WP2 Empirical statistical downscaling and statistical analyses
WP Lead: Dr. Lalu Das, contribution from met.no.
Objectives:1) To identify suitable predictors for downscaling precipitation, collect the set of predictands, and prepare model output from GCMs. 2) To develop the statistical models, evaluate against independent data. 3) To construct high-resolution monsoon precipitation scenarios from existing CMIP3/CMIP5 GCMs outputs using empirical statistical downscaling (ESD) techniques over the Himalaya region of India. 4) Use state-of-the-art gridding tools to make maxs of projected rainfall statistics based on the ensemble of GCMs and downscaled pdfs. 5) Comparison between ESD-methods and output from one available RCM run (WP1).
We will first investigate the change in precipitation over the Himalaya region in the last century by analysing available precipitation data from the IMD using trend statistics (Brooks and Carruthers, 1953), Mann-Kendall test statistic (Sneyers, 1990) and Cramer’s tk-test (WMO, 1966). Next, high resolution past and future downscaled monsoon precipitation scenarios will be constructed using ESD following the methods of Benestad (2004, 2011) and Das and Lohar (2005). Common EOFs will be used as a reference system, matching GCM results with the reanalyses used to calibrate the downscaling models. It is planned to analyse observed IMD data, ERAINT and ERA40 for developing downscaled models and cross validation methods will be applied to validate the models. The models’ ability to capture past trends will be assessed. The performance of ERAINT is especially important for the hydrological analyses in WP1. Once the downscaled models are validated for study regions, they will be used to develop downscaled scenarios for all available IPCC GCMs (CMIP3/5) outputs. These future downscaled scenarios will be compared with the selected RCM-projection in WP1. Computation of extreme precipitation, based on new ESD-based techniques being developed in collaboration with National Center for Atmospheric Research (NCAR) in USA, will also be compared with the extreme values of daily runoff of hydrological scenarios from WP1.
Deliverables: 1) The model set-up on Indian stations is tested and a report on the evaluation of the ESD-method. 2) Report describing the downscaled results, with evaluation of the results from different GCMs. 3) Compile a description of the 5th, 50th and 95th percentiles of the downscaled results. 4) Report describing the comparison of EDS and RCM output.