Estimating runoff from ungauged catchments using regional modelling

Abstract

Establishing runoff series from ungauged catchment is a central challenge in hydrology. In Norway this is particularly evident in the design of small hydropower. The location of the power plant often is in small catchments with little little or no data available. The application of regional modeling for estimating runoff in ungauged catchments is one of the promising methods. The main goal of this study is to use ENKI hydrological model to calibrate free set of parameters which can be applied everywhere within selected region in order to estimate runoff in basins where Sweco Company was carrying out their own measurements.The study region located in central Norway and contain Sør-Trøndelag, Nord-Trøndelag and Møre og Romsdal Counties. The ENKI model has been set-up and all necessary hydro-meteorological and geographical input data for the period from 2000 to 2012 have been collected and processed. Three cases of calibration were carried out to obtain the best regional set of parameters for the entire study area.The first case of calibration was done including all the catchments and calibration period is from 2001 to 2005. The results showed variability of R2 from -0.24 (Farstadelva v/Farstad catchment) to 0.85 (Eggafoss catchment) for individual calibration. The regional set of parameters for the first case of calibration resulted in Nash efficiency of 0.15 which is comparatively very low.Second calibration run was done to improve Nash efficiency results. In this case the catchments which were giving poor R2 values have been excluded and calibration has been carried out over 9 catchments. The variability of R2 for second case is from 0.375 (Vistdal catchment) to 0.85 (Eggafoss catchment). The regional set of parameters has been improved and resulted in R2 = 0.55. Validation of the model has been carried out for the period from 2005 to 2011 and resulted with regional value of R2 = 0.583 indicating that the model is applicable for predicting runoff. Second case of calibration has been used for extraction runoff data series for Sweco catchments.The third case of calibration differs from the second one by performance of PcorrRain and PcorrSnow parameters as raster maps with best values from first case of individual calibration. The regional set for the third case of calibration resulted in R2 = 0.452 which is lower than the result from second calibration.The simulated runoff for Sweco catchments was extracted and compared with observed values. Applicability of ENKI regional modeling has been compared to scaling approach. The overall results were not satisfactory for small catchments which can be improved in further studies.