Prediction of evaporation with the CONSEVB simulation-model - an experimental evaluation

Abstract

The CONSERVB simulation model was evaluated by comparing predictions of evaporation, net radiation, and water and temperature profiles with measured values from a bare, tilled soil in New Zealand conditions. No model calibration was used whatsoever. The assessment included tests of the sensitivity of output to variations (uncertainty) in input parameters. On 18 of the 24 days simulated, evaporation was within 0.5 mm day-1 of the measured means, although that represented less than 20% error on only 6 days. The mean difference between measured and simulated surface soil temperature for the two drying cycles was 1.2 and 1.3°C; most of this error occurred in warm conditions around noon each day. Simulated soil temperature at 0.05 m exceeded measured values during the warmest parts of the day and fell below them at night. Soil temperature prediction was sensitive to the air temperature and solar radiation climatic inputs. Evaporation prediction was sensitive to initial profile water contents, to the soil moisture characteristic and to the unsaturated hydraulic conductivity inputs. An increase in the water content at each matric potential step in the soil moisture characteristic input by a factor of 1.10 resulted in an increase in simulated cumulative evaporation of over 40%. For the prediction of evaporation and soil water content, field effort in the measurement of the soil hydraulic properties and initial water contents (where simulations are short) for model parameterization is likely to give the highest cost-benefit. Very accurate determinations of these input parameters and functions are required for model evaluation. Given the sensitivities, and the uncertainties associated with measurement and prediction of model input parameters, the predictions from CONSERVB have large uncertainties associated with them. Field-measured values of cumulative evaporation were within the range of variation in predicted values that resulted from uncertainty in determination of initial soil water contents alone. CONSERVB is more applicable to an operations research modelling approach than to prediction of evaporation and surface water contents in specific conditions.