Technical Assistance Consultant’s Report

Project Number: TA 8166 December 2013

India: Climate Adaptation through Sub-Basin Development Investment Program

Cauvery Delta Zone: Climate Data and Future Scenarios —Final Report

This consultant’s report does not necessarily reflect the views of ADB or the Government concerned, and ADB and the Government cannot be held liable for its contents. All the views expressed herein may not be incorporated into the proposed project’s design.

IND (44429): Climate Adaptation through Sub-Basin Development Investment Program Cauvery Delta ,

Final Report

Cauvery Delta Zone: Climate Data and Future Scenarios

1 Table of Contents

1. INTRODUCTION 8

1.1. PROJECT BACKGROUND 8 1.2. GEOGRAPHICAL CONTEXT 8 1.3. CLIMATOLOGY 8 1.4. REVIEW OF EARLIER STUDIES 9 1.5. ORGANIZATION OF THIS REPORT 11

2. APPROACH AND METHODOLOGY 12

2.1. STRATEGY 12 2.2. SPATIAL DOMAIN 12 2.3. HISTORICAL CLIMATE DATA 13 STATION DATA 13 GRIDDED IMD DATA 14 APHRODITE GRIDDED PRECIPITATION DATA 14 2.4. CLIMATE MODELS USED 14 CMIP5 GCMS 14 ASSESSMENT OF MONSOON SIMULATIONS 16 SELECTION OF GCMS 16 REGIONAL CLIMATE MODEL (RCM) 17 2.5. DELIVERED DATA SETS 17 DATA EXTRACTION 18 BIAS CORRECTION 19 2.6. TROPICAL CYCLONES ANALYSIS 19

3. BASELINE CLIMATE AND EVALUATION 20

3.1. OBSERVED CLIMATOLOGY 20 SEASONALITY 20 TRENDS 22 RAINFALL 26 3.2. BASIC STATISTICS 26 3.3. EVALUATION OF GCM AND RCM 31 TEMPERATURE SEASONAL CYCLE 31 RAINFALL SEASONAL CYCLE EVALUATION 33 RAINFALL DAILY CLIMATOLOGY 35 SPATIAL PATTERNS IN SUMMER MAXIMUM TEMPERATURES 37 SPATIAL PATTERNS IN RAINFALL 38 3.4. BIAS CORRECTION 42

4. PROJECTIONS 43

4.1. CLIMATE CHANGE SCENARIO 43

2 SEASONAL CYCLE 43 DAILY RAINFALL PROJECTIONS 47 4.2. BASIC STATISTICS 49 4.4. SPATIAL CHANGES 51

5. DAILY RAINFALL ANALYSIS 55

5.1. RETURN PERIOD ANALYSIS FOR BASELINE 55 5.2. RETURN PERIODS FOR PROJECTED RAINFALL DATA 56 5.3. PROBABILITY OF RAINFALL 57 BASELINE 57 FUTURE PROJECTIONS 58

6. TROPICAL CYCLONES 60

6.1. CURRENT TRENDS AND VARIATIONS 60 6.2. TROPICAL CYCLONES AND POSSIBLE INFLUENCE OF CLIMATE CHANGE 62

7. CONCLUSIONS 64

8. REFERENCES 65

ANNEX I 67

STATION OBSERVATIONS 67 GRIDDED DATASETS 69 REANALYSIS DATASETS 69 CYCLONIC STORMS TRACKS DATA 69 LIST OF IMD DISTRICT RAINFALL MONIOTRING SCHEME (DRMS) STATIONS 71 SUMMARY OF DATA PROVIDED BY PWD, TAMIL NADU 76

ANNEX II 79

INVENTORY OF DAILY DATA FROM CIMIP5 GCMS 79

ANNEX III 81

DESCRIPTION OF DATA EXTRACTED FOR CDZ 81 DATA STRUCTURE OF DISTRIBUTED CLIMATE DATASETS 82 DATA SETS PRODUCED FOR CDZ 83 DATASETS FOR STATION LOCATIONS 84 DERIVED DATA 85 FILE STRUCTURE 86

3 LIST OF ACRONYMS

ADB Asian Development Bank APHRODITE Asian Precipitation—Highly-Resolved Observational Data Integration Towards Evaluation CDZ Cauvery Delta Zone GCM General Circulation Model (also Global Climate Model) GHGs Green House Gases msl Mean sea level INCCA Indian Network for Climate Change Assessment IPCC Inter -governmental Panel on Climate Change IPRC International Pacific Research Center (IPRC), University of Hawaii IPRC-RegCM IPRC Regional Climate Model IWRM Integrated Water Resources Management LLNL Lawrence Livermore National Laboratory, California, USA RCM Regional Climate Model PCMDI Program for Climate Model Diagnosis and Intercomparison PPTA Project Preparatory Technical Assistance TN Tamil Nadu

4 EXECUTIVE SUMMARY The project was aimed to prepare an assessment of current climate and future climate change over the Cauvery delta of Tamil Nadu, and to provide related data, analysis and interpretations, based on the latest science. This was mainly intended to support hydrologic analysis, particularly to build climate resilience in future designs for drainage improvement, flood control and irrigation structures being planned under the over-arching Climate Adaptation through Sub- basin Development investment Program (CASDP) of the Asian Development Bank (ADB). Impacts of future climate change are expected to be more pronounced in areas that are already vulnerable due higher population densities and exposure to natural hazards. In such contexts, effective use of climate information in planning strategies are of greater relevance now, than ever before. The present project effort brought together a variety of observational data sets, including long- term site-specific climate data to enable better characterization of local climate. Climate change scenarios up to the 2050s have been generated and analyzed from state-of-the-art GCMs and downscaled using a Regional Climate Model (RCM) to higher resolutions appropriate for water sector adaptation strategies. The importance of the Cauvery River and the Cauvery Delta Zone (CDZ) to the culture and livelihoods of the people of Tamil Nadu cannot be overstated. The tropical climate of the Tamil Nadu region is characterized around seasonal rainfall contributed by both South West (SW) and the North East (NE) monsoons influencing the Indian sub-continent. Set within this larger climatological context, the deltaic region of River Cauvery comes mainly under the influence of NE monsoon. Cyclonic disturbances, while providing important additional water resource to the CDZ in their benign forms, can cause substantial damage to life and property when they reach severe intensities. Although the mean annual temperatures over the Cauvery delta area is around 30 C, summer peaks can go up to 43 C with consequences to both water demand and evaporative losses. The task of the climate component of the project involved initial scoping of available data from the point of view of requirements of the water sector groups. An inventory of observed climate data sets for the CDZ was prepared from this initial scoping and some details are included in the this report. Observed climate data for project location available from the meteorological stations of the India Meteorological Department (IMD), data from available Government of Tamil Nadu Public Works Department (PWD), and gridded data sets of both IMD and other global sources were inventoried and assessed. A sub-set of these observed data sets were used to characterize the baseline climatology of the project location. General Circulation Models (GCMs) are the main tools available to project future climate and its change in response to various greenhouse gas emission scenarios. GCMs, projections are however still too coarse (~200 km) for use in sub-basin level climate change assessments. This study has therefore used Regional Climate Model (RCM) - downscaled high-resolution (~30 km) scenarios generated from coarse resolution GCMs projections, in addition to GCMs results. Results available from current genre of CMIP5 GCMs (Coupled Model Inter-comparison Project phase 5) models and dynamically downscaled RCM results forced with CMIP5 and CMIP3 lateral boundary conditions were used for preparing baseline and future climate change scenarios for the study area.

5 Both the GCMs, and the RCMs capture the seasonal evolution in temperatures but with a cold bias in certain cases pertaining to the GCMs and warm bias in case of the CCSM4 driven RCM. Among the model results compared temperature biases are of the order of about 2-3 C for maximum temperatures. Overall the models are able to closely capture temperature variations during the season as well as spatial patterns. The temperature evaluation indicates that there is more confidence in temperature projections, which is probably higher during the drier months of the year. Simulation of rainfall accurately is difficult for climate models, as rainfall is an end product of many inter-linked processes in the earth’s climate system. This gets even more challenging when projections are required over a small area like the CDZ because rainfall’s natural variability increases as we go from larger to smaller domains. However, It is encouraging to note that the GCMs and RCMs used in this assessment pick up the seasonal cycle of rainfall over the CDZ quite well. Significant biases do exist, which in one of the regional model results has been corrected using a bias correction procedure. Model results compared over a daily time scale showed quite a good similarity with respect to observed rainfall. Due to higher variability of rainfall there is lesser confidence in rainfall estimates and projections. Analysis of temperature observations from stations in the CDZ area shows an increasing trend in both maximum and minimum temperatures. The annual mean maximum temperatures are increasing at a rate of about 0.13 to 0.33 C/10 years. Both maximum and minimum temperature trends are most prominent during the cooler months of January and February. Historical observations of rainfall from both gridded and station observations show very little trends with predominance of year-to-year variability. The climate change scenarios for CDZ have been prepared for time horizons up to 2050 keeping in view the planning needs of water sector projects. Particular focus of our analysis was on temperature and rainfall variables as they are the key variables for the hydrological analysis and have also been extensively evaluated with relatively higher reliability as compared to other variables. However, data sets have been provided for other variables like humidity, wind speed, solar radiation, evaporation and mean sea-level pressure. Maximum temperature change over the CDZ projected by the models show a range from about 1.0 to 1.5°C by the 2050s. Minimum temperatures show a larger increase with changes ranging from 2-3°C. Spatially, the temperatures changes show a large variation over the TN region with the range going from 2-6°C for maximum temperatures in the 2050s, with the RCM driven by GFDL showing higher temperature increase. These results are consistent with the earlier climate model projection studies undertaken by other research groups. These rising temperatures will have consequences to the water sector, in terms of higher water demand and enhanced evaporative losses. Higher mean temperatures may also translate into longer spells of heat waves in summers. In the projected temperature changes over the CDZ, the differentiation between high and medium emission scenarios is not seen clearly in changes up to 2050s. That means different emission scenarios do not impact temperatures differently over time-horizons until the 2050s considered for this study. The projected rainfall changes over the region vary, but most of the model results show increase in rainfall during the rainy season months. Seeing the lack of clear agreement among the various models used, it seems that this conclusion may not be as robust as the temperature

6 increase. Further analyses of the rainfall return-periods and assured rainfall amounts also indicate increase in rainfall, and extreme rainfall amounts. From the climate resilience perspective it could be interpreted as planning for a slightly longer return period design standard than the one being presently used. Although the rainfall projections are not clearly indicative of clear increase in magnitude, there seems to be a shift towards more a more variable behavior in daily rainfall amounts. Addressing such changes would perhaps require better management strategies that can dynamically adjust to a wider range of climate situations. Enhanced use of climate information in short-term management practices is soft option that me require consideration. In view of the range of bias in the model results, it would be perhaps better to use the delta method, which involves estimation of model projected changes with respect to their respective baselines. These changes can be imposed on observed climate data series at any required location to produce a future projection of a model scenario. Future variations in the number of storms show year-to-year variations, but the total number of detected storms for the entire period of simulation remains the same in both the base line and future scenario integrations both cases. In other words, no measurable change in the total number of storms is projected.

In summary, the climate change projections for the CDZ indicate a warmer regime with possibility of a wetter northeast monsoon season that is variable with higher rainfall extremes. Considering the uncertainties in future rainfall changes, it would be advisable to consider a mix of measures that include institutional capacities in adaptive management. One of the key aspects for better management will be information system that is supported by robust monitoring networks, good data archival and interface to enable decision-making.

7 1. Introduction

1.1. Project background 1. The project aims to enhance resilience of communities to climate change in the Cauvery delta of Tamil Nadu, India. Outcome of the project will be improved integrated water resources, flood and coastal management in the delta area. Reliable information of current climate variability and future climate change scenarios, along with their respective range of uncertainties and confidence levels are required to support the hydrologic analysis for sub-basin Integrated Water Resources Management (IWRM). 2. This climate component of this Project Preparatory Technical Assistance (PPTA) would aim to provide assessments of current climate and future climate change over the Cauvery delta of Tamil Nadu, and to provide related data, analysis and interpretations, based on the latest science to support the hydrologic analysis and design for drainage improvement, flood control and irrigation structures to be planned under this project. 3. The work has been carried out in fulfillment of the formal contract under the framework of the ADB UNESCO-IHE Knowledge Partnership initiated on June 11, 2012. The work was originally scheduled to be completed by November 2012, but since then has been accorded a cost neutral time extension until end of June 2013 in view of delays in accusation of data. 4. Climate change scenarios based on “climate models that demonstrate skill in simulating current climate and high-resolution regional model simulations forced by these climate models” are being prepared along with limitations and uncertainties associated with these projections. These data and results will be made available in a popular format, along with explanatory notes, for further use, particularly to the PPTA team commencing further work on this ADB project.

1.2. Geographical context 5. Cauvery is one of the largest Rivers of southern India, flowing from northwest to southeast draining a basin of about 81,155 sq. km straddling the States of Karnataka and Tamil Nadu. The Cauvery Delta zone lies between 10.0 N to 11.30 N Latitude and 78.15 E to 79.45 E longitude at the end of the Cauvery river basin in the eastern part of Tamil Nadu. With an area of about 14,470 sq. km comprising of 28 revenue taluks falling within four districts of Nagappatinam, and Thrivarur and parts of the district Trichy (5 taluks), (2 taluks) and Puddukkottai (one thaluk), CDZ represents an equivalent 11% of the area of Tamil Nadu State (ADB, 2011). 6. The terrain of the Cauvery basin is composed of ridge and valley topography on the western and central parts with plateaus in between and undulating terrain, rolling plains, fluvial plains, delta plains and coastal plains on the east. The maximum elevation of about 2,637 m is in Nilgri hills. The Karnataka plateau has an average elevation of about 900 m above mean sea level (msl), while the eastern Tamil Nadu plains including the CDZ lie below 300 m (Gopalakrishnan and Rao, 1986).

1.3. Climatology 7. Rainfall is the most important climatological resource for the Tamil Nadu state, and is contributed by both South West (SW) and the North East (NE) monsoon. The

8 typically tropical climate of the region is characterized around these seasonal current of moist winds from over the adjoining seas. Seasonal distribution of temperature in this region is basically a modulation of sub-equatorial temperatures by the monsoonal currents and local topography. Due to its proximity to the , the state experiences frequent tropical cyclones, some of which landfall with fiery intensity causing substantial damages to life and property, particularly to the CDZ coast, due to both high speed winds and storm surge. 8. Set within this larger climatological context, the deltaic region of River Cauvery is mainly under the influence of NE monsoon, whereas the river basin in the upper reaches is controlled mostly by SW monsoon (June to September) in its prominent source areas of Western Ghats. Rainfall in the upstream catchments, principally due to the SW monsoon, fills the reservoir on the Cauvery Basin enabling paddy cultivation outside the main rainy season of the catchment, which is during the October to November NE monsoon season. 9. The surge of water from the tributaries to the main course of Cauvery during SW monsoon brings a lot of sediment into the deltaic region from the uplands. The tributaries are often dry during the rest of the year. The region experiences a semi- arid tropical climate with mean annual temperature of 25°C and the maximum summer (March to May) temperature reaches occasionally up to 43°C. A number of dams constructed across the river in the recent past have modified the water discharge and sediment accumulation rates in the deltaic region. An average sediment accumulation rate between 0.4 and 4 mm/yr for the recent past has been reported in the Cauvery River basin with less sedimentation rate in the tributaries (Ramanathan et al., 1996).

1.4. Review of earlier studies 10. The climatological analysis of two representative stations Tiruchchirapalli (Trichy) (western region of the delta) and Nagappattinam (coastal eastern end of the delta) based on FAO CROPWAT have been carried out earlier as reported in ADB, 2011. This analysis shows mean daily temperatures at Trichy to range from a low of about 25C in December to a high of about 32C in May. April to June is the hottest period with mean daily maximums close to 31C. The mean daily temperature range is typically about 8C. At Nagappattinam mean daily temperatures range from a low of about 25C in December to a high of about 31.4C in May. May and June are the hottest months with mean daily maximums close to 31C. The mean daily temperature range is typically about 8C. The slightly lesser temperature maxima at Nagappattinam is due to its proximity to the coast where the land-sea breeze bring down the daytime maximum temperatures. Similarity in mean temperatures observed at the two stations indicates that Tamil Nadu Cauvery Delta Zone (TNCDZ) is a homogenous zone in terms of its temperature climate. 11. Trichy, located away from the coast records mean annual rainfall of 902 mm, while at Nagappattinam the annual rainfall is 1421 mm. Figure 1-1 gives the seasonal rainfall variation for the whole CDZ illustrating the stronger influence of the Northeast monsoon over the Cauvery Delta.

