Multi-scale Evaluation of Global Satellite Precipitation Products over Taiwan

Utilizing satellite-based remote sensing presents an effective remedy for addressing the scarcity of available precipitation data across oceanic and mountainous regions, where establishing ground-based observation facilities at an optimal density is impractical. To ensure comprehensive global coverage, multiple satellite precipitation products (SPPs) are generated by blending precipitation estimates acquired from diverse low Earth orbit (LEO) satellite platforms equipped with passive microwave (PMW) sensors. Due to the limited sampling frequency at which a single LEO satellite can revisit a specific geographical area, although adopting a multi-satellite approach alleviates this issue by leveraging data from all accessible PMW sensors, achieving seamless global coverage within acceptable latency windows is hardly guaranteed. To bridge gaps arising from infrequent PMW passes, cloud motion (morphing) vectors calculated from Geostationary (GEO) satellites are harnessed to propagate and interpolate PMW-derived precipitation estimates onto a spatial grid at 30-minute intervals for NOAA Climate Prediction Center (CPC) morphing technique (CMORPH V1) and NASA Global Precipitation Measurement (GPM) Integrated Multi-satellitE Retrievals for GPM (IMERG V06 Final).

However, the aforementioned blended SPPs have difficulties in estimating precipitation patterns in complex terrain and they are seldom studied in Taiwan with the high-quality ground-based data from its latest infrastructure. Therefore, this study aims to evaluate the performance of CMORPH and IMERG over Taiwan with well-known strong land-ocean interactions under its seasonal precipitation pattern using high-resolution ground-based reference data from the operational Quantitative Precipitation Estimation and Segregation Using Multiple Sensors (QPESUMS) system. The analysis encompasses spatial and temporal fluctuations in error statistics for CMORPH and IMERG, offering insights into their performance across diverse scenarios. By examining rainfall detection and estimation scores based on varying seasons and precipitation intensity thresholds, the study seeks to quantify errors exhibited by these products in relation to distinct seasonal conditions and precipitation magnitudes.

In detail, the evaluation of CMORPH and IMERG is conducted yearly, seasonal, and daily, respectively. Various statistical metrics are calculated to score the performance of CMORPH and IMERG against QPESUMS including relative bias (RB), Pearson correlation coefficient (CC), mean error (ME), normalized mean error (NME), mean absolute error (MAE), normalized mean absolute error (NMAE) and root mean squared error (RMSE). We inspected the precipitation time series of 2019-2021 using RB, CC, and MAE to quantify the relative bias, spatial pattern consistency and actual value differences between SPPs and QPESUMS at daily scale. To analyze the performance of sensitivity of SPPs at different precipitation intensity levels, the probability of detection (POD), false alarm ratio (FAR) and Heidke skill score (HSS) are tested on CMORPH and IMERG in terms of time series and each geographical location conditioned on seasons.

Building upon previous research centered on bias correction for SPPs over the US and Taiwan, this study extends its scope by focusing on delineating errors in CMORPH and IMERG that are dependent on scale and environmental factors.