Development of a trend forecasting model for environmental pollution monitoring

Abstract

A complex model for forecasting time series of environmental pollution indicators is described, considering the aggregation of various forecasting models, which are formed based on predictive statistical analysis of pollution indicators and have an adaptive nature. The model differs from known models by providing the possibility of adapting the model parameters to changes in the state of the environment, which is especially important in the conditions of using such models in monitoring systems. Th e complex forecasting model includes higher-order exponential smoothing, Holt, Winters, moving average, weighted moving average, and autoregression models. All the parameters set in these models are related to the Hurst index, which is calculated based on predictive fractal statistical analysis of the time series. Relevant descriptions and justifications are given. Using such a model as part of the econometric system will help predict and respond more effectively to possible changes in the values of pollution parameters. In particular, the persistence of the time series of pollution parameters can mean a stable trend of increasing or decreasing pollution. Suppose the time series becomes close to random or ergodic. In that case, this may mean an emergency or additional erratic emissions in the region that must be monitored. The described model is a forecasting model that is part of the system for monitoring environmental pollution parameters. In the future, a model for forecasting the pollution level in different regions of the People's Republic of China will be implemented.