Long-range forecasting, the use, benefits and reliability
Your organization could also benefit from using long-range weather forecasts. This Q & A explains how seasonal and sub-seasonal forecasts can help businesses plan for the future.
An interview with Dr Jan Dutton and Matt Dobson. Dr Jan Dutton is CEO at Prescient Weather Ltd, which is headquartered in the United States. Matt Dobson is a senior meteorologist at MeteoGroup, in the United Kingdom, and specializes in long-range forecasts. MeteoGroup and Prescient Weather partner to offer the World Climate Service seasonal and sub-seasonal weather forecast services.
What is long-range forecasting?
Jan: “Traditional weather forecasting extends to around 10- to 14-days. Looking into the future beyond that time frame is considered long-range, or long-term forecasting. The World Climate Service product provides forecasts and information for weeks, months, and seasons ahead. It’s not possible to predict the exact temperature 3 weeks ahead, let alone 3 months ahead, so we provide customers with probabilistic outlooks. That means we put both probability and confidence measures on our forecasts.”
Why long-range forecasting?
Jan: “It is all about managing risks and opportunities. Lots of companies have to monitor for potential disruptive events, or risks; long-term forecasts provide a means to do that. We supply long-term forecasts for the UN FAO, for example. Our data helps them to prepare for wetter than normal conditions and the possibility of locust outbreaks in sub-Saharan Africa. If we look at opportunities, energy traders can gain an advantage from probabilistic views of future conditions when buying and selling contracts. Opportunities are also available for totally different segments, such as retail shops. Forewarning of conditions detrimental to sales allows you to plan better.”
Matt: “For utility companies, it can be crucial to know the likelihood of strong and sustained climate anomalies well in advance. Here in the UK, one of the water companies was very happy with our seasonal outlook in 2018. Already in March/April we notified them of a strong signal for a much warmer and drier summer season. They were prepared, could strategically plan and were therefore able to meet the water demand throughout the whole summer. Energy traders and analysts are well-established clients. Knowing if there could be a major cold spell coming this winter, or a significant heatwave in summer, can significantly impact the demand, and therefore the price of gas and electricity.”
What kind of products do we offer?
Jan: “The World Climate Service offers not only long-range forecasts, it also has a web-based set of tools enabling commercial/industrial meteorologists to improve their seasonal and sub-seasonal forecasting processes. We pull in all the necessary information to make world class long-term forecasts. We process and present it in such a way that the time users would have had to spend surfing other websites is greatly reduced.”
Matt: “At MeteoGroup, we produce seasonal and sub-seasonal forecast summary reports to make the wealth of expert information provided by the World Climate Service forecast tools accessible to a diverse range of customers. Effectively communicating the key forecast messages, the confidence level and the risk factors, is as important as producing the forecast in the first place. For our forecasts to be useful, it is vital that our customers understand the weather “story” we are trying to convey. We take into consideration all the parameters crucial to our different clients. In addition to written reports, we provide one-on-one forecaster briefings, which helps us further communicate the long-range forecast effectively to our users.
Jan: “Meteorologists and the companies they work for get the most value from our products. The MeteoGroup/Prescient Weather relationship works so well because we provide critical tools for MeteoGroup meteorologists to interpret and create additional products. Matt Dobson adds value to the process by interpreting a vast amount of information and boiling it down to the impacts that customers will find most important. Based on his experience, he draws on the information to create a best estimate of what will happen. We have also created specific products for clients with specialized requirements; for example, the need to have the information presented in terms of a single “risk variable”. So, MeteoGroup and the World Climate Service strive to meet very specific customer requirements.
What sets us apart from other long-range forecasters?
Jan: “One of the things that sets us apart is that the World Climate Service is a one-stop shop for climate analyses, monitoring and prediction. Our analogue forecasting tools tend to be more comprehensive and more advanced than our competitors’. Within the dynamical modelling, forecast model calibration improves accuracy, as does combining several of the world’s leading forecast models to create one single solution. We have extensively analysed the past performance of the models to make corrections which increase accuracy. Calibration is a critical process to ensure forecast probabilities are correct: if we forecast a 60% probability of warmer than normal weather for a certain location, we have to be sure it occurs 60% of the time.”
Matt: Our basic starting point for long-range forecasting is the World Climate Service website. It is very complete, has a giant database of historical data and shows a large range of global teleconnections, meaning it provides insight into the relationships between climatological anomalies worldwide. Prescient Weather and MeteoGroup meteorologists have a wealth of knowledge on this matter, knowing where and how we can add value to the raw model data, and we learn more every day.
Jan: “I have to emphasize that our website provides maximum physical insight, so it’s absolutely no black box or magic. It’s the best and latest forecasting science in action. That is very important to mention.”
What are you very proud of?
