Banca de DEFESA: PAULA DA SILVA FERREIRA

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : PAULA DA SILVA FERREIRA
DATE: 27/04/2026
TIME: 14:30
LOCAL: Remoto:meet.google.com/dwv-xjqm-erh
TITLE:

Functional Monitoring and Impact Assessment of Ecological Restoration in the Amazon: Biophysical Indicators Derived from PySEBAL and Counterfactual Evaluation

KEY WORDS:

surface energy balance; evapotranspiration; ecological restoration; remote sensing; protected areas.

PAGES: 170
BIG AREA: Ciências Biológicas
AREA: Ecologia
SUBÁREA: Ecologia de Ecossistemas
SUMMARY:

This dissertation assessed the functional trajectory of the Bom Futuro National Forest, a protected area located in Porto Velho, Rondônia, using biophysical indicators derived from the PySEBAL algorithm and counterfactual scenarios based on Random Forest models. The first chapter analyzed the temporal trajectory of aboveground biomass production (AGBp), biomass deficit (AGBd), and biomass water productivity (WPb) during the dry season from 2013 to 2025, using Landsat 8 imagery. The Mann–Kendall test with autocorrelation correction, Sen’s slope estimator, Pettitt’s change-point test, and the Wilcoxon test with Cohen’s d estimator were applied, segmented across seven edge zones. The results indicated consistent functional decline: AGBp showed a significant negative trend in approximately 30% of the analyzed area, with a median rate of −2.889 kg ha⁻¹ day⁻¹ year⁻¹ and a more intense deterioration gradient in edge zones; WPb exhibited a structural break in 2020 in more than 50% of stable pixels, with no recovery in subsequent years; AGBd confirmed the intensification of water stress after the break. The second chapter employed Random Forest models with spatial block cross-validation to construct counterfactual impact assessment scenarios across seven areas subjected to ecological restoration interventions. The models showed satisfactory performance for biomass and water productivity variables (CV R² between 0.925 and 0.963), and the endogeneity diagnosis, based on mean-shift Z-scores, did not identify violations of the exogeneity assumption in any of the 52 tests performed. Impact estimates were predominantly negative in the first six years post-intervention, interpreted as a reflection of non-stationarity caused by the 2020 regime shift, which compromises the extrapolation of the counterfactual baseline. A positive reversal trend observed in years 7 and 8 post-intervention is consistent with late-stage functional recovery patterns described in the literature. It is concluded that the Bom Futuro National Forest exhibits progressive metabolic deterioration, aggravated by a functional regime shift beginning in 2020, and that the integration of PySEBAL, non-parametric time series analysis, and machine learning-based counterfactual modeling constitutes an operationally viable approach for large-scale ecological restoration monitoring in the Amazon.

COMMITTEE MEMBERS:
Presidente - ***.701.522-** - MARCELO LUCIAN FERRONATO - UNIR
Interna - ***.324.392-** - MARCELA ALVARES OLIVEIRA - UNIR
Externo ao Programa - 1361358 - JHONY VENDRUSCOLO - UNIRExterno ao Programa - 1658780 - EMANUEL FERNANDO MAIA DE SOUZA - UNIRExterno à Instituição - SEMIRIAN CAMPOS AMOÊDO
Externo à Instituição - HERNANI FERNANDES MAGALHÃES DE OLIVEIRA