Validation of Harvest Date Estimator Mobile Application For Banana (BanaTech-Version 1): A Way Towards A More Efficient Farm Planning

Abstract

Harvest date estimation helps farmers in the decision making which contributes to efficient farm planning. This helps to reduce losses due to untimely harvest of immature produce while ensuring that the market receives higher-quality bananas harvested at their optimum physiological maturity. A harvest date mobile application (BanaTech) was developed for Saba and Lakatan bananas based on accumulated heat units during bunch development stages. For this study, the forecast accuracy of estimating the harvest date of Saba bananas were validated and was conducted from August 2022 to December 2022 in Los Baños, Laguna, Philippines. Ninety (90) samples were tagged and categorized according to their phenological stages, and BanaTech was used to forecast the phenophases and harvest date. BanaTech’s calendar day and growing degree days (GDD) were compared to actual field observations and GDD calculations derived from actual weather data. This validation process involved the computation of forecast accuracy and bias. Results indicated that, on average, it takes 125.6 calendar days and 2269.6 °Cd (heat units) from the emergence of flowers to reach the harvest maturity of Saba bananas. In this validation, BanaTech accurately predicted the duration of bunch development in terms of calendar days 90.96% of the time, demonstrating a high accuracy level while accumulated GDD predictions were 80.12% accurate. For Lakatan, 87 days and 1340.6 °Cd was needed to complete bunch development while BanaTech predicted 88 days (92.43% accuracy) and 1228.3 °Cd (90.96% accuracy) which indicates highly accurate forecasting. Further delineation of early flower development stages of Saba (PS 60 to PS 63) and the latter fruit development stages (PS 75 to PS 79) is recommended to eliminate subjective biases on identification of actual phenological stage during observation and increase the overall forecast accuracy of the application. The results of this validation is useful for further improvement of the application to increase reliability of harvest prediction.