Comparison of GPR and multispectral imaging methods for soil moisture estimation in a Hautes Fagnes peatland
Files
Verbois_Gabriel_84071600_2025.pdf
Closed access - Adobe PDF
- 2.7 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- Peatland restoration assessment requires moisture information that is both spatially explicit and physically interpretable. We test a joint UAV strategy that couples airborne ground penetrating radar (GPR, 105–125 MHz; inversion to dielectric permittivity and Volumetric Water Content (VWC);∼ 2 m2 footprint; 30 cm to 40 cm depth) with multispectral imaging (MSI: NDVI, red-edge NDWI, TVDI). Cross-method agreement is evaluated at the campaign scale using field medians, errors-in-variables regressions (Deming, Passing–Bablok), Lin’s concordance, and Bland–Altman diagnostics, with in-situ TDR/TEROS and local meteorology for context. Across two fields, we find coherent seasonality (mid-summer VWC minima; autumn recovery) and persistent wet–dry zoning. In Field A, TVDI is strongly and inversely related to GPR VWC (R2 = 0.91; Deming slope ≈ −1), enabling a near-bias-free TVDI→VWC calibration; 95 % limits of agreement are ±0.109 m3 m−3 (N = 5 campaigns). In Field B, NDVI and NDWI remain negatively related to GPR (R2 ≈ 0.63/0.43), whereas TVDI becomes positively related (R2 ≈ 0.75) under late-season, energy-limited conditions that decouple surface temperature from subsurface storage; limits of agreement are ±0.115 m3 m−3 (N = 5). The joint MSI–GPR approach reduces interpretative ambiguity by combining high-cadence surface mapping with depth-integrated storage constraints. It supports simple linear calibration in Field A during the growing season and seasonal stratification in Field B, subject to sample-size and thermal-data limitations.