Spatial- Temporal- TES: Reanalysis Dataset based Short- Term Temperature Forecasting System

In this paper, we propose a Spatial-Temporal-TES system for short-term forecasting of temperature using reanalysis datasets. We use the ERA5 reanalysis dataset generated by European Center for Medium-Range Weather Forecast (ECMWF) over the 2015–2020 period and take advantage of Triple Exponential Smoothing (TES) algorithm which is one of the most well-known algorithms in time series prediction. The evaluation is done on an hourly value basis. Our experimental results show that the forecasting based on ERA5 exhibited good performance, and ensured a promising and reliable forecasting. Additionally, we compare our proposed approach to two other commonly used forecasting techniques: persistence forecasting and average forecasting. Our results reveal that our proposed method achieves a Root Mean Square Error of 0.57K and 1.01K for a forecasting interval of 12 and 24 hours in the future, respectively. It outperforms the other benchmarking methods.