Explore the AI/ML sub-seasonal forecast portal to view all forecasts submitted to the AI Weather Quest. Participants are challenged with providing quintile-based probabilistic forecasts for near-surface air temperature, mean sea level pressure, and accumulated precipitation, targeting lead times of days 19 to 25 (week 3) and 26 to 32 (week 4). Use the guide below to navigate the portal and better understand its functionalities.
Please note, whilst we show forecasts globally, evaluation scores for near-surface air temperature and precipitation only take place across land-dominated regions (>= 50% land cover).
Participants have four days to submit their forecasts (days 1 to 4). On day 5, all submitted forecasts are extracted and pushed to the forecast portal. In other words, forecasts initialised on a Thursday will not be viewable on the forecast portal until the succeeding Monday.
The homepage displays all forecast products submitted by participants. Each product corresponds to a specific variable from an individual model and is represented by a large thumbnail on the right. For instance, if the model ‘bestS2S’ submits forecasts for three variables, three separate products will appear.
Use the left-hand panel to filter products by:
Clicking on a forecast product opens all forecasts submitted to the AI Weather Quest for the corresponding model and variable combination. Each product is titled using the format:
teamname: modelname forecast variable quintile probabilities
For example: bestS2S: Model 1 Mean sea level pressure quintile probabilities.
Below each forecast visualisation, a brief description provides additional context about the forecast shown.
In the left-hand panel, you can refine the displayed map by selecting:
Next to each filter, the overview icon (four cyan blocks) allows you to display all corresponding forecasts within that selection filter simultaneously. For example, clicking the overview icon next to “Quintile interval” will show maps for each quintile range.
In addition, the forecast map includes several interactive features:
Finally, below the forecast visualisation you will find a brief description of the forecast shown.
To highlight quintile probabilities, we’ve developed a custom colorbar centred around climatological values. A light grey band spans the range from 0.15 to 0.25, representing probabilities close to the climatology of 20%.
Below 0.15: Shades transition from grey to black in 0.03 intervals, indicating lower-than-normal probabilities.
Above 0.25: The palette shifts through cyan, blue, and purple in 0.05 intervals, capturing increasingly higher probabilities.
This colour scheme ensures intuitive interpretation of deviations from the expected climatological probability, making it easier to spot significant anomalies.
The AI Weather Quest would like to thank all those in the Forecasts and Services Department who supported the development of this forecast portal. Special thanks go to Cihan Sahin for his comprehensive guidance on the portal creation process and for developing new code to facilitate the integration of new teams and products.