We have seen in earlier lectures how the scientific basis for seasonal climate prediction has arisen from the demonstration of coupling between the slowly evolving ocean and the atmosphere, leading to predictable large-scale seasonal atmospheric circulations that extend over continental regions. Over continental regions, year-to-year changes in rainfall are of particular relevance to many aspects of society. Taking into account the above two considerations, it was perhaps natural for early scientific research on seasonal prediction to often work with rainfall indices that were averages of the seasonal rainfall total over a large continental region. It was known that there were sub-regional aspects to climate, but to make the demonstration that there was any predictability at all, the large-scale area averages were used in pioneering research papers. As seasonal forecasts became more widely available through the 1990s, so interest grew in how they could be better used in decision-making strategies to better manage climate risk at various societal scales. It became clear that while the area-average seasonal rainfall total may have "something to do with" the risk of flooding during a season or the risk of crop failure, there was a need to know exactly what that relationship was, and furthermore, if at all possible, to provide forecast information on climate features that were more closely connected to the problem and that integrated into decision support frameworks.

Regional Climate Outlook Forums (RCOFs) developed in the late 1990s and contributed to enhancing a dialogue between users of climate information, operational producers of information, and climate researchers. Such regional forums have now met on a regular (usually at least annual) basis in many regions in advance of the rainy season to review the outlook for that season, and have evolved their own regional perspective and mode of operation. In most cases, they bring together the three groups mentioned above to consider the possibilities, limitations and needed improvements of the available climate forecast information. Another way in which dialogue between users, producers and researchers of climate information has started to identify the types of detailed information needed by users, is through surveys of user needs. Examples of the information needs that have emerged cover a range of sectors and levels of users from regional policy to smallholder farmers. For example, water resource managers can potentially benefit from reliable forecasts of stream flow into a reservoir. This requires reframing a seasonal rainfall forecast in the context of rainfall over a river basin catchment - which in many instances will quite closely match the spatial scale of demonstrations of predictability already made. However, a manager concerned with flood risk may be interested in the likelihood of flows greater than a certain threshold level occurring at some time during the coming season. This is related to the risk of extreme rainfall events occurring through the season, and, for a quantitatively informed decision, requires the generation of forecast information beyond a seasonal rainfall total. Agricultural applications have needs on a variety of different spatial scales depending on the problem being addressed. It includes the evaluation of information at the farm scale to identify farming risks, and includes information on the risk of a major moisture deficit occurring during the growing season, which in turn is related to the likelihood of dry spells during a season. Health specialists are interested in changes in the environmental suitability for certain diseases during the forecast season. For example, in some situations, changes in the amount of pooled surface water that prevails across the landscape during the season can alter the propensity for mosquitoes and associated diseases such as malaria. In some circumstances, temperature can also influence the distribution of mosquitoes.

The scientific basis for seasonal predictability established a degree of skill in the large scale circulation features. It is therefore reasonable to assume that many of the features described above also have a degree of predictability, though that predictability is currently largely unquantified and depends on how the predictable information in the large scale circulation cascades through the environmental system into the features that matter. The research challenge is to study that cascade - for example, from the seasonal mean large scale circulation (say, change in strength of the average monsoon circulation), to the changes that implies for the intensity and frequency of storms, and the change that implies for peak river flows. The creation of forecast information for such features is commonly called downscaling - here we are downscaling seasonal to interannual climate forecasts.

Downscaling leads us into studying the statistics of synoptic weather events during a season, such as the number of intense storms or the number of dry spells. It is anticipated that these statistics can be modified by the large-scale seasonal mean circulations and (in some special situations over tropical oceans) directly by the SST itself. This is why we believe there will be a degree of predictability of the statistics of the weather features However, since primarily it is only the seasonal mean atmospheric circulation that is modified by the SST, no information is possible regarding the timing of weather events during the season. For example, even if we can say there is an enhanced likelihood of dry spells during the season, we cannot say when they will occur.