IEEE Region 8 Climate Challenges - AI in Enhanced Weather Forecasting

Abstract

Upcoming Webinar - 23-Oct-2024 at 13:00 UTC (1 pm) - "IEEE R8CC AI in Enhanced Weather Forecasting - Overview and Guidelines"

Webinar is focused on providing scope, overview and possibilities on the competition, including explanation of datasets, APIs and giving examples of basic algorithms. It will be given by Hrvoje Novak, PhD, one of the authors of the technical task. The seminar will provide guidelines on using IEEE DataPort to submit your solutions. Additionally, it will have a dedicated Q&A session. The recording will be posted on this page afterwards. You can join it at:

Join Zoom Meeting: https://us05web.zoom.us/j/87589275035?pwd=cSONZUKyyl5ujBALAPkpgbpODFr9rk.1

Meeting ID: 875 8927 5035

Passcode: 8ypTy9

Instructions on how to submit your solution by using IEEE DataPort will be uploaded after the webinar.

As a result of climate change, extreme weather events are becoming increasingly frequent and resulting in a growing need for more accurate real-time updated weather prediction where short term weather forecasting (nowcasting) is gaining critical importance. With availability of real-time open-source data such as Numerical Weather Prediction (NWP) forecasts, satellite and weather radar imagery, and localized weather measurements, new and interdisciplinary possibilities are emerging in the way weather forecasts are generated.

Multi-modal real-time data can now be paired with machine learning approaches to improve the accuracy and reliability of weather predictions. Similar approaches are already being recognized with example initiatives by world’s leading companies and associations in the domain of meteorology and artificial intelligence. The purpose of the challenge is to gather all the experts in the domains, exchange approaches and algorithms, and pinpoint guidelines towards worldwide coverage of improving the accuracy of weather forecasting.

The aim of this competition is to leverage multi-source and multi-modal weather data and combine them with statistic and machine learning algorithms to generate accurate and reliable short-term weather forecasts. Main task of the competition is to generate 7 days ahead weather forecasts, on hourly resolution, for 2 variables in each of the 3 case studies selected as three biomes of Europe, Middle East and Africa (IEEE Region 8): Savanna Preservation, Clean Urban Air and Resilient Fields.

Deadline for submissions - 1 November 2024 at 12:00 (PM, noon) UTC

The competition and IEEE Dataport site is envisioned as an ongoing effort to contribute with learings and datasets in multiple cut-off dates. The first set of cut-off dates is:

Phase 1 - Competition, 1-Nov-2024: Using oepn datasets to generate 7 days ahead weather forecasts for case studies.

Phase 2 - Scoring, 11 November 2024: After one month of competition, the final evaluation will be performed on the real weather forecast, announcing the highscore and competition winners.

Phase 3 - Live Presentation, 27-30 November 2024: Top 5 teams will be invited to the IEEE Humanitarian Technologies Conference 2024 in Bari, Italy, to present their approaches and form a report as 4-page IEEE-style formatted manuscript.

The IEEE Region 8 Climate Challenges is a program of Region-wide competitions aiming at climate change mitigation by technological initiatives. As series of large-scale international competitions from different technical areas, challenges are focused to specific objectives, member focus groups and geographic areas to address the local perspectives. The three challenges are identified to put focus on during the program:

1. Challenge of utilization of AI in enhanced weather forecasting

2. Challenge of disaster-resilient communication

3. Challenge of technical innovations

IEEE Region 8 Climate Challenges is initiated by IEEE Region 8 and supported by IEEE Foundation and IEEE Humanitarian Technologies Board.

Instructions:

Contestants are allowed to and encouraged to utilize all other available open-source data. However, it is the contestant’s obligation to ensure that the data is publicly available and under no commercial licenses. No guidance is given on the method for data fusion and the forecasting algorithm; it can be based on a statistical approach, machine learning, or a combination of different approaches.

The task is completed by submitting 6 vectors of 24 x 7 values, corresponding to predictions of 6 variables, in hourly resolution for the next 7 days. The task will be evaluated over 7-day period after the submissions, on future values in the measurement dataset that occur in the real-world future, and unknown at the time of submission. The most accurate predicted variable will provide 50% of points, and the remaining 50% of points will be provided by joint contribution of all 6 weather variables.

Details on the datesets and competition are given in the documentation.