Main content area

Forecasting strawberry yields ('eascape') three weeks in advance: validation and optimisation of a method developed under commercial production conditions in Quebec

Deschenes, P., Boivin, C.
Acta horticulturae 2017 no.1156 pp. 465-472
correlation, cultivars, fruit yield, harvesting, marketing, models, plant development, strawberries, Quebec
Marketing Quebec strawberries has remained problematic given that producers must guarantee to the main food distribution chains, on a weekly basis, the volume of fruit they will harvest three weeks in advance. Those predicted volumes are based on previous years' yields but can be ineffective. This represent an impediment to Quebec strawberry producer's competitiveness. An effective and easy-to-use three weeks in advance forecasting tool for the 'Seascape' cultivar was developed between 2011 and 2013 and validated across Quebec in 2014 and 2015 on the site of 10 agribusinesses. A weekly green fruit enumeration total over 60 strawberry plant for a given field was used to produce an optimistic and conservative three weeks in advance total (marketable + culls) yield forecast. Both optimistic and conservative three weeks in advance forecasted strawberry fruit yields were compared to producer-assessed actual total yields for the corresponding period. Conservative forecasts outperformed optimistic forecasts for eight of ten sites in 2014 (optimistic root-mean-square error (RMSE) > conservative RMSE). In 2015 the conservative and optimistic forecasts showed similar accuracies. This difference is attributable to the greater actual total yields in 2015 (vs. 2014), which resulted in the optimistic yield forecasts being closer to actual yields that season. Mean absolute percent error (MAPE) differences were on average 35 (2015) and 40% (2014), indicating that a greater forecasting accuracy was achieved for the 2015 season. Strong sample correlation coefficient (r) values were obtained (0.33≤r≤0.91 optimistic 2014, 0.32≤r≤0.92 conservative 2014, 0.62≤r≤0.97 optimistic 2015, 0.53≤r≤0.96 conservative 2015). These values indicate that the forecasting models were adequately in describing variations of the actual yield three weeks in advance. Both forecasting models evaluated show promise. Better forecasts were obtained for sites where crop production occurred under ideal conditions. External factors adversely affecting strawberry plant development, adversely affect forecasting accuracy.