Complexity of rainfall dynamics in India in the context of climate change

Global climate change has become one of the major environmental issues today. Climate change impacts rainfall (and other hydroclimatic processes) in many ways, including its temporal and spatial variability. Hence, understanding the impact of climate change on rainfall is important to devise and undertake more effective and efficient adaptation and management strategies. The present study attempts to determine the temporal dynamic complexity of monthly historical and future rainfall in India at a spatial resolution of 1º × 1º. The historical and future rainfall data are simulated from 27 General Circulation Models (GCMs) from the Coupled Model Intercomparison Project Phase 6 (CMIP6). The historical rainfall consists of the rainfall data simulated by GCMs for the period 1961–2014, and the rainfall simulated by the GCMs under shared socio-economic pathway scenarios (SSPs) constitutes the future rainfall. Four scenarios (SSP126, SSP245, SSP370, and SSP585) and two different timeframes (near future (2015–2060) and far future (2061–2099)) are considered to determine how the rainfall and its dynamic complexity vary across the scenarios and timescales. The false nearest neighbor (FNN) algorithm is employed to determine the dimensionality and, hence, the complexity of the rainfall dynamics. The algorithm involves two major steps: (i) reconstruction of the single-variable rainfall time series in a multi-dimensional phase space; and (ii) identification of “false” neighbors in the reconstructed phase space and estimation of the dimension of the rainfall time series. The results suggest that the FNN dimensions of both the historical rainfall and future rainfall simulated by the 27 GCMs across India under all scenarios range from 3 to 20, indicating low to high-level complexity of the rainfall dynamics. However, only less than 1% of the study area shows high-level complexity in historical and future rainfall dynamics. Moreover, around 20 GCMs exhibit low to medium-level complexity of rainfall dynamics in 80% of the study area, with the dimensionality in the range from 3 to 10. Therefore, considering both the historical rainfall and future rainfall under all the four scenarios and the two timeframes considered in this study, the number of GCMs simulating rainfall that exhibits dimensionality in the range 11 to 20 are few. This may be an indication that the complexity of rainfall dynamics in India in the future will be low-to-medium dimensional.