Natural ecosystem monitoring is essential to identify changes in vegetation conditions that may indicate threats to biodiversity. This study evaluates the utility of Sentinel-2A satellite time series to extract growth strategy patterns to assess the condition of a heterogeneous native grassland impacted by invasive species. Our study demonstrates great potential for mapping and quantifying perennial graminoids and contribute to the conservation and monitoring of native grasslands.
Natural ecosystem monitoring is essential to identify changes in vegetation conditions that may indicate threats to biodiversity and conservation value. Invasive species are one of the major threats to natural systems. But detecting invasive species in natural heterogeneous environments is extremely difficult because changes in surface reflectance at the early stages of invasions are subtle. Moreover, a background of high spatial and temporal variability in environments (i.e. canopy, soils, and rainfall) limits statistical detectability.
Differences in growth strategies may help distinguish native from invasive vegetation, presenting an opportunity to use high spatiotemporal resolution imagery to monitor changes in plant community composition. This study evaluates the utility of Sentinel-2A satellite time series to extract growth strategy patterns to assess the condition of heterogeneous ecosystems impacted by invasive species.
This study focuses on the “Iron-grass Natural Temperate Grassland of South Australia”, an ecological community of high conservation value severely affected by the invasive annual wild oat (Avena barbata). Field survey data from 2020 provided counts of iron grass (Lomandra effussa) tussocks in circular areas with a radius of 10 m for 255 locations. Abundance observations ranged between 0 – 73.21 tussocks/100 m2. Locations invaded by wild oats presented a relatively lower NDVI during the dry season and a relatively higher wet-season NDVI compared to areas dominated by iron grass, providing the notion for a simple index. We divided the NDVI time series into dry and rainy seasons and generated a regression model to predict iron grass abundance from dry and rainy season NDVI.
Our model explained 61.38% of the variation, revealing the dry season data as the most important predictor. The predicted abundance map verified with the test data points indicated an accuracy of R2 = 0.55, with an RMSE of ±13 tussocks/100 m2. We tested the model using 2019 NDVI time series data with an RMSE of ±7.73 tussocks/100 m2 and an accuracy of R2= 0.8. The methodology proposed in the current study demonstrates the potential to use time-series satellite imagery to extract quantitative plant community characteristics. Our study demonstrates great potential for mapping and quantifying perennial graminoids and contribute to the conservation and monitoring of native grasslands.
Mr Diego Guevara
Diego is a Peruvian biologist from La Molina Agrarian University with a Masters in Applied Ecology from the University of East Anglia, UK. Diego’s PhD in the Spatial Sciences Group, University of Adelaide, is studying the monitoring and restoration of native temperate grasses of South Australia.
Mr Diego Guevara (PhD Candidate at University of Adelaide)