A Study on Regional Drought Analysis using Teleconnections and Suitable Global Climate Models for Telangana State

Abstract

Drought is a multifaceted natural climatic hazard that significantly affects both ecosystem and society. In comparison to other natural hazards, such as floods, storms, hurricanes, the spatial extent of droughts is usually much larger. Climate variability precipitates a heightened occurrence and severity of droughts on a global scale. Notably, India stands as one of the most drought-prone regions worldwide, experiencing drought events approximately every three years across distinct geographic areas, with a particularly profound impact observed in semi arid regions. As a result, it is important to investigate drought at regional and local scale with climatic conditions and its variations. Global Climate Models (GCMs) are the only models available for projecting climate systems at any timescale. Occurrence and distribution of drought characteristics can be analysed by using GCMs datasets. Understanding the dynamics of drought and its impacts in the context of climate change on a regional scale is therefore a vital area of research, so in this study the regional frequency analysis of droughts using suitable GCMs for Telangana state is carried out in view of recent catastrophic events. In the first section, selection of suitable GCMs of Precipitation (P), Maximum Temperature (Tmax), Average Temperature (Tavg) and Minimum Temperature (Tmin) entails the application of Standard Statistical Performance Metrics (SSPMetrics) over the period 1975-2005 in Telangana State, India. Skill Score (SS), Normalized Root Mean Square Deviation (NRMSD) and Correlation Coefficient (CC) SSPMetrics are utilized to evaluate 36 Coupled Model Intercomparison Project 5 (CMIP5) dataset models against observed data. Weights assigned to SSPMetrics are determined from entropy and sensitivity analysis. Compromise Programming (CP) is subsequently employed to rank the GCMs for each variable at individual grid point using distance measure method. The Group Decision-Making Approach (GDMA) is then applied to derive a combined ranking at each grid point. The ensemble climate models deemed suitable for each variable are identified as follows: for P, FGOALS-g2, CMCC-CMS and INMCM4.0; for Tmax, BCC-CSM1.1(m), CanESM2 and MIROC5; for Tavg, MIROC5, CNRM CM5 and BCC-CSM1.1(m); for Tmin, CanESM2, BCC-CSM1-1(m) and ACCESS 1.0 from historical simulations of CMIP5 GCMs. The computed net strength of each GCMs aligns with the ensemble model results. The determined ensemble GCMs are suitable for utilization in subsequent climate impact assessment studies that focus on precipitation, temperature or both such as studies on drought, flood, temperature extremes and other regional scale climatic phenomena. v In the second section of the study, regionalization of the study area and evaluating of drought indices, namely Standardized Precipitation Index (SPI), Standardized Precipitation Evapotranspiration Index (SPEI), Self-Calibrated Palmer Drought Severity Index (SC_PDSI) and their properties(Severity and Duration) are assessed. Regionalization process involved the analysis of hydro-meteorological data to discern homogeneous regions with akin characteristics. Application of Fuzzy-C-Means - Unsupervised classification is used for clustering and optimum number of clusters are identified as three for Telangana state. Subsequently, Drought indices are computed over a 12-month time scale for three identified homogenous regions from the year 1975 to 2017. The findings reveal that, in the context of the SPI, Region 1 exhibited severe drought period during August 2011 to September 2013, registering with -34.9 severity. The lengthiest drought episode in this region spanned from September 1984 to September 1987, encompassing a duration of 37 months. In the case of Region 2, exhibited severe and prolonged drought between June 2001 and August 2005, recording -44.43 severity and 51 months of duration. Region 3 experienced severe drought from August 1984 to August 1986, with a severity of -38.5, while the lengthiest drought event spanned from July 2014 to September 2016, encompassing 27 months of duration. SPEI, Region 1 exhibited severe and protracted drought event from August 2006 to September 2013, manifesting a severity of -81.07 and persisting for a duration of 86 months. For Region 2, experienced severe and extended drought spans during August 2006 to June 2013, registering -75.1 of severity and persisting for 83 months. Region 3 witnessed its most severe drought event during the period from March 2014 to December 2017, characterized by a severity of 68.02, while the lengthiest drought event occurred from March 2007 to May 2011, encompassing a duration of 51 months. For SC_PDSI, region 1, most severe and longest drought event occurred during September 2000 to June 2003 with severity -79.77 and duration 33 months; Region 2 exhibits its most severe and extended drought event, spanning from February 1984 to September 1985, characterized by -38.92 of severity and lasting for 19 months of duration. Conversely, Region 3 confronted an exceptionally severe drought event extending from May 1979 to June 1988, registering -227.75 of severity and persisting for 110 months of