Detection of soil compaction by the spatial analysis of vegetation characteristics via UAV and clustering algorithms

Abstract

Soil compaction due to field traffic can reduce yields and increase a fields susceptibility to surface runoff and soil erosion. However, detecting and monitoring compacted soils is time and labor intensive. Additionally, most detection methods focus on few points within a field and usually do not provide information on the spatial distribution of soil compaction and its effects on a field scale. In this study, we aim to present a method for detecting potentially compacted areas on field scale using an unpiloted aerial vehicle (UAV). Using an UAV enable the non-inversive collection of field vegetation patterns want, which can be linked to differences in soil properties and soil structure.In two study regions, we monitored six fields for four years. For each field, at least four UAV-flights were carried out per season. The UAV was equipped with a RGB and a multispectral sensor. Spatial corrections and spectral calibrations were performed with ground control points and calibration targets. To analyze the data for patterns in the vegetation, models of plant height and various vegetation indices (NDVI, SAVI, WDRVI, EVI) were calculated and combined with three different clustering algorithms (k-means, fuzzy k-means, CLARA). To validate the identified vegetation patterns, several field campaigns were conducted to analyze soil texture, saturated hydraulic conductivity, air permeability, bulk density, and yield. Our study shows that patterns in the vegetation can be distinguished by their geometry and orientation. Linear patterns running parallel to the tramlines and planar patterns in the headland indicate potential compaction, whereas rounded patterns are indicators for topography changes. Soil samples within the detected linear patterns overall showed an increase in bulk density, and a decrease in saturated hydraulic conductivity and air permeability. Yield samples were also reduced in those patterns for all studied crops. Thus, we can conclude that UAV analysis is suitable method for soil compaction detection. However, detected patterns can be overlaid by other causes.