A Spatio-Temporal Approach to Evolution of Spatial Homogeneity of Monsoon Extremes Over India

• Activity Number: 247
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Subhomoy Ghosh* and Buddhananda Banerjee
• Companies: National Institute of Standards and Technology and Indian Institute of Science Education and Research
• Keywords: spatio-temporal cluster ; hidden markov model ; GEV ; change point detection

Abstract

There is a growing literature on climate change w.r.t. severe changes in extreme events related to the Monsoon rainfall over India. Evidence of increasing spatial variability of extreme rainfall triggers several aspects such as: how are the spatially homogeneous regions or clusters changing w.r.t. extreme rainfall over time? Are there detectable evidences of changes in cluster behavior? Can we arrive at a forecast? Purpose of this study is twofold: firstly, we introduce a novel discrete-time finite state-space hidden Markov models with non-constant transition matrix depending on a set of exogenous covariates including cluster level temporal information. We present a space-time varying dynamic with conditionally independent Generalized Extreme value (GEV) distribution as modeling extremes. Secondly, we posit a model based spatio-temporal clustering algorithm based on the latter model. We illustrate this technique based on Monsoon rainfall data and show that the cluster characteristics are significantly changing along with the number of clusters. We also propose a change point detection method for the non-constant transition matrices.