The WWRP/WCRP S2S Database includes sub-seasonal to seasonal real-time and re-forecasts (up to 60 days lead time) from 11 operational centres around the world. This dataset is available from the ECMWF data portal since May 2015. The data can be retrieved directly from the web interface or programmatically via a Web-API.
To assess the user satisfaction with the S2S dataset and the quality of the service provided by ECMWF to access the S2S data, we issued a survey in March 2017 to more than 700 users who registered to access (but not necessarily retrieved) S2S data. The survey was open for 3 weeks, during which 16% of invitees responded. The full survey summary report is available as a PDF document.
About 78% of users said they are overall satisfied or very satisfied with the S2S dataset. Looking at specific aspects of the S2S dataset and it's delivery, 76% are at least satisfied with the Web interface to browse and explore the S2S dataset, 74% are at least satisfied with the Web API and almost 80% are at least satisfied with the documentation. Dissatisfaction (6.6%) was highest with the Web API service. (The statistics in this paragraph exclude respondents who answered "N/A".)
Below you find a first response to some of the issues raised.
A common complaint was the speed of getting data. Based on these findings we worked on improving the system performance and its monitoring and proactively contacted users to help them to make their retrievals more efficient. These measures have resulted in more data being delivered, as can be seen from usage statistics now publicly available.
Some dissatisfaction with the S2S dataset can be attributed to its complex structure and the number of re-forecast year/ensemble members/start dates available. The structure of the S2S is outside our control and could only be addressed through recommendations by WMO to operational centres. We will continue to improve the S2S documentation which includes descriptions of the available models, retrieve examples and more.
A number of users would like to retrieve the data in netCDF rather than in GRIB format. We have started working on a conversion. However, because of the complexity of the dataset an automatic conversion has not been completed yet.
Regarding the content of the dataset, the survey results don't show a consensus on the main priority for future developments of the S2S database. For the S2S project, this survey triggered recommendations for the next 5-year plan (S2S Phase 2) to have more 6-hourly surface variables (e.g. 10m wind).