9

Figure 1-1 Mean monthly rainfall (mm) over the Cauvery Delta Zone (CDZ) (Source: present study team based on climate normals of IMD stations) 12. Analysis of spatial variation of mean annual rainfall over the Cauvery Basin based on the India Meteorological Department (IMD) gridded 0.5×0.5 data that covers the 35 year period 1971 to 2005 has also been presented in ADB, 2011. There is a significant variation of annual mean rainfall across the sub-basin, with about 700 mm in the interior western area to more than 1200 mm in the eastern region nearer to the coast (Figure 1-2).

Figure 1-2 Mean annual rainfall over the Cauvery river basin based on IMD gridded rainfall data (Source: ADB, 2011, based on IMD gridded rainfall data set)

10 13. Annamalai et al. (2011) examined future climate change scenarios for the Cauvery river basin using dynamical downscaling method. More specific scenarios studies for the CDZ were reported by Geethalakshmi et al. (2011) in their study on adaptation strategies to sustain rice production within the Indian Network for Climate Change Assessment (INCCA) initiative of the Ministry of Environment and Forests, Government of India. Both these studies were supported originally by the ClimaRice project funded by the Ministry of Foreign Affairs, Norway and the Royal Norwegian Embassy, New Delhi. Besides these, the ADB (2011) also examined downscaled results for the Cauvery river basin for mid-century (2021-50) and end-century (2070- 2100) for SRES scenario A1B; and A2 and B2 scenarios for the end-century. 14. The predominant signal for the CDZ from these studies is that maximum and minimum temperatures show increasing trends that are a basin-wide, while rainfall during the NE monsoon shows increasing trend.

1.5. Organization of this report 15. After a brief introduction, the report describes the methodology and approach used for preparing climate change projections for the CDZ. Different observational and model data sets used for the current project are also given. The next chapter details the baseline climate of CDZ derived from the different observational data sets and climate models, both GCMs and RCMs. This section also presents evaluation of the model baselines with respect to the observations. Chapter 4 on projections describes the different future scenarios based on CIMIP5 GCMs and RCM results. The concluding chapter provides a summary of the work and uncertainties in scenario projections, and further efforts required.

11 2. Approach and Methodology

2.1. Strategy 16. As suggested in the ToRs, future climate change scenarios were generated based on a two-tier approach using GCMs and RCMs: i. Scenarios of temperature and rainfall changes from two CMIP 5 a select set of GCMs. Selection of these two GCMs was based on their performance of simulating the current climatological features of the monsoons over the Cauvery delta. A limited subset of other climate variables like wind speed, relative humidity, radiation etc are also provided as per their availability at daily time scales. ii. Regional Climate model runs for available emission scenarios (IPRC-RegCM) and representative time horizons based on CMIP 3 boundary forcing were analyzed to provide regionally downscaled scenarios for the CDZ. 17. Figure Figure 2-1 below outlines the strategy adopted to prepare climate change scenarios for CDZ.

Figure 2-1 Schematic illustrating the climate scenario data generation strategy.

2.2. Spatial Domain 18. Two nested domains (Figure 2-2) – one larger, covering the whole Tamil Nadu (TN) state (8.0 – 13.5oN/76.0 – 80.0oE) and another smaller, covering the core Cauvery Delta Zone (CDZ) (10.0 – 11.5oN/78.00 – 80.0oE) were demarcated for this study. The larger domain will enable analyses to ensure climatological consistency and extraction of gridded data sets, to be used as required in water resources impacts studies. All the gridded datasets are compiled for the bigger box (TN) that includes

12 some areas from adjoining states besides whole of TN. The smaller box covering the CDZ (all delta districts) is treated as a single box to create time-series for some of the model evaluations.

Figure 2-2 Spatial domain of the study area with Rainfall station locations (red dots), Tamil Nadu (TN – green box) and Cauvery Delta Zone (CDZ – red box)

2.3. Historical Climate Data

Station data 19. Climate data relevant to the project area are available from meteorological and hydrological stations located in the Cauvery Basin. These observational networks are being maintained by different agencies like the India Meteorological Department (IMD), Central Water Commission, Tamil Nadu State Irrigation Department, Public Works Department (PWD), Department of Agriculture, Indian Space Research Organization and other private sector agencies. However, many of these observation stations were set up during different times and do not have long series of data (> 30 years) that are suitable for climatological trend analysis. 20. Annex I gives a more detailed inventory of observed data sets available. 21. Daily data from 16 IMD observation stations in the Cauvery basin for the period 1971-2000 were analyzed. After quality checks, 7 stations, falling in the vicinity of the CDZ in the TN part of the Cauvery basin having more than 90% data were selected (Figure 2-2). The list of these stations along with their percentage of missing records during the 1971-2003 period is given in Table 2-1.

13 Table 2-1 IMD climate stations in the CDZ area used

S. No. Station Name Data missing (%) 1. Adiramapatinam 3.6 2. Cuddalore 0.5 3. Metturdam 6.3 4. 6.6 5. Salem 3.3 6. Trichy 3.0 7. Vedaranyam 9.8

Gridded IMD data 22. IMD gridded data sets for both rainfall (0.5 degree resolution; Rajeevan et al., 2009) and temperature (1 degree resolution) have been obtained and data extracted for the TN domain and compiled for the CDZ for 1971-2000 period to match the baseline period for which that IMD station data were available. Gridded data represent spatial averages and are important for making comparisons with climate model results.

APHRODITE gridded precipitation data 23. APRODITE1 gridded rainfall dataset available for the Asian region at (0.25 degree resolution) for the period 1971-2000 is the second rainfall climatology dataset used to compile baseline rainfall climate for CDZ.

2.4. Climate models used 24. GCM simulations available from genre of CMIP5 (Coupled Model Inter-comparison Project phase 5) models and dynamically downscaled RCM results forced with CMIP5 and CMIP3 lateral boundary conditions have been used for preparing baseline and future climate change scenarios for the study area.

CMIP5 GCMs 25. Daily data for MIROC5 GCM & CCSM4 for each of these GCMs climate variables have been downloaded from PCMDI, LLNL CIMIP5 site. Variables included rainfall and temperature (maximum and minimum), mean sea-level pressure, wind-speed and direction, surface incoming solar radiation and relative humidity. Annex III gives all the details of the data availability. 26. Modelling groups from the world over have contributed data from more than one model versions to CMIP3 and CMIP5 initiatives. Table 2-2 gives a list of the recent CMIP5 models with daily data sets available. These simulations include the modelling group’s best estimates of natural (e.g., solar irradiance, volcanic aerosols) and anthropogenic (e.g., greenhouse gases, sulfate aerosols, ozone) climate forcing during the simulation period. In contrast to CMIP3, both direct and indirect effects of

1 The APHRODITE (Asian Precipitation—Highly-Resolved Observational Data Integration Towards Evaluation of Water Resources) daily gridded rainfall data set for Asia developed by Japanese research groups (Yatagai et al., 2012) is a result of the APHRODITE project supported by the Water Resources initiative of the Environment Research & Technology Development Fund, Ministry of the Environment, Japan. This dataset covers a period of more than 57 (1951–2007) years was created by collecting and analyzing rain-gauge observation from 5000 to 12,000 stations across Asia.

14 aerosols are represented in CMIP5. The availability of these large integrations provide a unique opportunity to assess the models’ skill in simulating aspects of the monsoons and then select few models that demonstrate high skill for further use in preparing climate change scenarios. Table 2-2 CMIP 5: List of Modelling Groups (with daily datasets)2

Modelling Centre Institute ID Model Name

1. Commonwealth Scientific and Industrial Research ACCESS1.0 Organization (CSIRO) and Bureau of Meteorology CSIRO-BOM

(BOM), Australia 2. Beijing Climate Center, China Meteorological BCC-CSM1.1 BCC Administration CanESM2 3. Canadian Centre for Climate Modelling and Analysis CCCMA CanCM4

4. National Center for Atmospheric Research NCAR CCSM4

CESM1(CAM5) 5. Community Earth System Model Contributors NSF-DOE-NCAR

6. Centre National de Recherches Meteorologiques / Centre European de Recherche et Formation CNRM-CERFACS CNRM-CM5 Avancees en Calcul Scientifique 7. Commonwealth Scientific and Industrial Research Organization in collaboration with Queensland Climate CSIRO-QCCCE CSIRO-Mk3.6.0 Change Centre of Excellence GFDL-CM3 8. NOAA Geophysical Fluid Dynamics Laboratory NOAA GFDL GFDL-ESM2G GFDL-ESM2M GISS- E2-H 9. NASA Goddard Institute for Space Studies NASA GISS GISS-E2-R 10. National Institute of Meteorological Research/Korea NIMR/KMA HadGEM2-AO Meteorological Administration HadCM3 11. Met Office Hadley Centre MOHC HadGEM2-CC HadGEM2-ES

12. Institute for Numerical Mathematics INM INM-CM4

IPSL-CM5A-LR 13. Institut Pierre-Simon Laplace IPSL IPSL-CM5A-MR 14. Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research MIROC-ESM MIROC Institute (The University of Tokyo), and National MIROC-ESM-CHEM Institute for Environmental Studies 15. Atmosphere and Ocean Research Institute (The MIROC4h University of Tokyo), National Institute for MIROC Environmental Studies, and Japan Agency for Marine- MIROC5 Earth Science and Technology MPI-ESM-LR 16. Max Planck Institute for Meteorology MPI-M MPI-ESM-P

2 As on April, 2013 accessed from PCMDI, LLNL, CMIP5 site.

15 17. Meteorological Research Institute MRI MRI-CGCM3

18. Norwegian Climate Centre NCC NorESM1-M

27. For all climate models, one of the most challenging aspects is the simulation of rainfall climatology during both summer (June-September) and winter (October- January) monsoon seasons over South Asia. A realistic simulation of the basic state of monsoon rainfall climatology is a key feature, for assessing the future changes due to anthropogenic forcing particularly over the Indian region.

Assessment of Monsoon simulations 28. The simulated rainfall has been validated against observed rainfall climatology for the period 1979-2005 constructed from Global Precipitation Climatology Project (GPCP). The metrics used are pattern correlation (PC) and root mean square error (RMSE). 29. Either over the broader Asian monsoon region (20S-40N, 40-180E) or over the core rainfall region (15S-30N; 50E-160E), only few models depict statistically significant values of PC and RMSE (Annamalai et al. 2007; Turner and Annamalai 2012: Sperber et al. 2012; Annamalai et al. 2012). Even in these few “best” models, a further assessment of the “spatial distribution of regional rainfall maxima” suggests limitations, particularly in the amplitude of rainfall (Annamalai et al. 2012). 30. Compared to CMIP3, higher horizontal resolutions employed in CMIP5 models capture the topographically induced rainfall over the Asian monsoon region. This is important since Cauvery basin is regarded as a “hot spot” with steep orography of the Western Ghats that receive copious rainfall. Similar results are obtained for the diagnostics performed for the winter (October-January) seasons. Note that a realistic simulation of summer monsoon rainfall is a necessary condition for models to capture the winter monsoon. Our analysis suggests that in models, the monsoon circulation is better captured than the rainfall, and the systematic errors in simulating the regional rainfall have not improved from CMIP3 to CMIP5 (Sperber et al. 2012).

Selection of GCMs 31. Based on the suite of metrics mentioned, from the pool of CMIP5 models, the latest versions of the Community Climate System Model (CCSM4), and MIROC5 from Japan were selected for this project. The simulations from these global models are performed at an unprecedented horizontal and vertical resolution, an added value for understanding and assessing “regional” changes in a warmer climate. 32. A new set of emission scenarios called the Representative Concentration Pathways (RCPs) have been introduced. A particular RCP represents radiative forcing reached by the year 2100, without being linked to any specific socioeconomic development storylines as in the case of the earlier SRES. 33. Annex III gives an inventory of daily climate variables available from the modelling groups that have contributed data to CMIP5 so far. Out of the identified CMIP5 GCMs based on their monsoon performance, only two (CCSM4 & MIROC5) have daily data available. These are the two CMIP5 GCMs identified to be used in this project. Baseline and future time horizons up to 2050s will be analysed for the

16 TNCDZ project area for two future (RCP) scenarios viz. RCP45 and RCP85, representing medium and high emission futures.

Regional Climate Model (RCM) 34. The International Pacific Research Center (IPRC) regional climate model (IPRC_RegCM) has been used to simulate the current and future climates over the project location. The regional model’s was set up with appropriate domain and resolution based on prior experimentation. It has been demonstrated to realistically simulate observed climate characteristics, particularly the seasonal rainfall features over the Tamil Nadu area and the Cauvery river basin when forced with lateral boundary conditions from the European Center for Medium-range Weather Forecasting (ECMWF) reanalysis data (Annamalai, 2011). It must be mentioned here that at the time of initiation of this current project, there were not many regional modeling results available for the South Asian region at below ~50 km spatial resolution, whereas such high resolution regional climate scenarios were required to capture the climate of the region influenced by steep topography of the eastern Ghats. 35. Our collaborating research partner IPRC, University of Hawaii, provided the regional modeling results. IPRC’s regional climate model was used for the dynamical downscaling study, and all the model data analyses were undertaken collaboratively with Dr. Annamalai, IPRC. 36. IPRC-RegCM used in this study has 28 vertical levels with high-resolution in the planetary boundary layer. The lowest model level is roughly 25 m above the surface. The model domain extends from 50ºS to 55ºN, 5ºE to 170ºE with a grid spacing of 0.25º (~30 km horizontal resolution), in both zonal and meridional directions. For RCM baseline two sets of data - runs downscaled from GCMs CCSM4 and GFDL_CM2.1 have been prepared. As CCSM4 GCM is from the current genre of CIMIP5 GCMs, it will enable comparisons with the parent GCM.

2.5. Delivered Data sets 37. The current climate was characterized using available observed data sets that included global, regional and national gridded data sets; and observations from individual stations. Model baseline climates were established using historical runs from comparable time periods from select CMIP 5 models and available RCM results. The select CMIP5 models were evaluated for their ability to simulate regional precipitation patterns and circulation features associated with both the summer and winter monsoons that influence the study area. 38. All relevant observed data sets and model results have been subjected to basic statistical analysis. Key features of the baseline (present-day climate from observed data sets and climate model twentieth century runs) and future climate change (as projected by selected GCMs and RCM simulations) for the period representing the 2050s - expressed in relevant statistics over the sub-basin were computed. 39. Table 2-3 below lists the climate data sets delivered by the project. Table 2-3 List of different baseline and climate model data sets produced**

Data Details Parameters Time slice Frequency

Quality Station observatory for Rainfall (mm/day), 1971 -2000 Daily

17 checked daily seven locations Maximum temperature climate data Adiramapattinam, (◦C), Minimum temperature from IMD Cuddalore, Mettur, (◦C), Stations in the Nagapattinam, Salem, CDZ area Trichy and Vedaranyam. IMD gridded Gridded data at 1x 1 Rainfall (mm/day), 1971 -2000 Daily & resolution for Maximum temperature Monthly temperature and 0.5 x (◦C), Minimum temperature 0.5 for rainfall fields (◦C), covering TN domain GCM MIROC5 Baseline Rainfall (mm/day), 1981 -2000 Daily Projection (RCP45) Maximum temperature 2006 -2100 Daily Projection (RCP85) (◦C), Minimum temperature 2006 -2100 Daily (◦C), Surface Downwelling Shortwave Radiation (rsds in wm-2), ), Sea Level Pressure (psl in Pa), Near- Surface Specific Humidity (huss in kg/kg), Eastward Near-Surface Wind (uas in ms-1), Northward Near- Surface Wind (vas in ms-1) GCM CCSM4 Baseline Rainfall (mm/day), 1981 -2000 Daily Projection (RCP45) Maximum temperature 2006 -2100 Daily Projection (RCP85) (◦C), Minimum temperature 2006 -2100 Daily (◦C Sea Level Pressure (psl in Pa), RCM GFDL Baseline Irradiance (MJ m-2), Min 1981 -2000 Daily Projection Temperature (oC), Max 2021 -2050 Daily Temperature (oC), Early Morning VP (kPa), Mean Wind Speed (m s-1) , Precipitation (mm d-1), Dew point temp ( oC), Relative Humidity (%), RCM CCSM4 Baseline Irradiance (MJ m-2), Min 1986 -2005 Daily Projection Temperature (oC), Max 2081 -2100 Daily Temperature (oC), Early Morning VP (kPa), Mean Wind Speed (m s-1) , Precipitation (mm d-1), Dew point temp ( oC), Relative Humidity (%), Bias corrected Baseline Rainfall (mm/day), 1981 -2000 Daily RCM GFDL Projection 2021 -2050 Daily **Blue text pertains to observed climate or model baseline data.

