS and SDD are shown in Figure 3-2. According to the boxplots, fifty percent which is from 25% to 75% of the Turbidity is ranged between 10 and 50. TSS and SDD are fifty percent distributed from 10 to 38 and from 0.25 to 0.5 respectively.

Figure 3-2 Distribution, Turbidity, TSS and SDD Because of the limits of the laboratory instrumentation, the size and the texture of the sediment matter can not be detected. But Chemical Oxygen Demand (COD) was tested by HACH DRB200 reactor. The COD is a measure of the oxidizability of a substance, expressed as the equivalent amount in oxygen of an oxidizing reagent consumed by the substance under fixed laboratory conditions (European Chemicals Bureau; http://ecb.jrc.it/Documents/Testing-Methods/ANNEXV/C06web1992.pdf). According the detection of the COD of three water samples in

9


different areas, the COD is a little higher in the west part than that in the east part. So there is some organic matter in the Dahuchi Lake which may have some influence on the water turbidity.

Point ID COD(mg/l) 1 9 2 13 3 37

Figure 3-3 Locations and COD numbers of the samples

3.2. Relation between SDD and water turbidity

3.2.1. Some related terminologies

The purpose of the study is to characterize spatial and long-term temporal trends in water turbidity in Poyang Lake NNR. It is used to evaluate the effectiveness of environmental factors. Water quality parameters related include fecal coliform, dissolved oxygen, nutrients, plankton, chlorophyll a, SDD, temperature, salinity, pH, Turbidity and TSS et al. A total of 38 parameters including TSS, chlorophyll a, dissolved organic matter, diffuse attenuation coefficient and calcite are estimated using the MODIS sensor onboard the Terra and Aqua satellites (Carder et al.). For inland, estuarine and near-shore ocean waters, reflectance is affected by site-specific factors, such as the geology of the watershed, which affects the type of suspended sediments. We included parameters like SDD, TSS and Turbidity in this study. Transparency (or clarity) is a measure of how clear or transparent the water is and is usually measured as SDD and depends on both colour and light scattering. Water clarity has often been used as an indicator of a lake's overall water quality for it correlates well with water quality (Udy et al. 2005). The Secchi Disk is one of the most commonly used tools to measure water clarity (http://www.iisgcp.org/water/wic/clarity.htm; Borkman and Smayda 1998; Kratzer et al. 2003). Secchi Disk depths in waters are largely controlled by suspended sediment load. The suspended sediments have a strong effect on SDD (Kratzer et al. 2003). However, this method may be costly for intensive sampling within water bodies in which water clarity fluctuates highly in time and space scale (Kloiber et al. 2002; Sawaya et al. 2003). Turbidity is a unit of measurement quantifying the degree to which light travelling through a water column is scattered by the suspended organic (including algae) and inorganic particles. The scattering of light increases with a greater suspended load (http://www.iisgcp.org/water/wic/clarity.htm). It depends on the amount, size and composition of the suspended matter such as clay, silt, colloidal particles, plankton and other microscopic organisms. Thus, indirectly it points on the presence of all other studied parameters. Turbidity is commonly measured in Nephelometric Turbidity Units (NTU). SDD and nephelometric turbidity are both optical measures of water quality and differ from suspended

10


sediment concentration, which is a measure of the weight of inorganic particulates suspended in the water column (Harrington et al. 1992). TSS includes all particles suspended in water which will not pass through a filter. Suspended solids are present in sanitary wastewater and many types of industrial wastewater. There are also non-point sources of suspended solids, such as soil erosion from agricultural and construction sites. As levels of TSS increase, a water body begins to lose its ability to support a diversity of aquatic life. TSS can also destroy fish habitat because suspended solids settle to the bottom and can eventually blanket the river bed (Sipelgas et al. 2006). The most appropriate measure to use will depend upon the aquatic environment that is being studied. For riverine systems, turbidity or suspended solids are the most suitable measures. For wetland, estuarine and marine systems, transparency, measured by the Secchi disk method, is the most appropriate measure, although this method has obvious limitations in shallow waters (http://www.nrm.gov.au/monitoring/indicators/turbidity.html).

3.2.2. The relation between SDD and TSS, SDD and Turbidity

000-1-2-2

LNSDD Ln(m)

5

4

3

2

1

0

LN

TU

R

Ln

(NT

U)

r=0.883

0.50.0-0.5-1.0-1.5-2.0

LNSDD Ln(m)

5

4

3

2

1

0

LNT

SS

Ln

(mg/

l)

r=0.776

Coefficientsa

.502 .114 4.413 .000-.450 .035 -.883 -12.738 .000

(ConstantLNTUR

Model1

B Std. Error

UnstandardizedCoefficients

Beta

StandardizedCoefficients

t Sig.

Dependent Variable: LNSDDa.

ANOVAb

11.455 1 11.455 162.248 .000a

3.248 46 .07114.703 47

RegressionResidualTotal

Model1

Sum ofSquares df Mean Square F Sig.

Predictors: (Constant), LNTURa.

Dependent Variable: LNSDDb.

ANOVAb

8.862 1 8.862 69.793 .000a

5.841 46 .12714.703 47

RegressionResidualTotal

Model1

Sum ofSquares df Mean Square F Sig.

Predictors: (Constant), LNTSSa.

Dependent Variable: LNSDDb.

Coefficientsa

.364 .155 2.339 .024-.429 .051 -.776 -8.354 .000

(ConstantLNTSS

Model1

B Std. Error

UnstandardizedCoefficients

Beta

StandardizedCoefficients

t Sig.

Dependent Variable: LNSDDa.

11


Figure 3-4 Relation between SDD and TSS, SDD and Turbidity Since we are testing a reported trend of SDD by using measurements of suspended sediments, it is important to establish that there is a relationship between these parameters. Figure 3-4 shows a graph of the inverse of SSD against the TSS measured and a graph of the inverse of SDD against the Turbidity according to the field surveyed data. The correlation between Ln(TSS) and Ln(SDD) is 0.766. An adjusted R22 of 0.594 exists with estimated standard error (s.e.) which is less than 0.001 for the relation: Ln(SDD)=0.364-0.429Ln(TSS) where SDD and TSS are measured in meters and mg/l respectively. Similarly, an adjusted R22 of 0.774 exists with estimated s.e. for the relationship: Ln(SDD)=0.502-0.45Ln(Turbidity) where Turbidity is measure in NTU. The correlation between Ln(Turbidity) and Ln(SDD) is 0.883 which means the relation between SDD and turbidity is also strong. According to the ANOVA test and the t-test about the coefficients of the models which is shown in Figure 3-4, the accuracy of the regression and residual for the models is well. We therefore conclude that any trend in SDD in the Poyang Lake NNR should be associated with a trend in suspended sediment load which is measured by TSS and Turbidity.

3.3. Mapping temporal and spatial dynamics of water turbidity by MODIS images

3.3.1. Remote sensed data acquisition and processing

MODIS is one of the primary land observation sensors on-board the Terra (EOS AM) and Aqua (EOS PM) satellite. Terra's orbit around the Earth is timed so that it passes from north to south across the equator in the morning, while Aqua passes south to north over the equator in the afternoon. Terra MODIS and Aqua MODIS are viewing the entire Earth's surface every 1 to 2 days, acquiring observations in 36 spectral bands (Lin Wenjing 2006). The main MODIS products are listed in Table 3-1. The MODIS Surface Reflectance Daily L2G Global 500 m ISIN Grid product, MOD09, is a seven band product computed from the MODIS Level 1B land bands 1-7 cantered at 645, 858, 470, 555, 1240, 1640 and 2130 nm respectively (Table 3-3). The product is an estimate of the surface spectral reflectance for each band as it would have been measured at ground level if there were no atmospheric scattering or absorption. The correction scheme includes corrections for the effect of atmospheric gases, aerosols, and thin cirrus clouds. Table 3-2 lists the pixel resolution and bandwidth of the visible and thermal bands. Bandwidth ranges are in nanometers (nm) for the optical bands and micrometers (µm) for thermal bands. The potential applications indicate key uses considered by the instrument design teams.

