Aymeric Chazottes

Simulation of yearly rainfall time series at microscale resolution with actual properties: Intermittency, scale invariance, and rainfall distribution

Rainfall is a physical phenomenon resulting from the combination of numerous physical processes involving a wide range of scales, from microphysical processes to the general circulation of the atmosphere. Moreover, unlike other geophysical variables such as water vapor concentration, rainfall is characterized by a relaxation behavior that leads to an alternation of wet and dry periods. It follows that rainfall is a complex process which is highly variable both in time and space. Precipitation is thus characterized by the following features: rain/no-rain intermittency, multiple scaling regimes, and extreme events. All these properties are difficult to model simultaneously, especially when a large time and/or space scale domain is required. The aim of this paper is to develop a simulator capable of generating high-resolution rain-rate time series (15 s), the main statistical properties of which are close to an observed rain-rate time series. We also attempt to develop a model having consistent properties even when the fine-resolution-simulated time series are aggregated to a coarser resolution. In order to break the simulation problem down into subcomponents, the authors have focused their attention on several key properties of rainfall. The simulator is based on a sequential approach in which, first, the simulation of rain/no-rain durations permits the retrieval of fractal properties of the rain support. Then, the generation of rain rates through the use of a multifractal, Fractionally Integrated Flux (FIF), model enables the restitution of the rainfalls multifractal properties. This second step includes a denormalization process that was added in order to generate realistic rain-rate distributions.