A New Framework for Quantification of Regional Spatial Variability of Mesoscale Sub-daily Extreme Rainfall for Subcontinent

Determination of the reliable estimate of risk associated with hydrometeorological extremes over a region requires discerning information on spatial variability of the associated at-site statistics/parameters. Extreme rainfall at finer spatio-temporal resolution allows for improved analysis of spatial variability, as local-scale statistical similarities (LSS) and heterogeneities are disclosed. The knowledge of LSS facilitates the use of information on regional spatial variability (in lieu of complex at-site spatial variability) for risk analysis. In addition, it is established in literature that geographical features influence the occurrence of extreme rainfall over an area. For a subcontinent with complex non-uniform patterns of geographical features, the regional spatial variability may be influenced by the geographic composition. To quantify this regional spatial variability, statistically homogenous regions need to be deciphered. Most studies on the regionalization of sub-daily extreme rainfall (SDER) are limited to a smaller spatial extent, and none was focused on a subcontinent. Furthermore, there are no prior studies focused on the analysis of regional spatial variability of SDER. To study the role of geography in modulation of the regional spatial variability of mesoscale SDER, the present study proposes a framework. It involves (i) dividing the study area into subareas based on geographical features, as they are deemed to influence the occurrence of extreme rainfall, (ii) the delineation of each subarea into statistically homogenous SDER regions using a novel regionalization technique, (iii) quantification of the regional spatial variability of SDER in each subarea using the delineated regions and a proposed novel index, and (iv) identifying the role of geographic features in modulating the regional spatial variability. The efficacy of the proposed framework is demonstrated by application to Indian subcontinent (66.5-100^o E, 6.5-38.5^o N) considering 0.12^o resolution SDER data corresponding to different durations (1,2,3,6 and 12-hour) for the period 1981-2020. The data were prepared by bias correcting the 0.12^o resolution NCMRWF IMDAA hourly gridded rainfall (at 20,717 grids) to be consistent with the widely used 0.25^o resolution IMD (India Meteorological Department) daily rainfall. The Indian subcontinent is divided into seven subareas based on geographic features. On application of the framework, it has been found that the regional spatial variability of SDER in a subarea is regulated by its geography and that of its neighbouring subareas. Insights are obtained on the effect of factors such as orography and coastal width on regional spatial variability of SDER. The study is of significance as the knowledge discerned on potential covariates/attributes has wide applications including identification of similar extreme rainfall sites for regional frequency analysis for extreme rainfall and risk assessment of consequent floods at ungauged/sparsely gauged hotspots such as water control (e.g., dams, barrages, levees) and conveyance infrastructure (culverts) in river basins under various climate change scenarios. The inherent physio-geographic features of the catchment may not be enough to analyze the similarity with neighbouring catchments. The boundary conditions around the catchment also plays a role.