abhi8893 / Global-vs-Regional-Climate-Models

Contains my codes for the submitted poster at the National Space Science Symposium 2019

Potential advantages of using a dynamically downscaled higher resolution regional scale model PRECIS over India

Date: May 2019 DOI: 10.13140/RG.2.2.20933.70886 Conference: The National Space Science Symposium (NSSS) 2019 Project: Preparing for the Future: designing climate products and models for India (PREPARE) Authors: Abhishek Bhatia, Santosh Kumar Muriki, Saurabh Bhardwaj

Click here for the pdf file of the poster.

Background

Downscaling is a method for obtaining high resolution climate information from relatively coarse-resolution global climate models (GCMs). RCM takes account of much smaller scale terrain effects than GCMs due to better resolving of coastlines, which is particularly important for simulating orographic precipitation. It is important to assess the performance of RCMs over its driving GCM to explore the efficacy of dynamical downscaling. Such dynamically downscaled information is imperative in assessing regional climate risks which assists in much better planning for climate impacts at that scale. We present the potential improvements of PRECIS over its parent model HADGEM2-ES in simulating regional precipitation.

Data & Method:

• The PRECIS model is run over the Indian domain at a spatial resolution of 0.22° using the large-scale information from the parent GCM HADGEM2-ES (1.875° x 1.25°). The historical simulation of precipitation (1971-2000) is validated using IMD observation gridded dataset available at 0.25° resolution. (Pai et al. 2013).
• The performance of the two models are compared over homogeneous regions (India) by analyzing the patterns of JJAS mean precipitation, inter-annual variability, extreme precipitation, and active and break spells of ISMR.
• Summary statistics of errors for the historical simulation are presented for the parameters mean, max and min temperature validated using IMD temperature observation gridded dataset available at 1° resolution. (Srivastava et al. 2009)
• Core Monsoon region: 18°N – 28°N, 73°E – 82°E (Rajeevan et.al., 2010)
• Definition of Active/break spells over CMR: (Abhilash et. al. 2014) Active Spell: Area averaged rainfall > 40% of long term mean Break Spell: Area averaged rainfall < 40% of long term mean Normal Spell: Area averaged rainfall between -40% to 40%

Concluding Remarks:

• Although both the models show a significant dry bias in monsoonal mean precipitation over all the regions, PRECIS shows a smaller RMSE error in almost all regions except NE and NW and a better spatial pattern of inter-annual variability across all regions.
• Extreme precipitation (90th and 95th percentile) over WC and CMR is much better represented in PRECIS.
• PRECIS performs much better in simulating the wet days in the CMR, IP, WC and EC regions, and has a lower presence of the drizzle effect which is inherent in the CMIP5 GCMs.
• The active spells of the ISMR are much better represented in PRECIS in terms of the intensity (PDFs), and spatial correlations of the composite. The break spells are too dry in both the models, however PRECIS has a better spatial pattern over the CMR.
• Lower errors (RMSE) are noted across almost all seasons and regions for mean and max temperature, however the errors are significantly worse for min temperature.