H212-08
Improving pedotransfer functions: A new framework to predict soil saturated hydraulic conductivity

Wednesday, 16 December 2020: 17:51
Virtual
Alejandro Cueva, University of California Riverside, Riverside, CA, United States, Daniel Hirmas, University of California Riverside, Department of Environmental Sciences, Riverside, CA, United States, Attila Nemes, Norwegian Institute of Bioeconomy Research, Ås, Norway, Sharon A Billings, Univ Kansas, Lawrence, KS, United States, Alejandro N Flores, Boise State University, Boise, ID, United States, Li Li, Pennsylvania State University Main Campus, Department of Civil and Environmental Engineering, University Park, PA, United States, Hang Wen, Pennsylvania State University, State College, PA, United States, Xi Zhang, Oregon State University, College of Earth, Ocean, and Atmospheric Sciences, Corvallis, United States and Pamela L Sullivan, Oregon State University, College of Earth, Ocean, and Atmospheric Science, Corvallis, OR, United States
Abstract:
Pedotransfer functions (PTFs) are widely used tools to predict soil properties across different spatial scales. However, PTFs suffer from several known limitations. For example, PTFs rarely account for the contiguity of contrasting soil materials within a single pedon or the influence of depth on their predictions. Here we propose a new framework, which we refer to as a “pedontransfer function” (PnTF), that predicts the values of parameters that describe response variable depth functions rather than predicting the response variables themselves from a series of predictor variables. To build the PnTF, we used field-based pedon information from the Pedogenic and Environmental DataSet (PEDS), encompassing approximately 2,000 pedons and >13,000 soil horizons from across United States. We estimated saturated hydraulic conductivity (Ksat) using a generalized form of the Kozeny-Carman equation that incorporated measured values of soil bulk density and soil moisture at -33 kPa (field capacity) and -1500 kPa (permanent wilting point). Saturated hydraulic conductivity was used as the dependent variable and soil horizon midpoint depth (Z) was used as the independent variable. A linearized power model was fitted to the data from each pedon within PEDS. The intercept parameter (β0) of this model represents the Ksat at the land surface, while the slope parameter (β1) describes the rate at which Ksat changes with depth. Our results show a strong negative linear relationship between β1 and β0 (r2 = 0.80; P < 0.01), suggesting that β1 can be predicted from β0. To demonstrate the framework, we used multiple linear regression (MLR) and commonly measured soil properties from the uppermost horizon as well as climatic variables as input to predict β0 (R2 = 0.24; P < 0.01). The estimation of β0 may be improved by replacing MLR with machine learning, broadening of soil data to global datasets, and the inclusion of soil pore-network characteristics as inputs, where available. Our results suggest that soil properties at the surface, along with climatic variables, can be used to develop PnTFs that ultimately predict the rate at which Ksat changes with soil depth. We anticipate that PnTFs have significant functional advantages over PTFs that should be of interest to the community of developers and users of Earth system and community land models.
<< Previous Abstract | Next Abstract >>