DATE2022-06-22 20:14:47
TITLEComparison between statistics of observed and the ERA-Interim reanalyses for the northern region of Tunisia at rainfall event scale
AUTHORSSabrine Derouiche (1,2) ,C├ęcile Mallet (1) ,Zoubeida Bargaoui (2)
  1. 1) Latmos, Uvsq, Paris Saclay University, Paris (France) ,2) Lmhe, Enit, Tunis El Manar University, Tunis (Tunisie)
ABSTRACTIn general, the rainfall data sets produced by the numeric models have the advantage of being complete over a long period of time. Therefore, it is important to assess their quality in relation to the observations before using them as an alternative source of data. We aim to analyse the ability of ERA-Interim rainfall data to reproduce the structure of spatial and temporal variability of rainfall Northern Tunisia. Thus, a comparative statistical analysis of daily precipitations observed by a dense rain gauge network and aggregated into rainfall events with those produced by the ERA-Interim reanalysis model is undertaken for the 30 years period (1980-2009) and the winter season (December to February). As a first step, the rainfall events (or wet spells) are extracted from DJF daily timeseries from 70 rain gauge stations and equally from 9 ERA-Interim grid boxes covering the study area. Then, DJF seasons are characterized by six descriptors indicating the structure and the frequency of rainfall events over the season. DJF descriptors are: 1) event number, 2) rainy day number, 3) total accumulation, 4) average accumulation per event, 5) average duration per event and 6) average accumulation per rainy day. We compare sample histograms as well as classification results obtained by using self-organizing maps (SOM) method combined with a hierarchical classification. The preliminary statistical analysis of the obtained descriptors from the two data sources highlights very significant differences for specifics rainfall descriptors. Indeed, the average accumulation and the duration of rain events, the number of rainy days over the season are overestimated by Era-interim data. Clustering approaches yield to four typical classes of rainfall events in both cases (observed and reanalysed). The comparison of the spatial and temporal distributions of the four classes between observations and Era-interim data give a good coherence of the temporal structures. However, the spatial representation quality varies from one region to another of the study domain. The rainfall regime is better represented by Era-interim data in the southern region of the study area but is missing representation in the northern one, i.e., in the Mediterranean coastal part.