Real-Time Data Assimilation for Improving Linear Municipal Solid Waste Prediction Model: A Case Study in Seattle

AbstractA commonly used data assimilation (DA) algorithm, Kalman filter, is integrated with the seasonal autoregressive integrated moving average (SARIMA) model to make a one-step forecast of monthly municipal solid waste (MSW) generation in Seattle. The DA solves the problem that parameters of the forecasting model need to be updated in every forecasting process. The performances of prediction models are compared using mean absolute percentage error (MAPE), root-mean-square-error (RMSE), and 95% confidence interval. The MAPE of the SARIMA model with DA is 0.0422, whereas the MAPE of the SARIMA without DA is 0.0914. A 95% confidence interval of SARIMA without DA keeps increasing, whereas SARIMA with DA remains constant, which means DA raises the stability of SARIMA as time progresses. Results show that DA enables the same MSW prediction model with more accurate and more robust forecast results. The SARIMA parameter updating cycle can be prolonged, which saves time and effort.