Rural IoT soil data

Measurement campaign was performed across the entire growing season of 2023. Seven Arduino+LoRaWAN sensors were measuring soil moisture, temperature and pH, as well as solar irradiance in the radius of about 35 km around Gdansk, Poland. Raw data were being collected with a nomadic gateway aboard an UAV and transferred to the cloud for analysis.Based on the physics informed analysis of the underlying measurement processes anomalies were classified and parametrized. Heuristic values of their parameters were used by the anomaly detector implemented on the original constrained end (sensor) devices to clean transmitted data from incorrect, misleading or otherwise redundant data. Independently, the heuristic values of the above were optimized with simulated annealing to further improve the anomaly detector. Improvements were assessed by measuring distances between reference, heuristically cleaned and optimally cleaned time series.