Ecological restoration is increasingly applied in wetlands to rehabilitate depleted ecosystem functions and increase faunal habitat. A wide range of post-restoration trajectories have been reported in the literature, but factors constraining wetland recovery are still not fully understood. A consistent and standardized monitoring of wetland habitats is integral to identifying how constraining factors interact to divert projects away from restoration goals. However, field assessments can be logistically challenging and costly as a result of wetlands’ high spatio-temporal heterogeneity. Remote sensing can offer a low-cost monitoring of restoration projects yet it is currently underutilized in ecological restoration.
To address this need, we leveraged public aerial images to characterize vegetation dynamics in restored wetlands. Using 20 restored sites of the Sacramento-San Joaquin Delta of California as a case study, we assessed the impact of site design, landscape context, and environmental fluctuations on the post-restoration trajectory of plant communities. We conducted an object-based analysis on a 10-year dataset of aerial images captured by the National Agriculture Imagery Program (NAIP) to identify vegetated patches within each study site. From this patch classification, a series of spatial metrics were generated to describe spatio-temporal variation in the structure, distribution, and complexity of these vegetated patches.
Results/Conclusions
Our study sites commonly experienced a rapid initial increase in both total vegetation extent and individual patch growth followed by a stabilization in subsequent years. Preliminary results suggest that temporal changes in total vegetated extent are affected by site connectivity as well-connected sites showed prolonged lateral growth. The positive impact of connectivity on vegetation growth is apparent in recently restored wetlands which likely benefit from an increased propagule availability as the density of restored wetlands increases in the region. Across the study area, we observed a negative relationship between patch area and growth, with smaller patches showing faster growth than larger ones. Hydrological and climatic conditions in the initial years of site development appear to impact vegetation growth and dispersal as evidenced by fluctuations in the rate of apparition of novel patches. Lastly, the classification accuracy of our object-based analysis varied between 75% and 90% - this demonstrates a potential to use free aerial images for the low-cost monitoring of wetland properties. Our study enhances the current understanding of factors promoting wetland recovery at both site and landscape scale. This knowledge is critical to an improvement of restoration design and increased resilience of future projects.