A Multi-Temporal Analysis of Archaeological Site Destruction using Landsat Satellite Data and Machine Learning, Moche Valley, Peru

Abstract

The destruction of archaeological sites and the loss of archaeological landscapes remains a global concern as populations and urban areas continue to expand. Archaeological sites are not only significant to local communities, national identities, and modern tourist economies but also provide critical knowledge of past sociocultural interactions, settlement patterns, human-environment relationships, and risk mitigation strategies. While archaeological landscapes and site destruction have remained outside of traditional land use land cover change (LULCC) studies, they are a form of urban and agricultural land use. By conceptualizing archaeological site destruction within land change science, this study provides an innovative approach for assessing “what's left” of historically surveyed archaeological landscapes. Using a Random Forest algorithm and Landsat satellite data, this study quantifies archaeological site destruction attributed to LULCC in Peru's lower Moche Valley between 1985 and 2020. More than 400 archaeological sites previously recorded during the Chan Chan-Moche Valley Project (CCMVP, 1969–1974) are analyzed. Results indicate that less than a quarter of the original CCMVP sites remain on the landscape. The primary drivers of LULCC in the lower Moche Valley include population growth, migration, and government policies, while secondary drivers include heritage values. Positioning archaeological survey data within land change science and integrating machine learning techniques can benefit historic survey reassessments globally and provides significant knowledge of archaeological site destruction and the socioeconomic conditions that underly dynamic landscape changes.