The main purpose of this study is to realize the rapid and non-destructive determination of soil urease activity, so as to provide guidance for soil nitrogen transformation in time.
In this study, five gradient experiments of water regulation were set up under the conditions of multiple cropping of winter wheat and summer soybean. The data of soil urease activity and hyperspectral reflectance were collected. We explored the influence of water regulation on soil urease activity. And based on a variety of spectral transformation algorithms and modeling algorithms, hyperspectral monitoring models of soil urease activity were constructed.
Soil urease activity increased first and then decreased with the aggravation of drought stress. FD, CR, MSC, and SNV transformation can improve the correlation between spectral reflectance and soil urease activity. The accuracy of the models constructed by PLSR and SMLR was high. In the nonlinear algorithm, SPA-ANN based on SNV had the highest accuracy. Among all the models, the PLSR model based on FD had the highest accuracy, with R2v of 0.8564, RMSEv of 0.4013, and RPD of 2.5667. This study can provide technical support for the rapid determination of soil urease activity and provide a theoretical basis for further rational management of farmland.