The use of high resolution and high frequency satellite images as a support for precise and sustainable agriculture

This PhD thesis investigates the application of high-resolution and high-frequency satellite imagery to enhance precision agriculture, with a particular emphasis on rice cultivation in Northern Italy. Precision agriculture aims to optimize crop growth by gathering and analyzing data related to soil conditions, crop health, and weather patterns to enable informed decision-making in farming practices. The study uses multispectral imagery to create vigor maps for precision agriculture. By focusing on "Red," "NIR," and "Red Edge" bands, it helps agronomists improve prescription maps for chemical product quantities. Rice cultivation in Northern Italy faces significant challenges, including water scarcity and the increasing costs of chemical products due to geopolitical influences. This research assesses the potential of Sentinel-2 and PlanetScope imagery in mitigating these issues. Although Sentinel-2 is widely utilized in precision agriculture, it has certain limitations such as a five-day revisit period and edge effects. PlanetScope, with its superior spatial and temporal resolution, presents a viable alternative, provided the experience gained with Sentinel-2 can be effectively applied. The research objectives are to assess the availability and usability of satellite images for both sensors, evaluate the internal and external congruence of PlanetScope (in comparison with Sentinel-2), examine the edge effect, and calibrate PlanetScope images to align with Sentinel-2 index values. Techniques such as resampling, linear regression, histogram shifting, histogram matching, and the use of buffer zones are employed for analysis and calibration over indices and single spectral channels. Field activities were carried out with agronomist G.N. Rognoni at the "Riserva San Massimo" test site, focusing on rice cultivation using drone imagery and prescription maps for fertilization. Although not a primary research goal, drone acquisition was compared and calibrated for agronomist use. Results show PlanetScope images are more internally consistent and have fewer edge effects than Sentinel-2. Calibration using histogram matching has proven to be particularly effective over indices, and linear regression is useful for aligning bands prior to the application of indices. Consequently, a new technique has been developed that involves linear regression, histogram shifting, and histogram matching, even for worst-case scenarios. The study concludes that high-frequency and high-resolution satellite imagery, such as PlanetScope, can address issues related to cloud cover and provide reliable data for precision agriculture. Future research may apply this methodology to other crops and explore machine learning for calibration and new calibration methodologies.