Digital solutions for coffee and terroir studies: pedological perspectives and soil-enviroment relationships

Abstract

Terroir refers to the interaction between environmental factors and human action that influence the final quality of agricultural products. Although traditionally associated with viticulture, the concept has increasingly expanded into coffee production systems. In this context, soil properties, parent material, climate, relief, organisms, and crop management directly influence coffee fruit maturation and beverage sensory quality. From a pedological perspective, these factors can be interpreted as proxies for the stratification of agricultural areas into management zones, enabling the understanding of how different areas impact beverage quality and distinct coffee sensory profiles. In this context, digital tools may play a fundamental role in coffee terroir characterization by enabling the integration and mining of soil-environment data. Among these tools, Geographic Information Systems (GIS), proximal sensors, laboratory analyses, scientometric approaches, and machine learning algorithms stand out as important techniques for terroir studies. Therefore, this thesis aimed to apply digital solutions to coffee terroir studies in order to evaluate pedological perspectives and identify how soil-environment factors explain coffee maturation and beverage quality. The thesis was divided into three chapters: (i) the first chapter applies digital tools to support pedological perspectives by using scientometric approaches to identify the scientific panorama, advances, and research gaps related to the soil-terroir interface, in addition to a SWOT analysis containing strengths, weaknesses, opportunities, and threats associated with the integration of pedology into terroir studies; (ii) the second chapter employs GIS, proximal sensing, and conventional laboratory analyses as digital tools for data acquisition and terroir characterization in a coffee-producing area located within the Campo das Vertentes Geographical Indication region, subsequently applying association rule mining algorithms to investigate how agro-environment attributes influence different coffee ripness stages; (iii) the third chapter uses the same digital tools for terroir data acquisition and characterization in the same study area, combined with the Random Forest algorithm, to predict which soil-environment factors explain distinct coffee sensory profiles. The results demonstrate the potential of digital tools and data mining techniques for agricultural environment stratification and for understanding the influence of terroir on coffee maturation uniformity and beverage sensory profiles. Furthermore, this research contributes to consolidating a pedological approach to coffee terroir studies by providing methodological and analytical support for future investigations in this field.