This study presents a geomorphology based semi-distributed methodology for prediction of runoff of a catchment. In this proposed methodology, the catchment area is divided into a number of sub-catchments using the Thiessen polygon method. The rainfall records of particular rain-gauge station are considered as uniformly distributed over the entire sub-watershed. Four different weighting factors are proposed to obtain the sub-catchment’s contribution towards runoff. The weighting factors are calculated based on the geomorphological parameters of the catchment. The geomorphological parameters of the sub-watersheds are obtained from SRTM digital elevation data. The weighted contributions from all the sub-watersheds at current and previous time steps and the previous time step discharge are used to develop an Artificial Neural Network (ANN) for predicting the discharge at the basin outlet. A lump model considering average rainfall of the catchment is also developed using ANN for evaluating the performance of the proposed distributed model. For the lump model, average rainfall is calculated using Thiessen polygon method. The historic rainfall and runoff data recorded at the Dikrong basin, a sub-catchment of the river Brahmaputra is used to evaluate the efficiency of the developed methodology. The evaluation results show that the presented model is superior to the lump model and has the potential for field application. A comparative study is also carried out to obtain the most influential combination of geomorphological parameters in predicting the catchment’s runoff.