A Medium-Range Forecast Contest to Bridge the Gap between Academia and the Private Sector

Participants from both academia and the private sector have joined together to develop a new forecast contest focused on the medium range. Forecasters are tasked with predicting population-weighted heating or cooling degree days (depending on season) summed over Days 16–20, where Day 1 is the day the forecast is made. Forecasts are made once a week, and include both a mean and sigma value. Forecasts are evaluated based on the continuous Ranked Probability Skill Score, with the 30-year climate normal used to construct the reference forecast.

One of the main goals of the contest is to provide a forum for students and industry to learn, compete, collaborate, and have fun. Indeed, industry partners, by providing essential degree day data that is otherwise proprietary, make the contest possible, and their forecasts are included in the contest as well. Other forecast inputs include a variety of alternative climate normals and moving averages not necessarily based on a 30-year period, persistence-based approaches, and other objective techniques.

Collaboration is encouraged by requiring all forecasters to include a statement describing how they made their forecasts, in addition to the forecasts themselves. This transparent, “open-source” approach is intended to accelerate learning and simulate a workplace environment. In addition, a weekly forecast discussion for participants is held via videoconference. Those discussions often include special guests from industry who share some of their views on how they tackle forecasts at this lead time.

With approximately a year of forecasts made, this presentation will dive into the preliminary results, showing which techniques seem to be performing well, and providing some anecdotes regarding how student participation, and the subsequent networking that occurs while taking part in the contest, has paid off.