EVAPORATION FROM BARE ARID SOILS: PARAMETER ESTIMATION USING MONTE CARLO SIMULATIONS

The ecology and hydrology of desert environments – especially evaporation – is poorly understood. Here we use evaporation data from a weighing lysimeter to gain insight into the mechanism of evaporation from bare, arid soil. The combined process-based evaporation model by Shokri et al. (2009) and Or et al. (2013) was examined using Monte Carlo simulations. Evaporation rates after three storms were simulated with the model and then compared to measured evaporation rates from the weighing lysimeter. A sensitivity analysis was performed to consider model sensitivity to simultaneous variations in the depth of the drying front, ξ, total porosity, φ, and soil moisture content above and below the second drying front, Δθ. The values of ξ and φ were randomly selected from uniform distributions bounded based on laboratory studies; the value of Δθ was determined as a fraction of φ; and each Monte Carlo simulation comprised 10,000 realizations of the model. Diffusion-controlled evaporation began 2-3 days after precipitation for all events. RMSE and absolute percent bias (APBias) were determined for the 10,000 realizations of each event. For each event, a set of Pareto optimal parameterizations (a realization with RMSE and APBias such that no other realization improved both metrics) was identified. Within each of the three storm events, the value of φ is remarkably consistent. The Pareto optimal values of ξ and Δθ were much less consistent, both within the individual events and from event to event, than the values of φ. A second Monte Carlo analysis was run with expanded parameter ranges, two events converge on consistent values for all three parameters. Like many soils processes, it appears that the parameterization of this evaporation model is scale dependent, and assumptions based on laboratory data may not directly apply when the system is upscaled. The model itself, when parameterized without the laboratory-established limitations, performs well for two of the three events.