APPLICATION OF REMOTELY SENSED IMAGERY TO WATERSHED ANALYSIS A CASE STUDY OF LAKE KAROUN CATCHMENT, EGYPT

Transcription

1 APPLICATION OF REMOTELY SENSED IMAGERY TO WATERSHED ANALYSIS A CASE STUDY OF LAKE KAROUN CATCHMENT, EGYPT Noha Donia Lecturer, Institute of Environmental Studies and Researches, Ain Shams University, Cairo, Egypt ABSTRACT This study emphasized various remote sensing and geographic information system techniques, such as digital image processing and geographic overlay, to fill gaps using satellite imagery in the process of analysis the hydrological and environmental characteristics of Lake Karoun catchment. Environmental characteristics including Agriculture NDVI (ANDVI) temporally changed zones, Moisture index (MI), Land Surface Temperature (LST) and land use classes were obtained from high spatial resolution images (Landsat TM). Moreover, hydrological parameters, drainage flow directions, drainage networks and catchments from digital elevation model have been delineated using the Advanced Space-borne Thermal Emission and Reflection Radiometer (ASTER) remote sensing images. As results, lake's watershed characteristics including the environmental and hydrological factors for each watershed zone were presented and analyzed. This will help decision maker to take suitable actions for watershed management and lake restoration. Keywords: Lake Karoun; remote sensing; catchment delineation; GIS 1. INTRODUCTION Accurate delineation of a watershed plays an extremely important role in the management of the watershed. The delineated boundaries form the nucleus around which the management efforts such as land use, land change, soil types, geology and river flows are analyzed and appropriate conclusions drawn (Meijerink, 1988a). Current geographic information system (GIS) and remote sensing (RS) technologies provide ways for rapid collection of field data and prompt data processing (Meijerink et. al., 1994). The objective of the study is to present a new technique for extracting drainage flow directions, contributing (upslope) areas, and catchments from digital elevation models in Lake-dominated areas. Sustainable water resources planning and management require data to enable quantification of water quantity and quality. Information is required on the rates of transfers and storage of water within a catchment. Lack of adequate hydrological data introduces uncertainty in both the design and management of water resources systems.

2 Consequently, there is a need to develop methods for predicting flow characteristics at ungauged sites. 2. FAYOUM CATCHMENT Fayoum Governorate is a green Oasis located in the north of the Western Desert, 90 kilometers southwest of Cairo and it is directly linked to the Nile through Youssef Sea which derives its waters from Al Ibrahimiah Canal at Dairout up to Al Lahoon barrages 284 km from Dairout. Water is distributed through Youssef Sea and Hassan Wassef Sea. Youssef Sea serves the northeastern and central parts, whereas Hassan Wassef Sea serves the southern and western parts. It is a Governorate of the province of North Upper Egypt and is surrounded by the desert on all sides except from the southeast, where it is connected to Beni Suef Governorate. The total surface area of the Governorate is approximately 6000 km2. This Governorate is of a particular nature, differing from the Delta and Upper Egypt, and from the Oasis as well. The differences are not limited to agriculture; they extend to geographical and topographical features as the environment is a mix of agricultural, desert and coastal. The Governorate is composed of five districts, Fayoum, Senores, Ebshway, Itsa and Tamiya (UNDP, 2003) Irrigation system Fayoum Governorate has a special irrigation system due to the nature of its land which slopes downward steeply from south to north for 67 meters along a distance of 35 km towards Karoun Lake, with a sloping average of 2 m/km. Therefore, most of the Governorate s land (93%) is being irrigated directly through waterfalls with openings and irrigation continues round the clock. Water is distributed to farmers using a rotating system as it is shared by beneficiaries through irrigation canals letting water enter the fields of each farmer. The waste water flows from south to north towards the slope where it pours into Karoun and Wadi Al Rayan Lakes. Because of the special nature of Fayoum, it is possible to control the irrigation water coming from Youssef Sea to cover local needs, on the condition that the water level in Karoun Lake should not get too high so as to preserve the land bordering the Lake. Fayoum receives fresh water from one source, Bahr Youssef. Fayoum's share of irrigation water amounts to nearly 2.5 billion cubic meters annually. Karoun Lake s surface area amounts to 55,000 feddans and Wadi Al Rayan Lake s surface area amounts to 35,000 feddans. Both are considered the main source of water drainage for the agricultural lands in the Governorate (UNDP, 2003). The topographic map of Fayoum area is shown in Figure 1.

