Abstract 220

Abstract ID: 220

Subseasonal prediction of the 21-25 November 2016 heavy rainfall event over northwestern Italy

Lead Author: Daniele Mastrangelo
Institute of Atmospheric Sciences and Climate, CNR-ISAC, Italy

Keywords: heavy precipitation events, subseasonal forecasting, CNR-ISAC, ECMWF, Globo

Abstract: The Mediterranean basin is prone to intense precipitation events, particularly during late summer and autumn. Although a great effort is continuously provided to correctly predict this kind of events, numerical prediction of heavy precipitation remains a difficult task at all scales, including the subseasonal scale.
Between 21 and 25 November 2016, in the midst of a drought that lasted into 2017, northwestern Italy was hit by a prolonged event of intense precipitation. In some locations, total accumulations reached about half of the annual mean, with peaks up to 700 mm in 5 days. Floods and landslides were recorded in several areas of Liguria and Piedmont. The mean synoptic configuration shown by the ERA5 reanalysis for the period 21-27 Nov 2016, features a through in the west, with cut-off formation centered over the Iberian Peninsula, surrounded by areas of high 500 hPa geopotential height east of it. Slowly evolving, this west/east dipole, typically observed during heavy precipitation Mediterranean events, caused the continuous inflow of humid southerly currents yielding intense precipitation especially on the the exposed slopes.
In this work we evaluate the predictive ability beyond the second week of two operational sub-seasonal forecasts and highlight the large-scale dynamical features contributing to the good predictive performance. The subseasonal forecasts analyzed were those operationally issued by CNR-ISAC and ECMWF and available in the S2S project database. These forecasting systems predicted underestimated positive precipitation anomalies, and higher probabilities for the upper precipitation tercile, over the area under investigation in the runs initialized on 7 Nov 2016, thus for forecast days 15-21.
The analysis of the ensemble members shows that only few members are responsible for the good forecast performance on week 3: for instance, correctly reproducing the low/high dipole of geopotential anomaly over the Mediterranean basin, 6 ECMWF ensemble members account for ~ 25% of the predicted precipitation anomaly, whereas the 6 less rainy members account for just ~ 2%. The CNR-ISAC ensemble, based on the Globo model, shows a similar behavior. The best performing ensemble members are then compared to reanalyses to identify the dynamical events of the previous weeks possibly providing enhanced predictability for the selected precipitation event in the Mediterranean area. A key ingredient is traced back to the atmospheric dynamics occurring at the beginning of the month at northern American latitudes.
Although limited to a single case study, the analysis of the ensembles of two different forecasting systems suggests the need for methods that can highlight the useful information of few ensemble members hidden by the less skillful members.

Paolo Ghinassi (Institute of Atmospheric Sciences and Climate, CNR-ISAC)