Clustering Methods for Statistical Downscaling in Short-Range Weather Forecasts

Abstract

In this paper an application of clustering algorithms for statistical downscaling in short-range weather forecasts is presented. The advantages of this technique compared with standard nearest-neighbors analog methods are described both in terms of computational efficiency and forecast skill. Some validation results of daily precipitation and maximum wind speed operative downscaling (lead time 1-5 days) on a network of 100 stations in the Iberian Peninsula are reported for the period 1998-99. These results indicate that the weighting clustering method introduced in this paper clearly outperforms standard analog techniques for infrequent, or extreme, events (precipitation 20 mm; wind 80 km h 1). Outputs of an operative circulation model on different local-area or large-scale grids are considered to characterize the atmospheric circulation patterns, and the skill of both alternatives is compared.

Propagation of uncertainty from observing systems into NWP: COST‐731 Working Group 1

“…probabilistic global-scale QPF typically delivered on a grid with a mesh size of tens of kilometres for use in a small-scale river catchment. A number of statistical downscaling techniques were developed with the aim of producing precipitation fields with given statistical properties and consistent with the NWP QPF value based on multifractal cascades (Deidda, 2000;Ferraris et al, 2003), autoregressive models (Rebora et al, 2006) and analogue methods (Gutierrez et al, 2004;Wetterhall et al, 2005). Because there is a significant scale gap between ensemble prediction systems (EPSs) derived from NWPs and typical hydrological models, the EPS QPFs must be downscaled or disaggregated for scale correction.…”

The COST 731 Action: A review on uncertainty propagation in advanced hydro-meteorological forecast systems

“…To deal with this dilemma, a number of statistical downscaling techniques were developed with the aim of generating precipitation ensembles which are consistent with the NWP QPF value but have prescribed statistical properties, which give, most importantly, indications on how intense precipitation can expected to be on any area smaller than the NWP model mesh. These statistical approaches are based on multifractal cascades (Deidda, 2000;Seed, 2003), autoregressive models (Ferraris et al, 2003), and analogue methods (Wetterhall et al, 2005;Gutiérrez et al, 2004).…”

Snow trends in Northern Spain: analysis and simulation with statistical downscaling methods

“…To this aim, we focus on the binary variable DSO and analyze the performance of a statistical downscaling method based on analogs (or nearest neighbours) following the work by Gutiérrez et al (2004).…”

“…Furthermore, although most of the studies which link local snow observations to atmospheric patterns use large spatial patterns in order to account for large-scale flow, these studies deal with seasonal or monthly-averaged observations (Scherrer and Appenzeller, 2006). In our case, we work with daily observations and for such a purpose, we found that a smaller grid (including a temporal component by considering the fields both at 12 and 24 UTC) performed better than a larger pattern, in agreement with Gutiérrez et al (2004).…”

“…This is an important result if one wants to apply this method to climate models which have lower temporal and spatial resolutions. Finally, we also tested an alternative statistical downscaling method recently introduced as an extension of the standard analog methodology based on weather types (more details can be found in Gutiérrez et al, 2004). In this case, the method works by first applying a clustering algorithm to the reanalysis patterns in order to obtain homogeneous groups, or weather types.…”