Ad Hoc Modeling of Root Zone Soil Water with Landsat Imagery and Terrain and Soils Data

Agricultural producers require knowledge of soil water at plant rooting depths, while many remote sensing studies have focused on surface soil water or mechanistic models that are not easily parameterized. We developed site-specific empirical models to predict spring soil water content for two Montana ranches. Calibration data sample sizes were based on the estimated variability of soil water and the desired level of precision for the soil water estimates. Models used Landsat imagery, a digital elevation model, and a soil survey as predictor variables. Our objectives were to see whether soil water could be predicted accurately with easily obtainable calibration data and predictor variables and to consider the relative influence of the three sources of predictor variables. Independent validation showed that multiple regression models predicted soil water with average error (RMSD) within 0.04 mass water content. This was similar to the accuracy expected based on a statistical power test based on our sample size (n = 41 and n = 50). Improved prediction precision could be achieved with additional calibration samples, and range managers can readily balance the desired level of precision with the amount of effort to collect calibration data. Spring soil water prediction effectively utilized a combination of land surface imagery, terrain data, and subsurface soil characterization data. Ranchers could use accurate spring soil water content predictions to set stocking rates. Such management can help ensure that water, soil, and vegetation resources are used conservatively in irrigated and non-irrigated rangeland systems.