Preliminary results of irrigation management for mango using LSTM neural networks and IoT

Abstract

Mango cultivation in Colombia faces the impact of regional climate variability. To improve fruit development and minimize environmental and economic effects, it is necessary to implement efficient irrigation and appropriate water management technologies. In this study, we developed a trend forecasting system based on an LSTM neural network and technologies such as ThingsBoard, LoRA, and MQTT. The aim was to improve mango irrigation practices through informed decisions based on monitoring and predicting matric potential and evapotranspiration variables. This article describes the development and application of the system for mango irrigation management. Results validate the effectiveness of the proposed system for mango cultivation, with RMSE indices of 1.56 and 0.0019 and determination coefficients (R2) of 0.9989 and 0.9971 for matric potential and evapotranspiration, respectively. These findings support enhancing growth conditions and promoting sustainable practices. Despite data availability limitations, the system's efficacy in prediction and irrigation management demonstrates significant potential to maximize productivity and reduce the environmental and economic impacts of inadequate water management.

Keywords

• IoT
• irrigation management
• LSTM neural networks
• mango cultivation
• trends forecasting