Orbital, aerial, and proximal sensing applied to monitoring the spatial variability of coffee plantations

• Autor(a) principal: Martello, Maurício
• Data de Publicação: 2022
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/11/11152/tde-04012023-154932/
• topic: Agricultura de precisão; Colheita mecanizada; Geoprocessamento; Geoprocessing; Mechanized harvesting; Precision agriculture; Remote sensing; Sensoriamento remoto

Brazil is the largest coffee producer and exporter in the world. In recent years, the need to modernize agricultural production has grown, based on the spatial variability of soil and plant attributes, in order to increase crop efficiency and rural producer profitability. In this sense, different sensing techniques (orbital, aerial and proximal) have been tested to map the spatial variability of plant and soil characteristics in different production systems, allowing to obtain high frequency data quickly and at low cost. However, for the coffee culture, studies that approach these techniques together to generate information that subsidize improvements in the spatialized management of the coffee plantation are scarce. As it is a perennial crop, coffee production is the integrated result of the various factors involved in the management of the crop, soil, climate and the plant itself, such as the behavior of productive biennials, with years of high and low yields. The creation of a system that allows obtaining this information quickly and non-invasively is essential for an efficient management of the crop. In this sense, this study sought to evaluate the potential use of data obtained by different sensors, seeking to identify attributes and features of the variability present in coffee plantations that express a direct relationship with yield and, consequently, identify the biennial yield (temporal variability). To this end, the study was divided into five chapters that address these gaps from different perspectives. Initially, a brief review of the main works published in the area is presented, then, in the second chapter, we sought to evaluate the quality of the data obtained through a yield monitor embedded in a coffee harvester. The data obtained by the monitor, of the volume of coffee harvested, showed a high correlation with data obtained with the load cells, validating the method and allowing the mapping of three consecutive harvests, which made it possible to advance in the understanding of the spatial and temporal variability of coffee yield in commercial areas. Aiming to explore the potential of mapping production variability before harvest, chapters three, four and five used remote sensing techniques to predict and map the spatial variability of a commercial coffee crop, with different sensors and at three levels of data acquisition. In chapter three, active optical sensors (AOS) embedded in a tractor, at a proximal level, were explored. In the fourth chapter, the use of aerial images obtained by remotely piloted aircraft (RPA) was addressed and in the fifth chapter, high spatial resolution orbital images were used. In general, data from active optical sensors showed a high correlation with yield, as well as allowing the temporal monitoring of the spatial variability of the study area, identifying regions that present inversion of yield (biennial). The results presented with the use of aerial images obtained by RPA demonstrated the potential of images in the extraction of biophysical parameters from coffee trees. They also allowed the individualization of plants aiming at extracting height and volume data, allowing the observation of the space-time relationship of the variables studied with yield data during three consecutive seasons. High spatial resolution orbital images showed the potential to predict coffee yield with high assertiveness one year before harvest. In general, the results found in this work reinforce the importance of knowing the productive spatial variability in coffee areas, as this type of information helps in the search for possible causes of this variability so that the culture can be managed considering these spatial and temporal differences. This study presented the spatio-temporal variations of data from orbital, aerial and terrestrial sensors in a commercial coffee area, as well as their relationship with yield maps generated with high data density, allowing to estimate productivity and identify production variations caused by biennial.