Presented by: Javed Akhter1 Contributors: Lalu Das 2and Monami Dutta2 How well do the CMIP5 Models simulate Summer Monsoon Rainfall over the Core Monsoon Region of India? Presented by: Javed Akhter1 Contributors: Lalu Das 2and Monami Dutta2 1Department of Physics, Jadavpur University Email :akhterexpressju@gmail.com 2Department of Agricultural Meteorology and Physics, Bidhan Chandra Krishi Viswavidyalaya

To compare the skills of different multi-model ensemble approaches. Objectives To investigate the performances of CMIP5 models in simulating spatio-temporal pattern of intra-seasonal monsoon precipitation over Core Monsoon Zone of India. To compare the skills of different multi-model ensemble approaches. To judge whether the bias correction can provides added values for model performances. Observational Data: IMD Gridded Data (0.25x0.25 degree) GCM Data : Historical simulations of 45 GCMs from CMIP5 archive. (http://pcmdi9.llnl.gov). Domain: 73-820 E, 18-280 N, within which the monsoon trough/ CTCZ normally fluctuates.[Mandke et al. (2007)] Time Period used: 1961-2005.

Thresholds for different indices Performances of GCMs to simulate Regional Averaged Precipitation Statistical Indices used for model evaluation: Agreement Indices- Nash-Sutcliffe efficiency (NSE), d-index Error Indices-ratio of RMSE to the standard deviation of the (RSR) and Percentage Bias(PB) For the error indices (RSR & PB),threshold is calculated as the 1st quartile of all the model values. For the agreement indices (NSE & d), threshold is calculated as the 3rd quartile of all the model values. Threshold for monsoon is calculated as the average of the threshold values of four months. Models crossing threshold values for each index gets score 1; otherwise zero. Thresholds for different indices NSE d RSR PB June -3.446 0.392 2.085 ±40.800 July -2.284 0.363 1.792 ±25.300 August -2.302 0.385 1.797 ±19.500 September -0.885 0.413 1.358 ±16.000 Monsoon -2.229 0.388 1.758 ±25.400 Better Perform-ance of Models

Performances of GCMs to simulate Regional Averaged Precipitation

Performances of GCMs to simulate Regional Averaged Precipitation

Performances of GCMs to simulate Regional Averaged Precipitation

Performances of GCMs to simulate Regional Averaged Precipitation

Performances of GCMs to simulate Regional Averaged Precipitation Model June July August September Overall ACCESS1-0 1 ACCESS1.3 bcc-csm1-1-m 2 bcc-csm1-1 3 BNU-ESM 4 15 CanCM4 CanESM2 CCSM4 14 CESM1-BGC 13 CESM1-CAM5 9 CMCC-CESM 8 CMCC-CM CMCC-CMS CNRM-CM5-2 CNRM-CM5 CSIRO-Mk3-6-0 EC-EARTH 10 FGOALS_g2 FIO-ESM 5 GFDL-CM2p1 11 GFDL-CM3 12 GFDL-ESM2G 6 GFDL-ESM2M Model performance Scores:

Performances of GCMs to simulate Regional Averaged Precipitation Model June July August September Overall GISS-E2-H-CC GISS-E2-H GISS-E2-R-CC GISS-E2-R HadCM3 1 3 4 HadGEM2-AO HadGEM2-CC HadGEM2-ES inmcm4 8 IPSL-CM5A-LR 2 IPSL-CM5A-MR IPSL-CM5B-LR MIROC-ESM-CHEM 10 MIROC-ESM 11 MIROC4h 7 MIROC5 5 MPI-ESM-LR MPI-ESM-MR MRI-CGCM3 MRI-ESM1 NorESM1-M NorESM1-ME Model performance Scores:

Multi-Model Ensemble: Multi-model ensembles constructed in 3 different ways: i)Taking mean of all models(MM-all) ii) Taking mean of 7 better models(MM7) iii)Using multiple regression of 7 better models(MM7-reg) JUNE NSE d RSR PB MM-all -1.88 0.44 1.68 -45.3 MM7 -0.36 0.56 1.15 -23.1 MM7-reg -0.22 0.35 1.09 0.5 Best Model(bcc-csm1-1) -1.14 0.53 1.45 -28.2 JULY NSE d RSR PB MM-all -4.29 0.37 2.27 -44.5 MM7 0.08 0.41 0.95 -1.8 MM7-reg -0.41 0.11 1.17 0.5 Best Model(CESM1-BGC) -0.44 0.53 1.19 1.2

