Ecological Systems and Devices Annotation << Back THREE-DAY FORECAST OF METHANE CONTENT IN THE ATMOSPHERE OF THE ARCTIC BELY ISLAND (ARCTIC REGION OF RUSSIA) USING DISCRETE WAVELET TRANSFORM AND ARTIFICIAL NEURAL NETWORKS I.E. Subbotina, A.P. Sergeev, A.V. Shichkin, A.G. Buevich, E.M. Baglaeva The paper proposes a hybrid model combined wavelet transform and artificial neural networks for short-term forecasting of time series of hourly average methane concentration. The paper aims to improve the accuracy of predicting the dynamics in the surface concentration of greenhouse gases. The model was tested on the data of greenhouse gas monitoring on Bely Island, Yamalo-Nenets Autonomous Okrug (Russia), obtained in July-August 2017. Using the discrete wavelet transform, the methane concentration time series were decomposed into five components (one approximating and four detailing). To train five autoregressive networks with exogenous input (NARX), these components were used simultaneously with several meteorological parameters - temperature, humidity, and pressure. The methane concentration forecast was calculated as the sum of the forecasts for each of the five components. The quality of the forecast was assessed using several indicators of model accuracy. The hybrid approach improves the forecast accuracy by more than 20 per cent compared to the neural network prediction. Keywords: artificial neural networks, time series, wavelet transform, greenhouse gases, methane, atmosphere. DOI: 10.25791/esip.3.2022.1290 Pp. 48-55.