FlowCast is a ‘stand-alone’ software application developed to generate and analyse empirically based seasonal climate forecasts for any type of meteorological, hydrological and agronomic time-series data. It has been developed over a period of ten years to fill a need to customize forecasts to support environmental and agricultural based systems’ modeling and decision-making. It has been designed for use by scientists and decision-makers who have sufficient background knowledge in climate and its drivers.
One could question the need for such software when many forms of seasonal climate forecasts are already available and updated regularly on the Internet. These forecasts are produced by different international meteorological organisations that use their own in-house scientific programming to process the climate information. Unfortunately, these Internet products are typically simplistic in nature; spatially coarse; of limited temporal scale; lacking quality assurance; restricted to rainfall and temperature variables; and not targeting specific decision-makers’ needs. Therefore, software such as FlowCast can be used to customize forecast studies and overcome these limitations.
FlowCast can also integrate with outputs from coupled ocean/atmosphere General Circulation Models (GCMs) by incorporating downscaled predictand data and enhanced/extended predictor information. Recent trends embracing the potential of GCMs would suggest that empirical methodologies might soon be obsolete. For example, Stone et al. (2003, p46) suggested:
“while there is some scope for extracting more information for current empirical forecasting techniques in the short to medium term… there appears to be little scope for major improvements in those techniques in the long term. The future of seasonal climate forecasting now appears to come from the adoption of ensembles of coupled general circulation models”.
However, empirical methodologies can provide comparison and post-processing support of GCM outputs enhancing the newer technologies accessibility to users. At present, the outputs of different GCMs are limited in their form, often contradictory, and are relatively inaccessible and inflexible. Also, dynamical forecasts are currently no more accurate than statistical forecasts, although they have more potential at higher lead times (>3mths), and their accuracy could improve markedly in the next five to ten years with improved technology, and through better capturing of the effects of climate change (Climag Ed. 15, 2008, p3). Even then, empirical tools like FlowCast should still play an important supporting role given their open post-processing nature.
This open nature is both a beneficial asset and a potential liability. There is a real danger of this software being misused by users that little understanding of climate science, and who interpret results based on correlation alone without consideration of the associated driving mechanisms. The software makes it easy to ‘troll’ for best results and to alternate between predictive systems violating conventions on spatial and temporal consistency. This results in artificially inflated forecasting skill leading to over-confidence in the results and poorer decision-making. Therefore, it has always been intended that FlowCast be used only by those with an understanding of these implications, and is not intended for release to the general public. Also, a significant part of the software’s design is in providing temporal and spatial evidence to minimize artificial skill and to empirically support the assumptions of the driving predictor/predictand mechanisms.