SoilX GPR Soil Moisture Loam

We conducted a field data collection campaign at Worcester Polytechnic Institute (WPI) to support machine learning models for predicting soil water content at multiple depths. The experiment took place in a controlled open-air testbed created by excavating a 30 \u00d7 20 \u00d7 3 ft section of ground and filling it with homogeneous loamy soil, representative of typical New England farm topsoil.To record radar data, we used an off-the-shelf AKELA Stepped-Frequency Continuous Wave (SFCW) Ground Penetrating Radar system. The radar transmits over 4096 evenly spaced frequencies ranging from 0.4 to 2.0 GHz, with each frequency step set to 40 kHz. This range was selected to achieve strong signal penetration and fine resolution for shallow soil analysis. The radar was mounted on a rail system and elevated at three heights above the soil surface: 34, 57, and 79 inches.Measurements were taken at six probe locations arranged in three pairs labeled A, B, and C. For each probe pair, the radar was positioned centrally and performed approximately 100 individual scans at each height. These scans were averaged to create a clean signal representation for each position and elevation.To obtain ground truth soil moisture data, we used a Dynamax PR2\/4 probe to measure volumetric water content at depths of 10, 20, 30, and 40 centimeters. Measurements were taken at each of the six probe locations, and the values were averaged across each pair to match the corresponding radar readings.Data collection occurred during two separate field campaigns, one in October 2022 and the other in May 2023. These campaigns captured a range of soil moisture conditions, especially due to significant rainfall prior to the October collection. In total, we collected 92 matched sets of radar signals and soil moisture readings, covering different depths, locations, and measurement heights.