3 2.2. Drainage system The drainage system in Fayoum consists of three major natural drainage systems and one man-made system. These are: El-Batts drainage system at the east side El-Wadi drainage system on the west side Minor drains in the central part of the Fayoum Wadi El-Rayan drain (man-made) The drainage systems discharge into Lake Qaroun. El-Batts and El-Wadi drains discharge by gravity. Wadi El-Rayan drain conveys the drainage through a tunnel to the Wadi Rayan depression at the south west of Fayoum. Lake Karoun is a closed Lake system (it has no natural outlet) and used for recreation and fisheries (UNDP, 2003). Figure 1: Topographic map of Fayoum 2.3. Environmental system The environmental condition of Karoun Lake is governed by pollution status of the drains discharged into the Lake. This is of significance because, for practical purposes, pollutants deposited in Lake Karoun s watershed ultimately end up in the

4 Lake. In other words, Lake Karoun and Lake Rayan serve as the final pollutant sinks in the Fayoum drainage basin. Evaporation, biochemical transformation, settling and volatilization are the only processes that can remove water and pollutants from these sinks. Sanitary conditions are negatively influenced by the inadequate organization of sewage and liquid waste disposal. The environmental system in Fayoum cathchment can be described in the following: Drain water quality violates water quality standards and poses health risks Solid waste dumping in Fayoum city and around district towns poses a health threat Karoun Lake levels and salinity - Lake levels are continuing to rise at 0.1 m/year - Salinity rising 0.5 g/l each year Lake bottom silting up by average 10 mm each year 3. METHODOLOGY In this study, The Integrated Land and Water Information System (ILWIS) software was used to create the GIS and remote sensing images processing and analysis. 3.1 GIS database Four layers were digitized from topographic maps of Fayoum (scale 1:50,000), edited, built and geometrically corrected to UTM projection. These layers were administrative boundaries of districts, urban areas, vegetation and irrigation & drainage networks. 3.2 Watershed Hydrological characteristics Digital elevation model was created from the Advanced Space-borne Thermal Emission and Reflection Radiometer (ASTER) remote sensing images. Then GIS techniques within ILWIS software (Meijerink, 1988b) were used to delineate the watershed using the extracted digital elevation model (DEM) Creating a DEM from remote sensing images The ASTER sensor is part of the Earth Observation Satellite1, it offers nearly simultaneous capture of stereo images, minimizing temporal changes and sensor modeling errors. The visible and near infrared (VNIR) portion of the ASTER sensor includes two independent telescopes, a Nadir looking one and a backward looking one to help minimize image distortion during data capture. This simultaneous data capture provides true stereo coverage from which a DEM can be automatically extracted from using the stereopair option of ILWIS program.