Table 3-1 MODIS land products MODIS Products MODIS Products

MOD09 Surface Reflectance MOD15 Leaf Area Index/FPAR MOD10 Land Surface Temp./Emis. MOD16 Evapotranspiration/SR MOD11 Land Cover/Change MOD17 Primary Production MOD12 Vegetation Indices MOD43 BRDF/Albedo MOD13 Thermal Anomalies/Fire MOD44 Vegetation Continuous Fields

12


Table 3-2 MODIS visible, thermal bands and potential applications

Band# Pixel Resolution

(m) Bandwidth Range

(µm) Potential Applications

1 250 620-670 Absolute Land Cover Transformation, Vegetation Chlorophyll

2 250 841-876 Cloud Amount, Vegetation Land Cover Transformation

3 500 459-479 Soil/Vegetation Differences 4 500 545-565 Green Vegetation 5 500 1230-1250 Leaf/Canopy Differences 6 500 1628-1652 Snow/Cloud Differences 7 500 2105-2155 Cloud Properties, Land Properties

Table 3-3 MODIS/Terra Surface Reflectance Daily L2G Global 500m SIN Grid V004 products

SDS Units Fill Value Valid Range Scale Factor Band 1 Reflectance -28672 -100-16000 10000 Band 2 Reflectance -28672 -100-16000 10000 Band 3 Reflectance -28672 -100-16000 10000 Band 4 Reflectance -28672 -100-16000 10000 Band 5 Reflectance -28672 -100-16000 10000 Band 6 Reflectance -28672 -100-16000 10000 Band 7 Reflectance -28672 -100-16000 10000

QC Flags Bit field 787410671 0-4294966019 -- Orbit and Coverage -- 15 0-255 --

Number of observations -- -1 0-127 -- MOD09 was chosen to be used in this research. The data have already been corrected for the effect of atmospheric gases, aerosols and thin cirrus clouds and thus provide an estimate of the spectral reflectance as it would be measured at ground level in the absence of atmospheric scattering or absorption (http://modis-sr.ltdri.org/html/surfref.htm 2006), so after the acquisition of the MODIS images, we don’t need to do the atmosphere correction. At first, the images have to be re-projected to UTM according to WGS-84. Secondly, it’s better to resize the data and images to make them focus on the research area. Also changing them to a suitable format is necessary. And then, bands 3-7 should be re-sampled to the 250 m resolution and stacked with band 1 and band 2. At last, analyse the water body only by masking the other type of land cover. The Flowchart is showed specifically in Figure 3-5.

13


Figure 3-5 Flowchart of images processing

3.3.2. Empirical approach

The turbidity of water is highly variable in water and varies over a broad spectrum of time and space scales. Most traditional field sampling methods can not resolve sediment dynamics (Miller and McKee 2004). Many studies have demonstrated that remotely sensed data could monitor the concentration of suspended sediments. The most common techniques for analyzing the remotely sensed data to determine water constituent concentration are based on the brightness of reflectance. Three different approaches can be used to determine the reflectance: empirical approach, semi-empirical approach and analytical approach (Allan 1985). Empirical models are also known as statistical, numerical or data driven models. This type of model is driven from data, and in science is usually developed using statistical tools. In other words, empiricism is that beliefs may only be accepted once they have been confirmed by actual experience (Skidmore 2002). The empirical approach is based on the calculation of a statistical relation between the water constituent concentrations and reflectance (or radiance). The main concept of this approach is retrieving suspended sediment concentration from statistical relationships between spectral reflectance by the suspended sediment near the surface of the water and corresponding ground truth data according to water samples (Duane Nellis et al. 1998). The advantage is that the empirical algorithms are easy to use and they are straightforward. Literature review on empirical algorithm for estimating water turbidity parameters shows vast variety of algorithms used. They start from a simple linear regression between reflectance and water constituent concentrations to non-linear multiple regressions between combination of band ratio(s) and the concentrations. But the disadvantage are that spurious results may occur while using this method, because a causal relationship does not necessarily exist between the parameters studied and results of empirical algorithm always need in situ data validation because illumination, atmospheric conditions. However, spectrometry measurements performed at the flat water surface and sunny conditions along with direct water sampling are free from some of those limitations (Turdukulov 2003).

14


3.3.3. Empirical models related and validation

Reflectance data from MODIS images provide information on qualitative distribution pattern of suspended matter. Higher reflectance is an indicator of more turbid surface water (Liis Sipelgas 2006). To have quantitative measure of turbidity distribution, reflectance values should be converted to SDD. Instead of using sophisticated optical model, a simple linear regression between reflectance values and SDD determined from water sample was checked. These simple regression models are commonly used but are known to be site-specific, depending on optical properties of water column. Earlier study of sediment pattern of water in Poyang Lake revealed that the correlation between MODIS Band 1 and SSD which is SDD=-0.11 Ln(Red)-0.17 is significant. Thirty-one samples in three lakes were used (Liu Qian 2006). A good correspondence of SDD and the inverse of beam attenuation at 514 nm was found also by Pfannkuche (Pfannkuche et al. 2000). A strong relation between MODIS blue/red spectral radiance ratio and SDD was illustrated (Lillesand and Chipman 2001). Beginning in July of 2001, weekly to biweekly estimates of SDD for eleven lakes plus Green Bay, Lake Michigan have been made. The spectral sensitivities are the blue and red bands on MODIS. These estimates were produced using the MODIS blue/red spectral radiance ratio. The model was derived based on MODIS imagery acquired on September 8 and 17, 2000. Eventually, this approach can probably be extended to some 100 lakes state-wide, plus similar numbers in Minnesota and Michigan (Lillesand accessed June 8 2006). Both of these two models were validated by the field samples in this study. Only band 1 is far away to explain the SDD of the field samples and the result is quite different. MODIS blue/red spectral radiance ratio was validated also. Only 59% of the variation of the SDD can be explained with 95% confidence. This is because that empirical models are usually site-specific and locally and empirical models of the natural environment are not often applicable when extrapolated to new areas according to the climate or soil conditions (Skidmore 2002). Another research was also done by Guofeng Wu (2006). This yielded sufficiently accurate concentration estimations. The research gave better accuracy of the in-situ measurement done at the same time as the acquisition date of remotely sensed imagery. This multiple regression model created by an analytical approach given by Wu Guofeng was Ln(SDD) = 0.474 + 15.240* Blue – 21.130 * Red (Guofeng Wu 2006). The SDD was measured using a standard 20 cm Secchi disk on 5 points respectively in Dahuchi Lake, Shahu Lake and Meixihu Lake at about weekly intervals from April till October in 2004 and 2005 by the Bureau of Jiangxi Poyang Lake National Nature Reserve and the International Crane Foundation. After deleting the samples 500 m offshore, only 71 samples on 5 points were left to build this model. This model described the relation between the natural logarithm of SDD and MODIS bands. The model includes the red band which is band 1 and blue band which is band 3. The red band had a negative sign while the blue band was positively related in this case. The model explained 88% of the variation of the natural logarithm of SDD with estimated s.e. of 0.37. F-tests showed that regression models was statistically significant at p = 0.0001. An accuracy test should be done for this model according to the field samples taken in the field work before. By the assistance of some software such as ArcGIS, the reflectance of band 1 and band 3 on samples spots can be extracted from the MODIS image on the field work day. According to the model mentioned by Wu, SDD numbers on these spots can be calculated respectively. The accuracy analysis

15


result is shown in Figure 3-6. Then scatter plots in Figure 3-6 revealing the agreement of SDD predicted from MODIS images with SDD of the 48 samples measured by field survey were obtained. Using Field surveyed data as X-axis and Predicted data as Y-axis, Figure 3-6 shows the model can explain 79% of the variation of the SDD surveyed in field with estimated s.e. of 0.15. The correlation which is 0.892 is high when the confidence is 95%. F-tests showed that the regression model was statistically significant at p = 0.0001 (F=179.89, d.f. =1). Figure3-7 shows the distribution of the residual which equals to the model-induced SDD subtracting form field surveyed SDD. The regression residuals are normally distributed.

1.601.401.201.000.800.600.400.200.00

SDD m

1.60

1.20

0.80

0.40

0.00

SD

D_P

redi

cted

m

r=0.892 n=48

95% confidence

--- Y =X

Figure 3-6 Scatter plots revealing the agreement of SDD predicted from MODIS images with SDD by field survey

1.401.201.000.800.600.400.20

SDD m

0.40

0.20

0.00

-0.20

-0.40

Res

idua

l m

Figure 3-7 Distribution of the residual

3.3.4. Detecting the pattern of water turbidity by MODIS images

Water turbidity of Poyang Lake NNR varies during different season in one year. In this research, nearly monthly time series images were used, so extracting the SDD of Poyang Lake from the images are very important. Figure 4-7 shows the flowchart of assessing SDD variation with MODIS images.

16


Corrected MODIS09

images (Research area)

Erdas (Modeler)

Extract SDD on 5 quite different points

Calculate the average SDD of the nearest 9 pixes for each of the 5 points

SDD images (Research area)

Seasonal pattern of SDD on 5 pionts

Figure 3-8 Flowchart of detecting seasonal pattern of water turbidity A model should be built to extract band 1 and band 3. By using the ERDAS software, we obtained spatial distribution maps of SDD in the research area, according a suitable experimental model by Guofeng Wu (Guofeng Wu, 2006) between image reflectance and SDD. The derived SDD maps can describe the spatial pattern of water turbidity in the different lakes. So, the spatial pattern and the possible factors inducing these patterns can be analyzed. Five sample points where the water turbidity is different were selected in different parts of the Dahuchi Lake. The average of SDD in the nearest 9 pixels for each point was calculated and the average of SDD on these five points was derived. Seasonal pattern of the average SDD were obtained for each year from 2004 to 2006. The results are illustrated in the next section.

