Predictions at the seasonal to sub-seasonal scale are important for planning and decision-making in a variety of disciplines, and improving understanding and model skill at this timescale is a key research priority. An as yet underexplored approach to sub-seasonal prediction using data science and graph theory methods that are increasingly common to other fields outside of meteorology and climate science shows potential to improve predictions at this challenging timescale.
A recently accepted Journal of Climate paper by Lu et al., addresses this deficiency by exploring the application of correlation networks to sea surface temperature (SST) and sea level pressure (SLP) to determine if teleconnected patterns could be exploited to inform prediction of precipitation out to 30 days.
In this research, supported by CPO’s CVP and MAPP programs, the authors’ results demonstrate that their approach could potentially be more effective for extended range forecasting of precipitation using SST and SLP as compared to conventional principal component analysis. Additional testing of this method with an expanded CMIP data set is a suggested next step.
To access the paper, visit: http://dx.doi.org/10.1175/JCLI-D-14-00452.1
Novel data science approaches could drive advances in seasonal to sub-seasonal predictions of precipitation
