Evaluating the role of model choice in Sub seasonal to Seasonal predictions for forecast uptake over Southern Africa

Much of the Sub seasonal to Seasonal forecasting (S2S) is conventionally focused on either communicating information to users on model skill development and enhancement. For instance, in Southern Africa, farmers use S2S to aid their selection of planting preparations. However, there is little to no information on the difference that one data set over another has on real world decisions - such as choice of cultivar, choice of crop, planting date, amount of fertilizer, and type of equipment. In this study, we explored a range of individual GCMs and RCMs, and ensemble averages, and examined how farmers' real-life decisions are made, based on the data, where the S2S forecast is an influential factor. Our study used ERA5 as a reference to define natural statistics of variability of climate indices. Here, we treated the model data as perfect models, across multiple years, models, and seasons, and we generated predictions of decision outcomes, with the options of a further combinations of models. We then mapped out a "consequence space" to a range of choices and selections and projected their implications into a climate change context as well. Thus, our study addressed the question of what the impact consequences (such as choosing a cultivar) would be if the model were reality and completely trusted by users. Preliminary results show that prediction skills of models vary across regions and months. The implication of this study is that the information provided by a forecast is more trustworthy. For decision-makers, using this approach, the forecast has increased usefulness though a vague forecast may still be less valuable.