« Back

Abstract:

Nowadays, within the Carpathian region (Ukraine), the main factor characteristics determining the temporal prediction shall be the sunspot activity (the Wolf number) and seismic activity (the total energy released in earthquakes). Engineering-geological zoning, as a method of representative analysis of geological data, is used as a spatial component, which includes the grouped structural-tectonic, geomorphological, morphometric and climatic features of a specific engineering-geological area. The article proposes a new approach to the temporal prediction within individual engineering-geological areas. For the selected study area, we have performed a statistical analysis of factor characteristics, which are a measure of factor determinability to choose the optimal model of long-standing temporal analysis within the engineering-geological areas of the Carpathian region of Ukraine. To substantiate the methodology of spatial-temporal analysis, we have patterned the data regarding the landslide activity, sunspot activity (the Wolf number), groundwater levels, an average of annual temperature, seismic activity (total energy released in earthquakes), an average of annual rainfall within the certain engineering-geological area of the Transcarpathian inner depression. Statistical analysis of data based on the consistent study of the factor characteristics. At the same time, we have performed the correlation analysis with the allocation of autocorrelation functions as a basis for considering temporal dynamic features of each of the factor characteristics. Based on them, we have drawn up the periodograms of main climatic factors such as an average of annual rainfall, an average of annual temperatures and an average of groundwater level. Under the spectral analysis, we distinguished the main periodicities by summarizing the data at the point of landslide manifestation and three points in the form of a function of the temporal series. There was a comparison of the temporal series based on PREDICT function of MathCad integrated mathematical package and neural network prediction. We determined that the basic frequency components of temporal activation dynamism of landslide processes for certain engineering-geological area of the Carpathian region of Ukraine can be divided into the cycles having a period of small and large activation. Under the long-term prediction, the main cycle of the landslide activity is 28 years. In particular, the neural network prediction distinguished the highest probability of the landslides in 2020-2021 of the first half cycle (2019-2023) and 2028-2030 of the second half cycle (2027-2032). The prediction based on Predict function highlighted the period of 2020 – 2030. Therefore, the next massive activation of the landslides will be in 2020-2023 and in 2028-2030.