Identifying spatio-temporal trends in seagrass meadows to inform future restoration

Seagrass restoration requires information on a range of factors including site environmental conditions, appropriate planting techniques, and the identification of sites most likely to support seagrass. To address the question of where to focus restoration efforts, a key first step is to identify trends in the spatio-temporal distribution of seagrasses to identify areas of persistence, loss, and recent gains. Areas of recent recovery (and adjacent areas), can then be targeted by practitioners for assisted recovery and restoration, whilst areas of persistent loss can be avoided. Here we identified the contemporary distribution, density, and species composition of seagrass ecosystems (using Sentinel 2 imagery and supervised object-based imagery analysis) and integrated these data with historic extents to identify spatio-temporal trends in seagrass distribution in Western Port, Victoria, Australia. Contemporary classifications demonstrated acceptable accuracies (Overall Accuracy 0.77-0.85, User Accuracy 0.76-0.97) and predicted a contemporary seagrass extent of 222 km(2); with 48 km(2) of low-density recovery predicted to have occurred since 1999. Comparisons with historical seagrass extents indicated some seagrass recovery since large-scale losses in 1983, although some areas of loss were also present. Recovery included a net gain of approximately 95 km(2) in the past 20 years and an eastward range expansion; suggesting environmental conditions have improved and are now conducive for restoration efforts in some areas. Results demonstrate that accurate, low-cost, remote sensing of seagrass ecosystems is possible and show how understanding spatio-temporal trends can guide the spatial allocation of resources by prioritizing areas for restoration where recovery is beginning to occur.