Potential on the Use of GIS Watershed Modeling for River Basin Planning _ Case Study of Attanagalu Oya Basin, Sri Lanka

Careful management for identifying and allocating available water resources is important, to preserve water for present as well as future users. Modeling at basin scale can provide necessary information on river basin planning and support policy makers in their decisions on allocating water resources in the basin. This study focused on identifying the potential of GIS for spatially distributed modeling of watersheds for river basin planning. The spatially distributed GIS model represents the simulation of runoff in selected river basins. Selected river basins are characterized by nodal network approach and sub-catchment areas calculated using GIS tools. Runoff is assumed to be triggered by rainfall and mainly dependent upon land use, soil type and slope conditions. Spatially varied land use, soil and slope data for the selected catchments were digitized using GIS (vector format) and land parcels created by overlay operations in GIS showing different catchment characteristics were used in the model. All calculations were performed with standard GIS tools of ArcGIS software. A simple conceptual model was used to compute runoff from each land parcel. The model is capable of calculating runoff from catchment characteristics based on predetermined runoff coefficients at the presence of precipitation data. A nodal representation of case study basin was prepared using MIKE BASIN to estimate streamflow at each node. The runoff coefficient for Attanagalu Oya basin with the spatial variation of parameters was estimated as 0.51 using this approach. Model provides streamflow at any node provided in the model. The average streamflow identified to fluctuate between 10m3/s to 60m3/s without taking water extractions into consideration. Land use, soil, slope patterns can be changed in the GIS model to calculate new coefficients with new spatial variation and can be incorporated to the nodal modal to visualize the changes in streamflow patterns. Model results indicate the potential of the use of spatially distributed model for decision making when carrying out river basin planning.


Introduction
Rapid population growth, urbanization, drastic changes in land use and growing industrialization are threatening water resources with the increasing demand for water. Changes in aforementioned factors directly or indirectly affect the rainfall, and streamflow patterns. Hence, estimating streamflow with available rainfall to manage available water is important for water resources management.
Relationship between rainfall-runoff is a complex hydrological phenomenon. Rainfall runoff models have taken part in watershed modeling. During the past few decades rainfallrunoff modeling has attempted to establish major hydrologic concerns. For instance, Naden [10] and Troch et al. [16] present spatially distributed rainfall runoff models which include channel flow routing and overland flow; Jothityangkoon and Sivapalan [8] developed a distributed rainfall-runoff model for extreme flood estimation. GIS based rainfall runoff models have also received similar attention and have shown their importance. Jain et al. [6] had introduced a GIS based distributed rainfall runoff model capable of handling the catchment heterogeneity in terms of distributed catchment characteristics and rainfall which was applied to isolated storm events in several catchments. In addition, rainfall-runoff modeling for ungauged basins is of great importance due to the limited data. Catchment characteristics such as land cover, soil, and slope have been identified as the major parameters which can be used to represent generated streamflow in ungauged basins since they govern the runoff [11,12,21]. This study focused on developing a GIS model to identify the potential of GIS for modeling watersheds using spatially distributed parameters.
Water resources are scattered unevenly indicating abundance and scarcity ranging from regional level to global level. Most of the basins in the world are ungauged and even in gauged watersheds sub-watersheds are ungauged. To manage a river basin, it is necessary to analyse them at spatial locations where there are extractions, diversions and joining of branches. The major objective of this study was to model ungauged watershed with the extrapolation of results from gauged watersheds, and to identify the potential of using GIS for the incorporation of spatially distributed physical characteristics, inputs, extraction and diversions.

2.
Methodology and Study Area 2.1 Methodology 2.1.1 Model Description This paper discusses the spatially distributed modeling of Attanagalu Oya basin using MIKE BASIN software to assess the water resources availability at a downstream station of the basin. Figure 1 shows the methodology flow chart of the study; including model development for gauged watersheds to establish runoff coefficients and extrapolated model simulations for Attanagalu Oya basin in order to analyze streamflow.
Available data were collected and checked for adequacy and consistency as data is very important for reliable river basin modeling [13,15]. Details and methods given in Wijesekera et al, [20] were used as guidelines for data collection and processing. Rainfall data, streamflow data, evaporation data, demand data, and also land use, soil, contour and other digital data required for GIS layer preparations were collected for the study [11]. Data were checked for consistency and compatibility with different techniques such as visual examining, double mass curve method and statistical checks for homogeneity. In homogeneity checking it was assumed that the data are independent and with normal distribution though it may not always be satisfied by some data sets [3,7].
The model used for the study is as follows [11,12].
Governing catchment characteristics namely the land use, soil, and slope were grouped into classes of five, two and three for land use, soil and slope, respectively.

