Object-based Evaluation of Satellite Precipitation Retrievals: A Case Study of the Summer Season over CONUS

Abstract:

Satellite precipitation retrievals that have high spatial and temporal resolutions are suitable for various applications, such as hydrologic modeling and watershed management. Many validation studies have been established to understand the strengths and limitations of these satellite precipitation retrievals. In this study, an object-based validation approach is adopted to evaluate several satellite precipitation retrievals focusing on the spatial and geometric patterns of precipitation. This object-based validation approach identifies precipitation objects using an image processing technique referred to as watershed transform. Several object attributes are diagnosed and analyzed based on the distance measurement. Three object-based verification scores are summarized to determine the overall performances of satellite precipitation retrievals. The Integrated Multi-satellitE Retrievals for GPM (IMERG) and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) were evaluated using the object-based approach. The NOAA stage IV MPE multi-sensor composite rain analysis was utilized as the ground observations. The comparative assessments were conducted at 0.25° by 0.25° on a daily scale in the summer season of 2014 over the continental United States (CONUS). The results suggest that IMERG possesses the similar spatial pattern of local-scale precipitation areas against stage IV observations. In addition, IMERG depicts the sizes and locations of precipitation areas more accurately against stage IV.