| Abstrakt: |
Spatial normalized difference vegetation index finds various applications in crop monitoring and prediction. Although this index is mainly aimed to represent the state of vegetation cover, it is suggested that it could be utilized for other remote monitoring purposes, for example, soil humus content monitoring. The study was carried out in 2022-2023 fallow-field period in Kherson oblast, the South of Ukraine, to establish the relationship between the values of bare-soil normalized difference vegetation index and content of humus in the soils of the region. Statistical modeling was performed using the best subsets regression analysis in BioStat v.7 and artificial neural network with back propagation of error algorithm in Tiberius XL. The best performance was recorded for the combined model of cubic regression and artificial neural network, with moderate fitting quality (coefficient of determination is 0.29), and good prediction accuracy (mean average percentage error is 13.22%). The results approve the suggestion of possibility of spatial vegetation index use in soil state monitoring, especially, if further scientific work enhances the fitting quality of the model. [ABSTRACT FROM AUTHOR] |