Data extraction 40. Three/four broad categories of data sets have been developed for use keeping in view a range of users with different levels of experience in using climate model data sets. 41. The originally available GCMs and RCMs are in netcdf (*.nc) gridded binary format, with control files (*.ctl) that describe the data for use in Grid Analysis and Display System (GrADS) a popular open source tool for display and analysis of earth

18 sciences data. Using GrADS package, data has been extracted for zone wise and station wise. Zonal daily time series data (baseline and projection of GCMs and RCMs) were extracted by averaging all the grids in the spatial extent of the zones Tamil Nadu (TN) zone (76E-80E & 8N-13.5N) and Cauvery Delta Zone (CDZ) (76E- 80E & 10N-11.5N). Station wise daily time series of climate scenarios data were generated only for RCMs (baseline and projection of 2 RCMs) by averaging nearest four grid points for each of the seven IMD station locations. These station locations can be taken as representative points for the CDZ. Details of the grids averaged and spatial extent are available in Annex III. All the extracted daily time series datasets are in csv text format to enable easy import to most GIS and hydrological/hydro- dynamical modeling software. 42. Estimates of rainfall amount associated with 1 in 2, 5, 10, 20, 50, 100 and 200 year events, respectively; estimates of rainfall amount at median, 50%, 80%, 90% and 95% confidence levels, respectively for the three seasons; estimates of rainfall depths for duration of hourly (if data available), daily, 2-days, 5-days and 10-days, respectively; 43. Datasets were organized under four major sections (four main folders) 1. RAW (grid wise GCM and RCM output in text format) 2. Zone wise (GCM and RCM outputs averaged for two major zones Tamil Nadu and CDZ) 3. Station wise (RCM outputs averaged for seven selected IMD station location) 4. Derived (Rainfall return period, Probability of rainfall occurrence and rainfall depths for seven IMD station location), and details of each of the sections are explained in detail in Annex III.

Bias correction 44. Despite all efforts to improve regional climate simulations, uncertainties in GCM simulations cascade when downscaling is performed. Therefore strictly speaking, any direct comparison with observations can at best provide us guidance of the model uncertainties that then helps us to assign confidence levels to the climate change projections, both in GCMs and RCMs. 45. Rainfall field from one of the RCM runs has been bias corrected using statistical techniques. These data have also been extracted and analyzed for basic and derived statistical parameters.

2.6. Tropical Cyclones Analysis 46. Considering the importance of Tropical cyclones to the CDZ, trends have been examined using historical records, published research and diagnostics based on long-term numerical re-analysis data, RCM baseline and projections. 47. A recent cyclone tracking technique was applied to observations, and coarse- resolution climate model and high-resolution regional climate model simulations. The analysis was performed during both the southwest and northeast monsoon seasons, and for baseline and future climate scenarios. The dynamical downscaling runs used were conducted by forcing the IPRC regional model (IPRC_RegCM) with lateral and boundary conditions taken from the coarse-resolution climate models (CMIP3/5) that “best” capture the current monsoon rainfall climatology and its variations (Annamalai, 2013).

19 3. Baseline climate and evaluation

3.1. Observed climatology

Seasonality 48. Climate normal of a variable is represented as an average, typically for a 30-year period. The seasonality can be expressed as the temporal evolution of this climate normal and illustrates the timing of the maximum and minimum during a year. It is an effective indicator of climate of a location, area or region. It is useful to therefore compare the seasonal cycles of temperatures and rainfall over the whole of the Tamil Nadu (TN) state as well as the particular project area, the Cauvery Delta Zone (CDZ). Comparing observed present day climate with model baseline also provides a good indicator of the model performance. Temperature

Figure 3-1 Seasonal Cycle of maximum and minimum temperature IMD stations and IMD gridded data set over TN (top panel) and CDZ (bottom panel).

49. The highest values of maximum and minimum temperatures are encountered during the summer months of April-May, while the lowest values are observed during December-January months over the whole region (Figure 3-1). There are, however, clear differences between the stations near the coast and the more inland stations. Maximum temperature observed in interior places (Trichy, Salem and Mettur) show higher values compared to coastal stations (Cuddalore, Adhiramapatinam and Vedaranyam). However, further inland towards the western region of Tamil Nadu,

20 lower temperatures are encountered because of elevated orography of the “Ghats” or mountainous tracts. Influence of the SW monsoon is seen in the maximum temperature series of IMD gridded as lowered temperature during the June to September months. This east-west temperature gradient in maximum temperature during spring months (MAM) is clearly brought out by the baseline temperature analysis using the IMD gridded data shown in Figure 3-2. The coarser resolution of the IMD data is however unable to clearly differentiate the slightly lower temperatures of the CDZ due to the coastal influence.

Figure 3-2 Distribution of maximum temperatures over Tamil Nadu during March-April-May season –based on IMD gridded data Rainfall

Seasonal cycle of Rainfall - TN & CDZ Zone Trichy 450 Nagapattinam 400 Adhiramapattinam 350 Vedaranyam 300 Salem 250 Mettur 200 150 Cuddalore Rainfall mm 100 APH_TN 50 APH_CDZ 0 IMD_TN Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec IMD_CDZ Months

Figure 3-3 Seasonal cycle of Rainfall – IMD gridded, Aphrodite gridded &IMD stations for TN and CDZ zones

21 50. The peculiarity of rainfall seasonal cycle Tamil Nadu is the competing effect of the Southwest (SW) and the Northeast (NE) monsoon along an east-west direction. Over CDZ, the dominating influence of NE monsoon is clearly seen as highest monthly normal rainfall amounts during the Oct-Nov-Dec months (Figure 3-3). Part of the higher rainfall amounts recorded during the months of November and December at Coastal stations like Nagapattinam, Vedaranyam, and Cuddalore may be attributable to cyclonic activity. Moving westwards towards the inland areas of the Cauvery basin there is greater influence of the SW monsoon, as apparent from the June-September monthly normal of rainfall. Both IMD and Aphrodite gridded rainfall data sets are able to pick up the increased rainfall over TN compared to CDZ during the SW monsoon season (June-September), similar to the observed station data. Figure 3-4 shows the spatial distribution of rainfall over the Tamil Nadu region during two rainy seasons viz. SW (JJAS) and NE (OND). The higher rainfall amounts over CDZ during NE, as compared to SW monsoon season is quite clearly brought out on a spatial scale.

Figure 3-4 Distribution of rainfall during summer monsoon (JJAS) and winter monsoon (OND) seasons based on IMD gridded data

Trends 51. Basic statistics of the important climate variables namely maximum and minimum temperature and rainfall have been computed for the baseline period (1971-2000) for all available observed data sets. In the following section the observed trends are presented. The trends and basic statistics are presented for the different seasons as outlined in the terms of reference for the project.

22

Temperature

52. Figure 3-5 shows the time-series of yearly averages of maximum temperature anomalies3 for the whole CDZ zone from the IMD gridded data. Similar time-series are also plotted for two representative stations - Trichy (inland location) and Cuddalore (coastal location). Lower panel of Figure 3-5 shows the anomaly time- series for annual average of minimum temperature. A warming tendency can be seen in almost all the temperature time-series. Despite the warming trend year-to- year variations are also quite prominent. Similar time-series plots have been presented in Figure 3-6 & Figure 3-7 for the spring (March-April-May, MAM) and northeast monsoon (OND) seasons. shows the temperature trends over the CDZ during different seasons based on both IMD gridded data and station observations. Except for Vedaranyam station all other stations exhibited positive temperature trends in maximum and minimum temperatures.

Figure 3-5 Year-to-year variations of maximum and minimum temperature anomalies over the CDZ

3Anomalies are deviations of the maximum temperature averages of each year from the long-term mean calculated based on a 30-year period (1971-2000).

23

Figure 3-6 Year-to-year variations of summer season (MAM) maximum and minimum temperature anomaly in CDZ

24

Figure 3-7 Year-to-year variations of maximum and minimum temperatures over CDZ during the NE monsoon season (OND).

Table 3-1 Observed Temperature trends over CDZ using IMD gridded data

Trend in °C/10 year Data Annual JF MAM JJAS OND TMAX IMD gridded 0.20 0.32 0.25 0.12 0.18 IMD stations Trichy 0.19 0.22 0.24 0.10 0.23 Cuddalore 0.33 0.59 0.39 0.13 0.38 Adhiramapatinam 0.16 0.23 0.18 0.10 0.16 Vedaranyam 0.13 0.13 0.14 -0.15 0.15 TMIN IMD gridded 0.11 0.25 0.11 0.07 0.07 IMD stations Trichy 0.12 0.42 0.09 0.03 0.06 Cuddalore 0.23 0.42 0.21 0.17 0.20 Adhiramapatinam 0.07 0.33 0.06 0.01 -0.02 Vedaranyam -0.13 -0.31 0.16 -0.24 -0.20

25 Rainfall 53. Figure 3-8 shows the variations in the observed rainfall over the CDZ during the historical baseline period. There are no clear trends discernable but during the northeast monsoon season (OND), more number of weak monsoon years (negative anomalies) is observed. This, combined with a warming tendency (Figure 3-7) implies hardship for water resources.

Figure 3-8 Inter-annual variations of precipitation (IMD gridded, APHRODITE gridded and IMD station data during summer monsoon (JJAS) and northeast monsoon seasons

3.2. Basic statistics 54. Table 3-2 & Table 3-3 give the basic statistics for baseline temperatures (maximum and minimum) as computed from the different observed historical and model historical runs for the baseline period 1971-2000. The shaded rows in the Tables are CDZ stations.

26

Table 3-2 Maximum Temperature statistics for the CDZ

Season JF MAM JJAS OND

Data sources Mean SD CV Mean SD CV Mean SD CV Mean SD CV

IMD gridded 31.1 0.6 1.9 35.9 0.6 1.6 35.2 0.5 1.3 30.7 0.5 1.5 IMD stations Trichy 31.4 0.7 2.3 37.0 0.8 2.0 35.9 0.6 1.7 30.7 0.6 2.0 Adiramapatinam 30.7 0.7 2.2 33.8 0.9 2.6 34.0 0.6 1.7 30.4 0.5 1.6 Cu ddalore 29.5 0.7 2.4 34.1 0.5 1.6 35.2 0.5 1.5 30.0 0.5 1.6 Mettur 33.0 0.7 2.0 37.5 0.5 1.4 34.4 0.5 1.5 31.6 0.6 1.9 Na gapattinam 28.9 0.6 2.1 33.5 0.5 1.5 35.4 0.6 1.6 30.0 0.4 1.4 Salem 33.2 0.9 2.7 37.4 0.7 1.9 33.9 0.7 2.0 31.3 0.7 2.3 Vedaranyam 29.7 0.8 2.5 33.6 0.8 2.3 33.8 0.7 2.0 30.1 0.7 2.4 GC M CCSM4 27.5 0.5 1.7 33.2 1.3 3.9 30.5 0.7 2.2 28.6 0.3 1.0 GCM MIROC5 29.3 0.7 2.3 33.5 0.6 1.8 33.3 0.7 2.0 29.8 0.6 1.9 RC M CCSM4 31.5 0.8 2.4 37.3 0.7 2.0 36.0 0.6 1.7 30.9 0.5 1.7 RCM GFDL 31.3 0.8 2.5 38.1 0.4 1.1 36.4 1.0 2.6 34.0 2.2 6.4

Table 3-3 Minimum Temperature statistics for CDZ

Season/ JF MAM JJAS OND Data sources Mean SD CV Mean SD CV Mean SD CV Mean SD CV IMD gridded 21.1 0.7 3.5 25.3 0.5 2.0 25.5 0.3 1.1 22.8 0.4 1.7 IMD stations Trichy 20.8 0.8 3.8 25.3 0.6 2.4 25.8 0.3 1.2 22.7 0.4 1.8 Adiramapatinam 21.0 0.8 3.9 25.7 0.5 1.8 25.8 0.5 1.9 23.2 0.5 2.4 Cuddalore 20.7 1.0 4.7 25.1 0.5 2.0 25.6 0.5 1.9 22.7 0.5 2.4 Mettur 20.9 0.9 4.2 25.2 0.7 2.7 24.2 0.6 2.4 22.1 0.6 2.7 Na gapattinam 22.9 0.8 3.6 26.4 0.7 2.5 26.3 0.3 1.2 24.1 0.4 1.8 Salem 19.5 0.8 4.2 24.0 0.6 2.6 23.2 0.5 2.0 20.9 0.5 2.5 Vedaranyam 22.5 1.0 4.6 25.2 1.0 3.8 25.4 0.9 3.6 23.1 0.9 4.0 GC M CCSM4 21.8 0.8 3.4 24.4 0.8 3.3 25.6 0.4 1.4 24.4 0.4 1.8 GCM MIROC5 21.7 0.6 2.9 25.3 0.5 2.1 26.1 0.4 1.5 23.6 0.3 1.4 RC M CCSM4 21.0 0.6 2.8 25.5 0.6 2.3 25.8 0.3 1.2 22.7 0.4 1.9 RCM GFDL 15.7 0.8 5.2 19.4 0.6 3.3 20.4 0.6 2.8 17.4 0.6 3.2

55. From these tables it is clear that during coolest months there is a better agreement in minimum temperatures range, i.e., 21-23°C. The IMD gridded data set shows a

27 baseline average of 21.1°C for minimum temperatures during January-February (JF) for the CDZ and the corresponding GCM and RCM mean values fall within the range 21.0 – 21.8°C. The RCM forced by GFDL however seems to have a cold bias, about ~5.0 °C below observed values. 56. The maximum temperatures during the hottest months March, April & May (MAM) from various baseline datasets fall within a range 33.2 – 37.3°C, and the range is larger as compared to the minimum temperature range. The IMD gridded data set shows a baseline average of 35.9°C with CCSM4 value of 33.2oC and 33.5°C for MIROC5, indicating a probable cold bias over the CDZ area. Some of the coastal stations show lower mean temperatures perhaps under the maritime influence (Figure 3-9), which is may be pronounced in the GCMs. This aspect would require further research to clearly identify the reason. The RCM results however show a higher mean maximum temperature (37.3°C) that is closer to the IMD gridded baseline estimates. Figure 3-9 below compares maximum temperatures during the seasonal extremes. JF' MAM'

35# 38# 33# 37# 31# 36# 35# 29# 34# 27# 33# Maximum'Temp'(C)' Maximum'Temp'(C)' 25# 32# 23# 31# # # # # # # # # # # # # # # # # # # id y D T G 4 5 d# hy# M D# TT G M R 4# 5 M r h M U T A EM DR C SM c A U E A E D M C S G ic C E N L E SM O C Gri ri R C N L E S O C r M A V C R D# T I M A V C IR C D# T DIRA S C I #C D S C # A M M IM A M M IM C C R R Figure 3-9 Baseline Maximum temperatures from station observations, IMD gridded data and GCM/RCM simulations during cooler months (JF) and hot season (MAM) over CDZ. (Gridded data sets are presented in different colors: red IMD, brown CCSM4 & MIROC5 GCMs and red hash is RCM forced by CCSM4)

57. In general, both GCMs seem to be underestimating the maximum temperature evolution during all seasons as seen in Figure 3-10 below. The regional climate model (RCM), which is driven, by one of the GCMs (CCSM4) overestimates the maximum temperatures. The error bars in Figure 3-10 indicate the spread amongst the observed data sets. Both GCMs show 2-3°C cooler mean maximum temperatures during most of the seasons. The RCM baseline is 1-2°C warmer during summer and SW monsoon season and quite close to the observed temperatures during the NE monsoon and the cooler months of JF. Thus, RCM appears to capture the regional characteristics well.

28 IMD#OBS# CCSM4#GCM# MIROC5#GCM# RCM#CCSM4#

39#

37#

35#

33#

31#

Maximum'Temp'(C)' 29#

27#

25# JA# MAM# JJAS# OND# Figure 3-10 Mean maximum temperatures over CDZ during different seasons.

