Digital mapping and sensors to assist soil genesis and estimation of ecosystem services

Abstract

Guaranteeing food security and more efficient agricultural production is an ongoing endeavor. Therefore, studies related to the agricultural production basis, which is the soil, are fundamental, starting with pedogenetic and morphological analyses, as well as studies on larger areas such as watersheds and more detailed scales on agricultural properties, both for soil classes and key attributes like carbon and carbon stock. Thus, this work begins in Chapter I, studying a Paleosol profile with contrasting redoximorphic features. A Paleosol is a soil that retains attributes from when it was formed and has not evolved alongside the current environment, making it a potential tracer of climate changes. With the advancement of knowledge, different techniques have been used in soil studies, and proximal sensors have proven to be powerful tools that, combined with traditional analyses, can assist in pedogenic studies by providing robust data. Hence, this study used data from proximal sensors applied to the digital soil morphometrics technique, which aims at the detailed characterization of soil profiles in a quantitative and spatialized way, with the potential to provide insights into aspects related to mineralogy, hydrology, paleo, and current-soil forming processes involved. This study demonstrated that the digital morphometrics applied to the Paleosol have the potential to aid in the conservation of spring areas due to the position of this soil profile in the landscape. In Chapter II, we used fuzzy logic to conduct digital soil mapping on four watersheds of Pontal do Paranapanema-SP. The goal was to create more refined maps of these watersheds and evaluate the impact of a detailed scale on ecosystem services related to sediment dynamics and aquifer recharge. This analysis allowed us to assess how agricultural management and environmental planning could be improved to minimize erosive processes and siltation of water resources and to help preserve aquifer recharge areas. Additionally, we explored how increasing the level of detail in soil maps and ecosystem services can contribute to improving public policies applied to environmental conservation. In Chapter III, we conducted a study to predict and map carbon and carbon stock at a detailed scale at the property level. Proximal and remote sensor data, along with variables related to soil formation factors and the variability of its attributes, were tested for this purpose. The Random Forest algorithm was used for modeling. The study found that carbon stock is challenging to predict as it varies even within the same soil class, use, and parent material, which suggests that more attributes influence its distribution in the area. These findings encourage further studies to evaluate the prediction and mapping of carbon stock using other variables for different types of environments.