Accurate Measurements of Forest Soil Water Content Using FDR Sensors Require Empirical In Situ (Re)Calibration
Bruno De Vos, Nathalie Cools, Arne Verstraeten, Johan Neirynck
Abstract
Monitoring volumetric soil water content (θv) is the key for assessing water availability and nutrient fluxes. This study evaluated the empirical accuracy of θv measurements using standard and in situ calibrated frequency domain reflectometers (FDR) with gravimetric water content and bulk density measurements of 1512 samples gathered from 15 profiles across 5 ICP Forests level II intensive monitoring plots. The predicted θv, calibrated with standard functions, predominantly underestimated the real water content. The measurement error exceeded the threshold of 0.03 m3 m−3 in 93% of all soil layers. Layer specific calibration removed bias and reduced the overall prediction error with a factor up to 2.8. A simple linear regression often provided the best calibration model; temperature correction was helpful in specific cases. To adequately remove bias in our study plots, a calibration dataset of up to 24 monthly observations was required for topsoils (whereas 12 observations sufficed for subsoils). Based on estimated precision errors, 3 sensors per soil layer proved to be sufficient, while up to 16 sensors are needed to meet the required accuracy in organic topsoils. Validating FDR sensor outputs using in situ gravimetric measurements is essential for quality control and assurance of long term θv monitoring and for improving site specific instrumentalization.