|Morin, E. - UNIVERSITY OF ARIZONA|
|Krajewski, W. - UNIVERSITY OF IOWA|
|Gao, X. - UNIVERSITY OF ARIZONA|
|Sorooshian, S. - UNIVERSITY OF ARIZONA|
Submitted to: Journal of Hydrometeorology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: March 13, 2003
Publication Date: November 22, 2003
Citation: Morin, E., Krajewski, W., Goodrich, D.C., Gao, X., Sorooshian, S. 2003. Estimating rainfall intensities from weather radar data: the scale dependency problem. Journal of Hydrometeorology. 4:782-797. Interpretive Summary: Arid and semi-arid regions account for approximately one-third of the land mass of earth. These regions are experiencing continued pressure from population growth in many parts of the world. Water is a critical resource in these regions and is often in short supply. Detailed study of water resources and the hydrology of semi-arid regions is important if we are to continue to populate and use these regions. Rainfall estimates from National Weather Service radar shown daily on popular news forecasts are being used for water resource decisions and models. However, the methods to estimate rainfall from radar are not well tested for semi-arid regions. Rainfall observations from the Walnut Gulch Experimental Watershed, operated by the U.S. Dept. of Agriculture, Agricultural Research Service were used to better describe and understand the nature of the methods used to transform radar into rainfall estimates. It was found that the rainfall rates estimated from radar depend on the area covered by radar and period of time used to estimate the rainfall. This is important as these estimates are used throughout the Southwest to forecast floods. Radar-rainfall estimates are also increasingly being used for watershed hydrology and management models.
Technical Abstract: Meteorological radar is a remote sensing system that provides rainfall estimations at high spatial and temporal resolution. The radar-based rainfall intensities (R) are calculated from the observed radar reflectivities (Z). In this paper we explore scale-dependency of the power-law Z-R parameters when estimated from radar reflectivity and rain gauge intensity data. The multiplicative (a) and exponent (b) parameters are said to be "scale dependent" if applying the observed (gauge-based) and calculated (radar-based) rainfall intensities to the objective function at different temporal and spatial scales results in different "optimal" parameters. Radar (WSR-88D, Weather Surveillance Radar-1988 Doppler) and gauge data are analyzed from convective storms over a mid-size, semi-arid and well-equipped watershed. Using the Root Mean Square Difference (RMSD) objective function, a significant scale-dependency was observed, which resulted in a considerable increase of the (a) parameter and decrease of the (b) parameter with the increase of time and space scales. We examine how scale-dependency relates to two sources of uncertainties. The major source considered is observational uncertainties, which we examine both experimentally and with simplified models that allow representation of error in the observation. In order to identify an optimal scale for determining Z-R further investigation is needed to study the error structure of reflectivity and rain intensity.