58. Table 3-4 presents the basic statistics of rainfall over the CDZ from different baseline data sources. There are significant differences between different stations and the different observed datasets. As the main rainy months over the project area are the Southwest monsoon season (June, July, August and September: JJAS); and the Northeast monsoon season (October, November and December: OND) the discussion here will focus on these seasons.

Table 3-4 Rainfall basic statistics over CDZ

Season JF MAM JJAS OND Data sources Mean SD CV Mean SD CV Mean SD CV Mean SD CV IMD gridded 28.4 44.9 158.2 90.9 32.8 36.1 299.6 71.6 23.9 471.3 162.4 34.5 Aphrodite 41.2 58.3 141.6 108.2 42.2 39.0 354.6 106.0 29.9 589.0 203.2 34.5 IMD stations Trichy 21.0 41.1 196.2 99.5 67.6 67.9 338.9 119.4 35.2 390.4 197.5 50.6 Adiramapatinam 56.4 87.4 154.9 120.8 77.9 64.5 316.4 152.0 48.1 641.4 323.9 50.5 Cuddalore 51.0 90.4 177.1 50.7 50.2 99.0 375.4 136.2 36.3 763.7 296.7 38.9 319.9 173.1 54.1 Mettur 10.7 16.7 156.4 182.6 71.0 38.9 378.2 106.5 28.1 Nagapattinam 74.2 112.7 151.8 79.0 65.5 82.9 260.1 99.3 38.2 919.9 280.5 30.5 Salem 8.5 19.5 229.2 152.0 65.6 43.1 502.1 167.3 33.3 286.9 127.8 44.5 Vedaranyam 88.4 120.2 135.9 71.3 52.1 73.0 225.1 135.3 60.1 922.4 304.3 33.0 GCM CCSM4 100.1 105.3 105.2 142.4 95.8 67.3 718.8 166.6 23.2 711.0 176.3 24.8 GCM MIROC5 44.0 38.6 87.8 129.9 43.8 33.7 311.7 109.9 35.3 358.8 105.5 29.4 RCM CCSM4 89.2 40.0 44.8 139.4 102.5 73.6 575.0 115.2 20.0 621.8 312.6 50.3 RCM GFDL 25.7 19.1 74.1 23.9 23.2 97.0 268.3 97.5 36.4 111.2 105.5 94.9

29 59. During the SW monsoon season the range of rainfall observed is about 260 – 500 mm as compared to 320 – 900 mm observed during the NE monsoon season over the IMD stations in the CDZ area. The two gridded rainfall data sets also show differences, with APHRODITE showing higher rainfall as compared to IMD gridded data set. The difference is about 50 mm during the JJAS and about 118 mm during the OND season as estimated during the baseline period. These ranges indicate approximately the degree of uncertainty in rainfall observations. 60. Figure 3-11 below shows the differences in rainfall amongst the various baseline datasets during the two major wet seasons over the CDZ area. While the station averages roughly indicate the spatial variations, the differences amongst the two gridded datasets gives an idea of the range of uncertainty in rainfall estimation over the area.

Figure 3-11 Baseline rainfall from IMD stations, gridded datasets and GCM/RCMs

61. During the SW monsoon season the CCSM4 shows a wet bias while the MIROC5 GCM rainfall baseline is quite similar to observations. The RCM that is forced by the CCSM4 also displays a wet bias, but is lower than the parent GCM. In the NE monsoon season the models seem to be in better agreement with the observed rainfall, with values falling with the range of uncertainty of observations. Figure 3-12 below shows the seasonal rainfall for all the seasons along with uncertainty envelops computed using standard error estimates.

30 IMD"OBS" CCSM4"GCM" MIROC5"GCM" RCM"CCSM"

800"

700"

600"

500"

400"

300"

200"

100"

0" JF" MAM" JJAS" OND" Figure 3-12 Observed Seasonal rainfall from IMD stations and gridded datasets compared with GCM and RCM simulated baseline values

3.3. Evaluation of GCM and RCM

Temperature seasonal cycle 62. Both the GCMs, and the RCMs capture the seasonal evolution in maximum temperature as observed but with a cold bias in GCMs as already noted from the results shown in the Tables 3 & 4 in the preceding section. The CCSM4 GCM shows a cold bias in its simulations of maximum temperatures over the CDZ. The bias becomes particularly pronounced during the southwest monsoon months (Figure 3-13). Downscaled results of CCSM4 show a slight warm bias in maximum temperatures. The MIROC5 model seems to be simulating the observed temperatures fairly well.

31

Figure 3-13 Evaluation of GCM (CMIP5 MIROC5 & CCSM4), RCM (GFDL 2.1 & CCSM4) maximum temperature seasonal cycle with observations.

32

Figure 3-14 Evaluation of GCM (CMIP5 MIROC5 & CCSM4), RCM (GFDL 2.1 & CCSM4) minimum temperature seasonal cycle with observations.

63. In the case of minimum temperature evolution, both GCMs capture the observed seasonal cycle but the RCMs counterparts simulate too cold a bias (Figure 3-14). This seems to be persistent problem, particularly affecting the downscaling results from GFDL CM2.1 results.

Rainfall Seasonal cycle evaluation 64. Despite the evaluation is performed on a smaller region, both the GCMs and RCMs depict the seasonal cycle in rainfall that is remarkable. One possible interpretation could be - rainfall seasonal cycle, which is governed by large-scale processes, is perhaps well represented in the climate models. The difference is, however, in the amplitude of the simulated rainfall (Figure 3-15). Figure 3-16 shows a similar plot of bias corrected GFDL RCM rainfall time-series.

33

Figure 3-15 Evaluation of GCM (CMIP5 MIROC5& CCSM4), RCM (GFDL 2.1 & CCSM4) rainfall seasonal cycle with observations.

Figure 3-16 Evaluation of GCM (CMIP5 MIROC5& CCSM4), RCM (GFDL 2.1 bias corrected) rainfall seasonal cycle with observations

34

Rainfall Daily Climatology

Figure 3-17 Evaluation of GCM CMIP5 CCSM4 rainfall daily climatology with observations.

Figure 3-18 Evaluation of RCM CCSM4 rainfall daily climatology with observations.

Figure 3-19 Evaluation of RCM GFDL rainfall daily climatology with observations.

35 65. A further examination of the seasonal cycle in terms of daily rainfall climatology (Figure 3-17, Figure 3-18 & Figure 3-19) indicate the regional model, particularly the RCM forced by CCSM4 represents the evolution more realistically than the one forced by GFDL. Further analysis was carried out to introduce bias correction into the RCM rainfall simulations.

36 Spatial Patterns in Summer Maximum Temperatures

Figure 3-20 Spatial distribution of maximum temperatures during summer season (MAM): a) IMD gridded data, b) MIROC5 GCM and c) CCSM4 GCM.

66. Figure 3-20 shows a comparison of CMIP5 GCMs baseline maximum temperature climatology during summer (MAM) with observed data. CCSM4 baseline is fairly close to the observations, particularly over the CDZ region indicated by the red box. Model temperatures are however, lower as compared to the IMD observed baseline indicating a cold bias in the model. Further comparative analyses are being carried out to systematically quantify these biases. 67. The RCMs baseline spatial patterns of maximum temperatures during MAM (Figure 3-21) are closer depictions of the spatial variations shown by the observed IMD gridded plots (Figure 3-20). The higher spatial resolution of the regional model seems to enable better simulation of spatial variations in temperatures. The lower temperatures over the west (due to higher elevations of the terrain) and coastal areas are brought out. The IMD gridded temperature data based on observations is unable to capture these spatial variations due to its coarser resolution at 1 degree.

37 ! Figure 3-21 Spatial distribution of maximum temperatures during the summer season (MAM) in Baselines of RCMs forced by GFDL and CCSM4

Spatial Patterns in Rainfall 68. Figure 3-22 & Figure 3-23 show the spatial variations in rainfall over the Tamil Nadu region during the southwest (SW) monsoon season (JJAS) and the northeast (NE) monsoon season (OND). Both models capture the enhanced rainfall activity over the CDZ during the NE monsoon season, but MIROC5 produces lesser rainfall due to which reason the spatial variations of the observed climate are not seen.

38

Figure 3-22 Spatial plots of rainfall during the southwest (JJAS) and the northeast (OND) monsoon seasons in the MIROC5 baseline simulations.

Figure 3-23 Spatial plots of rainfall during the southwest (JJAS) and the northeast (OND) monsoon seasons in the CCSM4 baseline simulations.

39

69. The rainfall spatial climatology of CCSM4 clearly captures (Figure 3-23) the southwest monsoon season contributing higher rainfall over the western region while the CDZ region experiencing maximum rainfall during the northeast monsoon season.

Salient aspects of model evaluation 70. The evaluation results show that the global and regional climate models are successful in simulating the broad aspects of the climate of the CDZ, including the monsoons. But differences in details such as timing and intensity of rainfall are present. Some uncertainties also arise from the fact that there are differences among observed data sets owing to the different ways in which point observations are combined to produce spatial representation of climate. i. Seasonal cycle of temperatures over the project area is well simulated, but with biases that are more prominent in minimum temperatures when compared with gridded observational data set of IMD. To overcome the issue of temperature bias, the delta approach has been adopted wherein deviation of climate change projections are made with respect to corresponding model baselines. ii. Both the GCMs and RCMs depict the seasonal cycle in rainfall averaged over the project area remarkably well. Differences in rainfall amounts were corrected for one of the RCMs using bias correction approach. Although this resulted in significant improvement, some residual differences remain during some of the months despite the bias correction. iii. Regional models significantly improve the representation of spatial distribution of both temperature and rainfall over the Tamil Nadu region including the project area.

Climate variables

71. Our focus in this study has been on temperature (including maximum and minimum) and rainfall based variables. This is because reliable long-term observed data sets for these parameters are readily available, and climate model simulations these parameters have been evaluated more rigorously.

40 GFDL RCM

! ! ! CCSM4 RCM !

!

Figure 3-24 Spatial plots of rainfall during the southwest (JJAS) and the northeast (OND) monsoon seasons in the two RCM baseline simulations.

72. Figure 3-24 shows the seasonal rainfall patterns simulated over the whole of TN by the two RCMs. On comparing with IMD gridded data rainfall climatology in Figure 3-4, GFDL seems to simulate the spatial variation in rainfall better than than the CCSM4, but the quantity of rainfall is underestimated.

41 73. To improve the rainfall estimates of the GFDL forced RCM simulations for the TN a bias correction procedure was undertaken which is described in the following section.

3.4. Bias Correction 74. Rainfall variability largely depends on its frequency and amount and it is difficult to estimate average rainfall in particular region. A standard procedure in hydrological studies is used to rectify the systematic errors by projecting the model values to observations in the base-line period (Sharma et al, 2007, Van Pelt et al, 2009) and adjust the future values accordingly. Weibull distribution function (Weibull, 1951) method is used for bias correcting the GFDL forced regional model rainfall simulations and to make them as close as possible to those of the IMD observations. 75. Application of the bias correction procedure involves by transforming the distribution of the GFDL_CM2.1 baseline data to that of the IMD observed data. By calculating the slope of the ordered daily non-zero rainfall between IMD observations and the base period regional model rainfall indicates that the under estimation of rainfall by the GFDL model. This under estimation of rainfall is removed by the bias adjustment procedure, which is indicated by the slope values. 76. The bias is calculated by using the Weibull distribution scale and shape factor between observed and the regional model rainfall. The bias in the scale factor is calculated by subtracting the scale factor of base period regional model from the scale factor of IMD observations. Similarly, the shape factor is also calculated for the base period. The scale and shape factor calculated for the future scenario is by adding the bias in scale and shape factor to the original scale and shape factor of the GFDL model. Thus, the resultant regional model rainfall is expected to achieve a closer fit to the observations. In our results after bias correction, this improvement is well represented in below.

Figure 3-25 Seasonal cycle of rainfall compared with the bias corrected rainfall: IMD gridded and GFDL CM2.1 Baseline (1981-2000) and future scenario (2021-2050) for TN region.

42

4. Projections 77. CMIP5 projections from two GCMs – CCSM4 and MIROC5 based on two future RCP scenarios representing possible climate states of the 2050s over the TN have been presented as described in section 2. These GCM projections give the overall scenarios of climate change over the TN, which serves as a background to consider changes over the CDZ. 78. The second tier of climate change projections are from the regional dynamical downscaling results using RCMs. Here we have used a mixture of results available from two different RCM runs using the same model, but forced by two different GCMs viz. CCSM4 and GFDL CM2.1. The former model is from the CMIP3 experiments, while the latter is from the more recent CMIP5 runs. While the GFDL RCM runs are available for the 2050s time horizon, the CCSM4 projection results were only available for the 2080s time horizon (Table 4-1). Therefore, the projections results presented must be interpreted accordingly.

Table 4-1 Time-slices and scenarios used for the RCM simulations

S. No Models Baseline Period Future Scenario Emission scenario used 1. GFDL CM2.1 1981 -2000 2021 -2050 SRES Y1B 2. CCSM4 1986 -2005 2081 -2100 RCP 60

79. GFDL Y1B: In the Y1B scenario atmospheric CO2 concentrations reaches doubling in the mid century itself. The B1 storyline and scenario family describes a convergent world with the global population that peaks in mid-century and declines thereafter. In this scenario world is characterized by low population growth, high GDP growth, low energy use, high land-use changes, low resource availability and medium introduction of new and efficient technologies. 80. CCSM4 RCP 60 Scenario: The Representative Concentration Pathways (RCP) are based on selected scenarios from four modelling teams. The CMIP5 is used the four RCP scenarios in their future climate projection model runs. Four RCPs were selected and defined based on their total radiative forcing pathway and level by 2100. The RCP 6.0 is developed by the Asia-pacific integrated modeling (AIM) team at the National Institute for Environmental Studies (NIES), Japan. It is a stabilization scenario where total radiative forcing is stabilized to 6 W/m2 after 2100 without overshoot. The CO2 concentration also reaches the 850 ppm in 2100 and it stabilizes without overshoot after 2100.

4.1. Climate Change Scenario

Seasonal Cycle Temperature

81. Both minimum temperatures and maximum temperatures show increase over the CDZ in almost all the months and future scenarios of climate change projected by

43 CMIP5 GCMs and the RCM downscaled as seen in Figure 4-1, Figure 4-2 and Figure 4-3.

Figure 4-1 MIROC 5 GCM baseline and climate change projections for Maximum temperatures during the 2050s for two GHG scenarios.

Figure 4-2 CCSM4 GCM baseline and climate change projections for Maximum temperatures during the 2050s for two GHG scenarios.

44

Figure 4-3 RCM baseline and climate change projections for Maximum temperatures during the 2050s for two GHG scenarios

82. The projections from the RCM are perhaps able to better resolve the climatological features of the temperature, but their magnitude of change in response to future GHG increase scenarios seem to be smaller as compared to the GCMs. This could be at least partly attributed to the very nature of climate, which shows higher variability as we zoom into smaller areas looking for location specific signatures of climate change. Maximum temperature climatology in RCM also seems to bear a stronger influence of the SW monsoon resulting in a temperature dip during the monsoon season.

45

Figure 4-4 Baseline and climate change projections for minimum temperatures during the 2050s in different GCM and RCM projections of change in response to GHG scenarios

83. The minimum temperatures show more consistent increases across the different models used in this study for the CDZ. Rainfall

84. The rainfall climatology shows an increasing tendency in most of the model projections. Figure 4-5 shows the seasonal cycles of rainfall projected for CDZ.

46

Figure 4-5 Baseline and climate change projections for rainfall during the 2050s in different GCM and RCM projections of change in response to GHG scenarios.

Daily rainfall projections 85. The rainfall increase in the seasonal climatology seen in the Figure 4-5 as marginal increase manifests as higher variability in both the RCM projections of future changes. This is seen in the daily normals during the period representative of the 2050s. Figure 4-6Figure 4-7 of daily normals showing baseline and projections from the two regional downscaling results clearly illustrate this aspect of rainfall change.

47

Figure 4-6 Daily rainfall RCM GFDL baseline (before bias correction) and future projection during 2050s. Black dots are observations from the IMD gridded data set.

Figure 4-7 Daily rainfall RCM CCSM4 baseline (before bias correction) and future projection during 2050s. Black dots are observations from the IMD gridded data set.

