Intelligently filling sensory data gaps: Application of Machine Learning in irrigation systems

With global population growth and climate change, agriculture in the 21st century is facing significant and complex challenges. Increasing productivity has become a requirement, as has ensuring the sustainability of available resources, putting positive pressure on changes in production methods and the efficient management of these resources. Irrigation is one of the cultural practices with the greatest responsibility for vegetative growth, operating in parallel with sustainability. Producing more with less has become the motto for farmers these days, and it is crucial to minimize waste and ensure that, in this case, irrigation meets the specific needs of crops at all stages of their cycle. Advances in communication technologies and the adoption of IoT equipment (sensors, humidity probes and other monitoring tools) have enabled farmers to significantly improve their decision-making, promoting future insights based on the data collected. This is one of the objectives of the work carried out by the company HIDROSOPH, which plays a vital role in this process, offering innovative solutions in the optimization and efficient management of water. So, in partnership with HIDROSOPH, we proposed a project to validate and consolidate the data collected by the IoT equipment in operation, guaranteeing the uniformity, integrity and applicability of the data in the various services provided by the company. The case study focused on data from 12 highdensity olive grove farms from different locations in mainland Portugal, for the year 2023. There was a recurring problem on almost all of the farms, which consisted of interruptions in data recording due to technical failures in the sensors. To address this problem, the XGBoost Machine Learning model was used to impute predicted values into data gaps. Although the practical implementation was focused on XGBoost, other models were evaluated theoretically, and it was concluded that XGBoost is the most suitable solution for ensuring continuous data integrity and supporting the algorithms used in the company's services. Resolving these gaps is essential to maintain the effectiveness and reliability of the solutions provided by HIDROSOPH.