IoT and Ensemble Long-Short-Term-Memory-Based Evapotranspiration Forecasting for Riyadh.

Area of Science:

• Agricultural Science
• Environmental Science
• Data Science

Background:

• Evapotranspiration (ET) is vital for water resource management and agricultural efficiency.
• Accurate ET forecasting is challenging due to complex meteorological variable requirements.
• Internet of Things (IoT) enables real-time meteorological data collection for improved forecasting.

Purpose of the Study:

• To develop and evaluate an ensemble-learning approach for ET forecasting using limited meteorological data.
• To assess the performance of Long Short-Term Memory (LSTM) and ensemble LSTM models (bagged and boosted) for ET prediction.
• To demonstrate the efficacy of the proposed approach under limited meteorological conditions in Riyadh, Saudi Arabia.

Main Methods:

• Utilized daily maximum temperature (T), mean humidity (Hm), and maximum wind speed (Ws) for ET forecasting.
• Implemented Long Short-Term Memory (LSTM) and ensemble LSTM (bagged and boosted) models.
• Evaluated model accuracy using coefficient of determination (R2), root mean squared error (RMSE), and mean squared error (MSE) on data from 2001-2023.

Main Results:

• The bagged LSTM model achieved the highest accuracy in ET forecasting, with an R2 of 0.94.
• Boosted LSTM and off-the-shelf LSTM models showed lower accuracy (R2 of 0.91 and 0.77, respectively).
• The bagged LSTM model demonstrated superior efficiency with lower RMSE (0.42) and MSE (0.53).

Conclusions:

• Ensemble learning, particularly the bagged LSTM approach, effectively forecasts evapotranspiration with limited meteorological data.
• The proposed IoT-based and ensemble-learning method offers a viable solution for efficient water management in agriculture.
• The bagged LSTM model provides a reliable and accurate tool for ET prediction in data-scarce environments like Riyadh.