86. Further confidence in these results is lent by the analysis of daily rainfall data for the whole CDZ for baseline and future climate change conditions. In Figure 4-8 if we compare the number of days above 150 mm rainfall, a much higher number is seen in the case of projected daily rainfall over CDZ as compared to the baseline.

48

Figure 4-8 Distribution of daily rainfall for all the grids in CDZ for RCM GFDL Baseline (1981-2000) & projection (2021-2050)

4.2. Basic statistics Table 4-2 Maximum temperature basic statistics over CDZ for JJAS, OND, JF, MAM (Projections)

Season/ JF MA M JJAS OND Data sources Mean SD CV Mean SD CV Mean SD CV Mean SD CV GCM CCSM4 RCP45 (2021- 28.8 0.3 1.2 32.8 0.9 2.6 32.8 0.7 2.2 30.5 0.2 0.8 2050) GCM CCSM4 RCP85 (2021- 28.7 0.6 2.0 34.6 0.9 2.5 31.6 0.5 1.7 29.7 0.4 1.3 2050) GCM MIROC5 RCP45 (2021- 30.2 1.1 3.6 34.5 1.0 2.8 34.2 0.7 2.0 31.2 0.8 2.6 2050) GCM MIROC5 RCP85 (2021- 30.5 1.0 3.1 34.9 0.8 2.4 34.6 0.5 1.5 31.4 0.6 2.1 2050) RCM GFDL 32.9 1.1 3.3 38.8 0.8 2.1 36.1 1.2 3.2 33.9 1.9 5.5 2021-2050 RCM CCSM4 32.6 0.7 2.2 39.1 0.8 2.0 38.9 0.3 0.9 33.2 0.6 1.8 2081-2100

49 Table 4-3 Minimum temperature basic statistics over CDZ for JJAS, OND, JF, MAM (Projections)

Season/ JF MAM JJAS OND Data sources Mean SD CV Mean SD CV Mean SD CV Mean SD CV GCM CCSM4 23.0 0.5 2.3 25.6 0.6 2.4 26.6 0.3 1.1 25.4 0.3 1.2 RCP45 (2021-2050) GCM CCSM4 22.7 0.9 3.8 25.9 0.7 2.7 26.6 0.3 1.3 25.5 0.4 1.4 RCP85 (2021-2050) GCM MIROC5 24.3 0.7 3.0 27.7 0.9 3.1 28.1 0.5 1.8 26.1 0.4 1.7 RCP45 (2021-2050) GCM MIROC5 23.9 0.7 3.1 27.6 0.7 2.7 28.2 0.4 1.4 26.0 0.4 1.7 RCP85 (2021-2050) RCM GFDL 2021- 18.2 1.0 5.4 21.6 0.6 2.7 22.4 0.6 2.6 19.6 0.8 4.2 2050 RCM CCSM4 20.9 0.6 2.9 23.0 0.4 1.8 24.6 0.2 1.0 23.4 0.4 1.5 2081-2100

Table 4-4 Rainfall basic statistics over CDZ for JJAS, OND, JF, MAM (Projections)

Season/ JF MAM JJAS OND Data sources Mean SD CV Mean SD CV Mean SD CV Mean SD CV GCM CCSM4 RCP45 (2021-2050) 84.4 60.0 71.0 110.2 66.8 60.6 813.0 94.9 11.7 740.9 183.6 24.8 GCM CCSM4 RCP85 (2021-2050) 60.2 32.5 54.0 126.8 77.7 61.3 774.3 151.1 19.5 734.0 231.3 31.5 GCM MIROC5 RCP45 (2021-2050) 102.5 105.3 102.8 121.6 48.6 39.9 394.7 145.8 36.9 392.1 195.7 49.9 GCM MIROC5 RCP85 (2021-2050) 58.8 57.0 96.9 107.0 50.4 47.1 398.4 135.9 34.1 330.2 169.0 51.2 RCM GFDL 2021- 2050 54.1 59.3 109.6 154.9 64.6 41.7 355.1 98.0 27.6 289.5 139.0 48.0 RCM CCSM4 2081-2100 56.1 35.6 63.5 124.9 123.8 99.1 540.8 127.2 23.5 622.8 211.1 33.9

87. Basic statistics show an increase in temperature and rainfall over the CDZ.

50 4.4. Spatial changes

Figure 4-9 Maximum temperature deviation of RCM GFDL and RCM CCSM4 baseline from IMD gridded data (1971-2000).

88. Temperature changes over the CDZ projected by the two RCMs range from 2-6 °C, with the RCM driven by GFDL showing higher temperature increase (Figure 4-9). 89. In case of rainfall changes during both SW and NE monsoon were examined. Figure 4-10 shows the spatial variations in the projected rainfall changes as compared to respective model baselines. The RCM forced by GFDL projects a possible decrease in rainfall, whereas the other model shows an increase. 90. In the subsequent chapters more detailed analysis of daily rainfall has been presented.

51

Figure 4-10 JJAS and OND rainfall departure (%) of RCM GFDL and RCM CCSM4 baseline from IMD gridded data (1971-2000).

52 4.5. Summary of Changes

Figure 4-11 Summary of projected changes over the CDZ a) Maximum Temp. b) Minimum Temp. and c) annual Rainfall in the GCMs and RCMs used in this study

53 91. Summary of projections for the 2050s decade over the CDZ from the different GCMs and RCM downscaled results presented in Figure 4-11 clearly indicates the overall warming expected in the region. There is also a predominant agreement among model results that makes this conclusion a “high confidence” estimate. Although the rainfall amount show increase, as the amounts are small and there is a lower level of agreement this conclusion is of medium to low level of confidence. 92. We also made some preliminary comparison of the evaporation data sets from the MIROC that show an increasing trend in evaporation () consistent with the warming temperatures over the project area. The confidence in these estimates is very low as the model baseline data could not be evaluated due to insufficient length of the observations.

CDZ!

100.0! 80.0! 60.0! 40.0! 20.0! 0.0! 1982! 2002! 2022! 2042! 2062! 2082! Evaporation!mm/month! 1970! 1974! 1978! 1986! 1990! 1994! 1998! 2006! 2010! 2014! 2018! 2026! 2030! 2034! 2038! 2046! 2050! 2054! 2058! 2066! 2070! 2074! 2078! 2086! 2090! 2094! 2098!

CDZ_R1! CDZ_R2! CDZ_R3! ! CDZ!

80.0!

60.0!

40.0!

20.0!

0.0! 1982! 2002! 2022! 2042! 2062! 2082! Evaporation!mm/month! 1970! 1974! 1978! 1986! 1990! 1994! 1998! 2006! 2010! 2014! 2018! 2026! 2030! 2034! 2038! 2046! 2050! 2054! 2058! 2066! 2070! 2074! 2078! 2086! 2090! 2094! 2098! Figure 4-12 Base-line (1970-98) andCDZ_R1! projections CDZ_R2! of evaporation CDZ_R3! from MIROC5 GCM for the CDZ from three different model runs. !

54

5. Daily Rainfall analysis 93. Annual maximum one day accumulated rainfall for the period 1971 to 2000 has been fitted with Gumbel distribution using CFA (Cumulative Frequency Analysis) software package. Datasets has been tested using various distribution function in this tool, Gumbel (Fisher-Tippett type I) distribution (skew to right) is widely accepted for return period analysis and proves to be one of the best fitting distribution by having correlation of fitted data more than 0.9. 94. The formula for Gumbel (Fisher-Tippett type I) distribution (skew to right) are as follows: Fc = Exp[-Exp{-(A*X+B)}] Ft = -Ln{-Ln(Fc)} Ft = A*X + BFc = cumulative frequency X = stochastic variable Exp = exponent Ft = transformed Fc Ln(y) = natural logarithm (with base e) of y A & B are found from a linear regression of Ft on X (or Xt),

5.1. Return Period Analysis for Baseline The return period of rainfall amount calculated for seven locations are tabulated below (Table 5-1)

Table 5-1 One-day rainfall return periods from observations and baseline model data

Station Rainfall mm in various return periods 2 Years 5 Years 10 Years 20 Years 50 years 100 years 200 years IMD Surface Observatory record (1971-2005) Adiramapatinam 113 179 223 265 319 360 400 Cuddalore 125 192 235 278 332 373 413 Mettur 83 124 151 178 212 237 262 Nagapattinam 139 208 253 297 353 395 437 Salem 75 100 117 133 153 169 184 Trichy 92 153 193 231 281 318 355 Nagapattinam 142 199 236 272 318 353 387 RCM Baseline* – GFDL (1981-2000) Adiramapatinam 59 88 107 126 149 167 185 Cuddalore 72 125 160 194 238 271 303 Mettur 73 115 143 170 205 231 257 Nagapattinam 59 97 122 145 176 199 222

55 Salem 95 152 190 227 274 310 345 Trichy 73 110 134 158 188 211 234 Nagapattinam 56 89 111 132 159 180 200 RCM Baseline* – CCSM4 (1986-2005) Adiramapatinam 122 209 267 322 393 446 500 Cuddalore 92 144 178 211 253 285 317 Mettur 108 141 163 184 211 231 251 Nagapattinam 100 146 176 205 243 271 300 Salem 104 136 157 177 203 223 242 Trichy 109 156 187 217 256 285 313 Nagapattinam 107 149 176 203 237 263 289 *RCM Grids close to IMD surface observatory were chosen for return period analysis

5.2. Return periods for projected rainfall data The return period for projected rainfall data were calculated and tabulated below ()

Table 5-2 One-day rainfall return periods from model projections of rainfall changes

Station Rainfall mm in various return periods 2 Years 5 Years 10 Years 20 Years 50 years 100 years 200 years RCM Projection* – GFDL (2021-2050) Adiramapatinam 110 171 210 249 298 335 372 Cuddalore 127 190 232 273 325 364 403 Mettur 125 165 191 216 249 273 297 Nagapattinam 84 130 161 190 228 256 284 Salem 121 181 220 258 307 343 380 Trichy 103 143 170 196 230 255 280 Nagapattinam 88 136 168 198 237 266 295 RCM Projection* – CCSM4 (2081-2100) Adiramapatinam 121 166 195 224 260 288 315 Cuddalore 116 179 220 260 312 351 389 Mettur 126 165 192 217 249 273 298 Nagapattinam 116 185 231 274 231 373 416 Salem 121 175 210 244 288 320 353 Trichy 123 167 196 224 260 287 314 Nagapattinam 118 169 202 235 277 308 339 *RCM Grids close to IMD surface observatory were chosen for return period analysis

56 5.3. Probability of rainfall 95. Probability of assured rainfall quantum at various percentages (50%, 80%, 90% and 95%) has been calculated for three major seasons summer (MAM), south west monsoon (JJAS), north east monsoon (OND). CFA tool has been used to fit the seasonal rainfall time series data with probability distribution function. After testing various distribution function, Generalized mirrored Gumbel distribution method has been chosen because of its good correlation value (>0.9) for the fitted datasets.

96. The formula for Generalized mirrored Gumbel distribution (any skewness) are as follows Fc = 1 - Exp{-Exp{-A*X^E+B)} Xt = Ln(X^E) = E*Ln(X) Ft = -Ln{-Ln(1-Fc)} Ft = A*Xt + B Fc = cumulative frequency X = stochastic variable E = exponent Ft = transformed Fc Xt = transformed X Ln(y) = natural logarithm (with base e) of y A & B are found from a linear regression of Ft on X (or Xt), Where E is optimized in the tool automatically in the generalized distributions using a range of values and selecting the value giving the minimum sum of squares of deviations of calculated and observed cumulative frequencies.)

Baseline

The calculated rainfall probability of three seasons for seven locations are tabulated below (Table-3)

Table 5-3 Rainfall probability in three major seasons (Baseline).

Seasonal accumulated rainfall (mm) probability Season Location IMD Observation data RCM GFDL RCM CCSM4 95% 90% 80% 50% 95% 90% 80% 50% 95% 90% 80% 50% Adiramapatinam 11 26 51 118 2 4 8 20 16 34 65 151 Cuddalore 0 2 7 36 2 5 10 25 2 9 28 105 Mettur 67 91 122 183 1 3 9 33 4 12 30 94 MAM Nagapattinam 2 6 19 65 2 5 10 24 7 18 42 116 Salem 26 54 90 157 1 2 7 25 1 8 25 87 Trichy 11 22 41 92 1 2 6 21 6 17 39 110 Vedaranyam 2 10 27 73 1 2 5 18 14 32 66 165 Adiramapatinam 100 143 202 326 58 93 142 245 191 259 337 465 Cuddalore 150 197 258 378 40 75 135 290 248 317 405 572 JJAS Mettur 187 230 284 384 81 130 202 370 451 536 631 790 Nagapattinam 55 109 171 272 68 104 157 275 247 310 380 492 Salem 243 309 387 529 62 107 176 340 393 469 548 670

57 Trichy 111 162 226 346 83 123 176 277 339 407 489 632 Vedaranyam 57 88 134 236 57 87 131 229 180 247 323 445 Adiramapatinam 199 286 406 659 1 6 20 80 161 256 396 717 Cuddalore 242 342 478 762 1 7 27 120 114 185 292 540 Mettur 51 92 157 311 2 7 21 77 30 65 128 306 OND Nagapattinam 397 510 654 931 2 7 25 101 145 231 360 655 Salem 95 134 186 296 1 4 15 67 34 69 132 305 Trichy 60 115 199 386 1 7 21 82 117 187 291 531 Vedaranyam 463 576 714 967 1 5 20 90 82 255 395 716

Future projections The calculated rainfall probability of three seasons for seven locations are tabulated below (Table-3)

Table 5-4 Rainfall probability in three major seasons. (Projection)

Seasonal accumulated rainfall (mm) probability Season Location RCM GFDL RCM CCSM4 95% 90% 80% 50% 95% 90% 80% 50% Adiramapatinam 25 45 79 165 5 15 41 135 Cuddalore 20 36 61 127 1 4 16 77 Mettur 46 58 100 169 1 4 16 67 MAM Nagapattinam 13 31 63 139 2 8 26 103 Salem 35 54 81 144 1 3 13 57 Trichy 31 54 86 151 1 3 14 69 Vedaranyam 22 45 83 171 4 15 43 149 Adiramapatinam 155 196 247 343 203 258 327 457 Cuddalore 163 215 280 402 194 264 356 541 Mettur 285 353 432 567 402 491 599 798 JJAS Nagapattinam 149 192 247 350 217 272 339 466 Salem 208 263 332 463 353 429 521 689 Trichy 160 204 259 366 288 362 448 596 Vedaranyam 148 189 241 340 204 261 334 474 Adiramapatinam 53 88 140 262 144 315 494 768 Cuddalore 59 104 176 355 90 170 287 540 Mettur 51 81 124 222 80 122 184 321 OND Nagapattinam 62 105 172 334 112 285 464 734 Salem 46 74 116 214 73 115 176 313 Trichy 76 110 159 263 139 234 345 529 Vedaranyam 35 69 128 286 94 314 518 805

58 97. The analysis results of return periods and probability of assured rainfall indicate a disagreement between the two regional downscaled estimates. One of the models (RCM GFDL) indicates significantly higher return period amounts. Due to disagreement there is low confidence in these results. However, these data can be used to plan for building resilience infrastructure from a risk management perspective. Rainfall daily time series for the whole CDZ area has been provided at a ~30 km resolution.

59 6. Tropical Cyclones 98. Low-lying coastal areas of Tamil Nadu are prone to storm surges and very heavy rainfall due to passage of tropical cyclones. These cyclones typically form and intensify over the warm waters of the Bay of Bengal, and based on the observations maintained by the India Meteorological Department Cyclone Warning Center tropical cyclones hit Tamil Nadu preferably during the months of October to January coinciding with the northeast monsoon season. Some unique characteristics of tropical cyclones that hit Tamil Nadu include strong winds and heavy rainfall in a short period of time. Consequently, passage of cyclones impact local flooding, and water-logging in low-lying areas that cause heavy damage to standing crops as well road and rail traffics in Tamil Nadu. Specifically, an untimely cyclone during the mature stages of paddy crop over the Cauvery river basin results in measurable economic losses to small-scale farmers in the region. Therefore, understanding and documenting their expected changes in a global warming scenario are of socio- economic importance.

