Benefit of time-varying downscaling model for the assessment of urban temperature rise

Abstract

Temperature rise may exhibit time-varying characteristics due to its inherent non-stationarity. Whereas most of downscaling models inherently assume stationarity, this study establishes the benefit of Time-Varying Downscaling Model (TVDM) over time-invariant approaches, such as Statistical Downscaling Model (SDSM) and Bias Corrected Spatial Disaggregation (BCSD) model. Daily maximum (Tmax) and minimum temperature (Tmin) at some of the metropolitan cities in India, namely Delhi, Kolkata, Hyderabad, Mumbai, Chennai, and Bengaluru, are considered for demonstration. During the baseline period, the correspondence between observed and downscaled temperature is much better in case of TVDM as compared to BCSD and SDSM, particularly for Tmax. Next, the future temperature is downscaled for two Representative Concentration Pathways (RCPs), viz., RCP4.5 (medium–low-type) and RCP8.5 (high-emission) scenarios. The future period is divided into three epochs (30 years each) and results are compared with the baseline period. In general, the time-varying approach (TVDM) reveals more increase as compared to time-invariant approach (SDSM) during the future period. The drastic (~\u20093 °C) change in Tmin is noticed by the end of twenty-first century (epoch-3) during post summer months (June through January) as per TVDM. Overall, the time-varying (TVDM) approach shows more rise during the future period as compared to time-invariant approaches and the results from the former are more reliable, since its performance during baseline period is much better than the latter.