Archaeological Prospectivity and Terrace Morphometry Using Open Geospatial Data: A Reproducible GIS–Machine‐Learning Workflow for the Cuyes River Valley, Ecuador

Abstract

ABSTRACT This study develops and evaluates a reproducible archaeological prospectivity workflow for data‐scarce tropical landscapes by integrating restricted legacy archaeological documentation with openly accessible environmental geospatial datasets. The Cuyes River Valley in southern Ecuador, located on the eastern‐Andean foothills where montane environments descend toward the Upper Amazon basin, provides the test case. Archaeological reports produced in 2009–2010 were converted into a structured inventory of 78 terrace records from Espíritu Playa, San Miguel de Cuyes and Nueva Zaruma. Terrace morphometry was analysed using area, height, length–width ratio and elevation, whereas K ‐means clustering identified recurrent forms. A presence–background random forest model combined known terrace locations with Copernicus DEM‐derived elevation, slope and roughness and WorldClim temperature and precipitation covariates. Spatial cross‐validation yielded a mean ROC‐AUC of 0.753 and mean average precision of 0.604, indicating useful but spatially variable ranking performance. Annual precipitation, elevation and annual mean temperature were the strongest model predictors. The results distinguish higher, cooler and drier upper valley systems from a lower, warmer and wetter Nueva Zaruma context, whereas morphometric clustering separates moderate upper valley terraces, large construction units, lower valley compact terraces and small elongated features. Exact archaeological coordinates were withheld and public outputs spatially generalized. Beyond the case study, the workflow demonstrates how legacy archaeological records can be transformed into reproducible, ethically constrained prospectivity products for survey prioritization and heritage management where field data and high‐resolution remote sensing remain limited.

The authors' abstract, as published at the source. Archaeological Prospection, 2026 · DOI ↗