Sharing experiences on seasonal forecasting

Contributed by Louise Crochemore

It was at the confluence of the Rhine river and the Moselle river that international speakers gathered to share their experiences on seasonal forecasting and discuss.

The colloquium “Seasonal forecasting – Current Challenges and Potential Benefits for Decision Making in the Water Sector” took place this past week on October 15^th-16^th in Koblenz Germany. It was organized and hosted by the Federal Institute of Hydrology – Bundesanstalt für Gewässerkunde (BfG), the World Meteorological Organization (WMO) and the IHP/HWRP Secretariat.

The presentations from the two days gave us a wide range of current states of affairs on seasonal forecasting in several institutions spread over the world. From these presentations some disparities emerged but the general impression was a coherent approach to seasonal forecasting.

Presenters often agreed on several key recommendations on the use of seasonal forecasts:

• The use of multi-model approaches, whether in terms of climate or hydrological models, was often mentioned and seemed to have almost become a common practice. In addition, several presentations described both statistical and dynamical forecasting approaches, which were either applied along-side or merged (e.g. the Bureau of Meteorology in Australia or the University of Wisconsin-Madison in the United States).
• Along with the use of several climate or hydrological models, the use of several reanalyses to evaluate models was highlighted in Florian Pappenberger’s presentation.
• The need for post-processing of seasonal forecasts from GCMs for meteorologists and pre-processing for hydrologists was stressed by most presenters and presented as a necessary step prior to applying seasonal forecasts from climate models in any study.

Three main methods of dynamical seasonal streamflow forecasting were presented:

• Ensemble Streamflow Prediction,
• Ensemble Streamflow Prediction with a selection of members that are pertinent for the forecast period given external climate indices or states,
• Seasonal hydrologic forecasts generated from using seasonal meteorological forecasts as inputs to hydrological models.

Two key aspects were targeted when dealing with seasonal forecasting:

• Several presentations highlighted the economic objectives behind seasonal forecasting: e.g. reservoir operation in Russia, presented by the Hydro-meteorological center of Russia; hydropower in Sweden, by SMHI; water supply in the United States, by NCAR or navigation in Germany, by BfG.
• Other presentations focused on the expectations of end-users in terms of seasonal information and the communication of this information. This seems to depend on the objectives of the seasonal forecasting but also on common practices of the stakeholders. For instance, when Jean-Pierre Céron from Météo-France mentioned that the end users he works with prefer to receive computed probabilities based on forecast members rather than ensemble members, Jan Danhelka from the Czech Hydrometeorological Institute (CHMI) observed that the end users he is dealing with prefer to compute the probabilities themselves based on the forecast ensemble members they receive. Abdou Ali of AGRHYMET described a stakeholder preference in West Africa for forecasts formed by a consensus of experts, rather than direct forecast system outputs.

Moreover, it seemed that countries or continents are not at the same level in terms of seasonal forecasting acceptance and integration in common practice. For example, Narendra Tuteja from Bureau of Meteorology first set the climate context for the work on seasonal forecasting at BoM: the millennium drought from 1997 to 2009. In such a context, seasonal forecasting became a priority. This led to the completion of an integrated and fully operational tool which is now in use for climate and flow forecasting in Australia. In other parts of the world, it seemed that advanced tools for seasonal forecasting were not a priority either due to a lack of will from local stakeholders to update to newer tools or due to other hydro-meteorological priorities.

During the colloquium, several projects or systems focusing on, generating or making use of seasonal hydrologic forecasts were mentioned, among which EUPORIAS, SPECS, DEWFORA, CHPS, and GLOWASIS.

Interestingly, the presentations ended with a comment from Albrecht Weerts, who presented his main works with forecasts based on Ensemble Streamflow Prediction. He mentioned the hesitations one can have with seasonal forecasts from GCMs that need to be corrected by a factor from 1 to 2 during post-processing.

The presentations were followed by discussions with the objective to initiate a work group on seasonal forecasting. A possible mentioned output from such collaborations could take the form of a guidance document on boundaries for seasonal hydroclimate forecasting within WMO.

And of course, HEPEX was often mentioned and was proudly advertised by Florian and Andy. Johannes Cullmann also mentioned the possibility of a HEPEX-like seasonality initiative (“Seasonalex”? “Seasonex”?). To be continued!

I enjoyed a lot the colloquium, which was very relevant to my PhD research on seasonal forecasting for water supply, carried out at IRSTEA in Antony (France) and as part of the Project Interreg DROP. This post is just a personal impression of the colloquium and any comments, or complementary feedback from participants of the meeting, are welcome.