The rise of forest plantations in Chile's Mapuche's homeland: Four decades of land cover estimates from a CNN-RNN model and the Landsat program

Abstract

This paper documents how policy supporting the expansion of forest plantation has transformed the homeland of the largest indigenous group in Chile, the Mapuche. It presents the first decadal estimates of land cover since 1973 in a region that contains most Mapuche reservations to study how plantation forests have transformed Mapuche's landscape in the past four decades. The proposed CNN-RNN deep learning architecture combines low-resolution satellite imagery from the 1970s with modern contemporary satellite imagery to deliver state-of-the-art decadal land-cover maps. The results reveal that plantations sharply increased in the proximity of reservations since 1973. A large share has replaced native vegetation.

Keywords: Mapuche, Plantation forest, Remote sensing, Deep learning