The field of land system science aims to improve understanding of processes that lead to and stem from anthropogenic land use change, including social, economic, ecological and technological processes at scales ranging from local to global. Remote sensing assessments of land use and land cover change (LULCC) have sustained rapid progresses in ecology by measuring conversions between and within ecosystems (e.g., changes in forest canopy cover, leaf area index, carbon stock). However, fewer studies have focused on distinguishing the different agents associated with land use changes, and relating this with different social, economic and technological processes. In this study, we assessed land use conversions and modifications in Gurué District (589 km²), an area of Mozambique that has experienced recent growth in foreign investments in agriculture through large-scale land acquisitions, often resulting in land use conversions to mechanized agriculture. We built a flexible land classification system with Collect Earth software to gather training and validation data. We then examined the suitability of five image compositing techniques for generating cloud-free Landsat images for land use classification of spectral and textural features. We assessed three time periods between 2007 and 2017 and characterized land use change, including the expansion of large-scale mechanized cropland. Using Google Earth Engine and Collect Earth, this paper shows that some compositing methods adversely affect classification results. We present a new LULC assessment methodology for monitoring the expansion of cropland associated with different agents of change. This is an essential step toward understanding Gurué’s land system and where changes may be associated with positive and negative impacts upon small-scale Mozambican farmers and the natural environment.

Keywords:

Figures supplement:

https://drive.google.com/file/d/1_9ump-ZXH2bKDh79lUYdILWadM_ch_OO/view