Use of Spectral Reflectance of Arable Soil Surface for Soil Properties Detection

The objective of this study was to find the correlation between spectral reflectance of the arable soil surface and the main parameters of soil fertility in the visible spectral range based on proximal sensing data, to model relationship between soil fertility and spectral reflectance, and to determine whether the discovered correlations can be applied to satellite data. The majority of present methods for collecting information about the main soil agrochemical characteristics in the field are laborious and time-consuming. Moreover, as a rule, heterogeneity within the fields is not taken into account, which results in irrational fertilizer application, soil degradation, and environmental problems. Thus, it is crucial to develop methods that allow obtaining reliable information about soil parameters timely without applying extra efforts in the fields. We examined test field with gray forest soils (Phaeozems Albic). We have applied multiple regression modeling for soil properties prediction. The results demonstrated that each field requires its unique equation model. High correlation was shown for several soil properties. The best linear regression models have shown for exchangeable calcium Radj2 = 0.76, for exchangeable magnesium Radj2 = 0.79, for soil organic matter (SOM) content Radj2 = 0.87, for total Nitrogen, Radj2 = 0.89, for pH of salt extract, Radj2 = 0.859. Using cross-validation, we evaluated the predictive capacity of the spectral indices. A number of spectral indices were proposed to predict the properties mentioned. The use of these indices allows to receive the relevant data (on soil properties) quickly and reduce the fertilizer costs and optimize crop growth. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.