3.4. Analysing the temporal dynamics of water turbidity using a statistical multivariate model

3.4.1. Analyzing the related environmental factors

Suspended sediments have a strong effect on water turbidity (Scheffer 1998). It has been shown that wind is one of the main factors of lake sediment re-suspension in many cases (Booth et al. 2000; Kratzer et al. 2003; Cozar et al. 2005). The wind direction and wind speed in Poyang Lake NNR show seasonal changes. A seasonal trend can be seen clearly from the data which is collected from the mean wind speed record of the weather station. From the remote sensed images, a seasonal change of water turbidity can also be found. Based on the field observation and material analysis, it is assumed that wind is one of the main factors influence the water turbidity pattern. A wind-induced re-suspension model is going to be chosen to analysis the extent that wind influences the water turbidity. Lake depth influences sinking loss and re-suspension of suspended particulate matter, but also has very pronounced implications for the light climate related to the water clarity (Scheffer 1998). Rainfall has some relation with water clarity by influencing phytoplankton et al. (Tomasko et al. 2005). Relationships between lake morphometric parameters and night-time lake surface temperature were

17


investigated in North American temperate lakes using the ASTER kinetic temperature (AST08) product. Temperature is significant related with the SDD (Becker and Daw 2005). Water birds were counted during the breeding season and the winter months. The deeper the lake, the higher the threshold for Secchi disc depth, because clearer water is necessary for sufficient light to reach the bottom of the deeper lake to allow development of bottom-dwelling macrophytes (Rip et al. 2006) Of course, there will be still other factors that may influence the water turbidity in Poyang Lake NNR, such as fishing, dredging etc.. It is mentioned by Liu Qian that the bottom-feeding swans maybe cause turbid in winter. It is also mentioned that the fishing and dredging influence water turbidity (Liu 2006).

3.4.2. Statistic methods related

The value of each water turbidity parameter and environmental variables were derived based on the sample plots and historical data. Descriptive statistics such as box plots etc. were used to identify the distribution of these variables. All variables were detected to see if they are normally distributed or not. Statistical methods, such as regression analysis, can be used to analysis the relationship between the seasonal pattern of the lakes and related environmental factors. Regression model building is often an iterative and interactive process (Faraway 2002). Two main purposes of regression analysis can be used for as follows: To describe a relation between two or more variables; to predict the value of a variable (the predictand, sometimes called the dependent variable or response), based on one or more other variables (the predictors, sometimes called independent variables). A simple correlation or regression relates two variables only; a multiple correlation or regression relates several variables at the same time (Rossiter 2005). These two kinds of regression analysis are both used in the study to analysis the relation between SDD and environment factors. Regression diagnostics are necessary to evaluate the models designed. The first diagnostic is how well the model fits the data. This is the success of the model conditional on the sample data. The fitted method applied to a linear model object returns values predicted by a model at the observation values of the predictor. The residuals give the discrepancy of each fitted point from its observation. If any are unusually large, it may be that the observation is from a different population, or that there was some error in making or recording the observation. Leverage should also be identified and verified (Rossiter 2005). Accuracy assessment and comparison is necessary for building a model. The accuracy of these predicted models should be verification. Error matrix and ANOVA methods can be used. At the same time, the distribution do the residual should be test. Error matrix is the most suitable method to compare different classifiers (Congalton, 1991). The name ANOVA stands for Analysis of Variance is used because the original thinking was to try to partition the overall variance in the response to that due to each of the factors and the error. ANOVA is used to determine how much of the variability in some property is explained by one or more categorical factors. As with regression, this does not necessarily imply a causal relation from the factors to the response. However, it does supply evidence to support a theory of causation that can be justified with a conceptual model. The simplest ANOVA is one-way, where the total variance of the data set is compared to the residual variance after each observation’s value is adjusted for the mean for the one factor (Rossiter 2006).

18


3.4.3. A statistical multivariate water turbidity model

There exist some empirical models about the wind induced re-suspension in shallow lakes existed (Scheffer 1998), (Cozar, Galvez et al. 2005), (Bailey and Hamilton 1997), (Hamilton and Mitchell 1996). By comparing the similar hydrology situation, a suitable model could be built according to the specific situation in Poyang Lake NNR. According to the secondary data about the wind speed, water level and total precipitation and consulting the texture of the sediment and the wind direction situation, the situation of water turbidity which is figured by SDD can be predicted then by statistic analysis. Afterward, according to the historic data of the area, which is recorded by the bureau, the seasonal dynamics of the water turbidity can be predicted. The accuracy of the statistical multivariate model in the research area can be assessed by regression analysis of the data from the predicted model by MODIS and the field data. So finally it can be known how wind and other hydrological or meteorological variables in the area influence the water turbidity. An analysis flowchart is showed below. Wind speed

& wind direction

Water level

Total precipitation SDD

Statistic Analysis

statistical multivariate

model

Model Predicted by

MODIS

Accuracy assesment

Analysis

Accuracy is low

Accuracy is enough

How wind influence the water turbidity

Figure 3-9 Analysis the influence of wind on water turbidity There are more factors than wind which can also influence the water turbidity, such as bird, fish and rainfall. So more related environment factors can be added to the wind-induced model after detecting the influence of the wind on water turbidity. The wind-induced model can be developed to a more complicated model to detect the affect of the other factors. In the meantime, more scenarios can be considered. At the same time, field investigation and material analysis can be helpful to the analysis. Literature review on related information about dredging and other government policy collected also can help to analysis the man-made influence on the water turbidity.

19


20

(4) The spatial distribution of water turbidity in different lakes is different. In Dahuchi Lake, the SDD is a little higher in the south and west part in summer. While in Banghu Lake, it is much clearer in the southwest part than the northeast part.

(3) During 2004 and 2005, the SDD in different lakes of Poyang Lake NNR is different especially in winter. In winter, the SDD in Dahuchi Lake is obviously less than SDD in other lakes which means that the water in Dahuchi Lake is more turbid than other lakes. While in summer, the different is not that obvious.

(2) There is a clear difference in water turbidity between summer and winter. The SDD is quite high in summer when the lakes in Poyang Lake NNR are mixed together, especially between June and September while most of the lakes in Poyang Lake NNR are quite turbid in winter when the boundary of the lakes can be clearly distinguished.

(1) These figures give a qualitative illustration about the SDD trend during years. The water is clear in summer and gets turbid in winter. In 2004, the water is quite clear from June to August while in 2005 the time range which is from June to September is a little longer.

From the Figures，SDD distribution of Poyang Lake NNR during different season could be derived from MODIS images. The differences between seasons can be seen visually and the differences between different lakes are also shown clearly. Some characteristics of SDD spatial and seasonal variation can be described as below:

The MODIS images of different months from January 2004 to December 2006 were analyzed and the SDD maps of Poyang Lake for the different months were obtained. Figure 4-1, Figure 4-2, Figure 4-3 show the SDD variation of Poyang Lake of different months in 2004, 2005 and 2006 respectively. In the SDD maps, the blacker the spot is, the smaller the SDD is on the spot which means the more turbid the water is. Because of the cloudy weather in Poyang Lake Area, it is hard to find images without cloud, so the quality of some images is sometimes affected.

4.1. SDD distribution maps in Poyang Lake NNR using satellite data

4. Results

Mon

itorin

g va

riatio

n of

wat

er tu

rbid

ity a

nd re

late

d en

viro

nmen

t fac

tors

in P

oyan

g La

ke N

atio

nal N

atur

e R

eser

ve, C

hina

SDD

imag

es (y

ear

2004

)

20

0401

21

20

0402

14

200

4041

0

200

4050

9

20

0406

30

20

0407

22

20

0408

28

2

0040

916

21

Mon

itorin

g va

riatio

n of

wat

er tu

rbid

ity a

nd re

late

d en

viro

nmen

t fac

tors

in P

oyan

g La

ke N

atio

nal N

atur

e R

eser

ve, C

hina

20

0410

04

20

0410

28

2

0041

118

2

0041

212

Figu

re 4

-1 T

he S

DD

imag

es in

yea

r 200

4

SDD

imag

es (y

ear

2005

)

2005

0114

2005

0222

20

0504

21

20

0505

27

22

Mon

itorin

g va

riatio

n of

wat

er tu

rbid

ity a

nd re

late

d en

viro

nmen

t fac

tors

in P

oyan

g La

ke N

atio

nal N

atur

e R

eser

ve, C

hina

2

0050

603

200

5070

5

20

0508

17

200

5090

9

20

0510

18

200

5103

1

2005

1126

200

5120

7

Figu

re 4

-2 T

he S

DD

imag

es in

yea

r 200

5

23

Mon

itorin

g va

riatio

n of

wat

er tu

rbid

ity a

nd re

late

d en

viro

nmen

t fac

tors

in P

oyan

g La

ke N

atio

nal N

atur

e R

eser

ve, C

hina

SDD

imag

es (y

ear

2006

)

