Modelling of river flow in ungauged catchment using remote sensing data: application of the empirical (SCS-CN), Artificial Neural Network (ANN) and Hydrological Model (HEC-HMS)

Abstract

In the present study an attempt is made to provide empirical and deterministic modelling approach for deriving flood frequency curve in ungauged Keseke river catchment, South Nation Nationality and People (SNNP)-Ethiopia. The research work consists of (i) extracting of remote sensing data; (ii) evaluation and validation of remote sensing data; (iii) modelling of river flow using remote sensing data (climate and physiographic data) of the river catchment; (ii) three types of hydrological models validation and evaluation; (iv) developing of flood frequency model for each sub-catchment. The evaluation and validation of remote sensing data and river flow prediction is carried out on eight selected rivers in Keseke River catchment. The single gamma distribution quantile mapping is a good approximation to adjust satellite precipitation product data and the Pearson correlation function has shown a good correlation, mainly on heavy rain events. Results reveals that the SCS-CN and ANN approaches are suitable to predict river runoff in ungauged with reasonable accuracy in the investigated sub-catchments, and appears acceptable correlation between estimated and corrected satellite rainfall. A field campaign to obtain possible data was executed via interview and river cross section measures. The flood quantiles are compared with one time flow observation from field measured value (which is estimated from the river cross-section size) to identify the most representative hydrological model structure.