duration. Overall, these valuable insights of severity and duration of SPI, SPEI and SC_PDSI prove effective for analysing and assessing regional drought conditions. In the third section, the investigation delves into unravelling the impact of teleconnection on the relationship between drought indices (SPI, SPEI, SC_PDSI) and four prominent climatic indices: Southern Oscillation Index (SOI), Dipole Mode Index (DMI), Multivariate ENSO vi Index (MEI) and NINO3.4 are analysed for 1975-2017. Furthermore, interconnection between drought indices and climate indices is examined using Wavelet Coherence method. The results indicate that, drought pattern of DMI with SPI, SPEI and SC_PDSI is observed during 1984, 1985 and 1992. MEI and NINO 3.4 with SPI, SPEI and SC_PDSI during 1984 - 1986; SOI matched well during the year 1992 and 1993 with all the drought indices. Inter annual variability coherence for SPI with MEI is observed at 16 - 40 months (1982-1994 and 1995 2017 and SOI inter annual coherence is evident between 1975-1990 and 1995-2017; whereas Nino 3.4 intermittency is noticed at 1978-1992 and 2002-2015. Coherence is demoted with all the climate indices in the case of SPEI only SOI exhibited a highly significant influence at 14 to 40 months between 2002- 2014. Whereas significant coherence is smaller for SC_PDSI with DMI, MEI and NINO 3.4. SOI and MEI has significant coherence with SPI followed by SPEI and SC_PDSI compared to other climate indices. This reliable and robust quantitively results helps to understand relation between the climate and drought indices and new insights for further drought investigation. In section four, multivariate frequency analysis of Severity-Duration-Frequency (SDF) and Severity-Area-Frequency (SAF) curves are developed with SPI and SPEI at 12- month time scale at a threshold of -0.8. for three regions for time span of 1975-2017 (observed), 1975 2005, 2006-2035, 2036-2065, and 2066-2095 (projected datasets) using drought characteristics. The temporal evolution of drought entails a comprehensive examination of drought attributes through the analysis of SPI and SPEI within three homogeneous regions. This scrutiny encompasses both observed data and four distinct projected datasets. The objective is to discern the nuanced characteristics of drought over time, thereby contributing valuable insights to the understanding of drought dynamics. Increase of number of droughts are noticed in all regions for future periods compared to observed period of IMD. Mean interarrival time between droughts of SPI and SPEI is found to be maximum for Region 1 and Region 3 in historic period latter it is decreased further in projected periods. Maximum severity is showing increasing trend in all regions during 2036-3065 and 2066-2095 future periods. The incidence of moderate drought events exhibits an elevated frequency during both historical and anticipated future periods across all delineated regions i.e., nearly 30% of the droughts are moderate droughts for all the regions. The best fit copula for three regions is: for SPI – Clayton(region1), Gumbel((region2) and Frank(region3). SPEI, Gumbel(region1), Frank(region2), and Frank(region3). In later part of twenty-first century mean interarrival time is observed to be reducing and number of droughts are observed to be increasing for both SPI vii and SPEI. A possibility of high number of droughts with less mean arrival time is expected with high severity and duration in the future at Region 1 followed by Region 3 and Region 2 respectively. Projected drought SDF curves represent highest severity as noticed for Region 1 and duration for Region 3 for SPI whereas for SPEI highest severity and duration is noticed for Region 1. All the curves rise convex upwards for region 1 & 2 and concave upwards for Region 3 which represents increase in severity with increase in duration for SPI and SPEI. Projected SAF curves depict drought severity and its spatial extent in relation to the drought return period, elucidating the spatial and recurrent patterns inherent in drought occurrences. These curves prove instrumental in examining the anticipated annual severity of droughts in the future, encompassing the associated percentage coverage of the affected area. Moreover, SAF curves facilitate the comparative analysis of historical droughts against those projected from future climate scenarios derived from GCMs outputs. The temporal evolution reveals a discernible escalation in drought severity associated with varying durations over time. Projections indicate that drought hazard is poised to reach its zenith during the periods of 2036 2065 and 2066-2095, surpassing levels observed in other analysed epochs. Leveraging information gleaned from SDF and SAF curves concerning drought severity, duration, percentage coverage of the area and return period, allows for the precise calculation of drought severity within designated regions. This information proves valuable for addressing agricultural demand and formulating optimal crop management strategies. The results and findings based on the application of statistical techniques in this study gives insight to use suitable GCMs for drought related climate impact studies and this study offers a view on potential drought condition in Telangana state, India.