|Title||Specification of rain gauge representativity error for data assimilation|
|Year of Publication||2011|
|Authors||Lopez, P, Ryu, G-H, Sohn, B-J, Davies, L, Jakob, C, Bauer, P|
|Secondary Title||Technical Memorandum|
The comparison of precipitation fields produced by numerical weather prediction models (grid box means) with rain gauge observations (point measurements) can be problematic. Obtaining a reasonable estimate of the representativity error (RE) of rain gauges is a prerequisite to their future use in data assimilation. Here, RE is evaluated in terms of the spatial variability of precipitation over a typical model grid box. The large-scale component of RE is estimated from ground-based radar precipitation data, while its small-scale component is assessed from high-density rain gauge networks. A quantitative estimation of RE is obtained for rain rate (RR) as well as for its logarithmic transform (ln[RR + 1]). Results confirm that for a given rain rate, the RE of a single observation increases with the size of the target grid box and the occurrence of convective precipitation, and decreases with the accumulation period. The contribution to RE from the small scales turns out to be usually lower than that from the large-scales, but is not negligible. The relative total RE exceeds 100% for weak precipitation, but can drop down to 20% or less for heavier precipitation. This drop in relative RE is even more pronounced for ln(RR+1) than for RR, while the range of RE values is expectedly much reduced in terms of ln(RR + 1). A simple parametrisation of RE in terms of ln(RR + 1) has been developed. It only depends on target resolution and day of the year, with a distinction between mid-latitudes and tropical regions. A month-dependent parametrisation of precipitation spatial correlations as a function of separation distance has also been formulated. Finally, the reduction of RE due to spatial correlations of the rain field and to the availability of multiple nearby rain gauges is considered. These parametrisations are expected to be applicable to midlatitude and tropical rainfall over flat terrain and to 6-hour rain accumulations only.