2.1.2
Determination of Basin Parameters Spatially varied land use, soil and slope data which were collected and identified on maps were digitized using GIS (vector format). Three catchment characteristics for a given basin was overlaid using overlay operation in GIS. This activity creates different land parcels with different catchment characteristics. Required land parcel was selected using table operations for later applications. A simple conceptual model was used to compute runoff from each land parcel using the concept presented in equations 1 and 2. The model assumed a linear function incorporating land use, slope and soil as major catchment parameters contributing to convert rainfall into surface runoff as illustrated in equation 3.
The cumulative runoff contributed from each land parcel in a catchment was taken as the surface runoff generated from the catchment.
where, R = Rainfall Aijk = Area of concern with given factors i,j,k Pijk represents the coefficient for i th land use type, j th slope class and k th soil type where i varies from 1-5, j varies between 1-3 and k from 1-2. For instance, P231 represents the coefficient for steep slope garden areas with red yellow podzolic soil type. Parameters of the model assigned by the above criteria are given in Table 2. Model parameters (P values) were estimated (optimized) using Mean Ratio of Absolute Error as the objective function [18,19]. Parameter optimization was initiated with the literature values [1,2,4,5]. Parameters at which minimum MRAE were selected as finalized parameters of the model.

Kelani Ganga at Glencourse and Kalu Ganga at
Putupaulla gauging stations were used as supportive catchments to evaluate runoff coefficient soil, slope and land use relationships since the catchment at Karasnagala gauging station did not cover all the soil types of Attanagalu Oya basin. Attanagalu Oya basin soils fall into only two major soil groups as Red Yellow Podzolic and Alluvial (Table 4).
Twenty years monthly rainfall data were selected for the Kelani and Kalu Ganga sub basins. Data period from 1951-1960 were selected for calibration and 1961-1970 were selected for verification. Ten years of monthly rainfall runoff data from 1970-1980 were used for model calibration (1970)(1971)(1972)(1973)(1974)(1975) and verification (1975)(1976)(1977)(1978)(1979)(1980) at Karasnagala sub-basin. Initial runoff coefficient values were obtained from literature [4,14,17]. Runoff coefficients were then optimized by matching calculated and observed runoff. It was assumed that watersheds with similar catchment characteristics will produce similar pattern of runoff from rainfall. It was also assumed that for monthly time scale computations, the soil wetness would not have a significant impact. In order to minimize this factor, the case study selected wet zone basins only.
Representing river basins by nodal network approach is the common approach for computations of water allocation [9], where lines represent the stream segments and nodes represent water extractions and other necessary activities.
Calculated streamflows from initial GIS model were incorporated to MIKE BASIN nodal network of Attanagalu Oya basin in order to identify flow at different nodes of interest. In the developed representation, nodes are located at stream segment connection points. Model requires specific runoff at each network branch. Sub-catchment area, relevant nodal points, flow direction, etc are the other inputs to the MIKE BASIN model. Moreover, in case of water extractions, such water extraction locations, any prioritization of water extractions, lag times, etc. would need to be added as input data.

Alluvial (AL) -(2)
Forest (1) area at Kotugoda is approximately 539 km 2 . Figure 2 shows the Attanagalu Oya basin with the stream network of the sub catchment at the Karasnagala stream gauging point and at Kotugoda. Five rainfall gauging stations were selected to represent the spatial variability of rain in the basin. The station at Karasnagala maintained by the Irrigation Department of Sri Lanka, is the only available stream gauging station (non-recording).
The station has had a recorder at Karasnagala, but site visits revealed that this is not properly functioning. Department of Meteorology maintains a daily evaporation records at Katunayake, which is a principal meteorological station recording data once in every three hours.

Kelani Ganga basin at Glencourse (Kelani Ganga sub basin)
Kelani Ganga basin is located in the wet zone of Sri Lanka between the latitudes 6° 45' and 7° 15' N, Longitudes 79° 50' and 80° 45'E. Sub basin of the Kelani Ganga at Glencourse was selected for this study. Total catchment area is about 1537 km 2 at Glencourse (Kelani Ganga sub basin). Data from six rainfall-gauging stations and one streamflow gauging were available for the study [11,12].

