Grape canopy surface wetness: Simulation versus visualization and measurement

Surface wetness is commonly measured with electronic sensors but simulation is a promising alternative, although these methods have rarely been compared statistically using visual observations as a truth data set. In this study, these comparisons are made in two vineyards in New York and one in Australia using simulations from the Surface Wetness Energy Balance (SWEB) model run from canopy-collected atmospheric inputs. The fraction of canopy wet surface area was visually observed on 45 leaves (Wobs), measured using a 15-sensor array (Wsen), and simulated using the SWEB model (Wsim). Both the measurements and visual observations were made in five canopy positions. Observations of wetness included periods of rain, dew and near dew. Overall the SWB model was slightly more effective (r2 = 0.73) than the sensor (r2 = 0.6), although there was substantial variation between sites. Since most applications use surface wetness duration, an additional comparison was based on the Canopy Surface Wetness Duration (CSWD). The mean absolute error (MAE) of the SWEB model varied from 0.7 to 1.5 h at the three sites, while for sensors the MAE varied from 1.1 to 1.9 h. Both SWEB and the sensor performed more poorly in dew than in rain. The sensor did not perform as well as in past studies, possibly because it was validated under a wider range of conditions or because of degradation of the painted surface. In what is likely to be one of the first rigorous comparisons of measurement and simulation, the SWEB model performed as well as a commonly used sensor for estimation of canopy wet surface area and surface wetness duration.