Microbiological indicators of soil quality predicted via proximal and remote sensing.

This work sought to predict soil microbiological attributes based on soil fertility and texture, elemental contents determined by portable X-ray fluorescence spectrometry, and terrain attribute data with and without addition of season (dry or rainy) and phytophysiognomy as auxiliary predictors. Soil samples were collected in both seasons in four phytophysiognomies. Analyses for prediction of basal soil respiration, microbial biomass carbon, metabolic quotient, and microbial quotient were performed. Terrain attributes, total elemental concentrations obtained by portable X-ray fluorescence spectrometry, soil fertility and texture as well as phytophysiognomy and season were used as predictor variables. Prediction models were created via conditional random forest algorithm and validated with leave-one-out cross-validation through coefficient of determination (R2), root mean square error, mean absolute error and ratio percent deviation. The best results were delivered when phytophysiognomy and season were included as predictors. Metabolic quotient, microbial quotient, microbial biomass carbon and basal soil respiration achieved the best prediction using only soil fertility and texture data (R2 = 0.79, 0.66, 0.65, 0.91, respectively). Predictions of basal soil respiration and metabolic quotient using only terrain data achieved R2 values of 0.91 and 0.73, respectively. Elemental concentrations determined by portable X-ray fluorescence spectrometry reasonably predicted two microbiological attributes. It is possible to adequately predict these four microbiological attributes both locally and spatially through terrain and soil properties data. We encourage further investigations on prediction of these and other microbiological attributes under different environmental conditions and at shorter spatial and temporal scales. • Microbiological attributes were accurately predicted from soil and terrain data. • Phytophysiognomy and season (dry or rainy) data improved predictions. • Soil fertility and texture best predicted microbiological attributes. • Terrain data accurately predicted basal soil respiration and metabolic quotient. • Spatial variability map of microbiological attributes enhanced data visualization. [ABSTRACT FROM AUTHOR]