Water Supply Forecast Rodeo: Final Prize Stage

Water managers in the Western U.S. rely on accurate water supply forecasts to better operate facilities and mitigate drought. Help the Bureau of Reclamation improve seasonal water supply estimates in this probabilistic forecasting challenge! [Final Prize Stage] #climate

$400,000 in prizes
Completed jul 2024
34 joined

Final model report submission format

In addition to the cross-validation predictions, you must submit a detailed model report outlining your solution methodology in order to be considered for the challenge's Overall Prizes. Unless otherwise specified, your report should focus on the final version of your model that generates the cross-validation predictions that you are submitting to the Final Prize Stage.

For details on the evaluation criteria, please refer to the Evaluation Criteria section in the Problem Description. You can find the model report due date in the Final Stage Key Dates section of the Home Page. Model reports can can be submitted by clicking on "Report submissions" link in the sidebar.

Report Format

Basic Requirements

  • Up to 12 pages of content, not including references
  • 8.5x11 inch paper size with minimum margin of 1 inch
  • Minimum font size of 11
  • Minimum single-line spacing
  • PDF file format
Submissions are required to be in English, but will not be judged based on English fluency. Judgment will be based on the content and ideas communicated.

Diagrams and visualizations are encouraged wherever they can improve the clarity of your report.

Required Sections

  • Title

  • Abstract

    • Provide a concise summary of your solution, highlighting the key features and algorithm(s) underlying your solution, previous work that your solution builds on, and any novel or unique aspects of your solution.
  • Technical Approach

    • Clearly describe the technical details of the solution per the following subsections.

    • Algorithm and Architecture Selection

      • Describe the model(s) and architecture used in your solution and why you chose them. A block diagram is recommended.
      • What other models did you consider or experiment with? Why did you not choose them?
      • Does your solution incorporate any novel modifications to a "standard" use of the algorithm?
      • What simplifications or improvements could be made to run your model(s) more efficiently without sacrificing significant accuracy?
    • Data Sources and Feature Engineering

      • How did you approach the selection of feature/predictor data sources?
      • Which data sources does your solution use? Why?
      • What other data sources did you consider or experiment with? Why did you not choose them?
      • What preprocessing is done to the raw data to create the features you used?
      • Is there any physical explanation or intuition for the features you are using?
    • Uncertainty Quantification

      • What approach did you use to make predictions for the 0.10 and 0.90 quantiles?
    • Training and Evaluation Process

      • How did you train your model(s)?
      • What approach did you use to validate your model(s)?
  • Discussion of Performance

    • How does your model performance vary across conditions? Please provide associated visualizations and metrics to support your explanations. For example, consider:
      • Location (e.g., elevations, snow climate zones)
      • Time (amount of lead time)
      • Climate conditions (e.g., dry, normal, or wet years)
      • Amount of cumulative streamflow volume (e.g., high or low values)
    • Describe what conditions your model is most accurate and least accurate in.
  • Changes Between Stages

    • What changes, if any, did you make to your modeling approach between the Hindcast, Forecast, or Final Prize Stages?
  • Machine Specifications

    • List the hardware specifications and runtime needed to train your model(s).
  • References