Thermal-RGB imagery and in-field weather sensing derived sweet cherry wetness prediction model

Abstract

Rain-induced cracking of sweet cherry (Prunus avium L.) fruit causes substantial economic loss to tree fruit growers annually. Increased fruit surface wetness triggers absorption of water in maturing fruits and leads to fruit cracking. This study was undertaken to develop and evaluate thermal-RGB imagery and in-field weather sensing derived wetness prediction models as tools to help mitigate cracking. We developed two cultivar-specific cherry wetness prediction models, one with only the weather data and other combined with the imagery data derived fruit surface temperature (FST). The suitability and accuracy of such models was validated for two cherry cultivars (cv. ‘Selah’ and ‘Skeena’). The FST and weather data derived model indicated an improved wetness prediction for both the cultivars. Strong relations (R2 = 0.80 and 0.86) and marginal prediction errors (Root Mean Squared Error = 8.7% and 3.5%) were observed between measured and predicted wetness for ‘Selah’ and ‘Skeena’ cultivars, respectively. However, weaker relations (R2= 0.66 and 0.53) were observed, when a model for a particular cultivar was validated against other cultivar, indicating the need of cultivar specific models. Such models can be integrated with a decision support system and crop protection (e.g. rainwater removal, chemical spraying) techniques, for improved crop loss management.