Evaluating the water needs of potted gardenia plants via low-cost sensor measurements and a transpiration prediction model

The use of low-cost sensors and/or transpiration models to predict plant water needs has proven to be particularly useful for water and fertilizer saving, especially in greenhouse ornamental and vegetable crops established in the Mediterranean region. However, various challenges, such as sensor faults and insufficient model calibration, must be addressed when the above-mentioned methods are used to obtain reliable estimations. In this study, i) a specially designed multifunctional scale (manufactured using 3D printing) with a 1 g precision load cell connected to an Arduino microcontroller, ii) environmental sensors (light, temperature, and humidity), and iii) transpiration and evaporation models were combined through appropriate software to optimize the irrigation of potted gardenia plants grown in a Mediterranean greenhouse. The software was developed in Python to integrate all components and to enable data acquisition on the PC. The equipment was adequate for detecting the irrigation start and end times, as well as for measuring the irrigation dose, volume of drainage solution, irrigation period, drainage period, and evapotranspiration, while the transpiration and evaporation models predicted plant water needs. After each irrigation, the scale readings were utilized to i) assess the accuracy of the model prediction and ii) recalibrate the model when the accuracy was less than 10% while also taking into account the environmental parameters. Using the aforementioned technique to develop an irrigation schedule, the model's regression coefficient (R²) was increased from 0.89 to 0.99, resulting in a 0.35 L m⁻² day⁻¹ saving of nutrient solution. In conclusion, the use of weight sensor measurements can significantly improve transpiration model prediction through a frequent recalibration of the transpiration model.