6.1. Current trends and variations 99. Currently for the Tamil Nadu coast the India Meteorological Department (IMD) uses the wind speed classification system as given in Table 6-1 below. Table 6-1 Cyclonic storms categories and wind speeds used by IMD

Type of disturbance Wind Speed Low Pressure Area Less than 17 knots/31kmph Depression 17 -27 knots/~31-50 kmph Deep Depression 28 -33 knots/~51-61 kmph Cyclonic Storm 34 -47 knots/~62- 87 kmph Severe Cyclonic Storm 48 -63 knots/~88-116 kmph Very Severe Cyclonic Storm 64 -119 knots/~117-220kmph Super Cyclonic Storm 120 knots/221 kmph and above

100. The India Meteorological Department (IMD) archives the historical data on Tropical Cyclones, including data of cyclone tracks as e-atlas. Figure 6-1 below shows the frequency of cyclones (Cyclonic Storms and Severe Cyclonic Storm categories:CS+SCS) crossing various regions of the Bay of Bengal coast. Over 60 storms of both categories have crossed the Tamil Nadu coast during the 1891-2008 period of available observational records.

60

Figure 6-1 Landfall frequency of Cyclonic Storms/Severe Cyclonic Storms over the coastal regions adjoining Bay of Bengal during the period 1891-2008.

101. The long-term data does not show any significant increasing or decreasing trends in the frequency of CS+SCS categories of Tropical Cyclones crossing the Tamil Nadu coast. This is illustrated by the analysis of long-term records presented in Figure 6-2.

Figure 6-2 Frequency of Tropical Cyclones (CS+SCS & SCS) crossing Tamil Nadu coast during 1900-2012.

61 6.2. Tropical Cyclones and possible influence of climate change 102. In a warmer world, changes in sea surface temperature over the Bay of Bengal and enhanced atmospheric moisture content associated with the large-scale circulation of the northeast monsoon are likely expected to influence the statistical properties of tropical cyclones that hit Tamil Nadu. Global climate models are typically coarse in horizontal resolution (150-200 km) and are therefore unable to capture the genesis location, intensity, number density and landfall sites of tropical cyclones. An alternate approach is to perform high-resolution global or regional (~10- 60 km) climate model integrations that depend on the availability of computer resources. Any future projections of tropical cyclone characteristics, however, are model dependent and thus pose large uncertainties, particularly over the Northern Indian Ocean (NIO) for two basic reasons. They are: (i) the number of cyclones forming over the NIO is too small compared to other tropical ocean basins and (ii) climate models are unable to capture the present-day cyclone climatology. 103. To obtain statistical robust results, the constraint in the small number of cyclones requires a longer simulation with high-resolution models leading to large computer resources. Even if sufficient resolutions are met, due to limitations in model physics, robust climatology from climate model simulations is not attainable. Knutson et al. (2010) pointed out the diversity in the future cyclone characteristics’ projections. The authors reported that in their ensemble simulations of 19 cases, 14 experiments indicated a decrease while only 5 members showed an increase in the cyclone frequency over the NIO. Examining the output from a series of very high-resolution (10 km) climate model simulations, Murakami et al. (2012) reported for no significant changes in the over-all frequency but noted slight changes in the genesis over the Arabian Sea. 104. Recently, Annamalai (2013) applied the cyclone tracking technique employed in Stowasser et al (2009) to observations, and coarse-resolution climate model and high-resolution regional climate model simulations. The analysis was performed during both the southwest and northeast monsoon seasons, and for base line and future climate scenarios. The following criteria must be satisfied for a system that we identify and track: A local vorticity maximum at 850 hPa exceeds 5·10-5 s-1 (Annamalai et al. 1999). A local pressure minimum exists within a radius of 250 km of the vorticity maximum; this minimum pressure is taken as defining the center of the storm system. To be considered as a model storm trajectory, a storm must last at least 2 days. 105. Note that the regional climate model integrations were conducted by forcing the IPRC regional model (IPRC_RegCM) with lateral and boundary conditions taken from the coarse-resolution climate models (CMIP3/5) that “best” capture the current monsoon rainfall climatology and its variations. To validate our technique against the best track maintained by the India Meteorological Department the tracking program was applied for the unusual cyclone activity season of October 2011-January 2012. Here, the variable used is low-level vorticity obtained from ECMWF Interim reanalysis products. Figure 6-3 shows the instantaneous position of the storms over both the Arabian Sea and Bay of Bengal. Typically, from their genesis locations, the cyclones last for about 4-5 days, and in this particular season one named cyclone, Thane, made land-fall over Tamil Nadu during December end (yellow line in Figure 6-3). In summary, the technique employed here faithfully captures the tropical storms over the NIO (see Annamalai 2013 for more details).

62

Figure 6-3 Instantaneous position of the storms over both the Arabian Sea and Bay of Bengal

Figure 6-4 Year wise detection of tropical cylones in the IPRC RegCM model.

63 106. Figure 6-4 shows the year-wise detected storms in the IPRC_RegCM integrations for base line (blue) and future scenario (green). Note that in each case, a 20-year simulation was performed. While there are clear year-to-year variations in the number of storms in both the base line and future scenario integrations, the total number of detected storms for the entire period of simulation virtually remains the same in both cases. In other words, no measurable change in the total number of storms is projected. Our results are consistent with those of Knutson et al. (2010) and Murakami et al. (2012).

7. Conclusions

107. A combination of results from select GCMs and RCM were used to prepare climate change scenarios for the Cauvery Delta Zone. Climate change scenarios data sets have been prepared for this zone for further use in hydrological analysis. 108. The study started with a preparation of an inventory of climate data, both observed and climate model results. CMIP5 results from two models known to be robust in their simulation of monsoon circulation were used to broadly characterize changes expected in response to climate change. These results were considered along with high-resolution results from IPRC RegCM to characterize climate change over CDZ. 109. Climate change scenarios for the CDZ indicate that higher temperatures and heavy rainfall events may increase in the future. Regarding cyclones, the studies thus far have not shown any increase in frequencies. These results are consistent with findings of similar climate modeling studies undertaken earlier with slightly different climate modeling set up by Annamalai et al. (2011). 110. Datasets of climate change scenario results for the CDZ have been prepared for distribution. These data have been processed into formats that are easy to use in studies linked to hydrology and hydrodynamics. 111. Using climate modeling results and interpreting them to specific locations is highly uncertain. The climate change scenario results therefore need to be interpreted with caution. In this study bias correction has been used to improve rainfall simulation of one of the RCMs. Climate impact studies must possibly use a range of model results, rather than a single one. 112. The inherent complexities of the climate system makes it difficult to make confident estimates about rainfall changes at localize scales such as the CDZ. What the current modeling results provide are a plausible set of future outcomes. Robust water management practices to address current rainfall variability constitute effective steps in the direction towards building adaptive capacities. This will enable benefits from better resilience today and more effective adaptation in future. 113. Institutional systems to gather, analyze and take decisions on climate information getting available at seasonal and weather scale will enable better response to variable and changing climate. ______

64 8. References ADB, 2011, TN 7417 – IND: Support for the National Action Plan on Climate Change. Support to the National Water Mission – Final Report, September 2011, Appendix 4 Cauvery Delta Sub- basin, pages 116, September 2011. Annamalai, H., Prasanna, V., Hafner, J., Nagothu, U. and Kitterod, N. O., Modeling the current and future climate over the Cauvery river basin. In Sustainable Rice Production in a Warmer Planet, Macmillan Publishers, New Delhi, 2011, pp. 59–81. Annamalai H, , M. Mehari and K.R. Sperber (2012) A recipe for ENSO-monsoon diagnostics in CMIP5 models J Clim H. Annamalai, 2013: Quantifying climate model uncertainties in projecting regional rainfall and its variations over South Asia. J. Climate (in preparation) Chaturvedi, R. K., Jaideep, J., M. Jayaraman, G. Bala and N. H. Ravindranath (2012) Multi- model climate change projections for India under representative concentration pathways, Current Science, Vol. 103 (7), 10 October, 2012 Geethalakshmi, V., A. Lakshmanan, D. Rajalakshmi, R. Jagannathan, Gummidi Sridhar, A. P. Ramraj, K. Bhuvaneswari, L. Gurusamy and R. Anbhazhagan, 2011, Climate change impact assessment and adaptation strategies to sustain rice production in Cauvery basin of Tamil Nadu. Current Sci., Vol. 101(3): 342-347. Gopalakrishnan, K. S. and M. Sambasiva Rao, Environmental Assessment and Monitoring. In: Geography and Environment: Issues and Challenges, Prof. S. L. Kayastha Felicitation Volume, Eds. H. H. Singh et al., pages 289-298, Concept Publishing Company, New Delhi. 1986. Rajeevan, M. and J. Bhate, 2009: A high-resolution daily gridded rainfall dataset (1971-2005) for meso-scale meteorological studies, Current Science, 96(4), 558-562. Ramanathan AL, Subramanian V and Das BK, 1996. Sediment and Heavy metal accumulation in the Cauvery Basin. Environmental Geology 27(3): 155-163. Sharma, D., Das Gupta.A., and Babel, M.S. (2007). Spatial disaggregation of bias-corrected GCM precipitation for improved hydrologic simulation: Ping River Basin, Thailand. Hydrology and Earth System Sciences, 11, 1373-1390. Sperber, K. R., Annamalai, H, I.-S. Kang, A. Kitoh, A.Moise, A. Turner, B. Wang and T. Zhou (2012) The Asian Summer Monsoon: An Intercomparison of CMIP5 vs.CMIP3 Simulations of the Late 20th Century. Climate Dynamics. Stowaseer, M., H. Annamalai and J. Hafner 2009: Response of the South Asian monsoon to global warming: Mean and synoptic systems. J. Climate, 22, 1014–1036 TNAU, 2007, Cauvery Delta Zone – Status paper, Directorate of Research, Tamil Nadu Agricultural University, March, 2007, pages 88. (web link accessed on 1 June 2013, (http://www.tnau.ac.in/dr/zonepdf/CauveryDeltaZone.pdf) . Turner AG, Annamalai H (2012) Climate change and the south Asian summer monsoon. Nature Climate Change 2:1-9. Van Pelt, S. C., Kabat, P., Ter Maat, H. W., Van den Hurk, B. J. J. M., & Weerts, A. H. (2009). Discharge simulations performed with a hydrological model using bias corrected regional climate model input. Hydrology and Earth System Sciences, 13(12), 2387. Weibull, W., 1951 "A statistical distribution function of wide applicability."Journal of applied mechanics 18(3): 293-297. Yatagai, Akiyo, Kenji Kamiguchi, Osamu Arakawa, Atsushi Hamada, Natsuko Yasutomi, Akio Kitoh, 2012, APHRODITE (Asian Precipitation—Highly-Resolved Observational Data Integration Towards Evaluation of Water Resources): Constructing a Long-term Daily Gridded Precipitation

65 Dataset for Asia Based on a Dense Network of Rain Gauges, Bull. Amer. Meteor. Soc., 93, 1401-1415.

66 Annex I I. Inventory of Observed data sets

1. Observed climate data from meteorological and hydrological stations, gridded data sets of rainfall and temperature available from both national and global sources; and reanalysis data sets are proposed to be used to characterize the present climate of the Cauvery Basin area.

Station Observations 2. Climate data relevant to the project are available from meteorological and hydrological stations located in the Cauvery Basin that are being maintained by the India Meteorological Department (IMD), Central Water Commission, Tamil Nadu State Irrigation Department, Public Works Department (PWD), Department of Agriculture, Indian Space Research Organization and other private sector agencies. However, many of these observation stations were set up during different times and do not have long series of data (> 30 years) that are suitable for climatological trend analysis. Available details from, IMD, CWC and PWD stations have been compiled essentially with intention to scope out the observed data available. This scoping is more indicative rather than comprehensive as there are a many agencies involved in collecting short term and long-term climate and water related data. 3. The IMD has about 16 meteorological stations located in the Cauvery basin of Tamil Nadu (recording temperature, humidity, wind speed and rainfall) that are listed in Table 0-1.

Table 0-1 List of IMD Meteorological Stations located in the Cauvery Basin in Tamil Nadu

IMD Stations Districts 1. Peelamedu 2. Cuddalore Cuddalore 3. PortNovo Cuddalore 4. 5. Nagapattinam Nagapattinam 6. Vedaranyam Nagapattinam 7. Kudumiamalai Pudukottai 8. MetturDam Salem 9. Salem Salem 10. Adiramapattanam Thanjavur 11. Thanjavur Thanjavur 12. The Nilgiris 13. The Nilgiris 14. Trichy Trichy 15. Ariyalur 16. K.Paramathi

67

4. Some of the data for the IMD stations mentioned in Table 0-2, particularly temperature and rainfall is available from other projects. This is covers a limited period of 1971-2003. A preliminary quality control analysis of this daily data shows that we may be able to use only some of these stations for our long-term analysis. Significant lengths of missing data make it difficult to use all the stations for the list of available stations. Table 0-2 shows that for the 1971-2003 period the percentage of missing daily data, only ten stations are with less than 10% missing data and there are five stations with less than 5% missing data.

Table 0-2 Data availability for IMD stations - stations with less than 10% missing data are in bold.

S. No. Station Name Data missing (%) 1. Adiramapatinam 3.6 2. Ariyalur 42.7 3. Conoor 53.0 4. Cuddalore 0.5 5. Kodaikanal 3.0

6. K. Paramathi 29.5 7. Kudumiamalai 33.1 8. Metturdam 6.3

9. Nagapattinam 6.6 10. Ooty 8.1 11. Peelamedu 3.0

12. Portnovo 15.4 13. Salem 3.3 14. Tanjavur 44.7 15. Trichy 3.0 16. Vedaranyam 9.8

5. There are in all 233 existing rain-gauge stations (Reporting to India Meteorological Department under the District Rainfall Monitoring Scheme (DRMS)) inside the Cauvery basin. These rain-gauge stations are more or less uniformly distributed over the entire basin and their number is fairly adequate to characterize the current rainfall variability over the basin. Annex I provides a list of these rainfall monitoring stations from the IMD DRMS network located in the TN Cauvery Basin. 6. At 33 sites, gauge & discharge observation are made.4 There are 32 Water Quality observation stations and Sediment observations are made at 15 of these stations. Some of these data sets are expected to be available from the nodal officer in the implementing agency. The PWD also maintains a “State Ground and Surface Water

4 CWC, 2012, Integrated Hydrological Data Book – Non-classified river basins, Hydrological Data Directorate, Information Systems Organization, Water Planning & Projects Wing, Central Water Commission, New Delhi, India, March 2012, pp. 669.

68 Resources Data Centre. Public Works Department. Government of Tamil Nadu, India” from where additional data sets useful to IWRM could be accessed. 7. In response ADB request dated 20.7.2012, PWD Secretariat, have sent data for the Cauvery Basin obtained from the Water Resources Department, Tamil Nadu Government, Chennai. A summary of the furnished data is included at the end of this Annex. This data however could not be used in our climate analysis for CDZ because of its relatively shorter length and many missing values. 8. In addition to the network of regular meteorological stations there are newer networks of Automatic Weather Stations (AWS) being setup. Since 2009 the Department of Agriculture in collaboration with the Tamil Nadu Agricultural University (TNAU) has established 224 AWS5, one each at block level. Additional 161 are being established to cover the entire 385 Blocks. The AWS data is available from TNAU at a sub-daily scale and may be useful for hydrodynamic modelling studies. For climate analysis this data may be of limited as it is available only for a few years.

Gridded Datasets 9. Available high-resolution (about 50 – 100 km) gridded rainfall observations from two independent sources, namely the India Meteorological Department (IMD; Rajeevan et al. 2009) and APHORDITE (Yagati et al. 2012) will be used. Preliminary analysis indicates that the annual cycle of rainfall over the study area is realistically represented in both the products. Also, the year-to-year rainfall variations, particularly the flood and drought monsoons are robust between the two products. Global gridded rainfall data sets from the Global Precipitation Climatology Project (GPCP) have been used in the preliminary analysis. 10. For surface temperature both IMD’s gridded data and CRU (Climatic Research Unit, University of East Anglia, UK) data sets available at 100 km resolution and below are identified for use.

Reanalysis Datasets 11. To understand the circulation features, daily gridded data from (European Centre for Medium-range Weather Forecasting (ECMWF) Re Analysis) ERA-Interim reanalysis products will be examined. Consistent with the observations, the reanalysis products capture the associated circulations features during flood and drought monsoon years. From the reanalysis, the daily variables such as winds, sea level pressure etc. will be examined for the genesis and amplification of tropical cyclones that pass over the study area. The data rich period of 1979-2012 is chosen for all diagnostics.

