Remote sensing of plant community succession in a semi-arid grassland

Background/Question/Methods

Ecological applications of remote sensing in grasslands are typically linked to physiological parameters, i.e. leaf area index, chlorophyll content, and biomass production. However, these parameters are often tedious and expensive to collect, and therefore, focus on short-term or one-time studies. Instead, long-term vegetation studies tend to collect data on species diversity, richness, and plant community structure but fail to produce biophysical parameters that can be used as ground truth for remote sensing. Linking diversity measurements to remote sensing has been successfully attempted in forested areas but very few studies attempt to do so in grasslands. In this study, we used remote sensing and GIS techniques to monitor plant community succession in a semi-arid grassland, the Nebraska Sandhills. The Nebraska Sandhills occur in a semi-arid sand dune complex and is largely stabilized by mixed grass prairie. However, areas of sparse vegetation result from various disturbances, such as improper grazing, fire, or drought. Strong winds then mobilize the sparsely vegetated dune sand, creating denuded hollows termed blowouts. Little is quantitatively documented about the re-vegetation patterns of blowouts. We used permanent plot vegetation data from two blowouts, sampled annually over a 29- and 15-year period, to describe and model plant community succession. An annual late growing-season Landsat image, from 1981 to 2010, was used to calculate a normalized difference vegetation index (NDVI). Accuracy of seral stage classification using NDVI and how it relates to successional stages and diversity indicators [species richness (S), Shannon’s diversity index (H´) and Pielou’s evenness index (J´)] are reported.

Results/Conclusions

A cluster analysis revealed three distinct seral stages; early, transitional, and late. H’ and S were highest in the transitional seres, followed by the late, and then early communities. Species in the late seres were less evenly distributed than in the early and transitional communities. Early plant communities contained mostly rhizomatous grasses, which stabilized the sand and likely facilitated the establishment of later seres. As sample blowouts aged, the presence of early seres decreased while transitional and late seres increased. Markov chain predictions show that blowout areas will likely not return to a solely late seral stage, instead, the disturbed areas are predicted to stabilize in a mostly transitional (50%) and late (38%) seral state, with 12% remaining in an early seral state.  NDVI was 65% accurate in classifying seres and was significantly (p<0.05) correlated with seral stages (r=0.81), S (r=0.72), and H’ (r=0.73) in this semi-arid grassland.