A fundamental knowledge gap in understanding land–atmosphere interactions is accurate, high-resolution spatial representation of soil physical and hydraulic properties. We present a novel approach to predict hydraulic soil parameters by combining digital soil mapping techniques with pedotransfer functions (PTFs) and demonstrate that simple derived quantities are related to observed spatial patterns in ecosystem production during the North American Monsoon. Landsat reflectance and elevation data were used to predict physical soil properties at a 5-m spatial resolution for a semiarid landscape of 6265 ha using regression kriging. Resulting soil property maps were applied to the Rosetta PTF to predict saturated hydraulic conductivity and water retention parameters from which approximate water residence times were derived. Estimated idealized residence time for water lost to the deeper vadose zone and evapotranspiration corresponded to vegetation response. Antecedent precipitation was more important for explaining the relationships between modeled soil properties and vegetation response than the amount of monsoon precipitation. Increased spring precipitation before the monsoon produced stronger negative correlations with hydraulic conductivity and stronger positive correlations with plant available water. Modeled water residence times explained the patterns of vegetation and landscape morphology validating our approach as a method of producing functional spatial PTFs. Linking digital soil mapping with Rosetta led to predictions of hydraulic soil properties that were more closely related to vegetation dynamics than the data available in the Soil Survey Geographic (SSURGO) soil database.