Extreme precipitation events in the Mediterranean area: contrasting two different models for moisture source identification

Abstract

Concern about heavy precipitation events has increasingly grown in the last years in southern Europe, especially in the Mediterranean region. These occasional episodes can result in more than 200 mm of rainfall in less than 24 h, producing flash floods with very high social and economic losses. To better understand these phenomena, a correct identification of the origin of moisture must be found. However, the contribution of the different sources is very difficult to estimate from observational data; thus numerical models are usually employed to this end. Here, we present a comparison between two methodologies for the quantification of the moisture sources in two flooding episodes that occurred during October and November 1982 in the western Mediterranean area. A previous study, using the online Eulerian Weather Research and Forecasting (WRF) Model with water vapor tracer (WRF-WVT) model, determined the contributions to precipitation from moisture evaporated over four different sources: (1) the western Mediterranean, (2) the central Mediterranean, (3) the North Atlantic Ocean and (4) the tropical and subtropical Atlantic and tropical Africa. In this work we use the offline Lagrangian FLEXPART-WRF model to quantify the role played by these same sources. Considering the results provided by WRF-WVT as “ground truth”, we validated the performance of the FLEXPART-WRF. Results show that this Lagrangian method has an acceptable skill in identifying local (western Mediterranean) and medium-distance (central Mediterranean and North Atlantic) sources. However, remote moisture sources, like tropical and subtropical areas, are underestimated by it. Notably, for the October event, the tropical and subtropical area reported a relative contribution 6 times lower than with the WRF-WVT. In contrast, FLEXPART-WRF overestimates the contribution of some sources, especially from North Africa. These over- and underestimates should be taken into account by other authors when drawing conclusions from this widely used Lagrangian offline analysis.