A heavy convective storm that occurred over the eastern part of Puerto Rico (PR) was selected to conduct a preliminary validation of the Hydro-Estimtor (HE), which is a satellite rainfall retrieval algorithm. Puerto Rico (PR) has been selected as test bed since it is an extremely diverse terrain climatologically, has a dense rain-gauge network and during the rainy season severe rainstorms develops due to geographical location and complex orographical attributes. The easterly winds are coming from the eastern Atlantic during almost all year and play an important role bringing humidity into the island and stimulating orographical rainfall over the mountains of PR. Cold fronts dominate the weather pattern during wintertime. The tropical waves occur during the rainy season and frequently generate large amount of rainfall in the Caribbean basin. These tropical waves are largely the precursor of the tropical storms and hurricanes from June to November.

The studied tropical storm occurred during April 17, 2003 and generated a significant flash flood over the urban and rural areas causing four human dead, and $ 17,000,000 in economical damages. Puerto Rico has a rain-gauge network that provides a unique data set to conduct an accurate validation. The United States Geological Survey owns and operates the rain-gauge network that includes 125 sensors located through PR and records rainfall every 15 minutes. Estimation of precipitation was generated by the HE that provides rainfall rate every 15 minutes and has 4km spatial resolution and the closest grid to a particular rain gauge was used to perform rainfall comparisons. Continuous and discrete rainfall measurements were compared and the following scores were computed: hit rate, probability of detection, false-alarm rate, bias, mean absolute error, and the mean square error. Discrete preliminary results show that the HE algorithm provides 0.70, 0.30, 0.38 and 0.50 for hit rate, probability of detection, false-alarm rate, and bias, respectively. The continuous validation results are 0.66, 4.25, and 0.25 for the mean absolute error, mean square error, and bias, respectively. These preliminary results reveal that the HE exhibits a satisfactory hit rate, since 70% of the time the algorithm indicates correctly when a rainfall event is present or absent. The false alarm is relatively small indicating the 38% of the time the HE indicates the presence of rainfall when in fact there is no rainfall. However, the algorithm reveals the following drawbacks: the probability of detection is small indicating that only 30% of the time the occurred rainfall events were detected. The both bias the discrete and continuous are smaller than one, indicating underestimation on both the number of times when rainfall occurs, and when estimating the actual amount of rainfall.