资源环境科技发展态势分析平台

Rainfall is a continuous-time phenomenon typically characterized by precipitation states such as rain, showers, and dry whose dependence varies over a variety of space-time scales. Here attention is focused on the effective identification of rain and shower precipitation states over a region where these states have been determined by a hidden semi-Markov model of continuous-time precipitation. The states identified provide an accurate description of precipitation dynamics and can be regarded as close proxies to synoptic weather types of the same name. The stochastic properties and structure of these states (rather than precipitation amounts) are explored and delineated. A primary objective of the paper is to better understand the impact of conventional space-time aggregation on the dynamics of rainfall. What aggregation time scales result in more faithful descriptions of the space-time dynamics of continuous-time rainfall? While rain might be expected to be more spatially coherent than showers and involve longer time scales, dry periods involve much longer time and space scales again than either rain or showers. These issues are discussed and conclusions drawn which provide guidance and insights useful for the development of space-time precipitation models and, more generally, the design of rainfall observation networks and data archives.

Plain Language Summary The main objective of this paper is to better understand how the dynamics of continuous-time rainfall are preserved after aggregation into more conventional time periods (minutes, hours, or days). What aggregation time scales result in the most faithful descriptions of the space-time dynamics of various rainfall types (e.g., showers, rain, or dry periods)? While rain may be more spatially coherent than showers and involve longer time scales, dry periods involve even greater time and space scales than either rain or showers. These scale issues are addressed using a statistical model of continuous-time rainfall and, in so doing, this paper provides guidance and insights useful for the development of space-time precipitation models and, more generally, the design of rainfall observation networks and data archives.