Abstract
We aim to investigate the potential of using artificial neural networks (ANN) for the prediction of sea surfacetemperature anomalies (SSTA) at seasonal time scales in the North Atlantic. At these time scales, SSTAs havebeen linked to the intensity and genesis of extreme weather events and fluctuations of marine resources, whichhave the potential for significant socio-economic consequences. Thus, providing reliable predictions of seasonalSSTAs can be very beneficial. Here, we aim to evaluate the performance of ANN over traditional methods, aftertraining with both simulated and observed data. Traditionally, seasonal SST forecasts are based on persistence andcommon statistical methodologies, often showing low skill particularly in the subtropics. Among the parametersinfluencing SST variability, previous work has shown that in addition to heat content persistence, SSTAs are alsoinfluenced by convergence or divergence of northward transported heat. This has been shown to improve the SSTAhindcast skill in regions of the North Atlantic, in particular for summer seasonal means. Our first test will involvetraining the ANN to recover this correlation.
