The CERISE project team was well represented at the EMS conference EMS2025 - Home in Ljubljana, Slovenia, 7–12 September 2025.
Gabriel Narvaez Campo (Meteo France), Núria Pérez-Zanón (BSC) and Jonny Day (ECMWF) gave CERISE talks and Jonny Day and Nils Noll (DWD) presented posters.
Gabriel’s talk took place in the hydrology session UP2.2 named " Exploring the interfaces between meteorology and hydrology" (Thursday 11th Sept.) The presentation showed that assimilating observed LAI in a LDAS initialization run leads to reduced bias and increased correlation in the simulated river discharges with respect to observation, especially in Europe and South America. Seasonal hindcasts initialized thereof also produce more accurate streamflow prediction at all seasons over South America and Europe, with some local exceptions. Improvements are also found over North-America except for winter forecasts that exhibit a lower skill, in particular in semi-arid regions and in catchments dominated by snow processes.
Núria presented her work titled “Comprehensive Assessment of Seasonal Forecasts Across Multiple Models and Model Versions” (DOI: 10.5194/ems2025-194) at the conference. The study focuses on the verification of seasonal forecasts and their synthesis through scorecards that evaluate various aspects of forecast performance — including systematic error, association, accuracy, and reliability — using both probabilistic and deterministic metrics. The presentation drew particular interest from the audience regarding the SUNSET tool (DOI: 10.5194/ems2024-361), which facilitates consistent and transparent assessment of seasonal forecast systems.
Jonny presented his work linking an overestimation of soil-moisture atmosphere coupling in climate models limits to regions of poor seasonal forecast skill over land in the Forecasting on subseasonal to seasonal to decadal timescales session. Giving both an oral presentation and a poster led to interesting discussions with other conference participants.
Nils’ poster was in the session OSA3.5 - Forecasting on subseasonal to seasonal to decadal timescales. Nils comments: “ The poster "ICON-XPP in the CERISE project: towards an LDAS including seasonal prediction system" presents the latest efforts at DWD in adding land data assimilation to the seasonal forecast system. Phase 1 included a 2DVAR scheme for assimilating snow depth from SYNOP stations. Experiments were carried out to determine the impact of the assimilation cycle length, with 5-daily assimilation giving better results than monthly assimilation. In a comparison with Phase 0, the impact of the snow assimilation on snow water equivalent leads to more snow in winter and less snow in spring and summer." Nils also showed early results of the impact of leaf area index (LAI) assimilation (Phase 2) on LAI itself, 2m temperature and soil moisture in the first 3 layers.
