Probabilistic forecasting reveals a silent collapse of archaeological heritage in Peru's Chillón Valley

Unregulated land-use changes are accelerating the loss of archaeological heritage, yet quantitative forecasts of site disappearance remain rare. We present an open, Python-based pipeline that converts two decades (2000 – 2023) of satellite-derived loss trajectories for 56 pre-Hispanic sites in Peru’s peri‑urban Chillón Valley into probabilistic collapse forecasts. After filtering for adequate temporal coverage, 33 sites were modelled with ordinary least-squares regressions whose parameter uncertainty was propagated through 10 000 Monte-Carlo draws. Under a business-as-usual assumption—recent degradation rates persist—median disappearance years and 95 % credible intervals place six sites as already lost, eight at high risk (< 15 years), and the remainder in mid- and late-century risk tiers. Kernel-density visualization of the full 2000 – 2023 loss record reveals three valley-wide degradation phases: initial stability (2000 – 2005), acceleration (2005 – 2015) and late-wave intensification (2015 – 2025). The framework shows how retrospective mapping and forward-looking uncertainty jointly prioritize intervention, while the peri‑urban context cautions against uncritical generalization to dissimilar landscapes. All code and data are openly archived to foster replication and adaptation.