Vegetation patterns at the landscape scale reflect myriad ecological processes and historical events, making the influence of any one process difficult to isolate or quantify. We combined remotely sensed LiDAR and hyperspectral data from the Carnegie Airborne Observatory with detailed site information and simultaneous autoregressive modeling to assess the relative importance of different factors in the assembly of an island community recovering from decades of overgrazing. Possible factors driving the system included disturbance intensity, environmental filtering, dispersal limitation, and feedbacks. This work was focused on Santa Cruz Island (SCI), a 250 km2 island off the coast of southern California where repeating topoedaphic patterns, a well-known disturbance history, and a limited set of dispersal agents constrain the number of possible ecosystem drivers.
Results/Conclusions
We found that vegetation height was the most readily explained by known factors (r2 = 0.52), while patterns in the normalized difference vegetation index (NDVI) were more weakly linked (r2 from 0.27 to 0.45). In general, the most important drivers of vegetation patterns were incident solar radiation, substrate, and land use history. Importantly, patterns in spatial autocorrelation among fruit and nut bearing shrubs were similar to the home ranges of the two main dispersal agents on the island, the SCI fox (Urocyon littoralis santacruzae) and the SCI scrub-jay (Aphelocoma insularis), which has implications for restoration both of SCI and neighboring Santa Rosa Island. In addition, we found a positive feedback relationship between environmental gradients, vegetation structure, and grazing, whereby areas that could support dense woody vegetation but were difficult for ungulates to access appeared less impacted by grazing pressure, allowing the vegetation in those areas to persist and soils to remain intact, while other areas were completely denuded and eroded. Our results demonstrate that geostatistics and remotely sensed data, when combined with detailed site knowledge, can effectively be used to quantify the relative influence of the different drivers of ecosystem assembly.