Cogongrass (Imperata cylindrica) is a problematic invasive species in the Southeastern US that invades longleaf pine forests, an endangered ecosystem with high understory plant diversity. Drought may exacerbate invasions if cogongrass is more drought tolerant than co-occuring native species. Cogongrass produces a large amount of biomass, which periodically dies back, producing a large amount of senesced and flammable material. As a result, fires in invaded habitats may be particularly intense, thereby exacerbating invasion effects. Remote sensing provides a tool for not only mapping invasive species on large spatial scales, but monitoring the effects of invasive species on ecosystems, including changes in water content and buildup of flammable material that both promote fire intensity.
To examine how drought and cogongrass invasion impact species diversity and ecosystem processes in longleaf pine communities, we established a common garden experiment where cogongrass invasion and rainfall input were experimentally manipulated. Here we test the ability to use hyperspectral measurements above the vegetation canopy to quantify alive and senescent cogongrass, native species diversity, and plant water status, as a tool for (1) frequently measuring ecosystem properties, which are prohibitively time consuming and expensive to measure in the field and (2) developing relationships that can be used to interpret the distribution and effects of cogongrass on landscape scales from airborne or satellite images.
Results/Conclusions
Our data show high biomass of senesced material in invaded plots irrespective of drought treatment. Plots with experimentally-induced drought had the same percent cover of dead (48±21%) and alive (34±15%) cogongrass compared to control treatments (dead 49±26%; alive 37±25%). Partial least squares regression indicated that hyperspectral data in the range of 450 – 2500 nm can be used to accurately predict the percent cover of both live and dead cogongrass. However, compared to the predictive ability for living cogongrass (R2 ranges from 0.57 to 0.81), the model showed better performance for the dead cogongrass (R2 ranges from 0.81 to 0.82). Our study shows the potential for cogongrass to thrive in drought conditions, and increase the amount of dead, and thus flammable, material compared to native vegetation in early successional longleaf pine systems. Furthermore, hyperspectral reflectance measurement techniques can be used to detect changes in ecosystem composition induced by invasive species at landscape scales using the increasingly available hyperspectral image data.