Cyclonic Storms tracks data 12. IMD has brought out an electronic version of the atlas “Tracks of Storms and Depressions in the Bay of Bengal and the Arabian Sea”, published by IMD in the years 1964, 1979 and 1996. The electronic version called Cyclone e-Atlas –is an integrated package providing substantial information on the Tracks of Cyclones and Depressions that occurred over Indian Seas during the period 1891–2007. This data set was used in conjunction with reanalysis and published research papers to examine the climatology of the tropical cyclones relevant to the TNCDZ. Figure 0-1

5 Government of Tamil Nadu (GoTN), 2012, Agriculture Department Policy Note.5 – Agriculture 2012 – 2013, S. Damodaran, Minister for Agriculture, 2012.

69 illustrates the tropical cyclones tracks extracted from the website version of the atlas for the period 1971-2011.

Figure 0-1 Tropical Cyclones (Cyclonic Storms and Severe Cyclonic Storms) landfalls in the vicinity of the Cauvery Delta during the period 1971-2011 (Source: IMD Cyclone e-atlas)

70

LIST OF IMD DISTRICT RAINFALL MONIOTRING SCHEME (DRMS) STATIONS

SL. No DISTRICT STATION NAME Class Latitude Longitude Elevation Catchment Deg Min. Deg. Min. (m) 1 COIMBATORE 10 39 77 1 270 101 2 COIMBATORE UDUMALPET 10 35 77 15 368 302 3 COIMBATORE PERIYANAICKAMPAL 11 9 76 57 436 302 4 COIMBATORE METTUPALAYAM 11 18 76 57 326 302 5 COIMBATORE COIMBATORE OBSY 11 0 76 58 409 302 6 COIMBATORE 11 12 77 17 324 302 7 COIMBATORE SULUR 11 2 77 8 329 302 8 COIMBATORE ANNUR 11 14 77 7 102 302 9 COIMBATORE 11 6 77 21 299 302 10 COIMBATORE PEELAMEDU/COI.AP OBSY 11 2 77 3 399 302 11 COIMBATORE HOSSANUR COM.UNV 12 CUDDALORE KATTUMANNARKOIL 11 17 79 33 11 302 13 CUDDALORE 11 46 79 34 24 303 14 CUDDALORE CUDDALORE OBSY 11 46 79 46 12 303 15 CUDDALORE PORTO NOVO 11 30 79 45 303 16 CUDDALORE 11 24 79 42 4 303 17 CUDDALORE SRIMUSHNAM 11 24 79 25 59 303 18 CUDDALORE VRIDDHACHALAM 11 30 79 20 29 303 19 CUDDALORE VRIDDHACHALAM AN 11 29 79 25 43 303 20 CUDDALORE TOLUDUR/TOZHUDUR 11 25 79 0 88 303 21 CUDDALORE P.PETAI/PORTO NO OBSY 11 30 79 46 3 303 22 DINDIGUL NILAKOTTAI 10 10 77 51 276 301 23 DINDIGUL KODAIKANAL OBSY 10 14 77 28 2343 301 24 DINDIGUL NATHAM 10 14 78 8 247 301 25 DINDIGUL DINDIGUL 10 21 77 58 280 302

71 26 DINDIGUL VEDASANDUR REV 10 31 77 57 76 302 27 DINDIGUL CHATRAPATTI 10 28 77 37 323 302 28 DINDIGUL 10 27 77 31 321 302 29 10 44 77 32 274 302 30 ERODE MULANUR 10 47 77 42 220 302 31 ERODE SATYAMANGALAM 11 30 77 15 258 302 32 ERODE GOBICHETTIPALAYA 11 28 77 26 213 302 33 ERODE BHAVANI 11 27 77 43 166 302 34 ERODE ERODE 11 21 77 43 302 35 ERODE PERUNDURAI 11 17 77 35 294 302 36 ERODE KODUMUDI 11 5 77 52 147 302 37 ERODE 11 0 77 34 318 302 38 ERODE BHAVANISAGAR 11 27 77 41 357 302 39 KARUR ARAVAKURICHI 10 46 77 55 198 302 40 KARUR 10 55 78 25 90 302 41 KARUR KARUR REV 10 58 78 5 119 302 42 KARUR MAYANUR 10 56 78 15 104 302 43 KARUR K. PARAMATHI OBSY 10 57 78 5 181 302 44 KARUR KADAVYR REV 10 39 78 14 119 302 45 KARUR PANCHAPATTI REV 10 49 78 8 134 302 46 KARUR THOGAMALAI REV 47 NAGAPATTINAM NAGAPATTINAM OBSY 10 46 79 51 9 301 48 NAGAPATTINAM VEDARANNIYAM 10 22 79 50 4 301 49 NAGAPATTINAM VEDARANNIYAM OBSY 10 22 79 51 4 301 50 NAGAPATTINAM SIRKALI 11 15 79 44 4 302 51 NAGAPATTINAM M.THURAI/MAYURAM 11 6 79 39 13 302 52 11 23 77 54 268 302 53 NAMAKKAL KUMARAPALAYAM REV 11 26 77 42 187 302 54 NAMAKKAL 11 27 78 11 274 302

72 55 NAMAKKAL NAMAKKAL 11 13 78 10 261 302 56 NAMAKKAL PARAMATHI 11 10 78 2 125 302 57 NAMAKKAL SENDAMANGALAM 11 17 78 14 261 302 58 NAMAKKAL MANGALAPURAM REV 11 20 78 33 302 59 NILGIRIS GUDALUR 11 30 76 30 1052 101 60 NILGIRIS NADUVATTAM 11 28 76 33 1950 302 61 NILGIRIS OOTACAMUND OBSY 11 24 76 44 2249 302 62 NILGIRIS KUNDHA %KAILKUND 11 16 76 39 1631 302 63 NILGIRIS KETTY 11 23 76 46 2042 302 64 NILGIRIS COONOOR OBSY 11 21 76 48 1745 302 65 NILGIRIS KOTHAGIRI 11 26 76 52 1908 302 66 VEMBAVUR REV 10 29 78 34 840 302 67 PERAMBALUR PERUMBALUR 11 14 78 52 840 302 68 PERAMBALUR JAYAMKONDAM REV 11 13 79 22 245 302 69 PERAMBALUR ARIYALUR OBSY 11 0 79 0 75 302 70 PUDUKOTTAI ARANTANGI 10 11 78 59 45 301 71 PUDUKOTTAI 10 23 78 49 197 301 72 PUDUKOTTAI TIRUMAYAM 10 15 78 45 93 301 73 PUDUKOTTAI PERUNGALUR 10 29 78 55 80 301 74 PUDUKOTTAI ALANGUDI 10 22 78 59 45 301 75 PUDUKOTTAI ARIMALAM AGRI 10 15 78 54 301 76 PUDUKOTTAI KARAMBAKKUDI 10 27 79 9 45 301 77 PUDUKOTTAI ILUPPUR 10 31 78 37 132 302 78 PUDUKOTTAI KULATTUR/KEERANU R 10 42 78 33 37 302 79 PUDUKOTTAI VIRALIMALAI 10 36 78 33 179 302 80 PUDUKOTTAI KANDARVAKOTTAI REV 81 SALEM SANKARIDURG 11 28 77 52 268 302 82 SALEM OMALUR 11 45 78 3 65 302 83 SALEM SALEM OBSY 11 39 78 10 278 302

73 84 SALEM YERCAUD 11 47 78 13 1402 302 85 SALEM THAMMAMPATTY 11 27 78 30 240 303 86 SALEM 11 36 78 37 226 303 87 SALEM VAZHAPADI REV 11 40 78 24 301 303 88 TANJAVUR AGRI 10 26 79 19 23 301 89 TANJAVUR ADIRAMAPATNAM OBSY 10 20 79 23 6 301 90 TANJAVUR ORATHANADU ANI.H 10 35 79 15 38 301 91 TANJAVUR TIRUKKATTUPALLI 10 51 78 58 54 302 92 TANJAVUR KUMBHAKONAM 10 58 79 22 25 302 93 TANJAVUR PAPANASAM 10 55 79 16 31 302 94 TANJAVUR TIRUVAIYARU 10 53 79 6 42 302 95 TANJAVUR VALLAM 10 43 79 4 46 302 96 TANJAVUR TANJAVUR OBSY 10 47 79 8 68 302 97 TANJAVUR THIRUVIDALMARTHU REV 98 TIRUCHIRAPALLI MUSIRI 10 57 78 27 289 302 99 TIRUCHIRAPALLI LALGUDI REV 10 52 78 50 192 302 100 TIRUCHIRAPALLI TIRUCHI.PALLI(A) OBSY 10 46 78 43 88 302 101 TIRUCHIRAPALLI 10 36 78 25 154 302 102 TIRUCHIRAPALLI MARUNGAPURI REV 10 26 78 24 156 302 103 TIRUCHIRAPALLI PULLAMBADI 10 58 78 55 206 302 104 TIRUCHIRAPALLI TIRUCHI TOWN OBSY 10 46 78 43 82 302 105 TIRUCHIRAPALLI TURAIYUR 11 9 78 36 137 302 106 TIRUCHIRAPALLI CHETTIKULAM 11 9 78 48 840 302 107 TIRUCHIRAPALLI THATHAIENGARPET 11 8 78 27 289 302 108 VALANGIMAN 10 52 79 23 24 301 109 THIRUVARUR KODAVASAL 10 51 79 29 26 301 110 THIRUVARUR NANNILAM 10 53 79 37 18 301 111 THIRUVARUR TIRUVARUR 10 46 79 39 14 301 112 THIRUVARUR TIRUTHURAIPNDI 10 32 79 39 11 301

74 113 THIRUVARUR 10 24 79 30 5 301 114 THIRUVARUR 10 40 79 27 16 301

115 THIRUVARUR NEEDAMANGALAM 10 46 79 26 21 301

75 Summary of Data provided by PWD, Tamil Nadu Meteorological Data at Monthly timescales Particulars Station District Period of data Latitude Longitude Altitude (m) 1. Rainfall (mm) Koradachery Nagapattinam 1973 -2009 100 46’ 10” 790 29’ 05” 61.687 2. Tmax (oC) 3. Tmin (oC) 4. Evaporation(mm) 5. Wind velocity (KMPH) 6. Relative humidity (%) 7. Sunshine hrs/day 1. Rainfall (mm) Kurunkulam Thanjavur 1981 -2010 100 42’ 0” 790 05’ 30” 61.66 2. Tmax (oC) 3. Tmin (oC) 4. Evaporation (mm) 5. Wind velocity (KMPH) 6. Relative humidity (%) 7. Sunshine hrs/day 1. Rainfall (mm) Thanjavur Thanjavur 1981 -2010 100 46’ 22” 760 08’ 09” 71 2. Tmax (oC) 3. Tmin (oC) 4. Evaporation (mm) 5. Wind velocity (KMPH) 6. Relative humidity (%) 7. Sunshine hrs/day

Rainfall Data alone recorded at many stations at Monthly timescales Station District Period of data Latitude Longitude Kollidam Nagapattinam 1989 -2011 110 19’ 46” 790 43’ 17” Manalmedu Nagapattinam 1983 -2011 110 12’ 02” 790 36’ 21” Myladuthurai Nagapattinam 1981 -2011 110 05’ 47” 790 39’ 11” Nagapattinam Nagapattinam 1981 -2011 104759 795019 Nagapattinam 1985 -2011 111359 794425 Tharangambadi Nagapattinam 1998 -2011 110138 795019 Thalaignaiyuru Nagapattinam 1981 -2011 103304 794558 Thirupoondi Nagapattinam 1981 -2011 103736 794840 Vedharanyam Nagapattinam 1986 -2011 102236 795107

76 Station District Period of data Latitude Longitude Adiramapattinam Thanjavur 1992 -2011 102026 792306 Authorial Thanjavur 1981 -2011 110026 792839 Boodalur Thanjavur 1989 -2011 104705 785848 Grand Anicut Thanjavur 1981 -2011 104942 784906 Thanjavur 1981 -2011 105806 792303 Kurunkulam Thanjavur 1984 -2011 103923 790553 Lower Anicut Thanjavur 1983 -2011 110843 792700 Manjalar Head Thanjavur 1989 -2011 110146 793151 NeithavasalThenpadi Thanjavur 1981 -2011 104151 791835 Orathanadu Thanjavur 1983 -2011 103703 791523 Papanasam Thanjavur 1981 -2011 105530 791614 Pattukottai Thanjavur 1981 -2011 102503 792003 Peravoorani Thanjavur 1981 -2011 101718 791205 Rajagiri Thanjavur 2000 -2011 105525 791439 Thanjavur Thanjavur 1989 -2011 104622 790809 Thiruvaiyaru Thanjavur 1981 -2011 105228 790617 Tirukattupali Thanjavur 1981 -2011 105121 785712 Vallam Thanjavur 1981 -2011 104307 790344 Vettikkadu Thanjavur 1984 -2011 103416 791142 Koradachery Thiruvarur 1989 -2011 104624 792921 Koraiyar Head Thiruvarur 1982 -2011 104724 792353 Kudavasal Thiruvarur 1989 -2011 105105 792847 Madhukkur Thiruvarur 1981 -2011 102929 792352 Mannargudi Thiruvarur 1987 -2011 104005 792644 Mullaiar Head Thiruvarur 1984 -2011 103918 793035 Muthupettai Thiruvarur 1993 -2011 102345 792923 Nannilam Thiruvarur 1986 -2011 115253 793634 Needamangalam Thiruvarur 1981 -2011 104610 792456 Pandayar Head Thiruvarur 1987 -2011 104543 792945 Thiruvarur Thiruvarur 1981 -2011 104627 793808 Tiruturaipoondi Thiruvarur 1981 -2011 103211 793805 Valangaimaan Thiruvarur 1998 -2011 105323 792331

77 Water Quality data Particulars Station District Period of data TDS, NO2+NO3, Ca, Mg, Na, K, Cl, SO4, CO3, In different villages of Thiruvarur 1982 -2011 HCO3, F, PH, EC, HAR, SAR, RSC Na (%) Thiruvarur district TDS, NO2+NO3, Ca, Mg, Na, K, Cl, SO4, CO3, In different villages of Thanjavur 1981 -2011 HCO3, F, PH, EC, HAR, SAR, RSC Na (%) Thanjavur district TDS, NO2+NO3, Ca, Mg, Na, K, Cl, SO4, CO3, In different villages of Nagapattinam 1982 -2011 HCO3, F, PH, EC, HAR, SAR, RSC Na (%)

Water Level data at Monthly timescale

Particulars Station District Period of data Water Level In different Taluks of Thiruvarur 1981 -2010 Thiruvarur district Water Level In different Taluks of Thanjavur 1981 -2011 Thanjavur district Water Level In different Taluks of Nagapattinam 1981 -2010 Nagapattinam district

Tide gauge data at Daily timescale

Particulars Station District Period of data Latitude Longitude Times and Heights of High and Nagapattinam 2009 -2011 10046’ 79051’ Low waters Thanjavur 1981 -2011

General Data for Nagapattinam, Thanjavur and Thiruvarur Particulars Station District Period of data Population - 2001 Nagapattinam 2001 Population - 2011 Thanjavur 2011 Land use pattern 1998 -2010

78 Annex II

INVENTORY OF DAILY DATA FROM CIMIP5 GCMs

Availability of Daily Data for period Availability of climate variables Historic RCP45 RCP85 RCP26 Tas Tas Pr Uas Vas Psl huss rsds Model H. Res.** al (Max) (Min) 1 BCC-CSM1-1 128 x 64 1850- 2006- 2006- 2006-         2.85  x 2.81 2012 2300 2300 2300 2 CAN-ESM2 128 x 64 1850- 2006- 2006- 2006-

