Weather prediction improved leaps and bounds during the satellite era, and in recent decades computational capabilities have grown immensely.
However, forecasting the details of weather (rainfall amounts and locations for example) has fundamental limits, meaning we are unlikely to be able to reliably look at the specifics much further than couple of weeks into the future.
So how do we gain value from seasonal forecasts that look further into the future?
We’ve all seen the tabloid headlines talking about predictions of oncoming cold winters or hot summers. These stories normally start with the words “forecasters say”, but often it is this first stage of communication that can lead to a misinterpretation of what a seasonal forecast is really telling us.
We are used to hearing weather presenters talk about short-term weather forecasting rather than longer-term monthly or seasonal forecasting.
The general public is often only exposed to a journalist’s interpretation of monthly to seasonal forecasts, which is in turn an interpretation of the modelled environment from which the forecasts were produced.
The production of these seasonal forecasts by some of the major powerhouses of weather and climate science—including
- National Oceanographic and Atmospheric Administration (NOAA) in the U.S.
- International Research Institute (IRI) in the U.S.
- Bureau of Meteorology (BOM) in Australia
- APEC Climate Centre (APCC) in Korea
- European Centre for Medium Range Forecasting (ECMWF) in Europe
- Met Office in the UK
—is a complex process, and each centre has its own nuance and individuality – therefore, intelligent interpretation and considered use of these forecasts is essential.
Short-Term vs. Seasonal Forecasts
Interpretation is the key here – seasonal forecasts are constructed and delivered differently to the normal short-term weather forecasts that we are used to seeing and using every day.
Short-term weather forecasts of up to a week or so have improved greatly since the satellite era and therefore we have high expectations of their accuracy.
They can determine whether it will rain in your back garden at a certain time of day with an impressive level of accuracy, and they are available in various forms with a swipe on the screen of a smartphone.
Seasonal forecasts need to be approached differently. They do not determine the weather at a set location, instead they tell us about the likelihood of shifts from the normal climate conditions.
This allows for the regional detail to be maintained as every location or region has its own expected climate. A classic quote springs to mind at this point, used by climatologists and meteorologists the world over to suggest the difference between weather and climate – Mark Twain posited “Climate is what we expect, weather is what we get”.
So our expectation being based on history is categorised in to climatological thirds for basic weather variables (such as temperature or precipitation).
The categories are generally called something similar to ‘increased likelihood’, ‘normal’, and ‘decreased likelihood’ and are relative to past climate.
Once we have this probability distribution, the seasonal forecast will only tell us whether there is a shift in the probability towards one or other of these categories.
If it describes an enhanced probability of wetter-than-average conditions, then we should also assume a lower chance of normal and drier conditions.
The shift in probabilities can also tell us about extremes—this is currently an area of fervent research.
Here Comes the Science
The scientific process behind seasonal forecasting involves using real observational data, homogenised and corrected to remove biases, merged into physically consistent general circulation models (GCMs) of the atmosphere and ocean to simulate the climate in as much detail as is computationally affordable.
GCMs today can capture variability in the climate system linked on a global-scale from different modes, such as El Niňo-Southern Oscillation (ENSO), North Atlantic Oscillation (NAO), Pacific Decadal Oscillation (PDO), etc.
Either alone or through combinations of these modes of variability we can start to glean what might happen, and when conditions collude in favour of wetter/drier/warmer/cooler conditions, essentially ‘loading the dice’ towards one or other of the categories in the probability distribution.
This means that although any outcome is still possible, there is an increased likelihood of a particular category.
Seasonal forecasts are best constructed from a consensus of different GCM results. Since no model can perfectly model every detail of the atmosphere, all models are essentially “wrong” but because they are “sub-perfect” in different ways, using an average or a consensus can smooth out the margins of error.
Seasonal Forecast Summary: Go-to Guide and Updates
A ‘go-to guide’ that explains some of the science behind seasonal forecasting has recently been produced by the Willis Research Network (WRN) and Willis Re to supply Willis and our clients with up-to-date seasonal forecast information.
The team will also summarise the latest available seasonal forecasts from the main centres around the world on a global and regional scale, to help inform a view on the likelihood of specific climate conditions in the next few months to a year.
The ‘go-to guide’ explains some of underpinning science, but also describes how we may be able to use this information in reinsurance and capital markets. As seasonal forecasting improves, opportunities to apply this information will develop.
Higher-resolution climate models will continue to develop and will more intricately represent the detail of the atmosphere, so it is realistic to expect improving forecast skill in the future.
Seasonal forecast summary updates will be made available periodically to furnish Willis associates and their clients with the latest thinking on strongest climate signals represented in the forecast for the coming months.