Downscaling of CMIP5 Models Output by Using Statistical Models in a Data Scarce Mountain Environment (Mangla Dam Watershed), Northern Pakistan

Abstract

In this study, statistical downscaling models were used to project possible future patterns of precipitation and temperature in the Jhelum River basin shared by Pakistan and India. In-situ meteorological data were used to downscale precipitation and temperature using different General Circulation Models (i.e., CanESM2, BCC-CSM1–1, and MICROC5) relative to baseline (1961–1990) under the Representative Concentration Pathway (RCP) scenarios RCP4.5 and RCP8.5. The downscaling models used were the Statistical Downscaling Model (SDSM), which uses multiple linear regression and weather generator methods, and the Long Ashton Research Station Weather Generator (LARS-WG), which uses weather generators. The results showed that the SDSM performance was slightly better than that of LARS-WG during validation and that the representation of the simulated mean monthly precipitation was more correct than that of monthly precipitation. The results also revealed that BCC-CSM1–1 performed better than CanESM2 and MICROC5 in the study region. The future annual mean temperature and precipitation are expected to rise under both RCP scenarios. The changes in the annual mean temperature and precipitation with LARS-WG were relatively higher than those with SDSM. Out of four seasons, winter and autumn are expected to be more diverse with regard to precipitation changes. However, although both models yielded non-identical results, it is certain that the basin will face a hotter climate in the future.