5 3.2.2 Creating a Depressionless DEM The first step in any of the hydrologic modeling tools in ILWIS is to fill the elevation grid. The model starts with a surface that has no sinks. Sinks are areas of internal drainage which do not drain out anywhere. The reason that sinks need to be filled in is because a drainage network is built that finds the flow path of every cell, eventually off the edge of the grid. If cells do not drain off the edge of the grid, they may attempt to drain into each other, which will lead to an endless processing loop Flow direction To calculate a drainage network or watersheds, a grid must exist that is coded for the direction in which each cell in surface drains. Flow direction is important in hydrologic modeling because in order to determine where a landscape drains, it is necessary to determine the direction of flow for each cell in the landscape. This is accomplished with the Calculate Flow Direction menu choice. For every cell in the surface grid, the ILWIS grid processor finds the direction of steepest downward descent. Flow direction is a focal function. For every 3x3 cell neighborhood, the grid processor stops at the center cell and determines which neighboring cell is lowest. Depending on the direction of flow, the output grid will have a cell value at the center cell, as determined by this matrix: If the direction of flow for a cell is due north, then in the output grid, that cell's value will be 64. These numbers do not have any absolute, relative, or ratio meaning, they are just used as numeric place holders for nominal direction data values (since grid values are always numeric) Flow accumulation Flow accumulation is the next step in hydrologic modeling. Watersheds are defined spatially by the geomorphological property of drainage. In order to generate a drainage network, it is necessary to determine the ultimate flow path of every cell on the landscape grid. Flow accumulation is used to generate a drainage network, based on the direction of flow of each cell. By selecting cells with the greatest accumulated flow, a network of high-flow cells can be created. These high-flow cells should lie on stream channels and at valley bottoms Watershed outlet (pour) points The next step in delineating watersheds is to select pour points. These are typically points at the edge of the grid, or just downstream of major confluences. Pour points are created by adding a new point layer to the project. Points should be added that are as close to the center of cells as possible. For this reason, it is good to have the highflow cells displayed and the data frame displayed at very large scale. Before watersheds can be delineated, the points need to be converted to a grid layer. The points must have an integer attribute that uniquely identifies each point, because the resultant watersheds will have the same value as the grid cells which act as pour points. Use that attribute as the value field in the output grid.

6 3.2.6 Delineating watersheds The last step in watershed delineation is to perform the function itself. The grid processor needs three grid layers: pour points, flow accumulation, and flow direction. Here are the contour lines placed atop the watersheds. The watershed boundaries do a fairly good job of the following ridge lines. Watershed delineation has to be performed on an iterative basis while moving upstream. The first watershed will contain the entire study area. The second round of watershed delineations will create preliminary sub-basins. Then smaller and smaller sub-basins will be continued until the management study objectives are met (e.g., maximum sub-basin size, representation of all pour points, or delineation of entire study area) Calculating of flow length One of the tools available in the surface hydrological toolset is flow length. Flow length is the distance traveled from any cell along the surface flow network to an outlet. This can be used to find areas that are closer to headwater locations or closer to stream outlets. In the image below, the red cells are farthest from the stream outlet, and the blue cells are closest to the outlet. Channel length is defined as the distance along the main channel, measured from the channel outlet to the farthest point inside the catchment: acs = h / L (1) where: h = Height difference between catchment outlet and the farthest point along the main channel L = Main channel length 3.2 Land Surface Temperature (LST) (Tran et al., 2001) LST was calculated using Thermal band of Landsat TM mosaic image. Thermal radiances was converted to at-sensor temperatures via the Planck equation and then to at-surface temperatures based on the following equations. L ( λ ) Lmax L ( λ ) min ( λ ) = Lmin ( λ ) + Qcal (2) Q cal max where: L Spectral radiance received by the sensor for the pixel in question. (λ ) L Minimum detected spectral radiance for the scene min ( λ ) ( mwcm -2 sr -1 µm -1 ).

