Accuracies and costs of prediction and mapping soil properties using proximal sensors: A systematic review

We provide an overview of the accuracy of soil property predictions using the most common proximal sensing (PSS) techniques in precision agriculture (PA), both standalone and in combination with one another or with environmental covariates. Based on 114 scientific papers, we evaluate the accuracy of soil property estimates by calculating the normalized root mean square error (NRMSE) using RMSE values and the range of the predicted soil property. Soil properties, PSS techniques, covariate types, and the model employed for predictions are the factors used to sort the accuracy results. We estimate PSS service costs using both the literature and a market study with questionnaires from private companies in PA. Our analysis indicates that diffuse reflectance spectroscopy (DRS) can estimate the greatest number of soil properties with high accuracy among the PSS techniques. Popular DRS applications include determining soil organic matter, nutrients, and soil texture. X-ray fluorescence (XRF) is the second-most popular technique for estimating soil properties. XRF is widely used in the field to determine elemental concentrations. On-the-go techniques such as electromagnetic induction (EMI) or gamma-ray spectroscopy (γ-ray) yield lower accuracy than point-based techniques. They are widely used by companies because they can delineate PA management zones in the field and are suitable for on-the-go mapping of soil properties such as mineralogy, texture, salinity, water content, cation exchange capacity, and soil depth. The combined use of PSS techniques generally doesn’t outperform the singular application, although the number of samples collected for calibration and the specific combinations of sensors, covariates, and modeling techniques, when correctly applied, may enhance the predictions of soil properties using PSS techniques applied singularly. These outcomes tend to depend on local site characteristics.

According to data, the estimated cost of surveying a hectare with PSS oscillates between 15.5€/ha and 130€/ha, whereas our company survey yielded an interval of 142-362€/ha. Price variability was influenced by personnel costs, fieldwork, data and reporting, sample analysis, and equipment. Increases in final prices can be attributed to accessibility and difficulties related to fieldwork and travel to the area of interest. This work aims to serve as a reference for the adoption of sensing technologies by farmers, policymakers, and companies, providing insights into the suitability of different PSS techniques for soil mapping, their associated costs, and what is available in the market. We foresee that PSS will become the standard approach for producing high-resolution maps and affordable soil property information in the future.

 

Acknowledgments

Work funded by the European Union’s Horizon 2020 Research and Innovation Program under Grant Agreement Nº 862695, and from the Tillämpninggsklivet Precisionsodling RUN 2021-00020 Region Västra Götaland