The global movement for ecosystem restoration has gained momentum in response to the Bonn Challenge (2010) and the UN Decade on Ecosystem Restoration (UNDER, 2021–2030). While several science‐based guidelines exist to aid in achieving successful restoration outcomes, significant variation remains in the outcomes of restoration projects. Some of this disparity can be attributed to unexpected responses of ecosystem components to planned interventions. Given the complex nature of ecosystems, we propose that concepts from Complex Systems Science (CSS) that are linked to non‐linearity, such as regime shifts, ecological resilience and ecological feedbacks, should be employed to help explain this variation in restoration outcomes from an ecological perspective. Our framework, Explore Before You Restore, illustrates how these concepts impact restoration outcomes by influencing degradation and recovery trajectories. Additionally, we propose incorporating CSS concepts into the typical restoration project cycle through a CSS assessment phase and suggest that the need for such assessment is explicitly included in the guidelines to improve restoration outcomes. To facilitate this inclusion and make it workable by practitioners, we describe indicators and methods available for restoration teams to answer key questions that should make up such CSS assessment. In doing so, we identify key outstanding science and policy tasks that are needed to further operationalize CSS assessment in restoration. Synthesis and applications. By illustrating how key Complex Systems Science (CSS) concepts linked to non‐linear threshold behaviour can impact restoration outcomes through influencing recovery trajectories, our framework Explore Before You Restore demonstrates the need to incorporate Complex Systems thinking in ecosystem restoration. We argue that inclusion of CSS assessment into restoration project cycles, and more broadly, into international restoration guidelines, may significantly improve restoration outcomes.

Eastern Rwanda consists of a mosaic of different land cover types, with agroforestry, forest patches, and shrubland all containing tree cover. Mapping and monitoring the landscape is costly and time-intensive, creating a need for automated methods using openly available satellite imagery. Google Earth Engine and the random forests algorithm offer the potential to use such imagery to map tree cover types in the study area. Sentinel-2 satellite imagery, along with vegetation indices, texture metrics, principal components, and non-spectral layers were combined over the dry and rainy seasons. Different combinations of input bands were used to classify land cover types in the study area. Recursive feature elimination was used to select the most important input features for accurate classification, with three final models selected for classification. The highest classification accuracies were obtained for the forest class (85–92%) followed by shrubland (77–81%) and agroforestry (68–77%). Agroforestry cover was predicted for 36% of the study area, forest cover was predicted for 14% of the study area, and shrubland cover was predicted for 18% of the study area. Non-spectral layers and texture metrics were among the most important features for accurate classification. Mixed pixels and fragmented tree patches presented challenges for the accurate delineation of some tree cover types, resulting in some discrepancies with other studies. Nonetheless, the methods used in this study were capable of delivering accurate results across the study area using freely available satellite imagery and methods that are not costly and are easy to apply in future studies.