Statistical downscaling model based on canonical correlation analysis for winter extreme precipitation events in the Emilia‐Romagna region

Abstract

Optimum statistical downscaling models for three winter precipitation indices in the Emilia-Romagna region, especially related to extreme events, were investigated. For this purpose, the indices referring to the number of events exceeding the long-term 90 percentile of rainy days, simple daily intensity and maximum number of consecutive dry days were calculated as spatial averages over homogeneous sub-regions identified by the cluster analysis. The statistical downscaling model (SDM) based on the canonical correlation analysis (CCA) was used as downscaling procedure. The CCA was also used to understand the large-/regional-scale mechanisms controlling precipitation variability across the analysed area, especially with respect to extreme events. The dynamic (mean sea-level pressure-SLP) and thermodynamic (potential instability-δQ and specific humidity-SH) variables were considered as predictors (either individually or together). The large-scale SLP can be considered a good predictor for all sub-regions in the dry index case and for two sub-regions in the case of the other two indices, showing the importance of dynamical forcing in these cases. Potential instability is the best predictor for the highest mountain region in the case of heavy rainfall frequency, when it can be considered as a single predictor. The combination of dynamic and thermodynamic predictors improves the SDM's skill for all sub-regions in the dry index case and for some sub-regions in the simple daily intensity index case.The selected SDMs are stable in time only in terms of correlation coefficient for all sub-regions for which they are skilful and only for some sub-regions in terms of explained variance. The reasons are linked to the changes in the atmospheric circulation patterns influencing the local rainfall variability in Emilia-Romagna as well as the differences in temporal variability over some sub-regions and sub-intervals. It was concluded that the average skill over an ensemble of the most skilful and stable SDMs for each region/sub-interval gives more consistent results.

Spatial and temporal variability of climate extremes in Romania and associated large‐scale mechanisms

“…The tested large-scale predictors have been mentioned in Section 2, which were found to be skilful in previous papers, either for Romania mainly in statistical downscaling models (Busuioc et al, 1999(Busuioc et al, , 2006(Busuioc et al, , 2010 or for various studies related to other European regions (e.g. Zorita et al, 1992;von Storch et al, 1993;Xoplaki et al, 2004;Tomozeiu et al, 2007;Busuioc et al, 2008). These predictors refer to dynamic variables (SLP) and thermodynamic ones (T850, SH700 and SH850), their combination is considered as optimum to find plausible connections with the regional climates (Huth, 2003).…”

Extreme events: dynamics, statistics and prediction

“…For example, weather typing was applied by to derive a climatology of severe storms in Virginia (USA), and by Plaut et al (2001) to characterize heavy precipitation events in several Alpine sub-regions. Busuioc et al (2008) used a regression-based approach to downscale indices of winter extreme precipitation in a climate context. In the same context, Haylock et al (2006) concluded that SDMs based on artificial neural networks are best at representing the interannual variability of heavy precipitation indices but underestimate extremes.…”

Weather regimes designed for local precipitation modeling: Application to the Mediterranean basin

“…We used Z500 to define WRs, like, for example, in work by Michelangeli et al [1995] or Yiou and Nogaj [2004]. However, information at much lower altitude has shown useful [e.g., Vrac et al, 2007b] to relate large-and local-scale features, and SLP is generally considered as valuable information for precipitation [e.g., Busuioc et al, 2008]. Indeed, Z500 is a spatially smooth variable and SLP is more sensitive to horography and land properties, and thus, may allow to capture more spatial variability.…”