Predicting the Campbell Soil Water Retention Function: Comparing Visible–Near‐Infrared Spectroscopy with Classical Pedotransfer Function

CORE IDEAS: The Campbell model was anchored at a soil water content of −1000 cm H₂O matric potential. The two model parameters were predicted using either soil fines or vis–NIR spectroscopy. Both methods accurately predicted the soil water retention curve. The soil water retention curve (SWRC) is essential for the modeling of water flow and chemical transport in the vadose zone. The Campbell function and its b (pore‐size distribution index) parameter fitted to measured data is a simple method to quantify retention under relatively moist conditions. Measuring soil water retention is time consuming, and a method to accurately predict the Campbell relation from either typically available soil parameters such as bulk density, clay‐size fraction, and organic matter content (soil fines) or from visible–near‐infrared (vis–NIR) spectroscopy may provide a fast and inexpensive alternative. However, the traditional Campbell model has a reference point at saturated water content, and this soil‐structure‐dependent water content will typically be poorly related to basic texture properties and thus be poorly predicted from vis–NIR spectra. In this study, we anchor the Campbell model at the water content at −1000 cm H₂O matric potential [log(1000)= pF 3]. Agricultural soil samples with a wide textural range from across Denmark were used. Soil water retention was measured at a number of matric potentials between pF 1 and 3. The soil water content at pF 3 and Campbell b were both well predicted using either a soil‐fines‐based pedotransfer function or vis–NIR spectroscopy. The resulting Campbell function anchored at pF 3 compared closely to measured water retention data for a majority of soils. The ability of the two methods to also predict field average SWRC was evaluated for three fields. Field average, predicted SWRC compared well with field average measured data, with vis–NIR overall performing better.