A debris-flow forecasting method with infrasound-based variational mode decomposition and ARIMA

Infrasound, known for its strong penetration and low attenuation, is extensively used in monitoring and warning systems for debris flows. Here, a debris-flow forecasting method was proposed by combining infrasound-based variational mode decomposition and Autoregressive Integrated Moving Average (ARIMA) model. High-precision infrasound sensor was utilized in experiments to record signals under twelve varying conditions of debris flow volume and velocity. Variational mode decomposition was performed on the detected raw signals, and the optimal decomposition scale and penalty factor were obtained through the sparrow search algorithm. The Hilbert transform, rescaled range analysis, power spectrum analysis, and Pearson correlation coefficients judgment criteria were employed to separate and reconstruct the signals. Based on the reconstructed infrasound signals, an ARIMA model was constructed to forecast the trend of debris flow infrasound signal. Results reveal that the Hilbert transform effectively separated noise, and the predictive model’s results fell within a 95% confidence interval. The Mean Absolute Percentage Error (MAPE) across four experiments were 4.87%, 5.23%, 5.32% and 4.47%, respectively, showing a satisfactory accuracy and providing an alternative for predicting debris flow by infrasound signals.