MODELING OF HYDROLOGICAL RESPONSES TO LAND USE/LAND COVER AND CLIMATE CHANGES

Abstract

The importance of rainfall-runoff modeling can hardly be over-emphasized. In this study, the popular Soil and Water Assessment Tool (SWAT) model has been used employing the data of Gidabo, Hombole, and Weyb watersheds of Ethiopia. The SWAT model employs water balance equation for computation of total runoff. In this equation, the rainfall-excess or direct surface runoff is computed using the popular Soil Conservation Service-Curve Number (SCS-CN) methodology. Thus, the determination of runoff heavily depends on land use/land cover (LULC) and climate change among several others. The present research work aims to apply multi-site calibration technique to evaluate the performance of the SWAT model, investigate the hydrological responses to individual LULC and climate changes, and propose SCS-CN-inspired models for simulation of long-duration rainfall-generated runoff, as follows. Multi-site calibration of SWAT model In hydrological studies, SWAT model is usually (i.e. conventionally) calibrated using the data of a single site, which is often the watershed outlet. It is, however, quite possible that a basin is gauged at multiple sites, and therefore, the model can be more reliably calibrated using a larger set of data as input besides allowing for compensating the instrumental/human errors associated with the measurements at one site by the data of other site(s). Following this concept, the popular SWAT model was calibrated/validated using the data of Hombole, Weyb, and Gidabo watersheds located in Ethiopia and measured for discharge at Hombole and Melka Kutire stations in Hombole watershed, Denbel and Alem Kerem stations in Weyb watershed, and Measa and Aposto stations in Gidabo watershed. To this end, the most influential parameters were selected first using the global sensitivity function of the SWAT CUP routine. Then, the model was calibrated following (i) simultaneous multi-site calibration (SMSC), where data of measuring stations were used simultaneously in a single calibration; and (ii) conventional calibration (CC), i.e. individual calibration in sequence from upstream gauging station to downstream gauging station. The calibration used five statistical error criteria involving NSE, KGE, PBIAS, RSR, and R2. The SMSC technique performed better than CC in all applications to three watersheds. The final SMSC calibrated parameter values of Gidabo watershed were used to explore the climate and land use/land cover change impacts on water balance components.Hydrological responses to land use/land cover and climate changes Land use/land cover (LULC) changes greatly affect the basin hydrology. Their lump-sum effect on various components of hydrologic cycle has been generally evaluated and reported in literature, except for a few studies which dealt with individual LULC changes. This study is a step forward in this direction to evaluate the hydrological response of individual LULC changes on different components of hydrological cycle of the Gidabo watershed, Ethiopia during the period 1988-2018 using the SWAT and Partial Least Squares Regression (PLSR) models. Change in LULC resulted in an increase in annual mean surface runoff by 23.6 mm during the period 1988–2018. Meanwhile, the annual mean baseflow, lateral flow, percolation and evapotranspiration (ET) were reduced by 9.1, 4.2, 10.2, and 12.2 mm, respectively. The changes in hydrological elements were much pronounced at sub-basin levels, which were mainly associated with uneven spatial LULC alterations. PLSR analysis indicated that expansion of cultivated and urban areas and contraction of forest evergreen, agroforestry, and grassland are the main contributors to the upsurge in surface runoff and decline in baseflow and percolation. Likewise, cultivated land affected the lateral flow and ET negatively in contrast agroforestry and grassland, which had a positive influence. Besides LULC change, climate change is also one of the key factors governing the watershed hydrology. Hence, the individual and combined impacts of future climate and LUCC changes on the water balance of Gidabo watershed have been studied. A climate ensemble of five regional climate models, derived from the Coordinated Regional Downscaling Experiment (CORDEX-AFRICA domain) under the Representative Concentration Pathways (RCPs) 4.5 and 8.5, have been used for climate assessment while Cellular Automata-Markov Chain model has been used to predict the future LULC maps corresponding to the years 2046 and 2075. The hydrological impact has been evaluated using the SWAT model at watershed and subwatershed levels while Pearson correlation matrix has been used to develop the relationship between change in rainfall and change in water balance