Uncertainty Analysis of Statistically and Dynamically Downscaled Precipitation Data: A Study of the Chaliyar River Basin, Kerala, India

Abstract

Although assessment of the anticipated impacts of projected climate change is very much required for many sectors, non-availability of climate data on the local scale is a major limiting factor. In this background, statistical downscaling has a lot of scope. However, the downscaled data have to be thoroughly analysed in order to assess the uncertainty associated with it. In this study, a detailed uncertainty analysis was performed on statistically and dynamically downscaled monthly precipitation data in the Chaliyar River Basin, in Kerala, India. The mean and variance of the downscaled and observed data for each month were compared. The Wilcoxon signed-rank test, Levene’s test, Brown–Forsythe test, and the nonparametric Levene’s test were performed on the downscaled precipitation data at 5 % significance level. Results showed that the error is not significant in the case of the statistically downscaled data using predictors generated from the reanalysis data. In the case of statistically downscaled data from the predictions of the general circulation model (GCM), error in the mean is significant in some months, probably due to uncertainty in the GCM predictors, whereas the error in the variance is insignificant. For dynamically downscaled data, the error in the mean as well as the variance is not significant. Uncertainty analysis is required to be performed on the downscaled data before its use in impact assessment.