Proximal Sensing Methods for Mapping Soil Water Status in an Irrigated Maize Field

Abstract

Approximately 80% of allocated freshwater in New Zealand is used for irrigation, and the area irrigated has increased by 55% every decade since 1965. The research described in this chapter therefore focuses on developing new techniques to map and monitor soil attributes relevant to irrigation water use efficiency. The apparent electrical conductivity (EC_a) of soils under a 33-ha irrigated maize crop was mapped using a mobile electromagnetic induction (EM) and RTK-DGPS system, and this map was used to select three contrasting zones. Within each zone, further EC_a values were recorded at a range of volumetric soil water contents (θ) to develop a relationship between EC_a, soil texture, soil moisture, and available water-holding capacity (AWC) (R ² = 0.8). This allowed spatial prediction of AWC, showing that these sandy and silty soils had similar AWCs (∼160 mm/m). High-resolution digital elevation data obtained in the EM survey were also co-kriged with TDR-derived θ to produce soil moisture prediction surfaces, indicating drying patterns and their relationship to topography and soil texture. There was a 12.5–13.1% difference in soil moisture to 45 cm soil depth between the wettest and the driest sites at any one time (n = 47). Spatial and temporal variability of soil moisture, indicated by these co-kriged prediction surfaces, highlights the need for a rapid high-resolution method to assess in situ soil moisture. The potential of soil spectral reflectance (350–2,500 nm range; 1.4–2 nm resolution) for rapid field estimation of soil moisture was therefore investigated. Soil spectra were pre-processed and regressed against known soil moisture values using partial least squares regression (R ² calibration = 0.79; R ² prediction using leave-one-out cross-validation = 0.71). These proximal sensing methods facilitate spatial prediction of soil moisture, information which could then be uploaded to a variable rate irrigator.