2

0060

109

2

0060

207

2006

0303

2

0060

417

20

0605

20

20

0606

11

200

6071

2

20

0608

14

24

Mon

itorin

g va

riatio

n of

wat

er tu

rbid

ity a

nd re

late

d en

viro

nmen

t fac

tors

in P

oyan

g La

ke N

atio

nal N

atur

e R

eser

ve, C

hina

25

Figu

re 4

-3 T

he S

DD

imag

es in

yea

r 200

6

2006

0926

200

6102

4


4.2. Spatial dynamics of water turbidity in Dahuchi Lake of Poyang Lake NNR

Spring (Apr10th, May9th) Summer (Jun30th, Aug28th)

Autumn (Oct4th, Oct28th) Winter (Dec12th Feb14th)

Figure 4-4 SDD images in Dahuchi Lake in 2004

UL (392457, 3228903) LR (401754, 3217347) Figures 4-4 shows the SDD situation in Dahuchi Lake in 2004. SDD images of four seasons were picked out to do the comparison. Some characteristics of SDD spatial variation in different seasons can be drawn as below: (1) In spring, water in the south part on the lake is much clearer than the north part. The difference

between north part and south part is not quite obvious. (2) The overall lake is getting much clearer in summer. The most turbid place in the lake is located in

the northeast part of the lake. The change between different places is a little distinct. (3) In autumn, the most turbid place turned to the east part of the lake clockwise as time passed by,

and the different between clear area and turbid are is quite obvious. (4) In winter, the SDD overall the lake is more or less the same, there isn’t quite visible difference

unless there is a little more turbid in the west part and the centre part. Figure 4-5 shows the SDD images of four seasons in Dahuchi Lake in 2005 (left) and 2006 (right). The overall situation in year 2005 and 2006 is more or less the same as that in 2004. The most turbid place starts from the north part in spring and turns clockwise until the southwest part in winter. In summer the most turbid place is more obvious and the difference between places is much larger.

27


Spring (Apr21th) Summer (Jul5th) Spring (Apr17th) Summer (Jun11th)

Autumn (Oct 18th) Winter (Feb22th) Autumn (Oct 18th) Winter (Feb7th)

Figure 4-5 SDD images in Dahuchi Lake in 2005(left) and 2006(right) UL (392457, 3228903) LR (401754, 3217347)

4.3. Temporal dynamics of water turbidity in Poyang Lake NNR

4.3.1. Temporal dynamics observed from MODIS in Dahuchi Lake

The historical images were analyzed using the algorithm calibrated from the field data. The situation in Poyang Lake NNR was a little different between different lakes especially in winter. So this sub-chapter was focused on the situation in Dahuchi Lake. Five representative points shown in Figure 4-6 were selected in the different parts of the Dahuchi Lake according to the spatial dynamics of water turbidity in different seasons. From the SDD maps of different times (Figure 4-1, Figure 4-2, Figure 4-3), the SDD numbers can be obtained, and the average of the SDD can be calculated as well. The temporal dynamics of SDD in these five points and the average SDD in 2004 and 2005 are got shown in Figure 4-6 and Figure 4-7 respectively. The curve graphs show the changes in the number of SDD over the period from January to December in 2004 and 2005. As can be seen, great changes have taken place in a whole year. Higher SDD appears in the summer season while in winter season the SDD is quite low. In 2004, over the period from January to April the SDD remained level. The number sharply went up to almost 2 meters from May to August. Then, there is a little decrease to around 1.5 m in September. After a little rise in late September, the number decreases steadily until the end of the year to almost zero. In 2005, the same dynamics can be seen clearly. But there are some differences. First, the overall SDD is less than that of 2004. The first peak showed up in June which is a little earlier than 2004, and the second peak which

28


showed up in late august is also earlier than 2004. The range stayed almost 3 months which is more or less the same as that in 2004. Second, the overall SDD number is less than that in 2004 and the peak around 1.3 meters is just 3/4 of the peak number in 2004. From the curves, the differences between five plots can be seen. In winter the difference is quite small which is almost hardly to tell. In spring and summer, the difference is less than 0.5m. The most obvious difference shows up in summer which is almost 1m in November in 2004. In 2005, the overall difference is a little smaller and the most obvious difference is in September which is about 0.5 m.

Temporal dynamics of SDD in 2004

0

0.5

1

1.5

2

2.5

Jan-04

Feb-04

Mar-04

Apr-04

May-04

Jun-04

Jul-04

Aug-04

Sep-04

Oct-04

Nov-04

Dec-04

date

SDD m

point1

point2

point3

point4

point5

average

Figure 4-6 Temporal dynamics of SDD derived from MODIS images in 2004

Temporal dynamics of SDD in 2005

0

0.5

1

1.5

2

2.5

Jan-05

Feb-05

Mar-05

Apr-05

May-05

Jun-05

Jul-05

Aug-05

Sep-05

Oct-05

Nov-05

Dec-05

date

SDD m

point1

point2

point3

point4

point5

average

Figure 4-7 Temporal dynamics of SDD derived from MODIS images in 2005

29


4.3.2. Temporal dynamics observed from MODIS in Shahu and Banghu Lake

A somewhat similar seasonal trend was observed in the neighbouring lakes which are shown in Figure 4-8. Because of the limited number of sampling points, however, the trends are less distinct from this preliminary analysis. These observations should be considered as indicative first observations, and can eventually be used to design future sampling strategies.

SDD variation(2005)

0

0.5

1

1.5

2

2.5

3

1-Jan-05

1-Feb-05

1-Mar-05

1-Apr-05

1-May-05

1-Jun-05

1-Jul-05

1-Aug-05

1-Sep-05

1-Oct-05

1-Nov-05

1-Dec-05

SDD m

datepoint_sha point_bang1 point_bang2

SDD variation(2006)

0

0.5

1

1.5

2

2.5

3

1-

Jan-06

1-

Feb-06

1-

Mar-06

1-

Apr-06

1-

May-06

1-

Jun-06

1-

Jul-06

1-

Aug-06

1-

Sep-06

1-

Oct-06

1-

Nov-06

1-

Dec-06date

SDD m

point_sha point_bang1 point_bang2

Figure 4-8 Temporal dynamics of SDD from MODIS in Shahu Lake and Banghu Lake in 2005 and 2006

4.4. Analysing the effects of environmental factors

4.4.1. Related environmental factors

Many studies confirm that the maximum water depth inhabited by macrophytes is positively correlated with water clarity. Obviously other aspects such as water depth are likely to affect the maximum inhabited depth (Scheffer 1998). There are reported correlations which includes a non-uniform turbidity distribution with respect to the depth of water (Wang and Seyed-Yagoobi 1995). Suspended sediments have a strong effect on SSD (Scheffer 1998). In many shallow lakes, inorganic sediment particles, but also algal cells or phytoplankton, go through a rapid cycle of sedimentation and re-suspension which induce the turbidity in water. Often, re-suspension is mainly due to wave action, but also fish searching for food in the bottom can stir up considerable amounts of sediment in some situations (Scheffer 1998). Re-suspension of lake sediments is a function of the bottom shear stress as

30


the result of local fluid motion just above the sediment and the local sediment characteristics. Although several hydrodynamic processes can intervene on sediment re-suspension, wind-induced waves are usually the dominant process in shallow lakes (Cozar et al. 2005).

4.4.2. The correlation between SDD and environmental factors

In order to identify the environmental factors which most affect water turbidity, their correlation with SDD was tested. The data in 2004 were used to analysis the correlation between SDD and environmental factors. The average SDD number was compared with water lever, total precipitation, mean wind speed and max sustained wind speed in the same day, and mean wind speed, max sustained wind speed two days before. From the figure, there are some strong relationships found. The correlation coefficient between SDD and water level is 0.80 and the coefficient between SDD and mean wind speed is 0.84. The relation between SDD and total precipitation which is only 0.35 is not quite high. At the same time, the max sustained wind speed is also quite related with the coefficient 0.76.

