نوع مقاله : مقاله پژوهشی
عنوان مقاله English
نویسندگان English
Satellite precipitation products provide an important source of rainfall information in regions with sparse gauge coverage. However, errors in precipitation occurrence and intensity, particularly over mountainous terrain, may limit their direct application. This study evaluated the GPM-IMERG V07 Final Run product and investigated the effectiveness of empirical quantile mapping in correcting daily precipitation estimates over northwestern Iran. A total of 205,860 paired gauge–satellite observations from 188 stations covering 2021–2023 were analyzed. Bias correction was performed using a zero-aware monthly empirical quantile mapping approach, in which wet-day frequency was first adjusted using a threshold of 0.5 mm day⁻¹, followed by correction of precipitation intensity on the remaining wet days. A leave-one-year-out temporal cross-validation scheme was employed to obtain an out-of-sample assessment and avoid overly optimistic performance estimates. The raw product overestimated both mean precipitation and wet-day frequency. The observed wet-day frequency was 11.51%, whereas the corresponding value for raw GPM-IMERG was 21.53%. Following correction, wet-day frequency decreased to 11.50%, and the frequency bias index improved from 1.871 to 0.999. Percent bias decreased from 20.865% to 4.493%, while mean absolute error declined from 1.249 to 1.197 mm. The corrected product also provided a closer representation of the 75th, 90th, and 95th percentiles of observed precipitation. Nevertheless, root mean square error increased from 3.721 to 4.182 mm, and the correlation coefficient decreased from 0.166 to 0.138. Although the false alarm ratio declined from 0.769 to 0.724, the probability of detection and critical success index decreased from 0.433 to 0.276 and from 0.178 to 0.160, respectively. The correction exhibited less consistent performance during 2023 and August, for intense precipitation events, and at stations located above 2,000 m. Overall, zero-aware quantile mapping effectively reduced cumulative bias and corrected wet-day frequency and the central part of the precipitation distribution, but it did not resolve large daily errors or the temporal mismatch of precipitation events. The correction reduced overall precipitation bias and adjusted wet-day frequency, but these improvements do not establish better performance for monthly or seasonal precipitation totals. Applications to flood warning, daily precipitation extremes, and hydrological modeling sensitive to precipitation sequencing require complementary evaluation.
کلیدواژهها English