AbstractPredicting harvest timing is a key challenge to sustainably develop soft fruit farming and reduce food waste. Soft fruits are perishable, high-value and seasonal, and sales prices are typically time-sensitive. In addition, fruit harvesting is labour-intensive and increasingly expensive making accurate phenological predictions valuable for growers. A novel approach for predicting soft fruit phenology and yields was developed and tested, using strawberries as the model crop. Seedlings were planted in polytunnels, and environmental and yield data were collected throughout the growing season. Over 1.2 million datapoints were collected by networked microsensors which measured spatial and temporal variability in air temperature, relative humidity (RH), soil moisture and photosynthetically active radiation (PAR). Fleeces were added to a subset of the plants to generate additional within-polytunnel variation. Cumulative fruit yields followed logistic growth curves and the coefficients of these curves were dependent on micro-climatic growing conditions. After 10,000 iterations, machine learning revealed that RH was the optimal factor informing the coefficients of these curves, perhaps because it is an integrative metric of air temperature and water status. Trigonometric models transformed weather forecasts, which showed a relatively low agreement with polytunnel air temperature (R2 = 0.6) and RH (R2 = 0.5) measurements, into more accurate polytunnel-specific predictions for temperature and RH (both R2 = 0.8). We present a framework for using machine-learning techniques to calculate curve coefficients and parametrise coupled weather models which can predict fruit yields and timing to a greater degree of accuracy that previously possible. Dataloggers measuring environmental and yield data could infer model parameters using iterative training for novel fruit varieties or crop types growing in different locations without a-priori phenological information. At this stage in the development of artificial intelligence and networked microsensors, this is a step forward in generating bespoke phenological prediction models to inform and support growers.
Machine learning techniques for daily solar energy prediction and interpolation using numerical weather models