Table 4-1 Correlation matrix between SDD and environmental factors

Correlations

1 .799** -.351 -.836** .352 -.764** .103.001 .218 .000 .238 .001 .726

14 14 14 14 13 14 14.799** 1 -.170 -.828** .301 -.719** .193.001 .560 .000 .317 .004 .508

14 14 14 14 13 14 14-.351 -.170 1 .502 .076 .540* .338.218 .560 .067 .804 .046 .238

14 14 14 14 13 14 14-.836** -.828** .502 1 -.192 .882** -.034.000 .000 .067 .530 .000 .908

14 14 14 14 13 14 14.352 .301 .076 -.192 1 .010 .614*.238 .317 .804 .530 .975 .026

13 13 13 13 13 13 13-.764** -.719** .540* .882** .010 1 .021.001 .004 .046 .000 .975 .942

14 14 14 14 13 14 14.103 .193 .338 -.034 .614* .021 1.726 .508 .238 .908 .026 .942

14 14 14 14 13 14 14

Pearson CorrelationSig. (2-tailed)NPearson CorrelationSig. (2-tailed)NPearson CorrelationSig. (2-tailed)NPearson CorrelationSig. (2-tailed)NPearson CorrelationSig. (2-tailed)NPearson CorrelationSig. (2-tailed)NPearson CorrelationSig. (2-tailed)N

average SDD

water level

TotalPrecipitation

MeanWindSpeed

2days before

MaxSustainedWindSpeed

2days before

average SDD water levelTotal

PrecipitationMeanWind

Speed 2days before

MaxSustained

WindSpeed 2days before

Correlation is significant at the 0.01 level (2-tailed).**.

Correlation is significant at the 0.05 level (2-tailed).*.

Considering the characteristics of each environmental factors, water lever, total precipitation, mean wind speed and max sustained wind speed in the same day were tested and linear regression analyses were used. Table 4-2 shows the results of the analysis. The regression analysis indicates that the highest correlation (R2 = 0.699) occurs in the regression analysis between SDD and mean wind speed. It appears that the relationship between SDD and total precipitation is not significant (F= 1.691; p=0.218) in a statistical sense. This calibrated relation was used for all further analyses.

31


Table 4-2 Summary of the regression analysis between average SDD and environment factors

Environment factor R2 Equation Significance Water level(WL) 0.639(s.t.=0.27) SDD=-0.424+0.516WL F=21.228; p<0.001Total precipitation(TP) 0.124(s.t.=0.43) SDD=1.059-00.513TP F=1.691; p=0.218 Mean wind speed(MWS) 0.699(s.t.=0.25) SDD=2.473-0.432MWS F=27.811; p<0.001 Max sustained wind speed (MSWS)

0.584(s.t.=0.30) SDD=1.893-0.147MSWS F=16.834; p<0.001

According to the regression relations established, an equation predicting the SDD was built as follows: SDD=-1.898+0.234WL-0.270MWS. This multivariate model only considers the influences of water level and mean wind speed, without total precipitation and max sustained wind speed. However, the R2 which is 0.736 is high with the standard error of the estimate 0.26. The ANOVA test shows the F of the regression is 9.273 and the significance is 0.003. If we take the max sustained wind speed into consideration, this equation becomes: SDD=-1.794+0.227WL-0.218MWS-0.026MSWS. SDD is predicted by the situation of water level, mean wind speed and max sustained wind speed at the same time. R2 is 0.739 with the std. error of the estimate 0.27. F is 6.385 with the sig. p equals to 0.01 which is a little higher, then the previous model, using water level and wind speed only.

4.4.3. Error analysis

A split data technique was used, where the 2004 data were used to derive the multivariate regression models and the 2005 data to validate the models. Therefore the 2005 data were used to evaluate the accuracy of the equation SDD=-1.898+0.234WL-0.270MWS. Figure 4-9 shows the model can explain 50.7% of the variation of the SDD surveyed in field with estimated s.e. of 0.38. The correlation which is 0.712 is high with a confidence interval of 95%. F-tests showed that the regression model was statistically significant at p = 0.002 (F=14.39, d.f. =1). Figure 4-10 shows the distribution of the residual which equals to the model-induced SDD subtracting form field surveyed SDD. The residuals are normally distributed.

1.4001.2001.0000.8000.6000.4000.2000.000

average SDD m

1.500000

1.000000

0.500000

0.000000

SDD

pre

dict

ed m

R =0.712 n=16

--- Y =X

Figure 4-9 Scatter plots revealing the agreement of SDD predicted using the multivariate model with field observed SDD

32


1.4001.2001.0000.8000.6000.4000.2000.000

averageSDD m

0.80000

0.60000

0.40000

0.20000

0.00000

-0.20000

-0.40000

-0.60000

resi

dual

m

Figure 4-10 Distribution of the regression residuals between modelled and observed data

4.4.4. The temporal variation of SDD predicted by the multivariate regression model

temporal variation of SDD in 2004

0

0.5

1

1.5

2

2004-1-1

2004-1-31

2004-3-1

2004-3-31

2004-4-30

2004-5-30

2004-6-29

2004-7-29

2004-8-28

2004-9-27

2004-10-27

2004-11-26

2004-12-26

date

SDD predicted

Figure 4-11 Temporal dynamics of SDD derived from the multivariate regression model in 2004

temporal variation of SDD in 2005

0

0.5

1

1.5

2

2005-1-1

2005-1-31

2005-3-2

2005-4-1

2005-5-1

2005-5-31

2005-6-30

2005-7-30

2005-8-29

2005-9-28

2005-10-28

2005-11-27

2005-12-27

date

SDD predicted

Figure 4-12 Temporal dynamics of SDD derived from the multivariate regression model in 2005

33


The similar temporal pattern of SDD in 2004 and 2005 is observed and is in agreement with the pattern we obtained from the MODIS images. Noteworthy is that the highest SSD levels show up almost the same time during the year. This of course can be explained by the higher lake water depths in these periods of the year. During the winter, there are obvious differences between the years. Here, we should think of a combination of factors like low lake levels, and periods with strong winds and the presence of feeding birds in some parts of the lake.

4.4.5. Accuracy validation between temporal pattern predicted by MODIS images and by the multivariate regression model

2.0000001.5000001.0000000.5000000.000000

SDD from the multivariate regression model

2.0000

1.5000

1.0000

0.5000

0.0000

SD

D f

rom

MO

IDS

imag

es

--- Y =X

R S q Linear = 0.139

Figure 4-13 Scatter plots revealing the agreement of SDD predicted using the SDD from MODIS In the x-axis, we show the SDD as predicted by the multivariate environmental factor model and in the y-axis the SDD predicted by MODIS images is shown. Although, the fit is not quite high, we try to explain the observed variations as follows:

Resuspended Sed.

Water Depth

Turbidity

Algae

ZooplanktonFish

Vegetation

Allelopathic Subs.

Nutrients

Waves

+

+

+ +

+

+

-

--

-

-

-

-

-

34


Figure 4-14 Environmental factors related with water turbidity and their effect It has long been noted that the water tends to be less turbid if there is aquatic vegetation. The first publications result from work in fish culture ponds. Schreither (1928), for instance, describes that the phytoplankton density of a pond was lowest in years with high aquatic macrophytes abundance. Many later studies report enhanced water clarity in the presence of vegetation. Especially spectacular are the observation of whole lakes that switch between a clear and vegetation state and a turbid state with few macrophytes. Analyses of the relationship between the transparency and the macrophytes abundance in large sets of lakes confirm that there is a systematic correlation (Scheffer 1998). A relation between factors and water turbidity is shown in Figure 4-14. Many shallow lakes can switch abruptly between a vegetated state with clear water and a turbid situation with high concentrations of phytoplankton and other suspended solids. It has been known that vegetation tends to enhance water clarity. A positive relation exists in the development of submerged vegetation which is once they grow, the water clears up. Water becomes turbid also because sediment disturbance by waves and benthivorous fish prevents plant settlement, and herbivory may help to prevent vegetation recovery. It has been shown that the distribution of juveniles of the 20 common fish species is generally statistically correlated with water turbidity (Cyrus and Blaber 1987). Also dredging plays a role in the water turbidity situation (Goodwin 1985). In Poyang Lake NNR, fishing and dredging in the Dahuchi Lake is allowed only during summer, while forbidden from October to April the next year in order to protect the birds passage, living here in winter while in Shahu Lake and Banghu Lake, it’s not restricted. So it will influence a lot for the water turbidity in winter between different lakes. It also seems to influence a lot that the water turbidity is turbid in winter. According to the field survey, there are a lot of algae in Shahu Lake, while there are almost no algae in Dahuchi Lake. This also can explain why the water in Dahuchi Lake is more turbid in winter than Shahu Lake. The water turbidity situation is correlated with birds action in shallow lakes (Rip et al. 2006). As the biggest freshwater lake in China with 3,130 km2 wetland coverage, which is 80% of the Poyang Lake region, the Poyang Lake is one of the first seven national nature reserves declared in Ramsar Sites providing habitats for rare migratory birds living through winter (http://www.iseis.cuhk.edu.hk/yuenyuen/project/poyang/poyang.htm 2005). There is a strong seasonal variation in bird grazing. Goose migrate from Siberia through the north of China to winter in the middle and lower Yangtze basin in search of suitable grazing lands (Si 2006). Most of the birds settle in Dahuchi Lake in winter. The everyday eating and living activity will induce the sediment re-suspension and increase the water turbidity. So we can expect that bird occurrence has an important effect on the water transparency and turbidity situation in Dahuchi Lake, and at least partly explain its higher winter turbidity than other lakes. There are no SDD measured records in winter because Dahuchi Lake is closed then. SDD got from MODIS is limit, so statistic is not quite possible. The number of birds in the winter of 2004 and 2005 can be seen from Figure 4-15. The figures below present the number of birds, observed during the autumn, winter and spring months in the lake area. The birds began to come since October and got the peak in the middle of November in 2004 and in the middle of December in 2005. The birds begin to leave as the weather gets warm and are almost gone until the beginning of April. In winter there is no SDD field record because Dahuchi Lake is closed then and SDD images got from MODIS images is limit, so statistic analysis is not quite accurate. It is also influenced by bird’s activities that the water situation is Dahuchi Lake more turbid than other lakes.

