<p>Essential to identifying regional precipitation dynamics, the delineation of homogenous regions has a range of applications in hydrology. Traditional approaches to regionalization have primarily focused on the magnitude and long-term variability of precipitation, impeding their ability to capture complex precipitation patterns. To achieve a comprehensive regionalization, it is pertinent to consider diverse precipitation characteristics (e.g., temporal distribution, seasonality, magnitude, and marginal distribution). Drawing on high-resolution (0.25° × 0.25°) gridded daily rainfall data (1951–2020), a parsimonious regionalization framework integrating these variables was developed. Information theoretical metrics (e.g., marginal entropy and apportionment entropy) were considered in quantifying the variability of rainfall in magnitude and over time. Further, a refined centroid is proposed to capture rainfall seasonality. Self-Organizing-Maps-based regionalization delineated the Indian Sub-continent into eleven distinct regions. The marginal and apportionment entropies, along with the seasonality, were major drivers of the clustering process. Compared to the results of similar studies, the proposed four-parameter regionalization framework proved robust. Temporal analyses of the rainfall characteristics indicate a marked rise in rainfall magnitude and its marginal distribution, along with a decline in its temporal distribution, indicating a possibility of increasing incidence of extreme events. Further, examining the temporal evolution of the clusters using a 30-year moving window revealed substantial changes in the Northwest and core monsoon regions attributable to variations in rainfall intensity and variability.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Revisiting homogeneous regions on the Indian subcontinent by accounting for entropy-based variability and precipitation seasonality

  • Siva Sai Syam Nandikanti,
  • Maheswaran Rathinasamy,
  • Jan Adamowski

摘要

Essential to identifying regional precipitation dynamics, the delineation of homogenous regions has a range of applications in hydrology. Traditional approaches to regionalization have primarily focused on the magnitude and long-term variability of precipitation, impeding their ability to capture complex precipitation patterns. To achieve a comprehensive regionalization, it is pertinent to consider diverse precipitation characteristics (e.g., temporal distribution, seasonality, magnitude, and marginal distribution). Drawing on high-resolution (0.25° × 0.25°) gridded daily rainfall data (1951–2020), a parsimonious regionalization framework integrating these variables was developed. Information theoretical metrics (e.g., marginal entropy and apportionment entropy) were considered in quantifying the variability of rainfall in magnitude and over time. Further, a refined centroid is proposed to capture rainfall seasonality. Self-Organizing-Maps-based regionalization delineated the Indian Sub-continent into eleven distinct regions. The marginal and apportionment entropies, along with the seasonality, were major drivers of the clustering process. Compared to the results of similar studies, the proposed four-parameter regionalization framework proved robust. Temporal analyses of the rainfall characteristics indicate a marked rise in rainfall magnitude and its marginal distribution, along with a decline in its temporal distribution, indicating a possibility of increasing incidence of extreme events. Further, examining the temporal evolution of the clusters using a 30-year moving window revealed substantial changes in the Northwest and core monsoon regions attributable to variations in rainfall intensity and variability.