Mobile app for analysing the heat damage risk in grape and apple fruit

The aim of this study was to enable farmers utilizing satellite weather data in the risk prediction for heat damage of fruit crops. To this purpose, a mobile app was developed, which allows users to set up their field data capturing fruit cultivar, production system, location, and connect to a public weather API or private weather station with custom network parameters. The user can track manual readings capturing fruit size, colour, and fruit damage during the season. The application uses a TensorFlow model trained with research data on several weather variables, fruit properties, and fruit damage observed. Combining weather and fruit data from own location, the user gains the predicted risk of heat damage forecasted in the coming days. The application is built on the flutter cross-platform framework, allowing distribution to main mobile operating systems in the future. All data are kept in an embedded database on the user’s phone and the application requires no login or account, putting farmers completely in charge of their own data. An optional, anonymized data upload via a channel secured by both TLS and client secret authentication is provided. By means of this secure data collection, the user can also upload own data on fruit damage in the field and contribute to the future work of advanced risk models. The according data server runs in a containerized setup and utilizes a simple, file based database to store all data. This citizen science approach will allow improved risk modelling.