7 L Maximum detected spectral radiance for the scene max ( λ ) (1.56 mwcm -2 sr -1 µm -1 ). Q cal max Maximum grey level (255). Q Grey level for the analyzed pixel. cal Once the spectral radiance ( L (λ ) ) was computed, it was possible to calculate radiant temperature directly by the following equation: K 2 T R = (3) K 1 ln + 1 L( λ ) where T R Radiant temperature in Kelvin for the pixel in question K 1 Calibration constant ( mwcm -2 sr -1 µm -1 ) K 2 Calibration constant ( K) L (λ) Spectral radiance for the pixel in question, calculated above in (1) From the radiant temperature T R, kinetic temperature T K, could be calculated using the equation: 4 T R = ε 1/ λ T K (4) where ε λ is the spectral emissivity, assuming to be 0.95 uniformly for the entire terrain/materials in this study. 3.3 Moisture Index (MI) (Tran et al., 2001) The soil moisture estimation method uses a dispersion diagram between T s and F r based on the triangle that defines the T s /F r space. The division of the F r image into areas with the same vegetation cover enables us to assume that, in each vegetation interval, the coldest areas correspond with the soils with the highest moisture content, and that the hottest areas have the driest soils. When interpreting the dispersion diagram it must be considered that soil moisture estimation from T s /F r space is only valid over the range in vegetation cover from 0% to 80%. In each F r category, the pixel with the lowest Ts is considered as the moisture limit and is given a soil moisture value of 100. The pixel with the highest thermal values constitutes the dry limit whose moist value is 0.1 (Not 0 to differentiate from the background of the image). The remaining pixels in each F r are scaled from 0.1 to 100 in relation to their F r. A set of intervals in the F r images are established, each 5% wide. Using these intervals, 16 binary masks were created and were subsequently used to divide up the T s image into 16 partial T s images, each of which corresponds to a certain range of vegetation cover. The most extreme pixels were not used as the dry and moist limits as they were considered to be outliers. To establish the dry and moist boundaries in

8 each Ts image, we define them in the interval of three standard deviations from the mean: the moist boundary being the lowest value and the dry boundary the highest value. 3.4 Agriculture NDVI (ANDVI) (Tran et. al., 2001) For Landsat mosaic image, NDVI is calculated from visible red (TM Band 3) and near infrared (TM Band 4) reflectance using the following equation: band 4 band 3 NDVI = (5) band 4 + band 3 NDVI was linearly scaled between limits for bare soil (NDVI o ) and 100% vegetation cover (NDVI s ). This linear scaling could also reduce errors in the process of sensor calibration and atmospheric correction (Hung et al., 2002). Transformation of NDVI to a scaled NDVI (N*) was accomplished using equation (6): N* NDVI NDVI o = (6) NDVI s NDVI o The fractional vegetation cover (F r ) was then calculated as Fr = N Land use/land cover map In this study, ISODATA algorithm was used to perform an unsupervised classification using for Landsat TM mosaic. A total of 50 unsupervised classes were created. The big number of classes (50 classes) was chosen to separate land class from urban class because of the big similarity between spectral values of both classes. The fifty classes were pared down to the 6 classes which are vegetation1, vegetation2, land, urban, desert and water. 4. RESULTS AND DISCUSSION 4.1 GIS Database The digital elevation model was developed from the ASTER satellite images intervals using ILWIS software as seen in Figures 3 and 4. Also, the DEM of Karoun Lake is shown in Figure 5. The Length of canals and drains were calculated from the derived irrigation and drainage network. The results are illustrated in Table 1. The length of canals ranged from 287 km to 396 km while length of drains ranged from 132 km to 299 km. Figure 6 shows the extracted irrigation network of Fayoum Governorate. The results of this study indicate that, determination of the watershed areas using ILWIS

9 software has the potential of providing accurate and reproducible results. This conclusion was derived from a comparison of the watershed characteristics developed using the GIS analysis approach outlined above with the data collected from the extensive on-site field survey data provided by USGS. The inspection indicates that, the total watershed areas calculated from the softwarebased delineation procedures are very close to those provide by USGS. The major benefit of using the software is to be able to develop these results in about three hours of computational time once the required data are compiled and rectified. Data collection and preparation took about three days once correlation of the interest area against an area where USGS had accurate information. To evaluate this area using conventional approach incorporating ground based measurement methodologies, it took over a month to collect and regress data. The catchment map of Fayoum area is shown in Figure 7. Table 1: Length of canals and drains in each district District length of Length of canals length of Length of drains canals (km) per unit area drains (km) per unit area Fayoum Senores Itsa Tamiya Ebshway Figure 3: Digitized map of Fayoum

