Modelling impact of land use change on sediment yield of river Yala, Kenya

Abstract

Increase in human population and the subsequent need for economic activities has led to degradation of catchments in most Sub-Saharan countries. Unsustainable land use coupled with management practices are the main causes of soil erosion, which leads to land degradation. Transboundary basins occupy about 60% of world fresh water in 192 countries. Of these, 310 lakes and rivers are shared by 153 countries and they serve 2.8 billion people, about 42% of the world population. The main objective of this study use SWAT model in simulating sustainable land use management practices of river Yala catchment . The specific objectives were to: determine the spatial and temporal land-use change of river Yala catchment (1973-2000), Set up, calibrate and validate the Soil Water Assessment Tool (SWAT) model to predict streamflow, sediment quantity, and concentration in river Yala , and Apply the SWAT model to analyze various management scenarios that may reverse the impacts of the land use changes of the Yala catchment . Land use/cover database for a period of 27 years representing the beginning, mid and end of the period for the years 1973, 1986 and 2000 were analyzed to determine changes. SWAT model integrated with Geographic Information System (ArcGIS, version 10.3) was used to analyze the images, simulate discharge and sediment yield. Other data required for modelling included soil, elevation, drainage, climate, and land use. The model was calibrated and validated using the SWAT-CUP and flow at IF02 gauge station (Tindinyo) on monthly time step for the years 1979-1983 and 1984 -1988 respectively. Sediment predictions and streamflow of the watershed was carried out by spatial resolution through watershed subdivision. Three scenarios were used to represent different patterns of LULC. Scenario A represented baseline, i.e. the original watershed conditions. In scenario B, 30% of pastureland was converted to the forest and for C, strip farming was introduced into the watershed. The findings of this study indicated that in the base year (1973), the largest LULC was occupied by vegetation - covering 56% of the entire area and then reduced to 30% in 1986, and 21% in 2000. The settlement area increased from 20% in 1973 to 67.9% in 2000.Bare lands that were 25% reduced to 20% then 11% in 1986 and 2000 in the three scenarios respectively. The highest sediment concentration was 3,552.4 mg/l in 1991 while the lowest was 612.71mg/l in 1985. Model performance measures coefficient of determination (R 2 ) was 0.72 and the Nash–Sutcliffe simulation efficiency (NSE) of 0.79 for calibration. For validation, R 2 = 0.80 while NSE was found to be 0.94%. These indicated a good performance of the streamflow simulation on the monthly time step. Flow prediction and soil loss are key tools for determining suitable land use and conservation measures. SWAT model integrated with GIS effectively simulated sediment transfer and water phenomena. It is therefore recommended that spatio-temporal land cover images of higher resolution based on future scenarios be analyzed to mitigate the negative effects and recommend appropriate management practices.