IMPACT OF CLIMATE AND CATCHMENT DYNAMICS ON RUNOFF GENERATION

Abstract

The Sub-Saharan region of Africa has been challenged by natural and man-made stresses extending from flood and prolonged drought to poor economic and institutional developments. Ethiopia, situated in the horn of Africa and its economy dominantly based on agriculture, is becoming a victim of such global challenges. The Rift Valley lakes and rivers system of Ethiopia has undergone major changes in recent past. Agricultural, water supply, and hydropower sectors are affected by variable climate patterns. The impact of changing climate condition is more profound in semi-arid regions of Rift Valley lakes basin where competition for water is immense. Quantitative assessment of the impacts of climate and catchment dynamics on runoff generation in the basin is vital. In the present study, an attempt has been made to explore the impact of climate change and catchment dynamics on runoff generation in the Rift Valley lakes basin of Ethiopia. The broad objectives of the present study are: i. Investigation of the spatial and temporal variability of annual and seasonal rainfall over Ethiopia, ii. Identification of non-stationarity and reasons of non-stationarity in hydro-climatic datasets in the Rift Valley lakes basin of Ethiopia using short and long-term time dependence analysis, iii. Assessment of the impacts of topographic, weather and catchment input parameters on runoff generation using Soil and Water Assessment Tool, iv. Analysis of the impacts of climate change on runoff generation using coupled atmospheric-ocean Global Climate Model (GCM) outputs for current and future climatic conditions under varied greenhouse gas emission scenarios, v. Evaluation of the impacts of temporal land use/land cover dynamics on runoff generation using distributed hydrologic model. Spatial and temporal rainfall variability analysis covers entire Ethiopia whereas investigation of non-stationarity in hydroclimatic variables is confined to Rift Valley lakes basin of Ethiopia. Assessment of the impacts of catchment and weather input parameters, GCM outputs under varied greenhouse gas emission scenario and land use dynamics on runoff generation are limited to Bilate (5330 km2) and Hare (166.5 km2) watersheds. Specific iii methodologies applied to achieve the intended objectives and major findings from the analysis are summarized as follows. Spatial and temporal variability of annual and seasonal rainfall over Ethiopia Monthly gridded rainfall data of 50 years (1951-2000) at 0.5° latitude x 0.5° longitude resolution covering entire Ethiopia was acquired from Global Precipitation and Climate Center (GPCC) and its validity for subsequent analysis is examined against nearby observed series. Trends in seasonal, annual and maximum 30-days extreme rainfall over Ethiopia are investigated using Mann-Kendall and Theil-Sen's slope estimator approaches. Spatial coherence of annual rainfall among contiguous rainfall grid points is examined for possible spatial similarity across the country applying Moran's spatial autocorrelation model. The association of Atlantic Multidecadal Oscillation index over Ethiopian rainfall pattern is also explored through statistical analysis. The main summer season and annual rainfall exhibit significant decreasing trend in northern, north-western and western part of Ethiopia. In most other parts of the country (approximately 77% of geographical coverage), the annual rainfall series remained without significant trend for the second half of 20th century. Based on the Moran's spatial analysis, annual rainfall for the total sampling points (381 grid stations) is divided into four zones of annual rainfall spatial similarity. Regions with high annual and seasonal rainfall distribution exhibit high indices of temporal (n and r2) and spatial (Moran index) autocorrelation coefficients. Atlantic Multidecadal Oscillation and annual rainfall indices over the last half century reveal modestly good correlation in the northern region whereas the association is weakly developed in other parts of the country. Investigation of non-stationarity in hydroclimatic variables Statistical analysis of short and long-term persistence in hydro-climatic variables such as rainfall, stream flow and lake level to detect possible time-trends over the historical period is undertaken. Mann-Kendall trend detection method, Theil and Sen's slope estimator, Hurst's coefficient, Spectral and Wavelet analysis approaches are applied to identify trends and periodic signals in hydro-climatic variables. Temporal land use/cover information and nearby Indian Ocean SST anomalies are further examined to study their association to hydro climatic fluctuations. iv Despite less statistically significant trend in seasonal and annual rainfall events and number of rainy days within the catchment, streamflow and lake level have showed significant increasing trend for more than 75 percent of events investigated. This observed non-stationarity is variable across hydro-climatic elements that could likely be attributed to the combined effect of global climatic variability on local climate and altered catchment condition over the years. The estimated Hurst's coefficient (H) is greater than 0.5 for all events of streamflow and lake level, which suggests a likely evidence of long term persistence in hydrologic variables. The deterministic