Prediction of Microclimate Parameters for Application in Precision Agriculture

Prediction of microclimate parameters such as temperature, wind speed, and humidity enables planning and management of agricultural production to improve crop yields and quality. This paper provides an overview of methods for predicting microclimate parameters based on past time series data as input. The paper provides insight into time series prediction methods using statistical models and neural network algorithms. The SARIMA statistical model and LSTM neural network model were implemented and evaluated on two case studies: the first one uses climate data from the Copernicus service CDS, while the other one is created using real sensor data streams from an agrometeorological station. Experimental results show that the proposed models perform well to predict future sensor readings in the short term for both datasets and demonstrate that practical prediction of microclimate parameters, which are critical for precision agriculture, is possible with both models.