components. The individual impact of climate change under RCP 4.5 shows a decline in mean annual surface runoff (Q), water yield (WY), and evapotranspiration (ET) during mid-century (2027-2056) by 12.88%, 14.41%, and 0.32%, and during late-century (2061-2090) by 18.15%, 22.63%, and 2.89% respectively with respect to the baseline period (1988-2018) which is often associated with a decline in rainfall. Likewise, under RCP 8.5 scenario, these water balance components have shown reduction during both these periods. The LULC change (often associated with an expansion in cultivated and urban area and a reduction in evergreen forest and grassland) alone leads to a positive synergy with an increase in mean annual Q and WY by 2.68% and 0.41%, respectively, and a negative synergy with a decrease in ET by 0.33% during the period 2018 to 2075. The combined effects of climate and LULC changes have shown a decline in Q, WY, and ET. Overall, the climate change impact is found to have a more prominent effect than LULC change alone as well as their combined effects on the future Q and WY in the study watershed. However, the combined effect of climate and LULC changes on ET is found to be more influential than their individual impacts. It is recommended that future climate and LULC change scenarios should be duly considered for water resources planning and management in the Gidabo watershed. SCS-CN-based rainfall-runoff modeling Rainfall-runoff modeling being highly complex, dynamic and non-linear exhibits temporal and spatial variability as it comprises several physical processes. Varying in complexity from lumped empirical to physical based space and time-distributed; several models are available in literature to model runoff. Although physically based models have proven very useful as a research tool but are of limited used in filed, especially in developing countries like Ethiopia as they require large amount of data. However, search is continuing for developing a new simpler model. Among many available rainfall-runoff models the SCS-CN method is one of the well-known simple rainfall-runoff models. Therefore, in this study an attempt was made to (i) clarify some issues related to the popular rainfall-runoff SCS-CN methodology and (ii) propose new SCS-CN inspired models for simulation of long-duration (viz., bi-monthly, monthly, seasonally, and annually) rainfall-generated runoff and compares its performance with the existing SCS-CN models and Mishra and Singh 1999 model. Issues of SCS-CN are: (a) both C (runoff coefficient) and CN (curve number) with P (rainfall) in contrast to that exhibited by field data, (b) F (infiltration) with Q (direct runoff), (c) Ia (initial abstraction) with S (potential maximum retention), and (d) CN that is also taken as an index of runoff potential. The performance of the proposed SCS-CN inspired models (M3-M8), the existing SCS-CN models (M1 and M2), Mishra and Singh 1999 model (M9), and a special case of M9 that hypothesizes that initial abstraction coefficient () = 0 (M10) are tested using rainfall-runoff data of four different agro-climatic river basins in Ethiopia. The performance of the models is evaluated using three statistical criteria involving NSE, RSR, and PBIAS. The resulting high NSE values and lowest RSR and PBIAS for the proposed models reveal that proposed models performed better to different (majority) duration datasets than the existing models. Similarly, the proposed models, when employed to the observed datasets of different time scales, per formed satisfactorily in both calibration and validation for all watersheds, underlining the efficacy of the proposed models in field applications. Besides the above, an attempt has also been made to show the existence of a relationship between SCS parameter CN, potential evapotranspiration (PET), and time duration (D) using the data of four Ethiopian watersheds. To this end, CN values were determined for various durations from observed rainfall-runoff data, and PET derived for different watersheds utilizing the Thornthwaite method employing temperature data. The results indicated that the CN values decline exponentially with increasing PET and duration D, and vice versa, for all three wetness (or antecedent moisture) conditions of the watersheds. It is largely attributed to the more opportunity time available for rainwater to infiltrate/evaporate with increase in duration. Thus, the proposed simple CN-PET-Duration relationship is also useful for estimation of PET from the available/published CN values. Keywords: Ethiopia; Multi-site calibration; SWAT model; hydrological component; land use/land cover (LULC) change; Partial Least Squares Regression (PLSR) model; climate change; SCS curve number method; long-duration hydrologic simulation; potential evapotranspiration.