Open AccessSubpixel-Scale Rainfall Variability and the Effects on Separation of Radar and Gauge Rainfall Errors
Author(s) -
Yu Zhang,
Thomas E. Adams,
J. V. Bonta
Publication year - 2007
Publication title -
journal of hydrometeorology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.733
H-Index - 123
eISSN - 1525-755X
pISSN - 1525-7541
DOI - 10.1175/2007jhm835.1
Subject(s) - subpixel rendering , environmental science , radar , remote sensing , spatial variability , meteorology , pixel , rain gauge , scale (ratio) , variance (accounting) , spatial correlation , kriging , statistics , precipitation , computer science , mathematics , geology , geography , accounting , cartography , telecommunications , business , computer vision
This paper presents an extended error variance separation method (EEVS) that allows explicit partitioning of the variance of the errors in gauge- and radar-based representations of areal rainfall. The implementation of EEVS demonstrated in this study combines a kriging scheme for estimating areal rainfall from gauges with a sampling method for determining the correlation between the gauge- and radar-related errors. On the basis of this framework, this study examines scale- and pixel-dependent impacts of subpixel-scale rainfall variability on the perceived partitioning of error variance for four conterminous Hydrologic Rainfall Analysis Project (HRAP) pixels in central Ohio with data from Next-Generation Weather Radar (NEXRAD) stage III product and from 11 collocated rain gauges as input. Application of EEVS for 1998–2001 yields proportional contribution of two error terms for July and October for each HRAP pixel and for two fictitious domains containing the gauges (4 and 8 km in size). The results illustrate the importance of considering subpixel variation of spatial correlation and how it varies with the size of domain size, number of gauges, and the subpixel locations of gauges. Further comparisons of error variance separation (EVS) and EEVS across pixels results suggest that accounting for structured variations in the spatial correlation under 8 km might be necessary for more accurate delineation of domain-dependent partitioning of error variance, and especially so for the summer months.