2.85 x 2.81 2005 2300 2100 2300 3 CCSM4 192 x 288 1850- 2006- 2006-     0.94  x 1.25 2005 2299 2300 4 CNRM-CM5 255 x128 1960- 2016- 2016- 2016-         1.4 x 1.4 2005 2100 2100 2100 5 CSIRO-MK3-6-0 192x 96 1850- 2006-         1.89  x 1.87 2005 2300 6 INMCM4 180 x 120 1850- 2006-         1.5 x 2.0 2005 2100 7 IPSL-CM5B-LR 144 x143 1850- 2006- 2006-         1.89  x 3.75 2005 2100 2100 8 IPSL-CM5A-MR 144 x143 1850- 2006- 2006- 2006-         1.89 x 3.75 2005 2100 2100 2100 9 IPSL-CM5A-LR 144 x143 1850- 2006- 2006- 2006-         1.89  x 3.75 2005 2300 2300 2300 10 MPI-ESM-LR 192 x 96 1850- 2006- 2006- 2006-        1.86 x 1.87 2005 2300 2300 2300 11 MRI-CGCM3 320 x160 1860- 2016- 2016-         1.12  x 1.12 2005 2100 2100 12 MIROC-ESM-CHEM 128x 64 1850- 2006- 2006- 2006-         2.85 x 2.81 2005 2100 2100 2100 13 MIROC5 256x 128 1850- 2006- 2006- 2006-         1.41  x 1.4 2012 2100 2100 2100 14 MIROC-ESM 128x 64 1850- 2006- 2006-         2.79 x 2.81 2005 2300 2100 15 HadGEM2-CC 192 x144 1860- 2006- 2006-         1.25  x 1.88 2005 2100 2100 16 HadGEM2-ES 192 x144 1860- 2006- 2006- 2006-         1.25 x 1.88 2005 2099 2299 2299

79 Availability of Daily Data for period Availability of climate variables Historic RCP45 RCP85 RCP26 Tas Tas Pr Uas Vas Psl huss rsds Model H. Res.** al (Max) (Min) 17 GFDL-ESM2G 144 x 90 1861- 2006- 2006- 2006-         2.0  x 2.5 2005 2100 2100 2100 18 GFDL-ESM2M 144 x 90 1861- 2006- 2006- 2006-         2.0 x 2.5 2005 2100 2100 2100 19 NOR-ESM1-M 144 x 96 1850- 2006- 2006- 2006-       1.89  x 2.5 2005 2300 2100 2100

Tas (Max.): Daily Maximum Near-Surface Air Temperature (K) Tas (Min.): Daily Minimum Near-Surface Air Temperature (K) Pr: Precipitation (kg m-2 s-1) Uas: Eastward Near-surface wind (m s-1) Vas: Northward near-surface wind (m s-1 ) Psl: Pressure at sea level (Pa) huss: Near surface specific humidity (l) rsds: Surface Downwelling Shortwave Radiation (W m-2)

**H. Res: Horizontal Resolution, first figure represent number of grid boxes along longitudes x latitudes and the second gives the approximate dimensions of the grid box in degrees. RCP: Representative concentration pathways

80 Annex III

Description of data extracted for CDZ 1. GCMs and RCMs outputs in netcdf format has been processed and extracted in to text format (csv) which can be viewed and analyzed easily in most of the packages by the users. 2. Datasets were organized under four major sections (four main folders) 1. RAW (grid wise GCM and RCM output in text format) 2. Zone wise (GCM and RCM outputs averaged for two major zones Tamil Nadu and CDZ) 3. Station wise (RCM outputs averaged for seven selected IMD station location) 4. Derived (Rainfall return period, Probability of rainfall occurrence and rainfall depths for seven IMD station location), and details of each of the sections are explained in this section.

Figure GCM & RCM data extraction – Averaged grid details.

3. The above figure shows the details of averaged points for the zones and stations. To create representative climate baseline and future scenario series for each station location points from four surrounding grid boxes were averaged. This approach provides a better representation of the climate over the location and also avoids errors that may arise due to artifacts of model parameterization schemes that could manifest at a single model grid cell. The spatial extent averaged for stations are as follows Adiramapattinam (79.2-79.6E, 10.2-10.6N), Cuddalore (79.4-79.8E, 11.7- 12.1N), Mettur (77.7-78.1E, 11.7-12.1N), Nagapattinam (79.7-79.9E, 10.7-11.1N),

81 Salem (77.9-78.3E, 11.4-11.8N), Trichy (78.4-78.8E, 10.7-11.1N), Vedaranyam (79.7-79.9E, 10.2-10.6N).

Data structure of distributed climate datasets The folder “RAW” has grid wise GCMs and RCMs projection outputs both baseline and projection. The details of the time slices, file structure and organization are presented in Table-1.

Table -1 : Grid wise GCM and RCM outputs

114. Sub Sub Sub folder 3 Description folder 1 folder 2 GCM 1.MIROC5 1.Baseline (1981 Time scale: Daily time series without 2.CCSM4 to 2000) leap 2.Projection No. of files: Each folder has 35 files (one RCP45 (2006 to file for each grid in whole TN domain 76E to 2100) 81E, 8N to 13.5N). Two GCMS (2), Three 3.Projection Categories – baseline (1) and projections RCP85 (2006 to (2). Total no. of files for 35 grids X 2 GCMs 2100) X 3 Categories is 210 files File format: Comma separated CSV format. File name: File names are organized as below LON-LAT_Base.csv LON-LAT_RCP45.csv LON-LAT_RCP85.csv For example: File names for grid location 76.25EX8.01N are as follows 76.25- 8.01_BASE.csv (baseline), 76.25- 8.01_RCP45.csv (projection rcp45), 76.25- 8.01_RCP85.csv (projection rcp85) Header: Each file for MIROC5 GCM has ten column year, day, rainfall (mm/day), maximum temperature (◦C), minimum temperature (◦C), Surface Downwelling Shortwave Radiation (rsds in wm-2), ), Sea Level Pressure (psl in Pa), Near-Surface Specific Humidity (huss in kg/kg), Eastward Near-Surface Wind (uas in ms-1), Northward Near-Surface Wind (vas in ms-1)

Each file for CCSM4 GCM has six column year, day, rainfall (mm/day), maximum temperature (◦C), minimum temperature (◦C), Sea Level Pressure (psl in Pa) RCM 1.GFDL 1.Baseline (1981 Time scale: Daily time series with leap 2.CCSM4 to 2000 for GFDL No. of files: Every year has one folder, so and 1986 to 2005 baseline (20) and projections (30 for GFDL for CCSM4) and 20 for CCSM4). Each folder have 896 2.Projection files (one file for each grid covering whole (2021-2050 for TN domain 75E to 81.75E, 7N to 14.75N).

82 GFDL and 2081- GFDL (baseline 20 folders, projection 30 2100 for CCSM4) folders) has 44800 files. CCSM4 (baseline Files are arranged 20 folders, projection 20 folders) has 35840 year wise files. Hence there are 80640 files in total. separately, so File format: Comma separated CSV format. one year has one File name: File names are organized as folder. below CLIVAR-LAT-LON in each folder (year) For example: File names for grid location 75.00EX07.00N are as follows CLIVAR- 0700-7500 Header: Each file for GFDL and CCSM4 RCM has ten column year, day, Irradiance (MJ m-2), Min Temperature (oC), Max Temperature (oC), Early Morning VP (kPa), Mean Wind Speed (m s-1) , Precipitation (mm d-1), Dew point temp ( oC), Relative Humidity (%), BIAS CORRECTED GFDL Baseline Time scale: Daily time series with leap RCM Projection No. of files: Each folder has 495 files (one file for each grid covering whole TN and adjacent areas 75E to 80.25E, 8N to 14.75N). Two Categories – baseline (1) and projections (1). Total no. of files for 495 grids X 2 Categories is 980 files File format: Comma separated CSV format. File name: File names are organized as below LON-LAT.csv For example: File names for grid location 76.25EX8.00N are as follows 76.25- 8.00.csv Header: Each file for MIROC5 GCM has four column year, day, rainfall (mm/day)

Data sets produced for CDZ Table 2: Zonewise daily time series data for GCMs and RCMs

Sub folder 1 Sub Sub folder 3 Description folder 2 GCM 1.MIROC5 1.Baseline (1981 to Time scale: Daily time series without leap 2.CCSM4 2000) No. of files: Each folder has 2 files (one file for 2.Projection each TN zone and one for Cauvery delta zone). RCP45 (2006 to Two GCMS (2), Three Categories – baseline (1) 2100) and projections (2). Total no. of files for 2 zones X 3.Projection 2 GCMs X 3 Categories is 12 files RCP85 (2006 to File format: Comma separated CSV format. 2100) File name: File names are organized as below Zone_Base.csv Zone_RCP45.csv Zone_RCP85.csv TN zone file starts with TN and CDZ file starts with CDZ. For example: CDZ zone files in each model folder appears like CDZ_BASE.csv (baseline),

83 CDZ_RCP45 (projection, CDZ_RCP85 (projection. Header: Each file for GCMs has six column year, month, day, rainfall (mm/day), maximum temperature (◦C), minimum temperature (◦C)

RCM 1.GFDL 1.Baseline (1981 Time scale: Daily time series with leap 2.CCSM4 to 2000 for GFDL No. of files: Each folder has 2 files (one file for and 1986 to 2005 each TN zone and one for Cauvery delta zone). for CCSM4) Two RCMS (2), Two Categories – baseline (1) and 2.Projection projections (1). Total no. of files for 2 zones X 2 (2021-2050 for RCMs X 2 Categories is 8 files GFDL and 2081- File format: Comma separated CSV format. 2100 for CCSM4) File name: File names are organized as below Files are arranged Zone_Base.csv year wise Zone_Projection.csv separately, so one TN zone file starts with TN and CDZ file starts with year has one CDZ. folder. For example: CDZ zone files in each model folder appears like CDZ_BASE.csv (baseline), CDZ_Projection (projection). Header: Each file for RCMs has six column year, month, day, rainfall (mm/day), maximum temperature (◦C), minimum temperature (◦C) BIAS GFDL Baseline Time scale: Daily time series with leap CORRECTED Projection No. of files: Each folder has 2 files (one file for RCM each TN zone and one for Cauvery delta zone). This is available for GFDL RCM only. Total no. of files for 2 zones X 1 RCM X 2 Categories is 4 files File format: Comma separated CSV format. File name: File names are organized as below Zone_Base.csv Zone_Projection.csv TN zone file starts with TN and CDZ file starts with CDZ. For example: CDZ zone files appears like CDZ_BASE.csv (baseline), CDZ_Projection (projection). Header: Each file has four column year, month, day, rainfall (mm/day)

Datasets for station locations The extracted dataset for each RCM and category are organized station wise a file. The details are presented in Table-3

Table 3: Station wise daily time series data for RCMs

Sub folder 1 Sub Sub folder 3 Description folder 2 RCM 1.GFDL 1.Baseline (1981 Time scale: Daily time series with leap 2.CCSM4 to 2000 for GFDL No. of files: Each folder has 7 files (one file for

84 and 1986 to 2005 each station). Two RCMS (2), Two Categories – for CCSM4) baseline (1) and projections (1). Total no. of files 2.Projection for 7 station X 2 RCMs X 2 Categories is 28 files (2021-2050 for File format: Comma separated CSV format. GFDL and 2081- File name: File name start with three letter code in 2100 for CCSM4) the beginning of the file Adiramapattinam (ADI) , Files are arranged Cuddalore (CUD), Mettur (MET) , Nagapattinam year wise (NAG), Salem(SAL), Trichy(TRI), separately, so one Vedaranyam(VED). year has one For example: Adiramapatinam files in each model folder. folder appears like ADI_BASE.csv (baseline), CDZ_Projection (projection). Header: Each file for RCMs has six column year, month, day, rainfall (mm/day), maximum temperature (◦C), minimum temperature (◦C) BIAS GFDL Baseline Time scale: Daily time series with leap CORRECTED Projection No. of files: Each folder has 7 files (one file for RCM each station). RCM (1), Two Categories – baseline (1) and projections (1). Total no. of files for 7 station X 1 RCM X 2 Categories is 14 files File format: Comma separated CSV format. File name: File name start with three letter code in the beginning of the file Adiramapattinam (ADI) , Cuddalore (CUD), Mettur (MET) , Nagapattinam (NAG), Salem(SAL), Trichy(TRI), Vedaranyam(VED). For example: Adiramapatinam files appears like ADI_BASE.csv (baseline), CDZ_Projection (projection). Header: Each file has four column year, month, day, rainfall (mm/day)

Derived data 4. Accumulated rainfall depths for 01, 02, 05, 10 days has been estimated for IMD observatory record, extracted RCM (baseline and projections) for seven station daily time series data. Rainfall return period and probability of occurrence results has been tabulated in a single text file for all the data sources. The files are organized station wise for each data source. The details are presented in Table-4 Table 4: Derived datasets

Sub folder 1 Sub folder 2 Sub folder 3 Description RAINFALL 1.IMD 01DAY Time scale: Daily time series with leap DEPTH STATION 02DAY No. of files: Each folder has 7 files (one file for each 2.RCM GFDL 05DAY station). IMD,GCM,RCM(5), Four Categories BASE 10DAY (01,02,05&10DAY). Total no. of files for 7 station X 5 3.RCM GFDL GCM, RCM & IMD X 4 Categories is 140 files PROJECTION File format: Comma separated CSV format. 4.RCM File name: File name start with three letter code in CCSM4 the beginning of the file Adiramapattinam (ADI) , BASE Cuddalore (CUD), Mettur (MET) , Nagapattinam 5.RCM (NAG), Salem(SAL), Trichy(TRI), Vedaranyam(VED) CCSM4 and followed by the no. of days accumulated rainfall PROJECTION For example: Adiramapatinam files in each folder appears like ADI_01DAY.csv (01Day accumulated)

85 ADI_02DAY (02day accumulated), Header: Files under 01DAY has four column year, month, day, rainfall (mm/day) SNO (1st Column), DATE (2nd Column), RAINFALL in mm (3rd Column) Files under 02DAY, 05DAY, 10Day has the following header SNO (1st Column),START DATE (2nd Column), END DATE (3rd Column), RAINFALL in mm (4th Column) RETURN Return periods for 2, 5, 10, 20, 50, 100, 200 years for PERIOD datasets IMD STATION, RCM GFDL BASE, RCM GFDL PROJECTION, RCM CCSM4 BASE and RCM CCSM4 PROJECTION are tabulated in one text file. (ReturnPeriod.txt) PROBABILITY Probability of rainfall occurrence at 50%, 80%, 90% OF and 95% for three major seasons (MAM, JJAS, OCCURENCE OND) for datasets IMD STATION, RCM GFDL BASE, RCM GFDL PROJECTION, RCM CCSM4 BASE and RCM CCSM4 PROJECTION are tabulated in one text file. (ProbabilityofOccurence.txt)

File structure Table 5: File summary

SECTION SUB FOLDER -1 SUB SUB –FOLDER3 No. of FOLDER- 2 files RAW GCM MIROC5 BASELINE 35 PROJECTION 35 (RCP45) PROJECTION 35 (RCP85) CCSM4 BASELINE 35 PROJECTION 35 (RCP45) PROJECTION 35 (RCP85) RCM GFDL BASELINE (20 sub 17920 896 folders) grids in PROJECTION (30 26800 1 folder. sub folders) CCSM4 BASELINE (20 sub 17920 folders) PROJECTION (20 17920 sub folders) BIAS CORRECTED GFDL BASELINE 497 RCM PROJECTION 497 ZONEWISE GCM MIROC5 BASELINE 2 PROJECTION 2 (RCP45) PROJECTION 2 (RCP85) CCSM4 BASELINE 2 PROJECTION 2 (RCP45) PROJECTION 2

86 (RCP85) RCM GFDL BASELINE 2 PROJECTION 2 CCSM4 BASELINE 2 PROJECTION 2 BIAS CORRECTED GFDL BASELINE 2 RCM PROJECTION 2 STATIONWISE RCM GFDL BASELINE 7 PROJECTION 7 CCSM4 BASELINE 7 PROJECTION 7 BIAS CORRECTED GFDL BASELINE 7 RCM PROJECTION 7 DERIVED RAINFALL DEPTHS IMD STATION 4 SUB FOLDERS 28 7 files in (01DAY, 02DAYS, each 05DAYS, 10 DAYS) folder - 4 RCM GFDL 4 SUB FOLDERS 28 sub BASELINE (01DAY, 02DAYS, folders 05DAYS, 10 DAYS) X7 RCM GFDL 4 SUB FOLDERS 28 PROJECTION (01DAY, 02DAYS, 05DAYS, 10 DAYS) RCM CCSM4 4 SUB FOLDERS 28 BASELINE (01DAY, 02DAYS, 05DAYS, 10 DAYS) RCM CCSM4 4 SUB FOLDERS 28 PROJECTION (01DAY, 02DAYS, 05DAYS, 10 DAYS) RAINFALL RETURN 1 PERIOD PROBABILITY OF 1 RAINFALL OCCURENCE Total no. of files 81972

87