Spatial pattern consistency and repeatability of proximal soil sensor data for digital soil mapping

Data from proximal soil sensors can facilitate digital soil mapping at high spatial resolutions. However, their use for predicting static soil properties, such as texture, is affected by spatio-temporal changes in environmental and measurement conditions. In this research study, seasonal changes in spatial patterns and repeatability of data provided by a platform that simultaneously measures the red (Red) and near infrared (NIR) reflectance, apparent soil electrical conductivity (ECa), temperature, and volumetric moisture content of topsoil (at 3–6 cm depth) were assessed. Test fields are located in Southern Finland with textures dominated by clay and fine sandy till. During single scans, mean relative differences between the data from duplicated measurement points ranged from ~4% to 6% and were the highest for temperature and Red values. The consistency of spatial patterns across seasons (spring and autumn 2021 and 2022) was the highest for ECa values, and the lowest for NIR. ECa and moisture were significant for predicting the clay contents at a cereal grain crop site, whereas temperature was significant at grass ley sites. Errors were generally lower when using spring data compared with autumn data (RMSE ranging from 4.8% to 11.1% for the data from different fields and measurement dates). For the fields, where static soil properties change at small spatial scales, spatially detailed moisture and temperature data support the understanding of seasonal changes in the spatial patterns derived from multi-sensor data, and the corresponding changes in the performance of calibration models.