Research often begins with a simple question. For this study, the question was: How reliable are satellite precipitation datasets in a region with complex terrain like Taiwan?
Taiwan experiences some of the most intense rainfall events in the world, particularly during typhoons, when daily rainfall can exceed 1000 mm. Accurate precipitation information is therefore critical for flood forecasting, landslide risk assessment, and water resource management. However, obtaining reliable precipitation data in such environments is not straightforward.
Rain gauges provide precise measurements, but their distribution is sparse in mountainous regions. Radar observations are useful for monitoring storms, yet they often suffer from terrain blockage in the Central Mountain Range. Because of these limitations, researchers and practitioners increasingly rely on satellite precipitation products (SPPs).
While working on hydrological and climate studies related to Taiwan, we repeatedly encountered an important issue: different satellite datasets often give different results. This raised an important research question—which dataset should we trust, and under what conditions?
To explore this, we conducted a 20-year evaluation (2000–2020) of four widely used satellite precipitation products—IMERG, CHIRPS, PERSIANN-CDR, and TRMM—using observations from 87 rain gauge stations across Taiwan. The analysis examined performance across multiple temporal scales, different Köppen–Geiger climate zones, elevation gradients, and long-term trends.
One of the key insights from this study is that there is no universally “best” satellite precipitation dataset. Instead, each product has its own strengths:
-
IMERG performs particularly well in detecting precipitation events, making it useful for hazard monitoring and event-based applications.
-
CHIRPS provides more reliable estimates of precipitation magnitude, which is important for hydrological modeling and water resource studies.
Another interesting finding is the strong influence of topography. Satellite precipitation accuracy generally decreases at higher elevations, highlighting the persistent challenge of monitoring rainfall in mountainous regions.
For us, this work was not only about evaluating datasets. It was also about understanding how satellite observations can be used more effectively in complex environments where ground observations are limited but accurate precipitation information is essential.
We hope this study contributes to improving the use of satellite precipitation products for hydrological research, climate analysis, and disaster risk reduction in mountainous and hazard-prone regions.