TY - JOUR

T1 - Counter-prediction approach to predict the missing values of a spatial series on the example of the dustiness in the snow cover

AU - Sergeev, A. p.

AU - Shichkin, A. v.

AU - Buevich, A. g.

AU - Baglaeva, E. m.

PY - 2023/6/1

Y1 - 2023/6/1

N2 - Currently, an increasingly important and complex task is to study and model the impact of human activities on the environment. Such studies are often founded on data from various screenings and monitoring. It is not always possible to get a complete set of data necessary for modeling. The paper proposes an original approach to predict the missing values of a spatial series. As a basis for testing the proposed technique, we used the nonlinear autoregressive neural network. The essence of the approach is that the final forecast of the model is the weighted average result of two forecasts, which are obtained by the model sequentially trained on the values preceding the predicted area on the left and right. Modeling data were obtained by monitoring the dustiness of the snow cover around the copper pit. To test the predictive accuracy of the approach, we created three spatial series. These were the raw series, the mixed series (randomly mixed values of the raw series), and the Gaussian series (independent values drawn from a normal distribution with the same standard deviation and mean as the raw series). The minimum errors were obtained for the original series.

AB - Currently, an increasingly important and complex task is to study and model the impact of human activities on the environment. Such studies are often founded on data from various screenings and monitoring. It is not always possible to get a complete set of data necessary for modeling. The paper proposes an original approach to predict the missing values of a spatial series. As a basis for testing the proposed technique, we used the nonlinear autoregressive neural network. The essence of the approach is that the final forecast of the model is the weighted average result of two forecasts, which are obtained by the model sequentially trained on the values preceding the predicted area on the left and right. Modeling data were obtained by monitoring the dustiness of the snow cover around the copper pit. To test the predictive accuracy of the approach, we created three spatial series. These were the raw series, the mixed series (randomly mixed values of the raw series), and the Gaussian series (independent values drawn from a normal distribution with the same standard deviation and mean as the raw series). The minimum errors were obtained for the original series.

UR - https://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcAuth=tsmetrics&SrcApp=tsm_test&DestApp=WOS_CPL&DestLinkType=FullRecord&KeyUT=000870654800001

UR - http://www.scopus.com/inward/record.url?partnerID=8YFLogxK&scp=85140253608

U2 - 10.1007/s40808-022-01577-2

DO - 10.1007/s40808-022-01577-2

M3 - Article

VL - 9

SP - 1523

EP - 1530

JO - Modeling Earth Systems and Environment

JF - Modeling Earth Systems and Environment

SN - 2363-6203

IS - 2

ER -