Development of a steady-state model to predict daily water table depth and root zone soil matric potential of a cranberry field with a subirrigation system

Abstract

Maintaining a steady water table depth (WTD) to ensure an optimal soil matric potential in the root zone (RMAP) is vital when growing cranberry under a subirrigation system; owing to losses and other hydrological processes. The excessive rising or falling of the WTD may threaten the plant transpiration either by saturation or lack of moisture in the root soil. A steady-state model was developed for a uniform soil column to predict WTD and RMAP under different weather conditions. The model is based on van Genuchten (VG) and Brooks and Corey (BC) analytical soil water retention functions coupled with Mualem, Brooks and Corey, and Gardner hydraulic conductivity models. The results show that the model is capable of predicting satisfactorily both WTD and RMAP. The VG model performed with a 78.13% accuracy for the WTD and an 88.59% precision according to the Kling Gupta Efficiency coefficient (r2 = 0.90, β = 1.00, and γ = 0.99) for the field storage. Meanwhile for the RMAP the Mualem and Gardner hydraulic conductivity models, predictions were successful 71.87% and 75.00% of the time, respectively. The BC model had a 78.13% success for the WTD, 86.93% accuracy in estimating the field water storage according to Kling-Gupta efficiency coefficient (r2 = 0.89, β = 1.00, and γ = 1.00) and the BC and Gardner hydraulic conductivity models had 65.63% and 71.88% success, respectively. A sensitivity analysis of the model, by means of the Morris method, reveals that for both models, the lower boundary condition impacts significantly both variables which are, however, less affected by the field capacity and the residual soil moisture content. The lower boundary condition interacts with the slope of the soil water retention functions, the height of the capillary fringe, the saturated conductivity, and the saturation moisture content which have non-linear effects. An uncertainty analysis shows that both variables for both models are normally distributed.