35


Number of birds in 2005

0

10000

20000

30000

40000

50000

60000

70000

80000

90000

100000

Oct-05

Oct-05

Nov-05

Nov-05

Dec-05

Dec-05

Jan-06

Jan-06

Feb-06

Feb-06

Mar-06

Number

Mar-06

Apr-06

Date Figure 4-15 The number of birds in 2004 (left) and 2005 (right) Figure 4-15 The number of birds in 2004 (left) and 2005 (right)

Number of birds in 2004

0

10000

20000

30000

40000

50000

60000

70000

80000

Oct-0

4

Oct-0

4

Nov-0

4

Nov-0

4

Dec-0

4

Dec-0

4

Jan-0

5

Jan-0

5

Feb-0

5

Feb-0

5

Mar-0

5

Mar-0

5

Apr-0

5

Date

Number

Turbidity had negative correlation with temperature and salinity, indicating that higher turbidity occurs in low temperature and salinity areas during winter (Islam et al. 2006). According the record of the weather stations, yearly trend of temperature in 2004 and 2005 are illustrated in Figure 4-16. The highest temperature showed up in July and the temperature stays low during January and February.

Turbidity had negative correlation with temperature and salinity, indicating that higher turbidity occurs in low temperature and salinity areas during winter (Islam et al. 2006). According the record of the weather stations, yearly trend of temperature in 2004 and 2005 are illustrated in Figure 4-16. The highest temperature showed up in July and the temperature stays low during January and February.

Figure 4-16 Temperature in 2004 and 2005 Figure 4-16 Temperature in 2004 and 2005

Temperature in 2004 and 2005

0

20

40

60

80

100

1 31 61 91 121 151 181 211 241 271 301 331 361

Julian day

Temperature ℉

year 2004 year 2005

5. Conclusions and discussion 5. Conclusions and discussion

The main objective of the study was monitoring lake turbidity using optical remote sensing in Poyang Lake NNR. Methods were developed and tested to quantify the water turbidity, SDD and the influence of environmental factors. Quantification of the water turbidity has been achieved by using a systematic approach through the establishment of the empirical relationship between the reflectance of MODIS data and in-situ SDD. In order to evaluate the influence of the environmental factors, a multivariate statistical model for predicting SDD by environmental factors was derived. The temporal patterns of SDD derived by MODIS were in good agreement with the statistical model, based on lake water levels and wind speeds as environmental factors. Accuracy test was done between these two obtained temporal patterns. For the investigation of the temporal and spatial pattern using optical remote sensing, the characteristics of the MODIS Terra instrument provide data well suited for the study of

The main objective of the study was monitoring lake turbidity using optical remote sensing in Poyang Lake NNR. Methods were developed and tested to quantify the water turbidity, SDD and the influence of environmental factors. Quantification of the water turbidity has been achieved by using a systematic approach through the establishment of the empirical relationship between the reflectance of MODIS data and in-situ SDD. In order to evaluate the influence of the environmental factors, a multivariate statistical model for predicting SDD by environmental factors was derived. The temporal patterns of SDD derived by MODIS were in good agreement with the statistical model, based on lake water levels and wind speeds as environmental factors. Accuracy test was done between these two obtained temporal patterns. For the investigation of the temporal and spatial pattern using optical remote sensing, the characteristics of the MODIS Terra instrument provide data well suited for the study of

36


dynamics suspended matter, and the MODIS data were found to be suitable to a limited extent for monitoring of water turbidity variation. Detailed conclusions and recommendations are given in the following sections.

5.1. Conclusions on the spatial and temporal variation of water turbidity

Monitoring, protecting and improving the quality of water in lakes and reservoirs is critical for targeting conservation efforts and improving the quality of the environment (Ritchie et al. 1994). The SDD of Poyang Lake NNR during different seasons and the water turbidity pattern of the changes have been derived successfully by using an experimental multivariate statistical model and MODIS images processing. High water transparency values or SDD are observed during the summer season while the most turbid situations always occur in winter. In different years, the trend is similar while the occurrence of detailed peaks (high or low SDD values) is a little different in the same lakes. Among the different lakes in the Poyang Lake NRR, we observe however specific situations. The lake turbidity situation during different seasons exhibited that the seasonal variations in water level in combination with wind speed variations were associated with the levels of water turbidity variation. This conclusion found in this research is similar to the ones found by the former researchers. A simple visual spatial analysis of water turbidity patterns in different seasons was carried out. Comparing the situation in different seasons, the most turbid places show in different directions. It is shown that turbidity estimation in the low-water season is less sensitive to the varying turbidity resolution than in the other seasons. This analysis also demonstrates that the lake turbidity estimation in the low-water season is indeed influenced by the water level, wind speed as well as other environmental factors. The current interest in regional scale phenomena requires data from sensors with coarse spatial resolution to meet both the demand for high temporal resolutions and the need for the demand for data sets of manageable size (Mary and Curtis, 1997). This study indicated that the MODIS medium spatial resolution data with high temporal frequency, coarse spatial resolution and free of charge has a good potential for the study of water turbidity estimation in seasonal lakes. The MODIS data was useful for mapping the spatial patterns of the SDD of seasonal lakes. Using simple processing procedures and suitable images processing methods, the water bodies can be extracted and the SDD can be calculated by the pixel information and spatial resolution. MODIS data can potentially be used in a variety of researches in water turbidity situation estimation, which is especially useful for monitoring natural reserves.

5.2. Conclusions on the effects of related environmental factors

According to the seasonal and temporal dynamics of the water turbidity situation in Poyang Lake NNR, many related environmental factors are considered. Some factors such as water level, wind speed, and rainfall which can be quantified were analysis by statistical methods. Some environmental factors such as fishing, dredging and bird’s influence are analyzed by theoretical reasoning, combined with field investigation and evidence. Also available sensus and other statistical data of the Poyang Lake NNR were used to assistant the analysis. The statistical correlation analysis between observed hydrological and meteorological variables, and Secchi disk depths and water turbidity, yielded a multivariate regression model, permitting to explain

37


seasonal SDD variations by lake water level and wind speed variations. A strong statistical significance between water level and wind speed and lake SDD and turbidity could be derived. The statistic model of water level and mean wind speed explain well the SDD situation in Poyang Lake NNR which means water level and mean wind speed influence a lot the water turbidity situation. It must be noted that also the effect of bird occurrence, during the winter months may contribute to the situation. Fishing and dredging affect the turbidity at the same time. As one of the most important settlement for cranes (Kanai et al. 2002), bird’s influence plays an important role for the water turbidity situation as well, especially in winter. However, the Dahuchi lake area being closed for access, during the winter, collection of ground sample data is not permitted, constraining somehow a more quantitative analysis of the effects.

5.3. Recommendations and future research on monitoring water turbidity

Some recommendations for future research on this topic can be formulated as follows: (1) The Poyang Lake area due to its status, offers limited to no access during certain periods of the

year. This makes full monitoring of certain variables difficult. We could recommend using and/or installing automated monitoring equipments (e.g. lake turbidity, water quality) which can operate over longer periods, and as such do not create human disturbance in the NRR. Based on the information produced in this work, and in combination with other researches, the spatial sampling designs and monitoring locations and scheme could be established.

(2) With some more quantitative data available (sediment and lake bottom basic physical and chemical characteristics), the effect of wind-driven re-suspension, at lower lake levels and/or the ancillary effects of bird occurrences could be better analyzed and deciphered, and also put into evidence, using mathematical models (e.g. lake numerical water quality models).

(3) Different types, grain sizes, shapes and compositions of the sediment may have different influence on the sediment re-suspension and the water turbidity. Taking it into account that different suspended sediment types will be contained in the water in various lakes will helpful to the analysis of the specific situation in different lakes.

(4) In this study, more field samples and a larger data range can be used to improve the accuracy. And more complicated model and methods can be tried also. In order to make the research more accurate, Use of the water level data and digital elevation model to calculate the water depth on different points in the lake is also feasible.

