Consecutive one-week model predictions of land surface temperature stay on track for a decade with chaotic behavior tracking

Abstract

Temperature prediction over decades provides crucial information for quantifying the expected effects of future climate changes. However, such predictions are extremely challenging due to the chaotic nature of temperature variations. Here we devise a prediction method involving an information tracking mechanism that aims to track and adapt to changes in temperature dynamics during the prediction phase by providing probabilistic feedback on the prediction error of the next step based on the current prediction. We integrate this information tracking mechanism, which can be considered as a model calibrator, into the objective function of the proposed method to obtain the corrections needed to avoid error accumulation. Experimental results on the task of global weekly land surface temperature prediction over a decade validate the effectiveness of the proposed method. Using an information tracking mechanism that provides probabilistic feedback on weekly predictions of temperature variations to calibrate a numerical weather prediction model helps avoid error accumulation over a decade.