Exploratory analysis of ground-based data in the field and multi-resolution satellite imagery to project land use and cover changes over deforested areas as detected by Terra-i

The Terra-i tool is a near-real time monitoring system which reports 250-m alerts of forest loss from 2004 to present at the pantropical scale. This study focused on a better understanding of the preceding and subsequent land use and cover (LUC) change trajectories across deforested areas as detected by Terra-i. In total, some 120 ground-based records collected during two field campaigns across the northern and eastern Peruvian Amazon were inspected and analysed at the alert extent and within a 1-km buffer zone. A set of variables were visually observed at (i) the time of field collection (occurrence of the change and types of LUC) and (ii) on an annual basis according to a multi-temporal analysis from 2004 to 2018 (types of LUC, percentage of intact/disturbed forest and fire activity). The use of multiple sources of geo-referenced data, including satellite data (from very high to coarse spatial resolution), drone imagery and field records, provided beneficial information to determine the consistency of the Terra-i tool as well as the spatial and temporal evolution of LUC trajectories over the sites analyzed. As a result, around 88% of the alerts presented any type of disturbance as Terra-i predicted. A remaining low proportion (12%) included false positives or areas which remained unchanged. The gathered annual LUC information indicated a gradual expansion of both cultivated and managed and mosaic of cultivated and managed/natural vegetation land cover types at the cost of tree cover from some years before Terra-i's forest loss predictions. In addition, it was found the LUC trajectories also varied according to their localization. While most of the sites at the northern side of the Peruvian Amazon followed a trajectory from forest to cropland, this resulted remarkably less evident for the sites located at the eastern side. The analysis suggested how we might improve the use of drones, high resolution imagery and online tools for the validation process.