Journal of Natural Science and Exploration
Development of a Quantile Regression Model for Replicating Extreme Precipitation in Different Climatic Zones of Morocco
Abstract
Oumechtaq Ismail, Oulidi Abderrahim and Bahaj Tarik
Given the dispersion and scarcity of meteorological stations across Morocco, as well as gaps in recorded data series, many researchers turn to satellite data to model flood risk. However, the direct use of these data can introduce biases, requiring corrections. Existing solutions primarily focus on the central values of the data, neglecting extreme events, which are actually responsible for flooding.
To date, no model has been developed to reliably reproduce extreme precipitation using satellite data while accounting for the specific climatic factors in each region that drive the generation of these extreme events. In this work, we aimed to fill this gap. To do so, we first evaluated the reliability of different satellite products. Then, we developed a quantile regression (QR) model tailored to each region. The Kullback-Leibler distance was used to identify the quantile that best reproduces ground-based precipitation.
The model was validated through a spatio-temporal approach, using extreme events recorded at stations not included in the model development, as well as considering events beyond the 2000-2018 period used to build the model.
The results show that the QR model, developed for each region, accurately estimates extreme precipitation. For example, the difference between the observed cumulative precipitation and that calculated by the model ranges from 1% to 13%. This addresses three major issues: the lack of information in areas without stations, the correction of biases associated with satellite precipitation, and the improvement in estimating extreme precipitation.