A Geostatistical Framework for Estimating Rain Intensity Fields Using Dense Rain Gauge Networks

Abstract:

Rain gauges provide direct and continuous observations of rain accumulation with a high time resolution (up to 1min). However the representativeness of these measurements is restricted to the funnel where rainwater is collected. Due to the high spatial heterogeneity of rainfall, this poor spatial representativeness is a strong limitation for the detailed reconstruction of rain intensity fields.

Here we propose a geostatistical framework that is able to generate an ensemble of simulated rain fields based on data from a dense rain gauge network. When the density of rain gauges is high (sensor spacing in the range 500m to 1km), the spatial correlation between precipitation time series becomes sufficient to identify and track the rain patterns observed at the rain gauge sampling rate. Rain observations derived from such networks can thus be used to reconstruct the rain field with a high resolution in both space and time (i.e. 1min in time, 100m in space).

Our method produces an ensemble of realizations that honor the rain intensities measured throughout the rain gauge network and preserve the main features of the rain intensity field at the considered scale, i.e.: the advection and morphing properties of rain cells over time, the intermittency and the skewed distribution of rainfall, and the decrease of the rain rate near the rain cell borders (dry drift). This allows to image the observed rain field and characterize its main features, as well as to quantify the related uncertainty. The obtained reconstruction of the rainfall are continuous in time, and therefore can complement weather radar observations which are snapshots of the rain field. In addition, the application of this method to networks with a spatial extent comparable to the one of a radar pixel (i.e. around 1km2) could allow exploration of the rain field within a single radar pixel.