Data Analysis
Out of the five gauging stations selected, for the Kotugoda sub-catchment, two stations are within the Karasnagala sub-catchment. Daily rainfall data were analysed to obtain the results. Highest annual rainfall has been observed in 1975 while minimum was observed in 1976. According to the computations the highest rainfall of 4322 mm has been observed at Vincit. Lowest rainfall of 1640 mm was observed at Henarathgoda when averaged over the 10 year analysis period.

Spatially Variable Basin Parameters
Data checks provided agreeable results with an error record at Karasnagala streamflow data in year 1975. Annual average rainfall for Attanagalu Oya basin is within the range of 2000mm-3500 mm. Typical dry month for the Attanagalu Oya basin is January. Average runoff coefficient was found as 0.40 for Attanagalu Oya basin at Karasnagala considering Karasnagala rainfall and streamflow without considering spatial variation over the catchment. Averaged optimized runoff coefficients from the study were obtained as 0.51 for Attanagalu Oya basin at Karasnagala under spatial considerations.
Comparing the covering area, latosol and regosol soil areas were neglected (0.3% of total catchment area). Hence, only red yellow podzolic soil (89.1% of total area) and alluvial soil (10.6% of total area) areas were considered in this study. Cultivations are the higher percentage of land use for the basin. Cultivations include coconut, rubber, paddy and other cultivation [4]. In case of the cultivations optimized runoff coefficients are 0.61, 0.57 and 0.20 for steep, average and flat slopes respectively for red yellow podzolic soils while for alluvial soils the runoff coefficients are 0.50 and 0.55 for average and steep slopes respectively. Residential areas generate higher amounts of runoff from rainfall. Such areas showed higher runoff coefficients values of 0.65, 0.60 and 0.55 for red yellow podzolic soils for steep, average and flat slopes respectively. Run off coefficients in residential areas with alluvial soils are 0.52, 0.56 in case of average and steep slopes respectively. Lowest runoff coefficient obtained for forest areas with alluvial soils followed by that for red yellow podzolic soils. Optimized runoff coefficients are given in Table  5.

Spatially Variable Stream Network
The nodal network prepared for Kotugoda sub basin with the incorporation of spatially variable inputs to MIKE BASIN model is given in Figure  5. In the Attanagalu Oya model simulation, water extractions could not be considered due to unavailability of data. Model demonstrations assessed water resources without extractions.
Model results provide the streamflow values at the Kotugoda outlet point as well as streamflow values at each node of interest. Calculated streamflow over time for Attanagalu Oya at Kotugoda outlet is shown in Figure 6. Flow duration curve for the calculated streamflow is given in Figure 7. Calculated zero flows are replaced by minimum flow of the basins at zero rainfall record periods because model results with zero streamflow with zero rainfall even though river flow has not gone zero.
According to the hydrograph, every year low flows can be observed in January while higher flows are observed in May and October. The average streamflow fluctuates between 10m 3 /s to 60m 3 /s. The model computations were done with the varied runoff coefficients [11,12]. However due to non-availability of streamflow, these values could not be verified. However the representative result showed that the GIS approach well demonstrated the potential of water resources management with the incorporation of the spatial variability of soil, slope and land use.

Conclusions
The present case study demonstrated the potential of the use of GIS to incorporate the spatial variability of catchment characteristics and rainfall to manage water resources of river basins. Such techniques require runoff coefficients, which have been verified for similar conditions.
This GIS application enable a river basin manager to analyze the behavior of surface water once the catchment characteristics are changed at various locations of given watershed.
The overall runoff coefficient for Attanagalu Oya basin with the spatial variation of parameters was 0.51. Model provides streamflow at any node provided in the model. The average streamflow identified to fluctuate between 10m3/s to 60m3/s without taking water extractions into consideration. Land use, soil, slope patterns can be changed in the GIS model to calculate new coefficients with new spatial variation and can be incorporated to the nodal model to visualize the changes in streamflow patterns.
To improve the model outputs, gauged streamflow data is needed for Attanagalu Oya basin. The study can be improved further by adding gauged water extraction scenarios, which is more helpful for river basin planning and management.