A selection of sensing techniques for mapping soil hydraulic properties

Abstract

Data on soil hydraulic properties are needed as input for many models, such as models to predict unsaturated water movement and crop growth, and models to predict leaching of nutrients and pesticides to groundwater. The soil physics database of the Netherlands shows several lacunae, and a substantial part of the data were collected more than thirty years ago and thus might not represent actual soil hydraulic conditions. There is a need to fill lacunae in the soil physics dataset, to make the dataset up-to-date and to aggregate soil hydraulic properties observed at point scale to larger spatial units. The relationship between the unsaturated hydraulic conductivity K or the volumetric water content θ and the pressure head h can be described for example by equations such as the Mualem-Van Genuchten equations, the parameter values of which can be predicted from soil properties by using pedo-transfer functions. These soil properties include clay content, loam content, organic matter content and the median grain diameter of the sand fraction (50-2000 µm). Application of proximal sensing techniques in observing soil physics properties in the field might reduce the costs of data collection. The accuracy with which proximal sensing techniques can predict clay content, loam content, organic matter content and the median grain size of the sand fraction (50-2000 µm) needs to be assessed as these are the explanatory variables in the pedo-transfer functions. The aim of this study was to select on-the-go proximal sensing techniques that are able to quantify soil variables that are used in the pedo-transfer functions for prediction of the parameters of the Mualem-Van Genuchten equations. We conclude that near infrared spectrometry, γ-ray spectroscopy and electromagnetic induction methods have a potential in spatial prediction of clay, silt and sand content, and that near infrared spectrometry has a potential in spatial prediction of organic matter content and soil moisture content. The use of pedo-transfer functions requires also the parameters bulk density and median of the sand fraction (M50). Bulk density can be measured by a device called RhoC. However, its performance in predicting bulk density of terrestrial soils has not been validated yet. Most instruments are applied to measure fractions clay, silt and sand. The application of these instruments in the determination of M50 needs to be developed. Possibly the combination of specific surface area and fractions of fine and course sand can give an indication. On the other hand new pedo-transfer functions can be developed neglecting the M50 value while increasing the contribution of other explanatory variables.

Mapping soil texture with a gamma-ray spectrometer : comparison between UAV and proximal measurements and traditional sampling : validation study

“…(Kempen, 2011;Kempen et al, 2011;De Vries et al, 2014). The digital soil mapping approach is more efficient due to the use of covariates (ancillary information such as elevation and land use maps) and advanced statistical methods (geostatistics and machine learning, Knotters et al, 2017).…”

“…When assessing possibly suitable covariates for digital soil classification mapping these should therefore be a proxy for one or more of these classifying properties. Knotters et al (2017) provide an overview of proximal soil sensors that are suitable to measure soil hydraulic properties. These include near and mid infrared spectroscopy, gamma-ray spectrometry, electromagnetic induction methods, ground penetrating radar and other methods.…”

“…The maximum elevation difference is 1.4 meter (Figure 7). The procedure to select the area is described in Knotters et al (2017) in Appendix 1. This Appendix is added to this report as Annex 2 since it directly describes the selection of the study area in this study and the selection of sampling locations for calibration and validation (Annex 1 and 3).…”

“…There are several ways to translate gamma radiation measurements to soil property information ranging from simple statistical methods like linear regression to more complex methods like Support Vector Machines (SVM) (Knotters et al, 2017;Viscarra Rossel et al, 2007;Heggemann et al, 2018;Van der Klooster et al, 2011). The data requirements and suitability of the method vary and depend on the research (area) at hand.…”

Noodzakelijke indicatoren voor de beoordeling van de gezondheid van Nederlandse landbouwbodems : Selectie van fysische, chemische en biologische indicatoren voor het meten van de bodemgezondheid

“…• (VIS)NIR spectroscopie: meer toegepast voor textuurbepalingen; vraagt veel dataverwerking en correctie voor watergehalte (Knotters et al, 2017).…”