The soil water retention curve (WRC) plays a critical role for the solution of various agricultural, hydrological and environmental problems such as the prediction of infiltration rate, surface runoff for irrigation design, modeling of contaminant transport in the unsaturated zone, estimation of soil water holding capacity for crop growth simulations, and the evaluation of soil strength and compressibility for trafficability assessment. These models require high vertical and horizontal spatially resolute data in order to perform well. However, the laboratory, labor, and time costs required to achieve good performance are cost prohibitive using currently available methods. Therefore, new methods are needed for accurately estimating the soil WRC in a cost effective manner. The purpose of the proposed study is to utilize a miniature digital camera, embedded in a soil probe, in combination with other soil sensors to enable the rapid, digital, objective assessment of the soil WRC at horizontal and vertical spatial resolutions not feasible using currently available methods. OBJECTIVES: The primary objective of the Phase I study is to demonstrate the feasibility of using soil image analysis in conjunction with measurements from other in situ soil sensors for the estimation of the soil water retention curve. The relative performance of three estimation methods will be compared. The most promising method will be developed during Phase II. Several questions will be asked, including: How can pore structure information, such as pore size distribution and porosity, be estimated with reasonable accuracy from soil images? How can pore structure information derived from soil images be incorporated into different indirect estimation methods? How would each method perform if input variables obtained from soil images are used? APPROACH: Our research will build upon the considerable body of research in the literature on the indirect estimation of soil water retention curves and upon our previous work on the extraction of quantitative soil metrics from in situ soil imagery and other in situ soil sensors. However, the research is unique in that we will use data extracted from the in situ soil images, and other sensors, as the input parameters for three indirect estimation methods identified from the literature. Our previous work with the Soil Imaging Penetrometer (SIP) did not address the estimation of the soil water retention curve. First, soil images and other penetrometer-based sensor data will be collected by taking multiple measurements at 30 locations from a 2000 acre research facility in De Forest, WI. At the same locations, duplicate soil samples will be taken at two discrete depths. The water retention curve of soil samples will be determined in a laboratory by standard methods. Second, several image processing techniques will be used to extract quantitative measures related to soil pore structure and fractal dimensions from the soil images collected in the first step. Third, the water retention curve will be estimated using three approaches: 1) pedotransfer function using artificial neural networks, 2) pore size distribution and capillarity, and 3) fractal concepts and percolation theory. Fourth, van Genuchten model parameters will be estimated using a non-linear optimization approach. The agreement between the models and laboratory measurements will be evaluated in terms of bias, mean difference, root of the mean squared difference, and the Pearson correlation coefficient, or similar metrics. Fifth, the feasibility of the overall approach will be discussed in relationship to currently available techniques for obtaining the same information, and the most promising estimation method will be recommended for Phase II
