Due to the lack of in situ observations in mountainous locations, the use of remote sensing data is an alternative to analyze rainfall distribution patterns during the passage of major hurricanes. In this work, gridded precipitation data from the CHIRPS database are evaluated by comparing with observations from weather stations during the passage of category 3–5 hurricanes for the period 1980–2024. The comparison between estimated and observed values is performed by regression analysis and the use of K and K0 coefficients. An advantage of using K-ratio and K0-ratio is the identification of overestimated or underestimated precipitation in the pixel records. The distribution of daily precipitation helped in a more concise way to better understand how well CHIRPS reproduced the observed rainfall patterns. Results show that correlations between observations and database estimates are in the range of 0.40–0.76, for eastern Pacific hurricanes, and 0.49–0.78 for Atlantic hurricanes, all of which are statistically significant; however, these results do not imply congruence between observations and estimates since CHIRPS fails to adequately reproduce the position of the highest precipitation core. In the initial stages of a tropical cyclone, near-zero correlations between observations and estimates indicate that CHIRPS is not able to reproduce the observed rainfall. It is recommended to use CHIRPS with caution when the focus is on analyzing rainfall patterns during the development of intense tropical cyclones.
Vega-Camarena et al. (Fri,) studied this question.