Jan: “I am proud that the World Climate Service is viewed in the industry as a market-leading, scientifically-oriented, long-term forecast solution. Customers love the website and the incredible amount of information we make available. For example, the 60 years of climate reanalysis data allows people to study climate and weather patterns and their impacts. I’m proud that we’ve created a computing capability that enables us to find relationships and test methodologies. I see a growing interest among organizations for long-range forecasts and I’m proud to know that we are one of the few companies ready to meet their requirements.”
Matt: “One specific situation comes to mind: the early winter forecast we made in late summer and autumn 2018. We were constantly predicting a mild first half of the winter over Europe, while some of our competitors were predicting early cold over Europe, most likely because of low sunspot activity and El Niño activity in the Pacific. Clearly there were some conflicting indicators. We did well and it was quite a big win. December and much of January were indeed mild.”
How do you make long-range forecasts?
Jan: “There are three main categories of forecasting tools on our portal. We use calibrated numerical weather predictions in our own Multimodel Ensemble, which means we can create worldwide maps showing the probabilities and anomalies of different meteorological parameters for the coming weeks and months (elements such as solar quantity, windspeed, precipitation or temperature). Next to the maps, we use and show analogues, which allow us to look back in time. We have a complex dataset in which you can contrast forecast scenarios with similar situations in the past, from 1951 onwards. The outcoming maps show signals and trends for the long-range. The third category offers the possibility to understand and compare worldwide teleconnections regarding meteorological, oceanographical or other environmental phenomena. All this data and the ability to compare, discover and understand teleconnections and global patterns, that is what’s improving our forecasts every time.”
Matt: “It all starts with the huge amount of World Climate Service data Jan is talking about. Subsequently, it is the meteorologist who understands and interprets this. We have been in this business for over 10 years, so we have our own added knowledge and understanding about what the season’s key influential drivers are likely to be. We also know which elements are more easily picked up by models and which are subtler. We make forecasts for different regions, for different clients and we brief them even if there is no signal or trend to be found. Informing the customer when confidence is low and there are weak and conflicting signals is as important as saying confidence is high and the forecast is very likely to come true. We always try to give our judgement and view of the forecast, rather than sitting on the fence.”
Jan: “We also offer our customers a monthly written document about the seasonal forecast. We have people working here that now have 10 years of experience with this subjective forecast. It’s a case of the human brain being the best machine learning tool on the planet.”
How reliable are long-range forecasts?
Jan: “This is an interesting question. In the science of weather forecast verification, there is a statistic called “reliability” which is a measure of how good a probabilistic forecast is. The idea is that if you forecast warmer than normal conditions with 60% probability, it should happen 60% of the time. With that in mind, we work very hard on the calibration process to ensure our dynamical model-based forecasts are statistically reliable. Most of the time, the models don’t agree on the likely conditions 3 or 4 weeks in advance. But, when they do, it’s a good sign the forecast contains useful information. So, I’d say World Climate Service long-term forecasts are reliable, but you don’t have to take our word for it: we participated in the US Bureau of Reclamation’s Forecast Rodeo, a sub-seasonal forecast contest, and I’m happy to say we ranked first overall. The table below shows the overall skill for seasonal forecasts. A climatological guess would be 33%; you will note we are definitely better than using a dice. Having said all that, seasonal forecasts are hard to objectively make and even more difficult to communicate.
Matt: “Long-term forecasts are obviously more challenging than predicting the weather 2-3 days in advance, so the forecasts are presented in a probabilistic way, as Jan has mentioned. Some customers are less experienced with interpreting probabilistic forecasts than others and so how the message is communicated is key, as is informing them about the level of confidence. I think a key asset of the MeteoGroup, Prescient Weather and the World Climate Service forecasts is the vast amount of forecasting experience we have, combined with industry-leading tools. Together these enable us to effectively inform customers when a long-range forecast is more likely to be correct and when confidence is low, so they are more likely to be able to take action at the correct time and reduce loss or make financial gain.”
Your slogan is: If you knew then, what we knew then …’ Then what?
Jan: “You could have taken action to mitigate risk. You could have made decisions that positively impacted the bottom line, or at least minimized damage to it. Users don’t necessarily have to wait for the short-term forecasts to see an opportunity or a risk developing.”
Will there ever be a time we can fully rely on long-range forecasts, just as we do now with short-range forecasts?
Matt: “I think we can fully rely on long-term forecasts today. The challenge is getting decision makers to think probabilistically and understand how to use the information contained in the forecasts. There is nothing absolute in long-range forecasting, there are only probabilities, which means there are also probabilities of business outcomes. Once you think in those terms, then long-term forecasts can be relied upon.”
“There is nothing absolute in long-range forecasting. There are only probabilities of what is most likely to happen which means there are probabilities of business outcomes. In those terms, long-term forecasts can be relied upon.”
- Matt Dobson, Senior Meteorologist at MeteoGroup.
Interested in long-range or short-range forecasts or have questions?
We are always happy to have a discussion.
To put our money where our mouth is, we participated in the US Bureau of Reclamation’s Forecast Rodeo, a sub-seasonal forecast contest. I’m happy to say we ranked first overall.