cyclic components of streamflow are legitimately represented as discrete finite Fourier series. The variance explained by the first two harmonics exceeds 96 percent in most cases and the monthly flows are approximated by the first two harmonics. Trend analysis carried out on various model combinations of discrete wavelet decomposed signals detected the prevailing trends in hydrologic variables efficiently. The average stations total rainfall is better correlated to summer season (June-September) SST whereas the association becomes weak for annual average SST. Impact of terrain, weather and catchment input parameters on runoff generation The impact of terrain, weather and catchment input parameters on runoff is assessed using process oriented Soil and Water Assessment Tool. The limitations of 30m resolution Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and 90m resolution Shuttle Radar Topography Mission (SRTM) DEM in watershed delineation are explored. Sensitivity of catchment input parameters while simulating runoff at Bilate and Hare watersheds is investigated and the most sensitive parameters are identified. Contrary to 90m SRTM DEM, the 30m ASTER DEM resulted in spurious flow accumulation path that subsequently reduced the watershed area by 29% and affected other basin parameters at Hare watershed. Soil and Water Assessment Tool effectively captured the underlying hydrologic processes while simulating runoff at both watersheds. The simulated annual water yield is within ±3.4% error to the observed series. Initial curve number for average soil moisture condition, deep aquifer fraction, minimum water depth in the shallow aquifer for flow and available soil water holding capacity parameters are found to either attenuate or accentuate the resulting runoff more significantly than other parameters in the watersheds. Simulating present and future runoff using Global Climate Model outputs The impact of large-scale atmospheric-ocean variables on local-scale hydrology is investigated through Global Climate Model (GCM) outputs under different greenhouse gas emission scenarios. Downscaled and subsequently bias corrected GCM outputs are applied to simulate present and future runoff at Bilate and Hare watersheds of Rift Valley lakes basin of Ethiopia. Future implications of extreme precipitation and runoff events are discussed from GCMs outputs for varied greenhouse gas emission scenarios. Large-scale GCM outputs are obtained from BCCR-BCM2.0 of Norway and CSIRO MK3.0 of Australia GCMs and subsequently reduced to local-scale weather variables (temperature and precipitation) using station weather data. Since GCMs are operating at coarser scales, the statistical downscaling model (SDSM) is employed to reduce large-scale atmospheric variables to local level weather condition. The statistical downscaling model, followed by bias correction, effectively reproduced the current climate (1990-1999) weather variables. Statistically downscaled and subsequently bias corrected daily temperature and precipitation variables are used to simulate runoff for present and two future (AIB and A2) greenhouse gas emission scenarios at Bilate and Hare watersheds. Simulated future runoff events are characterized by increased extreme events that ultimately resulted in increase in the gross annual runoff volume from the watersheds. The simulated runoff varies from -4% to 18 % at Hare watershed and is within the range of -4 % and 14 % at Bilate watershed. Simulated average annual water yield shows slight variation between GCMs. It lies within ±10 % at Bilate basin and ranges from -17% to 12% at Hare basin. Future water resources planning and management could likely be affected by such variability and hence existing design methods could expand their scope to account for these extreme events. Impact of temporal land use/land cover dynamics on runoff generation Temporally varying (1976/1986/2000) Satellite image acquired from landsat satellite system is processed and land use/cover classes are identified. Runoff is generated using SWAT model for temporally varying land use /land cover conditions and the ensuing results of land use dynamics on runoff is discussed. Statistical trend test is also applied to detect the unexplained natural climate variability. Joint analysis of watershed modeling for temporally varying land use/land cover condition and statistical time-trend analysis of streamflow is undertaken to explore the VI impact of altered land use/land cover condition on runoff generation at two watersheds. The method detected the underlying variability efficiently. The percentage of forest cover declined substantially at Bilate and Hare watersheds during 1976/2000 analysis period. The simulated surface runoff component increases progressively since 1970s. Percentage annual surface runoff varies from 10 to 23% at Bilate and 16% to over twofold at Hare watersheds. The increasing trend of observed daily maximum flow at Alaba Kulito and slightly raised slope of rainfall-runoff double mass curve since 1992 supports the attribution of climate induced changes at Bilate catchment. Future attempts could incorporate long record observed hydro-climatic variables to examine the accompanying spatial and temporal variability. Application of GCM outputs from multi-models better help to understand the future climate variability and its associated impacts on local Hydrology