Nowcasting of Precipitation in the High-Resolution Dallas–Fort Worth (DFW) Urban Radar Remote Sensing Network

Abstract

Short-term forecasting (nowcasting) of rainfall is widely used in providing early warning, and therefore, has very practical application. The CASA demonstration network deployed in the Dallas–Fort Worth (DFW) area consists of high-resolution X-band radars deployed around the National Weather Service Radar WSR-88D, and nowcasting of precipitation is an important product of this network. The current nowcasting technique DARTS was developed in frequency domain using fast Fourier transforms in order to provide nowcasts in a computationally efficient manner. Building on the earlier work on the CASA Project and the STEPS methodology, a stochastic nowcasting method is developed and tested for the first time in a small-scale urban environment using high-resolution radar data. The key idea is to decompose the reflectivity field into multiple spatial scales and generate stochastic perturbations in each scale to account for uncertainties in the nowcasts. It is shown that the proposed method can produce reliable nowcasts and uncertainty estimates up to 45 min and performs as well or better than DARTS in terms of the standard critical success index and mean error verification scores. Using the scale decomposition of STEPS, a power–law relationship is derived between the spatial scale and Lagrangian lifetime of precipitation on scales between 500 m and 50 km.