Introduction to seasonal forecasting
Seasonal forecasting is the attempt to provide useful information about the "climate" that can be expected in the coming months. The seasonal forecast is not a weather forecast: weather can be considered as a snapshot of continually changing atmospheric conditions, whereas climate is better considered as the statistical summary of the weather events occurring in a given season.
Numerical Weather Prediction (NWP) provides useful information for up to approximately 10 days in the future. It is based on solving a complex set of hydrodynamic equations that describe the evolution of the atmosphere , subject to the initial atmospheric state and initial conditions at the Earth's surface . Since the initial state is not known perfectly, all forecasts begin with estimates. Unfortunately the system is very sensitive to small changes in the initial conditions (it is a chaotic system) and this limits the ability to forecast the weather beyond 10-14 days.
Despite the chaotic nature of the atmosphere, long term predictions are possible to some degree thanks to a number of components which themselves show variations on long time scales (seasons and years) and, to a certain extent, are predictable. The most important of these components is the ENSO (El Nino Southern Oscillation) cycle which refers to the coherent, large-scale fluctuation of ocean temperatures, rainfall, atmospheric circulation, vertical motion and air pressure across the tropical Pacific. It is a coupled ocean-atmosphere phenomenon centered over the tropical Pacific but the scale of the fluctuations is quite vast, with the changes in sea-surface temperatures (SSTs) often affecting not just the whole width of the Pacific but the other ocean basins too, and the changes in tropical rainfall and winds spanning a distance of more than one-half the circumference of the earth. El Niño episodes (also called Pacific warm episodes) and La Niña episodes (also called Pacific cold episodes) represent opposite extremes of the ENSO cycle. The ENSO cycle is the largest known source of year-to-year climate variability.
Changes in Pacific sea surface temperature (SST) are not the only cause of predictable changes in the weather patterns. There are other causes of seasonal climate variability. Unusually warm or cold sea surface temperatures in the tropical Atlantic or Indian ocean can cause major shifts in seasonal climate in nearby continents. For example, the sea surface temperature in the western Indian Ocean has a strong effect on the precipitation in tropical eastern Africa, and ocean conditions in the tropical Atlantic affect rainfall in northeast Brazil. In addition to the tropical oceans, other factors that may influence seasonal climate are snow cover and soil wetness. When snow cover is above average for a given season and region, it has a greater cooling influence on the air than usual. Soil wetness, which comes into play most strongly during warm seasons, also has a cooling influence. All these factors affecting the atmospheric circulation constitute the basis of long-term predictions.
To summarize, seasonal forecasts provide a range of possible climate changes that are likely to occur in the season ahead. It is important to bear in mind that, because of the chaotic nature of the atmospheric circulation, it is not possible to predict the daily weather variations at a specific location months in advance. It is not even possible to predict exactly the average weather, such as the average temperature for a given month.
Statistical and dynamical approaches
The starting point for seasonal forecasting is a good knowledge of climate, that is, the range of weather that can be expected at a particular place at a particular time of year. Beyond a simple knowledge of climatology, statistical analysis of past weather and climate can be a valid basis for long-term predictions. There are some regions of the world and some seasons when statistical predictions are quite successful: an example is the connection between the rainfall in March-May in the Nordeste region of Brazil and the sea surface temperatures in the tropical Atlantic in the months before and during the rainy season.
In theory a very long and accurate record of the earth's climate could reveal the combined (and non-linear) influences of various factors on the weather, and analysis of many past events could average out the unpredictable parts. In practice the 50-100 year records typically available represent a very incomplete estimate of earth's climate. In addition seasonal predictions based on past climate cannot take full account of anthropogenic or other long term changes in the earth's system, such as the potential impact of global warming.
An alternative approach is to use the numerical weather prediction method by solving the complex set of hydrodynamic equations that describe the evolution of the Earth's climate system. For a seasonal forecast it is important to consider both the atmospheric and oceanic components of the Earth's system. In fact, the air-sea interaction processes that describe the complicated interchange between the atmosphere and ocean are essential to represent the ENSO cycle. Just as for synoptic range NWP forecasts, the calculation depends critically on the initial state of the climate system, particularly the tropical Pacific ocean for ENSO. Because of the chaotic nature of the atmosphere, a large number of separate simulations are made. They will all give different answers as regards the details of the weather, but they will enable something to be said about the range of possible outcomes, and the probabilities of occurrence of different weather events.
If the numerical models were very realistic, and if very large ensembles of such calculations could be performed, then the "climate" (i.e. the probability distribution of weather) to be expected in the coming months would be accurately described. To the extent that predicted "climate" differs from normal because of the initial conditions of the ocean/atmosphere/land-surface, the ensemble calculations could predict the correct seasonal forecast "signal". Unfortunately there are a number of problems that limit the seasonal forecast skill. Numerical models of the ocean and atmosphere are affected by errors, not all aspects of the initial state are well observed, and techniques for estimating the extra uncertainty that this introduces are still incomplete.
How reliable are today's seasonal forecasts?
The principal aim of seasonal forecasting is to predict the range of values which is most likely to occur during the next season. In some parts of the world, and in some circumstances, it may be possible to give a relatively narrow range within which weather values are expected to occur. Such a forecast can easily be understood and acted upon; some of the forecasts associated with strong El Nino events fall into this category. More typically, the probable ranges of the weather differ only slightly from year to year. Forecasts of these modest shifts might be useful for some but not all users.
The benefits of seasonal forecasting are most easily established in forecasts for some areas of the tropics. This is because many tropical areas have a moderate amount of predictable signal, whereas in the mid-latitudes random weather fluctuations are usually larger than the predictable component of the weather. The point at which seasonal forecasts become good enough to be useful to a particular user will depend on the user's requirements. In some cases, today's systems are already useful, although care should always be taken to interpret model outputs appropriately. As reliability continues to improve, a wider range of applications should become possible, and the value of seasonal forecasts will further increase. More work is still needed to relate probabilities of large-scale weather patterns to detailed impacts and applications. It must be remembered, however, that there are tight limits on what it is physically possible to achieve with a seasonal forecast system. It will only ever be possible to predict a range of likely outcomes. In many cases this range will be relatively large, and there will always be a risk of something unexpected happening. In many parts of the world, most of the variability in the weather will remain unpredictable.
Some seasonal forecasts available today are issued with probabilities (or error bars) which have been properly calibrated against past cases. An example is the Canonical Correlation Analysis (CCA) prediction of El Nino variability, which is regularly shown in the NOAA Climate Diagnostics Bulletin. Such forecasts are probably fairly reliable, but they have very wide error bars: they may state that in 6 months time there might be strong El Nino conditions, or fairly strong La Nina conditions, or anything in between. The outputs of seasonal forecast models generally have less spread but are also less reliable.
A proper calibration of a forecast system against data is not always easy to do. This is primarily because of the limited availability of past data. The problem is especially severe when the level of predictability is low so that many years of data are needed. Relatively low predictability on the seasonal time scale is a feature of much of the globe, but especially in mid- latitudes, and for smaller spatial scales (several hundred km, rather than several thousand). At the moment the ECMWF seasonal forecasts are not issued with calibrated probabilities. However, information about the reliability seen in past performance is available, in plots displayed together with the forecast products. The limited number of past forecasts means that we can only give a rough estimate of the reliability, particularly for smaller regions or local values. It is clear that the direct model output is still quite some distance from being perfectly reliable, although the level of reliability is improving.