Creating space for ecosystem restoration by increasing operational efficiency in sugarcane harvesting

Author(s)

Giulio Brossi Santoro

• Document type: Master's Dissertation
• Press: Piracicaba.
• Institution: Universidade de São Paulo (USP). Escola Superior de Agricultura Luiz de Queiroz (ESALA/BC)
• Defense date: 2022-07-06

Examining board members

Paulo Guilherme Molin; Adriana Cavalieri Sais; Leandro Reverberi Tambosi

• Advisor: Paulo Guilherme Molin

Abstract

The demand for ecosystem restoration has been growing globally, seeking to achieve different types of environmental and social benefits, such as biodiversity conservation, mitigation of climate change and the protection of water sources. However, the availability of areas for restoration of native ecosystems is hampered especially by agriculture. The efficiency increase of agricultural systems is an opportunity to free up marginal areas for restoration. In this way, improving the efficiency of sugarcane mechanized harvesting in the state of São Paulo stands out as such opportunity. The goal of this work was to identify marginal areas of mechanized sugarcane production, through the identification of Short Line Areas (SLA), aiming at expanding the restoration of native ecosystems to such regions. The SLA are characterized by short planting lines and in this project two minimum thresholds were considered: 50 and 100 meters. Thus, statistical regression models were developed using supervised machine learning algorithms to estimate statewide SLA. The models were based on a sampling of 120 agricultural landscapes of 25 km2 throughout the state (total of 7,553), in which the SLA were manually mapped and explanatory variables that could be correlated with the presence of the SLA were explored (mean slope; drainage density; percentage of sugarcane cover, and landscape metrics applied to the crop fields). Once the estimates were calculated and spatialized, the potential contribution of the restoration of native ecosystems in SLA was investigated in the context of reducing the Legal Reserve Deficit (LRD). The results showed the best performance of the models created by the Random Forest algorithm, for data adjustment (training) and also testing. The performances were evaluated through metrics such as the Coefficient of Determination (0.451 and 0.634), Mean Absolute Error (0.252 e 0.932) and the Root Mean Squared Error (0.323 e 1.195). With this approach, estimates for the entire state were 174.19 km2 of SLA considering the 100-meter threshold; and 39.78 km2 for the 50-meter threshold. The use of SLA to reduce the LRD showed the potential to mitigate 100% of the deficit in some landscapes (from 240 up to 2,479 according to the scenario), but in general it contributes to mitigating from 0.43 up to 4.83% of the sum of LRD from the landscapes, depending on the scenario considered. Although the estimates of SLA were numericly low towards the direct contribution to minimize the LRD, they represent an efficient approach to identify regions susceptible to restoration, which can still be addressed in the context of incentives for voluntary commitments; corroborating the efforts, commitments, policies and projects of ecological restoration undertaken in face of ecosystem restoration’s decade. (AU)

• FAPESP's process: 20/15792-3 - Creating spaces for ecosystem restoration by increasing operational efficiency in the harvest of sugarcane
• Grantee: Giulio Brossi Santoro
• Support Opportunities: Scholarships in Brazil - Master