Search

Water management at different scales in agriculture requires the evaluation of corresponding soil water content (SWC) information. Remote sensing technologies including satellite and unmanned aerial systems (UAS) with various sensors, provides a powerful method for monitoring the spatial-temporal variations of SWC at various scales. Few studies have applied UAS with different sensors to detect soil moisture for agricultural applications. The objectives of the study were to 1) estimate surface SWC with different algorithms using multispectral and thermal images from UAS; 2) compare the accuracy of SWC estimation between satellite images and UAS images; 3) determine the spatial variability of SWC indices derived from UAS and satellite images for decision support on image resolution requirements. Two indices, soil moisture index (SMI) and temperature vegetation dryness index (TVDI) were used to estimate SWC. UAS multispectral (4 cm resolution) and thermal images (8 cm resolution) and Sentinel-2A multispectral images (10 m resolution) were acquired on six dates in two fields in the Southern High Plains of Texas in 2018. The accuracy for soil moisture estimation was assessed at five spatial scales, 1, 3, 5, 10 and 15 m, under two surface scenarios, full scene, and soil surface only. A semivariogram was developed to determine the spatial variability of each UAS derived SMI and TVDI, and their corresponding Sentinel-2A image derived index. The accuracy of surface soil moisture estimation and the spatial variability of these two derived indices were compared between the UAS and Sentinel-2A images. Measured GWC and UAS derived SMI and TVDI showed strongest relationship at 1 m (R2 = 0.72 for SMI and; R2 = 0.74 for TVDI) spatial scale, and gradually decreased with distance. At the spatial scale of 10 m, Sentinel-2A and UAS images provided similar accuracy in soil moisture estimation using both SMI and TVDI. TVDI consistently performed better in estimating soil surface moisture at different spatial scales. Image-derived indices for both UAS and Sentinel-2A had a strong spatial dependency (nugget to sill ratio

A variety of empirical models have been developed to determine soil water retention and hydraulic and thermal properties of unsaturated soils. The evaluation of the applicability of these empirical models for a variety of soils, soil, water, and agricultural management systems under different climatic or environmental conditions continue to be an important concern. These fundamental soil properties are critical for many studies of vadose zone hydrologic processes especially in the semi-arid Southern High Plains (SHP), where the increasing pressure on the depleting Ogallala Aquifer has stressed to develop sustainable alternatives, for example, forage-livestock systems, to the irrigation-dependent crop monocultures for the SHP agriculture. The overall objective of this study was to evaluate: (1) four widely used closed-form unimodal soil water retention curve (SWRC) models, i.e., Brooks and Corey (the BC model), van Genuchten (the VG1 and VG2 models), Kosugi (the K model), and Fredlund and Xing (the FX model) models to describe and to combine with pore-size distribution models with which to predict the unsaturated hydraulic conductivity function, and (2) four soil thermal conductivity (λ) predictive models, i.e., Johansen, Côté and Konrad, Balland and Arp, and Lu et al. models for pasture unsaturated soils with. Soil samples (at soil depths of 0-5, 5-10, 10-15, and 15-20 cm) collected from two semi-arid pastures (referred to as the east and west pasture area) in the SHP region under a variety of long-term drip-irrigated and dryland forage management systems were used to quantitatively compare the performance of the selected SWRC and λ models for pasture soils. Model parameters were estimated by fitting these SWRC and λ models to the measured data. Measured SWRCs showed the enhanced soil water retention behaviors (i.e., saturated water content, field capacity water content, plant available water) in pasture unsaturated soils (0-20 cm) under different management practices. Differences in the measured SWRCs between the west and east pasture soils are primarily attributed to the higher clay content of the latter. Notably, the observed organic matter content at the east pasture areas was also relatively higher. Differences in mean water contents of the top 0-5 cm unsaturated soils between treatments for both pastures were significant at lower matric potentials, i.e., in the dry part of SWRCs. No effect of pasture management practices on the saturated water content (s) and field capacity water content (fc) was observed, while significant differences for r or wilting point water content (at -15000 cm H2O) values were found between treatments both in the east and west pasture areas. In contrast with the west pasture with mostly sandy clay loam soils, no significant differences in PAW (plant available water) between treatments were found in the clay loam east pasture likely due to its higher clay and organic matter contents as well as relatively lower soil bulk density. All the selected SWRC models agreed well with measured data in the 0 to -15000 cm H2O range. However, the results of the goodness-of-fit statistics (i.e., RMSE, R2, R2adj, and AIC) suggested that the VG models (VG1 and VG2) performed best across the measured range (wet and dry parts of the SWRCs) of the soil water content. The  values within a treatment in both pasture areas varied with the changes in the soil water content in the 0 to -15000 cm H2O; however, no significant differences in  values (in the measured water range) were observed among treatments in both pastures. The results of the goodness-of-fit statistics (i.e., RMSE, R2, R2adj, and AIC) suggested that all the  models agreed well with measured data. This suggests that all four  models could provide reliable results across the soil water contents in the 0 to -15000 cm H2O range (i.e., water contents both in the wet and dry range) for semi-arid pasture soils under different pasture management practices.