Monday 20 Aug 2018: Applications of hybrid seasonal forecasts
Erik Kolstad - Uni Research/Bjerknes Centre, Norway
Skilful seasonal forecasts would have been extremely useful tools for a wide range of users. Insurance companies could tweak premiums and instigate preventative measures if they knew in advance that a coming winter was expected to be colder than normal. Hydropower companies could sell more electricity in the summer if their reservoirs were likely to be refilled in the autumn. Erik leads a research project where the aim is to provide tailored seasonal forecasts to user partners in insurance, hydropower, wind power, solar power, ship routing, and commercial weather forecasting, as well as the Norwegian coast guard. He will talk about his experiences in co-producing forecasts, and in particular the two-way dialogue that has led to ideas for products and services. The project team consists of a mix of climate modellers, statisticians, computer scientists, and business experts. The evolving dialogue between this team and the users has also led to some interesting questions, such as: who owns the research, and where do we draw the line between proprietary and public knowledge? Should our algorithms reside in a "black box" or a "white/transparent box"? Erik will also give a brief outline of the plan for producing the forecasts, which will be based on a combination of dynamical model results and empirical relationships. The latter will be sought by both traditional and novel means, meaning through scrutiny of observations and reanalysis data and machine learning, respectively.