The livelihood of people in many regions of Ethiopia depends on rainfed agriculture. Accuraternprediction of crop yield could greatly improve potential famine and allow advancedrnplanning of intervention operations. This thesis explores the feasibility of a combined RegionalrnClimate Model (RegCM) and crop model for crop yield forecasting in Ethiopia, usingrnwheat yield for the Amhara region as a case study. An important focus in the investigationrnis to validate and asses the ability of RegCM-4 regional climate model to represent thernEthiopian summer rainfall. The ability of the RegCM-4 model in capturing temporal andrnspatial variability of precipitation over the region of interest is evaluated using metrics spanningrna wide range of temporal and spatial (Ethiopian domain average to local) scales againstrnGlobal Precipitation Climatology Project (GPCP) observational datasets. The simulated periodrnis 1995-2008. RegCM-4 shows a general overestimation of precipitation except thernhighlands part of the country. The precipitation bias over the Ethiopian highland, our mainrnarea of interest, is mostly less than 20%. The model captures well the observed interannualrnand inter-seasonal variability. On short time scales, simulated daily temperature and precipitationrnshow a high correlation with observations, with a correlation coefficient of 0.79rnfor kiremit season. It is therefore that RegCM-4 has sufficiently good quality to performrnclimate change experiments over Ethiopia, for application to impact and adaptation studies.rnRegCM-4 outputs are used to drive a process-based crop model, General Large Area Modelrnfor Annaula Crop (GLAM) for hindcasting zonal wheat yields in the Amhara region. Simulatedrncrop Radiation Use Efficiency (RUE) has been founded to be 1.81 which is expectedrnfor C3 crops. The yield in these simulations showed a negative bias (159-200kg/ha) withrnobserved yield over Souther(North Shew Zone) and South Western(Awi Zone) of Amhararnregional state. This is probably because at the field level the yield variability was mainlyrnaffected by field managements and diseases, pests and so on. GLAM does not predict therneffect of the detailed field management, diseases and pests on yield variability; and also inrnthis region there is overestimation of RegCM-4 precipitation, which might have lead to waterrnstress in GLAM model. At regional level(for all grid cells), there were higher correlationsrn(0.74) between observed and simulated yield. We therefore conclude that the GLAM modelrnis suitable to simulate crop yield at regional scale (approximately 50 km) using RegCM-4rnoutputs