Generation of 2D rain maps with realistic properties : methodology and results

International audienceThe precipitations are due to complex meteorological phenomenon’s and unlike other geophysical constituents such as water vapor concentration they present a relaxation behavior leading to an alternation of dry and wet periods. Thus, precipitations can be described as intermittent process. The spatial and temporal variability of this phenomenon is significant and covers large scales. This high variability can cause extreme events which are difficult to observe properly because of their suddenness and their localized character. For all these reasons, the precipitations are therefore difficult to model. This study aims to adapt a one-dimensional time series model previously developed by theauthors (Akrour et al., 2013, 2015) to a two-dimensional rainfall generator. The original time series model can be divided into 3 major steps: rain support generation, intra event rain rates generation using multifractal and finally calibration process. We use the same kind of methodology in the present study. Based on dataset obtained from meteorological radar of Météo France with a spatial resolution of 1 km x 1 km we present the used approach: Firstly, the extraction of rain support (rain/no rain area) allowing the retrieval of the rain support structure function (variogram) and fractal properties. This leads us to use either the rain support modelisation proposed by (Schleiss et al., 2014) or directly real rain support extracted from radar rain maps. Then, the generation (over rain areas) of rain rates is made thanks to a 2D multifractal Fractionnally Integrated Flux (FIF) model (Schertzer et al. 1987). This second stage is followed by a calibration/forcing step (forcing average rain rate per events) added in order to provide rain rate coherent with observed rain5rate distribution. The forcing process is based on a relation identified from the average rain rate of observed events and their surfaces. Our model has 12 parameters, 4 are universals and 8 depend on the climatic area. The presentation will first explain the different steps presented above, then some results illustrating the simulator’s capabilities will be provided. They show that the simulated two5dimensional fields have coherent statistical properties in term of cumulative rain rate distribution but also in term ofpower spectrum and structure function with the observed ones at different spatial scales (1, 4, 16 km2) involving that scale features are well represented by the model