Where agriculture expands into tropical and subtropical forests, social–ecological impacts are typically strong. However, where and how frontier development impacts on ecosystem functioning and services is often unclear, including which services trade-off against agricultural production. This constitutes a major barrier towards planning for more sustainable outcomes in deforestation frontiers. Here we assessed spatiotemporal change in multiple ecosystem services in the Argentine Chaco, a global deforestation hotspot. We modelled and mapped five ecosystem functions (i.e. carbon storage in biomass, carbon storage in soil, erosion control, excess rainfall retention by vegetation and soil fertility) which together provide three ecosystem services (i.e. agricultural suitability, climate regulation and flood regulation) for 1985, 2000 and 2013. We then employed this information to identify and map: (a) main trade-offs between ecosystem services and agricultural production, and (b) bundles of changes in ecosystem services through the use of Self-Organizing Maps. Our results highlight that land-use changes since 1985 have led to widespread and drastic declines in ecosystem functions and services across the Argentine Chaco. Mean losses of ecosystem services ranged between 6% and 10% for flood regulation, climate regulation and agricultural suitability. The largest losses occurred in the Dry Chaco subregion between 2000 and 2013. We find two main types of trade-offs between regulating ecosystem services and agricultural production. Increases in crop and pasture production occurred along with large and moderate losses, respectively, in flood regulation and climate regulation over 20% of the region. Our mapping of bundles identified five common patterns of change in ecosystem services, delineating areas of stable or degrading ecosystem service supply. This provides a powerful template for adaptive spatial planning. Synthesis and applications. Using the Argentinean Chaco as an example, we demonstrate how combining fine-scale land-use maps with biophysical models provides deep insights into the spatiotemporal patterns of changes in ecosystem services, and their trade-offs with agricultural production. The periodic updating of maps of trade-offs and bundles of change in ecosystem services provides key inputs for the adaptive management of highly dynamic and threatened landscapes, such as those in tropical and subtropical deforestation frontiers.