Assessing Forecast Quality of HII Flood Forecast Service in Chao Phraya River Basin

Abstract

Hydrometeorological forecasts are essential to water management plans including early warning and flood damage prevention. Forecasting models have varying levels of skill depending on the forecast location and period of the year. Measure of skills can have a strong influence on how forecasts impact decisions related to water management, and they must be communicated to the users of the forecasts. Various forecast verification methods are available for assessing the multiple facets of forecast performance including notions such as accuracy, reliability and sharpness. This paper describes a variety of complementary performance metrics to verify Hydro Informatic Institute (HII)’s flood forecasts in Chao Phraya River Basin. The accuracy of the forecasts is evaluated using the continuous rank probability score (CRPS) which quantifies the difference between a forecast distribution and observation. The sharpness of forecasts is calculated using the ratio of inter quantile range (IQRs) of streamflow forecasts and a historical reference. The reliability of forecasts is also considered using attribute diagrams and Kolmogorov-Smirnov (KS) test. In addition, this paper applies the traditional continuous verification methods and statistics such as Bias, Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Nash-Sutcliffe Efficiency Coefficient (NSE), Coefficient of determination (R²) and Pearson Correlation Coefficient (r). The comparison of the forecast and observed discharge indicate that the MIKE11 model can predict well. The trends are similar in almost all key stations and the overall correlation is acceptable. This study definitely answers the question regarding the correlation between the forecast and observed streamflow and the performance of the forecasts. Based on the verification statistics, it was demonstrated that HII’s flood forecasts are reliable.