JR
Members
First name (team leader)
Joyce
Last name
Leung
Organisation name
Individual
Organisation type
Other (please specify)
Organisation location
United States of America
First name
Reo
Last name
Sze
Organisation name
Individual
Organisation type
Other (please specify)
Organisation location
United States of America
Model
Model name
slowMamba
Number of individuals supporting model development:
1-5
Maximum number of Central Processing Units (CPUs) supporting model development or forecast production:
< 8
Maximum number of Graphics Processing Units (GPUs) supporting model development or forecast production:
< 4
How would you best classify the IT system used for model development or forecast production:
Single node system
Model summary questionnaire for model slowMamba
Please note that the list below shows all questionnaires submitted for this model.
They are displayed from the most recent to the earliest, covering each 13-week competition period in which the team competed with this model.
Which of the following descriptions best represent the overarching design of your forecasting model?
- Post-processing of numerical weather prediction (NWP) data.
- Machine learning-based weather prediction.
- Hybrid model that integrates physical simulations with machine learning or statistical techniques.
- Ensemble-based model, aggregating multiple predictions to assess uncertainty and variability.
What techniques did you use to initialise your model? (For example: data sources and processing of initial conditions)
Reanalysis inputs from ERA5 (daily fields: pr totals; tas/mslp means), bilinearly regridded to 1.5°, longitudes normalized to 0–360, and aggregated to the week3/4 windows. Basic QC on dims/coords before feature building.
If any, what data does your model rely on for real-time forecasting purposes?
ERA5 updates for the latest days used to build weekly predictors; AI-WQ 20-year weekly quintile climatologies for tas/pr/mslp for calibration and as a reliability anchor.
What types of datasets were used for model training? (For example: observational datasets, reanalysis data, NWP outputs or satellite data)
Reanalysis (ERA5) for predictors; AI-WQ weekly observations and 20-year quintile climatologies for targets/labels and verification alignment.
Please provide an overview of your final ML/AI model architecture (For example: key design features, specific algorithms or frameworks used, and any pre- or post-processing steps)
A lightweight Transformer backbone generates a multi-member ensemble of week3/4 forecasts for tas, pr, and mslp on a 1.5° grid. We derive quintile probabilities by comparing the ensemble distribution to 20-year climatology quintile thresholds, then apply a light reliability blend toward climatology.
Have you published or presented any work related to this forecasting model? If yes, could you share references or links?
No.
Before submitting your forecasts to the AI Weather Quest, did you validate your model against observational or independent datasets? If so, how?
Yes, we calculated internal RPSS against AI-WQ weekly observations mapped to quintile categories using the same 20-year climatology; scored on a 1.5° grid with a land–sea mask. Routine checks: probability sum to 1, no NaNs, range sanity, and basic reliability diagrams.
Did you face any challenges during model development, and how did you address them?
Precipitation reliability at week3/4: mitigated with stricter daily accumulation handling, increased climatology blending under low-signal, and additional local neighborhood features in the calibrator.
Dim/coord inconsistencies across files: fixed with a standardized loading/regridding pipeline.
Are there any limitations to your current model that you aim to address in future iterations?
Precipitation remains the most uncertain target at long leads. We plan to work on region-aware calibration, adaptive blending weights, and a lightweight MoE head for regime shifts.
Are there any other AI/ML model components or innovations that you wish to highlight?
Minimalist attention stack for efficiency; shared multi-task head across tas/pr/mslp; simple but effective confidence gate, blend-to-climatology for robust probabilistic reliability.
Who contributed to the development of this model? Please list all individuals who contributed to this model, along with their specific roles (e.g., data preparation, model architecture, model validation, etc) to acknowledge individual contributions.
Joyce: data preparation, model validation
Reo: model architecture
Submitted forecast data in previous period(s)
Please note: Submitted forecast data is only publicly available once the evaluation of a full competitive period has been completed. See the competition's full detailed schedule with submitted data publication dates for each period here.
Access forecasts data