Date on Master's Thesis/Doctoral Dissertation

5-2024

Document Type

Master's Thesis

Degree Name

M.A.

Department

Anthropology

Degree Program

Anthropology, MA

Committee Chair

Browne Ribeiro, Anna

Committee Co-Chair (if applicable)

Haws, Jonathan

Committee Member

Gaughan, Andrea

Author's Keywords

Amazonia; maximum entropy modeling; terra preta; Sentinel-2

Abstract

The results of a micro-regional predictive model of terra preta sites in the municipality of Gurupá are presented here. A Maximum Entropy approach was selected and modified to fit a presence-background model, rather than a presence-absence model typical of binary classifications. 13 known terra preta sites were used as input samples for the relative suitability model, with explanatory variables in the respective categories of topography, hydrology, soil type and vegetation index. A 30m spatial resolution DEM served as the basis for the former two categories, while additional TauDEM analysis was performed for the hydrology category. A Sentinel-2A mosaicked quarterly composite image of January 2021 served as the basis for derived data in the vegetation index category, NDVI and NDWI. Results of the model show the single most important predictor is the Gleysol soil type variable, with downslope at rivers being moderately contributive when riverine sites are used as test samples by the model. Notably, in addition to indicating presences in riverine fluvial contexts, this model displays strong presence signals at interfluves. When 2 out of 3 test samples used are interfluvial sites, slope, rather than downslope at rivers, is vi a major predictor of terra preta presence. In contrast to a basin-wide predictive model of terra preta sites (McMichael et al. 2014), elevation is a consistently absent variable when it comes to predictive power for this model. This is unsurprising, as Gurupá is generally low-lying with negligible increases in elevation towards its southeastern borders. This paper suggests that a basin-wide predictive model is unable to capture fine-scale environmental correlations with terra preta sites at the micro-regional scale. This research is a contribution to the growing body of work that includes the analysis of remote sensing data in the identification of terra preta sites in this region of Amazonia (Choi, Wright, and Lima 2020).

Download

Share

COinS