Journal of Agronomy and Field Innovations

Background: Advancements in precision agriculture are founded in continuous monitoring of environmental conditions. Although traditional agriculture is still a manual process, which causes delays in the operation of systems and difficulties in utilizing resources effectively and producing required yields.
Objective: To develop an integrated system architecture combining the environmental sensor networks and computing infrastructure of cloud technologies to implement real-time data-driven advisory during the crop management process.
Methods: The multi-layered structure of the system consisted of an environmental sensing layer, edge gateways, wireless transmission of data, storage of collected information in the cloud, and machine learning engine operations. The system was tested with the help of the simulation of an agricultural field data set for several crop cycles. The parameters monitored consisted of soil moisture, temperature, pH level, humidity of the air, rain, solar radiation, the moisture of leaves, windspeed, and concentration of carbon dioxide through low-power IoT sensors via Lora WAN and MQTT to the server based on the AWS technology. Random forest, Support Vector Machine, and Long Short-Term Memory approaches were used to develop the system.
Result: The combined method has resulted in precision in irrigation settings equal to 94.6%, as well as 91.2% in regards to disease onset detection, with average response time with regard to the cloud of 1.8 seconds, which incurs 22% less energy consumption in comparison with systems that constantly transmit data as a baseline. Water usage has been decreased by almost 27% when compared with conventional irrigation methods. 
Conclusion: The combination of environmentally-oriented sensors and cloud-based analytics allows improving the speed of farm-related decisions significantly.