(5) According to the field survey, the water capacity in the Poyang Lake NNR is getting smaller these few days which may have influence on the water turbidity situation and the birds’ activities. Related variation and influencing factors can be studied by further research. Especially the role of the SanXia dam could be considered carefully.

38


References

Allan, T. D. (1985). "Remote assessment of ocean color for interpretation of satellite visible

imagery : H.R. Gordon and A-Y. Morel. Lecture Notes on Coastal and Estuarine Studies, Springer-Verlag, Berlin, 114 pp., DM 58, US$ 22.50, ISBN 3-540-90923-0." Physics of The Earth and Planetary Interiors 37(4): 292.

Becker, M. W. and A. Daw (2005). "Influence of lake morphology and clarity on water surface temperature as measured by EOS ASTER." Remote Sensing of Environment 99(3): 288-294.

Booth, J. G., R. L. Miller, B. A. McKee and R. A. Leathers (2000). "Wind-induced bottom sediment resuspension in a microtidal coastal environment." Continental Shelf Research 20(7): 785-806.

Borkman, D. G. and T. J. Smayda (1998). "Long-term trends in water clarity revealed by Secchi-disk measurements in lower Narragansett Bay." ICES Journal of Marine Science 55(4): 668-679.

Carder, K. L., F. R. Chen, Z. P. Lee, S. K. Hawes and D. Kamykowski "Semianalytical moderate-resolution imaging spectrometer algorithms for chlorophyll a and absorption with bio-optical domains based on nitrate-depletion temperatures." Journal of Geophysical Research(104): 5403–5421.

Cozar, A., J. A. Galvez, V. Hull, C. M. Garcia and S. A. Loiselle (2005). "Sediment resuspension by wind in a shallow lake of Esteros del Ibera (Argentina): a model based on turbidimetry." Ecological Modelling 186(1): 63-76.

Cyrus, D. P. and S. J. M. Blaber (1987). "The influence of turbidity on juvenile marine fishes in estuaries. part 1. field studies at Lake St. Lucia on the southeastern coast of Africa." Journal of Experimental Marine Biology and Ecology 109(1): 53-70.

Duane Nellis, M., J. J. A. Harrington and J. Wu (1998). "Remote sensing of temporal and spatial variations in pool size, suspended sediment, turbidity, and Secchi depth in Tuttle Creek Reservoir, Kansas: 1993." Geomorphology 21(3-4): 281-293.

European Chemicals Bureau Degradation-Chemical Oxygen Demand. Faraway, J. J. (2002). "Practical Regression and Anova using R." Ann Arbor, MI: self-

published (web). Gomez, F., G. Gorsky, E. Garcia-Gorriz and M. Picheral (2004). "Control of the phytoplankton

distribution in the Strait of Gibraltar by wind and fortnightly tides." Estuarine, Coastal and Shelf Science 59(3): 485-497.

Goodwin, C. R. a. D. M. M. (1985). "Appearance and water quality of turbidity plumes produced by dredging in Tampa Bay, Florida." Deep Sea Research Part B. Oceanographic Literature Review 32(9): 774.

Guofeng Wu, e. a. (2006). "Comparison of MODIS and Landsat TM5 images for mapping tempo-spatial dynamics of Secchi disk depths in Poyang Lake national nature reserve, China."

Harrington, J. A., F. R. Schiebe and J. F. Nix (1992). "Remote sensing of Lake Chicot, Arkansas: Monitoring suspended sediments, turbidity, and Secchi depth with Landsat MSS data." Remote Sensing of Environment 39(1): 15-27.

Hellweger, F. L., P. Schlosser, U. Lall and J. K. Weissel (2004). "Use of satellite imagery for water quality studies in New York Harbor." Estuarine, Coastal and Shelf Science 61(3): 437-448.

39


Houborg, R., H. Soegaard and E. Boegh (2007). "Combining vegetation index and model inversion methods for the extraction of key vegetation biophysical parameters using Terra and Aqua MODIS reflectance data." Remote Sensing of Environment 106(1): 39-58.

Houborg, R. M. and H. Soegaard (2004). "Regional simulation of ecosystem CO2 and water vapor exchange for agricultural land using NOAA AVHRR and Terra MODIS satellite data. Application to Zealand, Denmark." Remote Sensing of Environment 93(1-2): 150-167.

http://ecb.jrc.it/Documents/Testing-Methods/ANNEXV/C06web1992.pdf. http://modis-sr.ltdri.org/html/surfref.htm (2006). Surface Reflectance User_Guide. http://modis-sr.ltdri.org/html/surfref.htm (accessed 16th February 2006). MODIS land surface

reflectance science computing facility. http://www.iisgcp.org/water/wic/clarity.htm Water Information and Concern (WIC) Fact Sheet

Series. http://www.iseis.cuhk.edu.hk/yuenyuen/project/poyang/poyang.htm (2005). Monitoring The

Dynamic Changes of Poyang Lake Wetland Through Remote Sensing, Yuen Yuen Research Center for Satellite Remote Sensing

http://www.nrm.gov.au/monitoring/indicators/turbidity.html Turbidity/Suspended Solids: Turbidity/Suspended Particulate Matter in Aquatic Environments. http://www.nrm.gov.au/monitoring/indicators/turbidity.html.

http://www.poyanglake.org/ Introduction to the Poyang Lake National Nature Reserve. Hu, C., Z. Chen, T. D. Clayton, P. Swarzenski, J. C. Brock and F. E. Muller-Karger (2004).

"Assessment of estuarine water-quality indicators using MODIS medium-resolution bands: Initial results from Tampa Bay, FL." Remote Sensing of Environment 93(3): 423-441.

Islam, M. S., H. Ueda and M. Tanaka (2006). "Spatial and seasonal variations in copepod communities related to turbidity maximum along the Chikugo estuarine gradient in the upper Ariake Bay, Japan." Estuarine, Coastal and Shelf Science 68(1-2): 113-126.

Jiang, Q. and D. R. Piperno (1999). "Environmental and Archaeological Implications of a Late Quaternary Palynological Sequence, Poyang Lake, Southern China." Quaternary Research 52(2): 250-258.

Kanai, Y., M. Ueta, N. Germogenov, M. Nagendran, N. Mita and H. Higuchi (2002). "Migration routes and important resting areas of Siberian cranes (Grus leucogeranus) between northeastern Siberia and China as revealed by satellite tracking." Biological Conservation 106(3): 339-346.

Kloiber, S. M., P. L. Brezonik, L. G. Olmanson and M. E. Bauer (2002). "A procedure for regional lake water clarity assessment using Landsat multispectral data." Remote Sensing of Environment 82(1): 38-47.

Kratzer, S., S. Buchan and D. G. Bowers (2003). "Testing long-term trends in turbidity in the Menai Strait, North Wales." Estuarine, Coastal and Shelf Science 56(2): 221-226.

Liis Sipelgas , e. a. (2006). "Operational monitoring of suspended matter distribution using MODIS images and numerical modelling." Advances in Space Research. Lillesand, T. M. (accessed June 8 2006). "Combining Satellite remote sensing and volunteer

secchi disk measurement for lake transparency monitor." Lillesand, T. M. and J. W. Chipman (2001). "-assisted lake water quality: Using satellite data to

observe regional trends in lake transparency." GIM International(15): 26-29. Lin Wenjing (2006). "Satellite based Regional-scale Evapotranspiration in the Hebei Plain,

Northeastern China." MSc. Thesis. Liu, Q. (2006). Monitoring Area Variation and Sedimentation Patterns in Poyang Lake, China

Using MODIS Medium-Resolution Bands. ITC. ENSCHEDE, THE NETHERLANDS, INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATION. MSC: 72.

Liu Qian (2006). "Monitoring Area Variation and Sedimentation Patterns in Poyang Lake, China Using MODIS Medium-Resolution Bands." MSc. Thesis.

40


Macia, A., K. G. S. Abrantes and J. Paula (2003). "Thorn fish Terapon jarbua (Forskal) predation on juvenile white shrimp Penaeus indicus H. Milne Edwards and brown shrimp Metapenaeus monoceros (Fabricius): the effect of turbidity, prey density, substrate type and pneumatophore density." Journal of Experimental Marine Biology and Ecology 291(1): 29-56.

Marten S. (1998). Ecology of shallow lakes, Chapman & Hall. Miller, R. L. and B. A. McKee (2004). "Using MODIS Terra 250 m imagery to map

concentrations of total suspended matter in coastal waters." Remote Sensing of Environment 93(1-2): 259-266.

Pfannkuche, J., J. Meisel and O. Mietz (2000). "Factors affecting clarity of freshwater lakes in Brandenburg, Germany." Limnologica - Ecology and Management of Inland Waters 30(4): 311-321.

Rip, W. J., N. Rawee and A. de Jong (2006). "Alternation between clear, high-vegetation and turbid, low-vegetation states in a shallow lake: the role of birds." Aquatic Botany 85(3): 184-190.