Multi-Model Ensemble: AUGUST NSE d RSR PB MM-all -4.55 0.39 2.33 -39.8 MM7 -0.12 1.04 -1.9 MM7-reg -0.33 0.23 1.14 -0.2 Best Model(CESM1-BGC) -0.58 0.56 1.24 6.8 SEPTEMBER NSE d RSR PB MM-all -0.53 0.43 1.22 -27.4 MM7 0.004 0.30 0.99 0.0 MM7-reg -0.15 0.27 1.06 1.5 Best Model(CESM1-CAM5) -0.26 0.55 1.11 12.7

Effect of Bias Correction: Quantile mapping used -PTF(Parametric transformations), DIST(Distribution derived transformations), QUANT(Empirical Quantiles), RQUANT(Robust Empirical Quantiles) Bias corrected values verified using leave-one-out-cross-validation(LOOCV). June NSE d RSR PB INTPL -6.74 0.33 2.75 -80.50 QUANT -1.49 0.26 1.56 0.30 RQUANT -1.47 0.27 1.55 0.00 PTF -7.41 0.29 2.87 -49.40 DIST -1.63 1.60 -0.50 One example:HadCM3 July NSE d RSR PB INTPL -3.71 0.35 2.15 -30.00 QUANT -1.28 0.29 1.49 -0.30 RQUANT -1.22 0.30 1.47 -0.20 PTF -3.20 2.03 -9.60 DIST -1.31 0.32 1.50 0.10

Effect of Bias Correction: August NSE d RSR PB INTPL -4.17 0.38 2.25 -29.60 QUANT -1.24 0.40 1.48 0.10 RQUANT -1.21 1.47 0.20 PTF -3.28 0.37 2.05 -4.70 DIST -1.09 1.43 0.00 One example:HadCM3 September NSE d RSR PB INTPL -1.94 0.44 1.70 -46.70 QUANT -0.72 0.50 1.30 0.60 RQUANT -0.69 1.28 PTF -1.71 0.48 1.63 -14.80 DIST -0.67 0.49

Performances of GCMs to simulate Spatial Pattern of Precipitation

Performances of GCMs to simulate Spatial Pattern of Precipitation

Performances of GCMs to simulate Spatial Pattern of Precipitation June Observed Raw-GCM BC-GCM

Performances of GCMs to simulate Spatial Pattern of Precipitation July Observed Raw-GCM BC-GCM

Performances of GCMs to simulate Spatial Pattern of Precipitation August Observed Raw-GCM BC-GCM

Performances of GCMs to simulate Spatial Pattern of Precipitation September Observed Raw-GCM BC-GCM

Conclusions: BNU-ESM,CCSM4,CESM1-BGC,GFDL-CM3 are better compared to rest of the models. Models perform relatively better in September than other months. Multi-model ensemble using all models is not so skillful compared to ensemble comprising smaller subset of better models. Bias correction provides additional improvement to the model performances.

References: Taylor KE (2001) Summarizing multiple aspects of model performance in single diagram. J. Geophys. Res., 106, 7183–7192 Rajeevan M, Bhate J, Kale J, and Lal B (2006) High resolution daily gridded rainfall data for the Indian region: analysis of break and active monsoon spells. Curr. Sci., 91, 296–306. Mandke S, Sahai AK, Shinde MA, Susmitha Joseph and Chattopadhyay R (2007) Simulated changes in active/break spells during the Indian summer monsoon due to enhanced CO2 concentrations: Assessment from selected coupled atmosphere–ocean global climate models; Int. J. Climatol. 27 ,837–859 Das L, Annan JD, Hargreaves JC, Emori S (2012) Improvements over three generations of climate model simulations for eastern India, Climate Research, Vol. 51: 201–216. Hua Chen, Chong-Yu Xu, Shenglian Guo (2012) Comparison and evaluation of multiple GCMs, statistical downscaling and hydrological models in the study of climate change impacts on runoff. Journal of Hydrology ,434-435, 36-45

Thank you