SkyTruth is developing tools to help land managers measure the impact of restoration activities from space.
SkyTruth just wrapped up the first year of a project to map vegetation trends along the Colorado River, so it seems like a good time to share some of the exciting progress we’ve made so far.
First, a bit of background: this project grew out of our work mapping mountaintop mining in central Appalachia. We’ve been tracking the extent of this disastrous activity for years, and we’ve been eager to turn our attention from measuring the extent of destruction from mining activity to measuring the extent of post-mining recovery. We’ve been mapping mining and vegetation in Appalachia for several years, so we decided to try out our approach on a vastly different – and far larger – landscape: the Colorado River watershed.
Second, a few useful clarifications and definitions. Technically, we’re mapping vegetation trends in the Upper and Lower Colorado River regions. The United States is divided and subdivided into ever-smaller hydrologic units, each with their own unique identifier called a hydrologic unit code. More commonly, they’re referred to simply by their acronym HUC. The system divides the country into 22 regions (referred to as HUC2s), 223 subregions (HUC4s), 387 basins (HUC6s), 2,303 subbasins (HUC8s), roughly 20,000 watersheds (HUC10s), and more than 100,000 subwatersheds (HUC12s).
The Colorado River is divided into two regions: the Upper Colorado River region and the Lower Colorado River region. The Upper Colorado region spans 113,347 square miles and contains 10 basins. The Lower Colorado region covers 163,842 square miles and contains 15 basins. Combined, the two regions are 8.6 times larger than SkyTruth’s mountaintop mining study area (about 32,000 square miles).
Similar to our mountaintop mining work, our goal is to create a 35-year time-series of Landsat Surface Reflectance-derived spectral indices for every subwatershed (HUC12) across the Upper and Lower Colorado River regions. For example, one of the spectral indices that we’re calculating is the Normalized Difference Vegetation Index (NDVI). It’s probably the most well-known and commonly used index for identifying vegetated areas and assessing vegetation’s greenness. In the context of this work, it’s a consistent and easy-to-compute proxy for the vegetation health in a watershed. Of course, NDVI has limitations, but with a 35-year time-series of measurements, we’ll be able to get a good idea of how vegetation health in a watershed has changed through time. You can find the specifics of all seven indices that we’re computing here.
Once we’ve got the analysis working, we’ll be able to plot those indices through time and identify interesting peaks and valleys at the subwatershed level. In theory, the peaks will be years when the landscape was greener than average and the valleys will be years when the landscape was less green than average. Just like with our mountaintop mining work, wherever we see a major deviation from the trend, we can investigate that subwatershed and that year in greater detail. Our hope is that watershed managers and river restoration specialists can connect vegetation trends over the last 35 years to different management practices. Ideally, they’ll be able to know what year they performed a particular treatment or management activity, and they can see how the subwatershed and the vegetation responded to that activity in the years that follow. We’re hoping that they’ll be able to identify which particular restoration techniques generate the landscape responses that they want, so they can make better choices in the future about targeting their efforts and resources
At this point, we’ve processed the most recent five years of Landsat imagery for all of the subwatersheds in the Colorado River regions and built a small, demo app using Earth Engine where users can visually compare indices across two different years in the San Miguel River watershed. We’re working to finish up the data crunching sometime in March, and then move on to building a fuller featured dashboard so that users can dive into the data themselves. The dashboard will allow people to select any subwatershed that they’re interested in and explore trends in all seven spectral indices over the last 35 years.
Make sure to keep checking the blog (or simply subscribe to get an automatic notification). Once we’ve finished processing the entire data set, we’ll publish another post that includes the code and details exactly what we’ve done. This will help others interested in replicating our work or adapting our approach for their own projects.
Featured image credit Amy Washuta, NPS/USGS Public Domain