The effect of subsurface temperature variability on the predictability of SST in the tropical Atlantic Ocean

Abstract

The role of thermocouple variability and its effects on tropical Atlantic SST is investigated through the generation of SST predictions using a linear inverse modeling technique developed by Penland (1989). In order to reflect the areas in which this dynamical process is active, the tropical Atlantic is divided into three regions. One is located along the equator in the central to eastern basin, one in the northern tropics along the coast of North Africa, and one in the western basin including both the western equatorial region and Caribbean. The equatorial region was chosen in order to capture the effects of thermocline variability along the equatorial wave guide. SSTs in the western and northern regions are not expected to be influenced by thermocouple variability. Previous and current studies have shown SST in these regions to be influenced by the Pacific ENSO and other atmospheric forcings. Three sets of experiments were conducted. In the first, predictions were generated using only SST as the predictor set, while the second used only heat content (as a measure of thermocline variability) as the predictor set. The third experiment combined both SST and heat content in hopes that adding subsurface information would improve SST predictions in the regions in which dynamical processes are active. Three different data sets were used: COADS observed SST, assimilated heat content and SST generated by Carton et al. (1999), and observed XBT data compiled by White (1995). Various combinations of these data were used in experiment 3. All variations reveal the same results; SST predictions are improved with the addition of subsurface variability in the equatorial region at longer lead times, while the north tropical and western regions show little to no improvement.