Evaluation of the impacts of ingesting TRMM data on the accuracy of quantitative precipitation estimates Obtained via the SCaMPR framework

The National Environmental Satellite, Data, & Information Service (NESDIS) has produced operational quantitative precipitation estimates (QPE) from GOES infrared measurements and some ancillary data for many years. The Self-calibrating Multivariate Precipitation Retrival Algorithm (SCaMPR) was developed to improve these infrared QPE's through real-time calibration with passive microwave rainrate estimates from lower-orbiting satellites. This study assesses the accuracy of two SCaMPR QPE products. The first was produced using only passive microwave data as the reference for calibration (referred to as SCaMPR), while for the second, TRMM Precipitation Radar data and Passive Microwave imager data were added to the reference data set (referred to as SCaMPR-TRMM). Data from the TRMM follow-on mission, the Global Precipitation Mission (GPM), will serve as a critical component in the next-generational NESDIS operational satellite rainfall algorithm, which will apply SCaMPR to the new Advanced Baseline Imager.

For the evaluation, SCaMPR and SCaMPR-TRMM data for the 8-year period 2000-7 were generated for the state of Texas. These data were compared with quality-assured reports of 242 precipitation gauges operated by the Lower Colorado Rive Authority. The evaluation was performed at both hourly and daily time scales, and the results were stratified by month. The evaluation metrics included bias, correlation; probability of detection (POD), and false alarm ratio. Quantiles of precipitation values were also compared to illustrate the distribution characteristics of satellite vs. gauge-based precipitation values. The results of the evaluation indicate that, on both hourly and daily scales, TRMM ingest tends to mitigate the low bias on an overall basis, and tends to slightly improve the correlation and the POD for majority of the months. Yet, in comparison to SCaMPR QPEs, SCaMPR-TRMM QPEs show considerable positive bias for the cool season (Oct-Feb), and contain fewer instances of high rainfall accumulations as indicated by the gauges. Based on these results, some refinements to the SCaMPR algorithm are now planned. Future work for the current study includes an evaluation of impact of the various satellite-based precipitation estimates on hydrologic models for basins in Texas.