Back to search

IKTPLUSS-IKT og digital innovasjon

Seasonal Forecasting Engine

Alternative title: Sesongvarslingsmotor

Awarded: NOK 15.9 mill.

Weather forecasts are rarely accurate more than five to ten days ahead. This is due to the chaotic nature of the weather, also known as the butterfly effect. Our project, Seasonal Forecasting Engine (SFE), has mainly revolved around using physical and statistical models to develop methods for estimateing the likelihood of different weather scenarios from 10 to 100 days ahead. Our overall aim has been to make such seasonal forecasts more relevant for users in the private and public sectors. From the start of the project, we worked closely with the user partners, including the insurer Tryg forsikring. In the spring of 2018, there was a large amount of snow in the mountains in southern Norway, and at the same time we had forecast that April would probably be colder than normal. This meant that there was a risk of a delayed snowmelt, and thus there was an increased risk of heavy flooding in eastern Norway. Based on this, Tryg sent out a press release in which they warned of a high risk of flooding, and in addition they sent an e-mail to all their customers in eastern Norway. We were also interviewed together on TV2, which made even more people aware of this information. This is an example of how long-range weather forecasts can be used. Through several such examples, we established good collaborations, and we decided to apply for funding for a new Centre for research-driven innovation (SFI), with a focus on predicting and managing climate risk. In 2020, we were granted Climate Futures, which is coordinated by NORCE and has 30-40 partners. Among these were several new research partners. In addition to the Norwegian Computing Center, the Nansen Center and the Geophysical Institute at the University of Bergen, we joined forces with the Meteorological Institute, the Norwegian School of Economics, the institute SNF, and Statistics Norway. Climate Futures has a more interdisciplinary composition than the original SFE project. The project group decided to merge the activities with Climate Futures, and since October 2020 we have operated under a common banner. In addition to some of the original partners in the SFE project (Tryg Forsikring, StormGeo, BKK, Agder Energi), we have recruited many new partners in four sectors: Resilient societies, Renewable energy, Sustainable food production and Smart shipping. A common feature for the partners is that they all need better forecasts from ten days to several years ahead. By the end of the SFE project, we had thus established a large and relevant user group, with opportunities to develop exciting and applied projects together with these in Climate Futures. In addition, in 2020 the project group was granted an EU project (Horizon 2020) for the development of climate services in Africa, with partners from Kenya, South Africa, the United Kingdom and Germany. CONFER (Co-production of climate services for East Africa) is coordinated by NORCE and builds on the knowledge that has been developed in the SFE project. At the end of 2021, there were around 20 ongoing projects in Climate Futures. All of these include at least one user partner. Within Sustainable Food Production, the following projects were underway: improvement of forecasts based on the dry summer of 2018; forecasts of sea temperature for the aquaculture industry; and forecasts of the first frost in autumn. Within Smart shipping, we researched, among other things, the use of sea ice distribution forecasts; the effect of sea temperature on fouling of hulls at port calls and subsequent fuel consumption; and forecasting of fuel consumption on shipping routes. In Renewable Energy we explored the use of seasonal forecasts in hydrological models; the use of multi-year climate forecasts for the hydropower sector; and the use of seasonal forecasts in projections of electricity prices. Within Resilient Societies, the following projects were active: How to understand and assess climate risk? Linking seasonal forecasts to insurance data; and climate risk management in finance. Furthermore, there were several cross-cutting projects underway, including the improvement of forecasts from NorCPM, our in-house model which performance has been improved in the SFE project; preparations for the use of monthly forecasts and weekly summaries on Yr; and research on language use - how do we talk about and understand uncertainty? The SFE project has also led to 16 peer-reviewed scientific articles and a total of 88 mass media contributions. In particular, our monthly seasonal forecasts, published on klimavarsling.no, a website we will continue to use in Climate Futures, has led to great interest. This website was viewed more than 23,000 times in 2021. The media's interest in the project has led to many useful contacts and has provided us with an excellent platform for dissemination.

Hovedpåvirkningen for forskningsorganisasjonene var at SFE-prosjektet innledet et svært fruktbart samarbeid som resulterte i påfølgende finansiering av blant annet CONFER og Climate Futures. SFE som plattform bidro også til at Bjerknes Climate Prediction Unit fikk støtte av Universitetet i Bergen og Trond Mohn-stiftelsen. Samarbeidet mellom organisasjonene og størrelsen på forskningsgruppene i organisasjonene har vokst betydelig. For eksempel har NORCEs klima- og miljøavdeling nå en dedikert gruppe som heter «Forecasting Engine», der det nå er mer enn ti forskere. For privat og offentlig sektor har arbeidet i SFE satt klimavarsling på kartet i Norge. Jevnlig formidling av sesongvarsler på klimavarsling.no har ført til omfattende kontakt med aktører fra begge sektorer. I flere tilfeller har dette ført til et formelt samarbeid i Climate Futures.

The aim of SFE is to develop a state-of-the-art operational seasonal climate prediction system for Northern Europe and the Arctic. Tailored seasonal predictions can be helpful tools for risk mitigation, and they can guide more efficient use of resources in many sectors of society, including agriculture, energy, water, transportation, and insurance. To our users, the SFE will be accessible through a flexible interface which can be queried to obtain predictions of relevant climate indices and variables. Under the hood, our "engine" consists of statistical algorithms that merge vast amounts of data into unified forecasts. As we increasingly understand the mechanisms that drive the enormously complex climate system, both dynamical and empirical models are steadily improving. At the same time, increased computational power, enhanced observations and remote sensing, and advanced statistical methods to blend models and observations, are driving a big data-fuelled revolution in climate prediction. What is urgently needed now is careful, but speedy, transformation of research into innovative practical applications and services. Our team consists of experts in handling big data, statisticians, climatologists, climate modellers, and climate service practitioners. Working in complementary fashion with the international research community, and guided by an international peer advisory committee, we will both improve our own models and, taking existing seasonal forecast ensembles, employ innovative empirical-statistical approaches to make the forecasts better and more relevant for our users.

Publications from Cristin

Funding scheme:

IKTPLUSS-IKT og digital innovasjon