es-doc for cmip6 Bryan Lawrence, Sebastien Denvil, Mark Greenslade, Eric Guilyardi, David Hassell, Sylvia Murphy, Charlotte Pascoe, Allyn Treshansky, the es-doc consortium; IS-ENES2 and NOAA partners, 17 Feb, 2016.
From GCMs to river flow in the midst of uncertainties: Can potential future plights could be alleviated with currently available forecasting skill ? The northwest part of India (NWI) known as the “wheat bowl” of the country, receives 20% - 25% of its annual precipitation during winter season. This precipitation is very important for the wheat crop, as it supplements the moisture and maintains low temperature during the reproductive stages. Most of the winter precipitation in the region is in the form of snowfall over western Himalayas. This precipitation, in turn, helps in maintaining the glaciers, which serve as the vast storehouse of freshwater supply to millions of people downstream throughout the year through rivers of western Himalayan origin. Therefore, for a country like India that gets more than 70% of its wheat production and fresh water from NWI region, the question arises whether strategies of winter-time precipitation prediction that have proved useful elsewhere can they be adapted to the exceptionally complex terrain of Himalayas as well? The aim of the present study is in two-folds. Firstly, it attempts to assess the seasonal predictive skill of six general circulation models (GCMs) (out of which four of them are from IRI Columbia University, one form NCEP and one from NCMRWF India at 1-month lead), for a period of 31 years (1982-2012). Secondly, an attempt has been also made to reproduce the information of the GCMs at higher resolution using downscaling approaches (both dynamical and statistical downscaling). The first part of the study reveals that the GCMs in general underestimate the observed climatology and inter-annual variability of precipitation and temperature and their skill is not satisfactory even with multimodel ensemble techniques. The second part of the work shows that dynamical downscaling (at 30 km resolution) with customized RegCM significantly reduces GCM biases. Further a comparison between Canonical Correlation Analysis (CCA) based statistical downscaling and Quintile Mapping (QM) based bias corrected dynamical downscaling has been made. The results suggest that the QM based bias corrected dynamical downscaling further improves the skill over domain of interest compared to CCA based statistical downscaling approach. Furthermore, the study reflects the model’s robustness at the event scale and paves a path for using dynamic downscaling methods in basin-scale studies. Finally, the plausible reasons of model failure and how DCMIP framework has played a key role in addressing such issues is highlighted.