As the study of terrestrial ecology moves toward discerning patterns at larger spatial and temporal scales, there is an increasing need for efficient and cost-effective ways to allocate sampling locations given multiple, often remote and geographically disparate, research sites. Remote sensing, combined with a stratified and spatially-balanced sampling design, provides one avenue through which this can be achieved. The National Ecological Observatory Network (NEON) comprised of sixty sites across the continent, is a unique case in point highlighting the utility of such an approach. Here we present a novel, standardized method for plot allocation and validation. It is comprised of a workflow which employs the Reversed Randomized Quadrant-Recursive Raster (RRQRR) algorithm, stratification and allocation by the National Land Cover Database (NLCD), and remote sensing of high-resolution aerial imagery (30cm-1m) to quantify stratification accuracy levels, all within a GIS environment.
Results/Conclusions
We have employed this method for multiple taxa at 29 NEON sites to date. To test overall NLCD stratification accuracy rates, we evaluated 100 RRQRR-generated points per dominant vegetation layer(s) at each site. Every point was assessed as a 20m radius circle, representing the potential plot scale and design in which many of NEON’s terrestrial measurements will be made. Accuracies ranged from 45-88%, lowest being Mixed Forest and highest Barren Land. An advantage of the RRQRR design is that a potential sampling location can be excluded due to logistical or science-related constraints (i.e. incorrect vegetation), and an alternate, predetermined sampling location is available without compromising the integrity of the sample design. As a whole, this approach has several advantages in that it is highly cost-effective, can be scaled depending on the scope of the project, and allows more data comparability through standardization regardless of geographic site location or taxa being studied. High-resolution aerial imagery is publicly available through various seamless servers and imagery providers; though it should be noted field reconnaissance may still be necessary in some cases prior to plot establishment to ensure desired accuracy levels.