Regional modelling for estimation of runoff from ungauged catchment, case study of the Narayani basin, Nepal

Abstract

Nepal is richly endowed with water resources.Since Nepal has huge potential in Hydropower and much more hydropower energy is to be extracted .There are feasibility studied carried out for numbers of project very before and has not gone to construction.These project and the new project should go through hydrological analysis.In underdeveloping country like Nepal the long time series data are not available for many intake sites.

Generally scaled data were used in planning of major hydropower project due to which ther is possibility of of incorrect forecast of flow as a result power plant has lower capacity than planned.

To overcome this situation ENKI could be a reliable model to find the flow of ungaged sites.The regional modeling concept was introduced for finding flow in unguaged station In. Regional model like parameterization is done in and are introduced to ungaged stations.

ENKI framework was setup by creating routines for the model and linking the model with the region.The input data required for the model were network, gridded and scalars precipitation,evaporation,temperature,runoff,catchment area,landuse etc.

All the data were taken from 1991-1994 for calibration and from 1995-1998 for validation. The river flow and other datas were found to be of good quality with no missing values.

The ENKI model needs some spatially distributed input which were preapered in ArcGIS and further reclasiification was done as per requirement then converted in IDRISI format.

After setting the model various adjustment of parameters was done to obtained good performance.Finally R2 for Devghat,Betrawati and Shisaghat was 0.7516,0.8118 and 0.5958 for calibration and .6,0.72,0.58 for validation respectively.Since there was negligible snow contribution on the model so significant low flow outflow from model was observed during first three month of the years.

The calibration could have been done for even long series but due to the size of raster being much larger than the catchment area ,the time series was minimized for fast simulation.

Finally, almost all parameters were obtained to acceptable range except for percolation which was much higher.