The science guiding design and evaluation of restoration interventions in tropical landscapes is dominated by ecological processes and outcomes and lacks indicators and methods that integrate human wellbeing into the restoration process. We apply a new systems approach framework for tree restoration in forest-agricultural landscapes to show how this shortcoming can be addressed. Demonstrating 'proof of concept', we tested statistical models underlying the framework pathways with data collected from a case study in Tanzania. Local community perceptions of nature's values were not affected by levels of self-reported wildlife-induced crop damage. But mapped predictions from the systems approach under a tree restoration scenario suggested differential outcomes for biodiversity indicators and altered spatial patterns of crop damage risk, expected to jeopardize human wellbeing. The predictions map anticipated trade-offs in costs and benefits of restoration scenarios, which we have started to explore with stakeholders to identify restoration opportunities that consider local knowledge, value systems and human wellbeing. We suggest that the framework be applied to other landscapes to identify commonalities and differences in forest landscape restoration outcomes under varying governance and land use systems. This should form a foundation for evidence-based implementation of the global drive for forest landscape restoration, at local scales. This article is part of the theme issue 'Understanding forest landscape restoration: reinforcing scientific foundations for the UN Decade on Ecosystem Restoration'.
Keywords: coupled human and natural systems; ecosystem disservices; ecosystem services; forest landscape restoration; predictive modelling; systems modelling.