10 Figure 4: Digital Elevation Model of Fayoum Figure 5: Digital Elevation Model of Karoun Lake

11 Figure 6: Extracted Irrigation network of Fayoum Governorate Figure 7: Catchment map of Fayoum

12 Land Surface Temperature As shown in Figure (8) Land Surface Temperature ranged from 302 Kelvin (28.85 C) to 318 Kelvin (44.85 C) (Kelvin = C ) with a mean of Kelvin (34.87 C) and standard deviation of The highest values exist in bare soil in the north while the lowest values exist in the vegetation areas Moisture Index (MI) As shown in Figure (9), the values of MI ranged from 0.38 to with a mean of and standard deviation of The values of MI decreased from south to north. Agriculture NDVI (ANDVI) As shown in Figure (10), the values of ANDVI ranged from 0.16 to 0.72 with a mean of 0.39 and standard deviation of The values of ANDVI decreased from south to north. Land use/land cover map Unsupervised classification was performed on Landsat TM image to generate 10 classes. However, the 10 classes were not able to discriminate between urban and vacant land classes due to the similarity of spectral signatures of the two classes. Therefore, the classification process was repeated to generate 50 classes. These 50 classes represented 6 classes; two classes of vegetation, vacant land, urban, water and desert. Water class appeared in one class, vegetation1 was divided into 6 classes, vegetaion2 was divided into 2 classes, urban class was divided into 5 classes, land class was divided into 13 classes and the rest (23 classes) represented the desert as shown in Figure (11).

13 Figure (8): Land Surface Temperature (LST) of Fayoum Figure (9): Moisture Index of Fayoum

14 Figure (10): Agriculture NDVI of Fayoum Figure (11): Land cover map developed from Landsat TM image, 9 August 1995 for the study area

15 5. CONCLUSION Current remote sensing and geographic information system technologies provide ways for rapid collection of field data and prompt data processing. This research presented a technique for extracting drainage flow directions, contributing (upslope) areas, and catchments from digital elevation models in Lake-dominated areas. Also it illustrates the methodology to derive other watershed environmental factors from remote sensing imagery. The methodology was applied in Lake Karoun area in Fayoum, Egypt as a case study. The results indicated that, this effort has successfully demonstrated that the used of remotely sensed data and open source available software provide an effective approach to develop accurate watershed analysis for various uses with a minimum amount of time, effort, and cost. Finally, more researches are needed for different terrain and different data to validate the current technologies for watershed modeling. REFERENCES Hung, L. Q., Dinh, N.Q., Batelaan, O., Tam, V. T. and Lagrou D., Remote sensing and GIS-based analysis of cave development in the Suoimuoi Catchment (Son La - NW Vietnam), Journal of Cave and Karst Studies, 64, 1, 23-33, Alabama. Meijerink, A.M.J., 1988a. Data acquisition and data capture through terrain mapping units. ITC Journal, 1: Meijerink, A.M.J., 1988b. Modelling in the land and water domain with a versatile GIS, ILWIS; experiences from a large tropical catchment. In: J. Boumaand A.K. Bregt (eds.), Land qualities in space and time. Proc. Symp. Int. Soc. of Soil Science, Wangeningen. The Netherlands, pp Meijerink, A.M.J., de Brouwer, H.A.M., Mannaerts, C.M. and Valenzuela, C., Introduction to the use of geographic information systems for practical hydrology. UNESCO, Div. of Water Sciences. ITC Publ. no. 23, 243 pp. Tran, H. and Yasuoka Y., Regional Climatic Effects of Urban Expansion in Northern Bangkok, Thailand, 7 th International Computers in Urban Planning and Urban Management Conference, July 18-20, University of Hawaii at Manoa, Honoloulou, Hawaii. UNDP, Fayoum Human Development Report.