Ritchie, J. C., F. R. Schiebe, C. Cooper and J. A. J. Harrington (1994). "Chlorophyll measurements in the presence of suspended sediment using broad band spectral sensors abroad satellites." J. fRESHWATER eCOL.(9): 197-206.

Rossiter, D. G. (2005). "Technical Note: An example of data analysis using the R environment for statistical computing."

Rossiter, D. G., Ed. (2005). Technical Note: An example of data analysis using the R environment for statistical computing. International Institute for Geo-information Science & Earth Observation (ITC). Enschede (NL).

Rossiter, D. G., Ed. (2006). Introduction to the R Project for Statistical Computing for use at ITC. Enschede (NL), International Institute for Geo-information Science & Earth Observation (ITC).

Sawaya, K. E., L. G. Olmanson, N. J. Heinert, P. L. Brezonik and M. E. Bauer (2003). "Extending satellite remote sensing to local scales: land and water resource monitoring using high-resolution imagery." Remote Sensing of Environment 88(1-2): 144-156.

Scheffer, M. (1998). Ecology of shallow lakes, Chapman & Hall. Si, Y. (2006). Mapping flood recessional grasslands grazed by overwintering geese: an

application of multi-temporal remote sensing, ITC. Sipelgas, L., U. Raudsepp and T. Kouts (2006). "Operational monitoring of suspended matter

distribution using MODIS images and numerical modelling." Advances in Space Research 38(10): 2182-2188.

Skidmore, A. (2002). Environmental Modelling with GIS and Remote Sensing, Taylor & Francies.

Tan, D., Y. Hu, F. Xia and H. Zhang (2006). "Application of MODIS in the environment monitoring in the Yangtze River." Yangtze River.

Tomasko, D. A., C. A. Corbett, H. S. Greening and G. E. Raulerson (2005). "Spatial and temporal variation in seagrass coverage in Southwest Florida: assessing the relative effects of anthropogenic nutrient load reductions and rainfall in four contiguous estuaries." Marine Pollution Bulletin 50(8): 797-805.

Turdukulov, U. (2003). Determination of water quality parameters using imaging spectrometry. MSC thesis.

Udy, J., M. Gall, B. Longstaff, K. Moore, C. Roelfsema, D. R. Spooner and S. Albert (2005). "Water quality monitoring: a combined approach to investigate gradients of change in the Great Barrier Reef, Australia." Marine Pollution Bulletin 51(1-4): 224-238.

Wang, J. and J. Seyed-Yagoobi (1995). "Effect of water turbidity on thermal performance of a salt-gradient solar pond." Solar Energy 54(5): 301-308.

Wang, Y. and Y. Yang (2004). Environmental Monitoring. Wolfe, R. E., M. Nishihama, A. J. Fleig, J. A. Kuyper, D. P. Roy, J. C. Storey and F. S. Patt

(2002). "Achieving sub-pixel geolocation accuracy in support of MODIS land science." Remote Sensing of Environment 83(1-2): 31-49.

41


Wu, Y. and W. Ji (2002). Study on Jiangxi Poyang lake National Nature Reserve, Forest Publishing House, China.

Yang, D., L. Wang, T. Ma and c. Cheng (2001). "Estimating the Semimetals Effect of Chaohe River to Miyun Reservoir Usine Multi-temporal TM Images." Scientia Geographica Sinica.

Younes, W. A. N., N. Bensoussan, J.-C. Romano, D. Arlhac and M.-G. Lafont (2003). "Seasonal and interannual variations (1996-2000) of the coastal waters east of the Rhone river mouth as indicated by the SORCOM series." Oceanologica Acta 26(4): 311-321.

42


6. Appendix

Appendix A: The equipments required in the field survey Equipments used in field survey Quantity

Motorboat 1 Plastic bottles Depend on the number of samples

GPS device and GPS receiver 1 Secchi Disk 1

Water sampler 1

Appendix B: Geographical coordinates of field samples No Latitude N Longitude E X Y

1 29°6.735´ 115°57.032´ 397890 3220877.1

2 29°6.878´ 115°56.673´ 397310 3221146.4

3 29°8.037´ 115°56.306´ 396734 3223292.1

4 29°8.281´ 115°56.699´ 397375 3223736.9

5 29°8.258´ 115°56.929´ 397748 3223691.1

6 29°7.812´ 115°57.583´ 398801 3222858.1

7 29°7.570´ 115°57.204´ 398182 3222416.6

8 29°7.390´ 115°57.087´ 397990 3222085.9

9 29°7.151´ 115°56.915´ 397707 3221647

10 29°6.914´ 115°57.028´ 397886 3221207.7

11 29°6.829´ 115°57.187´ 398143 3221048.4

12 29°6.681´ 115°57.313´ 398344 3220773.3

13 29°11.575´ 116°00.793´ 404064 3229762.4

14 29°11.929´ 116°00.830´ 404129 3230415.6

15 29°12.585´ 116°01.146´ 404651 3231622.8

16 29°14.314´ 116°00.131´ 403034 3234829.7

17 29°16.349´ 115°59.343´ 401790 3238598.7

18 29°17.128´ 116°00.572´ 403792 3240020.3

19 29°17.620´ 116°01.580´ 405432 3240915.3

20 29°18.625´ 116°03.009´ 407760 3242752.2

21 29°17.580´ 116°02.799´ 407405 3240825.2

22 29°17.161´ 116°02.976´ 407685 3240049

23 29°16.810´ 116°03.201´ 408044 3239397.9

24 29°14.947´ 116°04.680´ 410411 3235938.3

25 29°14.300´ 116°04.905´ 410766 3234740.7

26 29°13.947´ 116°07.207´ 414490 3234060.2

27 29°13.698´ 116°08.083´ 415906 3233589.8

28 29°13.485´ 116°08.708´ 416915 3233189

43


29 29°13.251´ 116°09.333´ 417925 3232749.6

30 29°12.975´ 116°10.654´ 420061 3232224.7

31 29°13.219´ 116°10.270´ 419442 3232679.7

32 29°13.383´ 116°09.201´ 417712 3232994.9

33 29°13.566´ 116°08.658´ 416835 3233339.2

34 29°16.656´ 116°03.387´ 408343 3239111.1

35 29°17.397´ 116°02.791´ 407389 3240487.3

36 29°18.164´ 116°02.898´ 407574 3241902.3

37 29°18.644´ 116°02.878´ 407548 3242789

38 29°18.015´ 116°01.901´ 405958 3241640.4

39 29°16.780´ 115°59.965´ 402804 3239386

40 29°11.357´ 116°00.385´ 403399 3229365.4

41 29°11.230´ 116°00.198´ 403094 3229133.4

42 29°10.876´ 116°59.682´ 402081 3228068.7

43 29°10´30.1´´ 115°55´10.9´´ 394953 3227860.2

44 29°10´20.1´´ 115°55´25.5´´ 395344 3227548.8

45 29°10´13.2´´ 115°55´54.2´´ 396118 3227329.4

46 29°10´20.8´´ 115°55´43.5´´ 395831 3227565.9

47 29°10´28.9´´ 115°55´33.6´´ 395566 3227817.7

48 29°10´31.0´´ 115°55´17.6´´ 395134 3227886.3

Appendix C: Field survey and laboratory analysis results No SDD(cm) Water Depth(cm) Turbidity(NTU) TSS(mg/L)

1 21 79 50 32

2 24 90 46 41

3 22 92 60 26

4 25 100 42 37

5 27 103 46.8 30

6 24 96 50 27

7 26 91 51 24

8 24 67 45 19

9 26 79 51 35

10 23 70 25 29

11 25 63 42.8 24

12 22 47 90 44

13 98 210 6.8 10

14 27 380 60 58

15 23 273 41 42

16 33 more than 300 61 67

17 26 more than 300 54.5 62

18 34 more than 300 85 59

19 32 more than 300 45 62

20 45 more than 300 10 11

21 39 more than 300 10.2 11

44


22 37 more than 300 10 9

23 46 more than 300 19.5 16

24 38 more than 300 26.4 24

25 42 more than 300 13.6 16

26 43 more than 300 14.5 12

27 48 more than 300 12 5

28 49 more than 300 16 8

29 48 more than 300 12.5 10

30 45 more than 300 12.5 11

31 50 more than 300 11.8 10

32 50 more than 300 10.2 10

33 46 more than 300 12.5 8

34 43 more than 300 26 22

35 44 more than 300 29 19

36 42 more than 300 21.8 10

37 31 more than 300 106 87

38 32 more than 300 81 81

39 24 more than 300 95 90

40 95 more than 300 14.8 13

41 88 more than 300 15.3 17

42 84 more than 300 12 8

43 110 132 4 4

44 132 175 1.1 2

45 135 142 1.6 1

46 125 136 2.8 18

47 108 125 5.3 5

48 124 135 2.5 2

45