Smart & Green: An Internet-of-Things Framework for Smart Irrigation

Abstract

Irrigation is one of the most water-intensive agricultural activities in the world, which has been increasing over time. Choosing an optimal irrigation management plan depends on having available data in the monitoring field. A smart agriculture system gathers data from several sources; however, the data are not guaranteed to be free of discrepant values (i.e., outliers), which can damage the precision of irrigation management. Furthermore, data from different sources must fit into the same temporal window required for irrigation management and the data preprocessing must be dynamic and automatic to benefit users of the irrigation management plan. In this paper, we propose the Smart&Green framework to offer services for smart irrigation, such as data monitoring, preprocessing, fusion, synchronization, storage, and irrigation management enriched by the prediction of soil moisture. Outlier removal techniques allow for more precise irrigation management. For fields without soil moisture sensors, the prediction model estimates the matric potential using weather, crop, and irrigation information. We apply the predicted matric potential approach to the Van Genutchen model to determine the moisture used in an irrigation management scheme. We can save, on average, between 56.4% and 90% of the irrigation water needed by applying the Zscore, MZscore and Chauvenet outlier removal techniques to the predicted data.

Sustainable Irrigation System for Farming Supported by Machine Learning and Real-Time Sensor Data

“…Compared to the results found in the literature, where [7] achieved an accuracy of 80% using Random Forest and [31], that used a similar dataset, also having results in the 84% accuracy for Random Forest, it is possible to conclude that Random Forest is indeed the best algorithm to predict irrigation scheduling. Other solution can be found using other models, such as Dynamic Neural Networks [32], that achieved a 10% margin of error for irrigation scheduling based on soil moisture sensors, and [33], that schedule irrigation based on soil moisture predictions using XGBoost, with a 13% margin of error. In this last study, it is also possible to check a study done on the same algorithms as our study, and between them, Random Forest was also the best solution, with a margin of error of 14%, similar to our 85% accuracy.…”

“…In [36] presents a solution capable of reducing 23% of water in irrigation using LoRa nodes that gather information about air temperature and humidity and soil moisture, that were analysed using fog computing. The Smart&Green framework, presented in [33], recommends the optimal irrigation management-based field configuration and soil moisture data, using computational models to reduce the water used in irrigation in 56%. Finally, [32] used Dynamic Neural Networks to predict the soil moisture in the fields and adjust the irrigation process accordingly, achieving up to 46% in water savings.…”

Development of Soil Moisture Monitoring by using IoT and UAV-SC for Smart Farming Application

“…In related work of smart irrigation, IoT framework for smart irrigation such as data monitoring, preprocessing, fusion, synchronization, storage, and irrigation management enriched by the prediction of soil moisture was presented in [20]. Smart&Green was used to preprocess soil moisture data with outlier removal criteria and techniques of Zscore, MZscore, GESD, and Chauvenet to provide a more precise irrigation water needed in irrigation management.…”

Soil color analysis based on a RGB camera and an artificial neural network towards smart irrigation: A pilot study

“…According to the Food and Agriculture Organization (FAO), the world contains about 1400 million km 3 of water, but only 45000 km 3 of this are fresh water resources [ 1 ]. The agricultural sector, particularly irrigation, consumes about 70% of global freshwater [ 1 , 2 ]. Globally, current water resources will be barely sufficient for agricultural communities by 2050 [ 1 ].…”