This study used remotely sensed mean normalized difference vegetation index (NDVI) data, which is a measure of vegetative greenness, to serve as a primary indicator of resilience. Mean NDVI for riparian vegetation was collected in 3 beaver complexes and 5 non-beaver developed sections along the Salinas River in California over a time span of July 2017 until December 2023.

Landsat Imagery was collected using the United States Geological Survey’s (USGS) web tool Earth Explorer (“EarthExplorer,”). Earth Explorer is an open-source tool that only requires a USGS account, which is a free process. In the search criteria tab, the date range was refined for the study’s desired time frame, which was 2017-06-30 to 2023-12-31. Cloud coverage was also modified so that only images with 10% or less cloud cover would be generated. In the data sets tab, Landsat C2 U.S. Analysis Ready Data (ARD) was selected. Using the spacecraft identifier dropdown menu under the additional criteria tab, I selected Landsat 8 and 9, which include the most recently launched satellites producing imagery with a resolution of 30 meters. The satellites have global coverage and revisit the same spot every 8 days. To get Landsat imagery that encompassed the entirety of the study area, the tile grid horizontal and vertical were set to 2 and 10, respectively, under the additional criteria tab. The fill (no data) dropdown menu from the additional criteria is also enabled for the specification of results with less than 10% of no data. Sometimes data is missing from captured images, so this option ensures that my search only produces images where less than 10% of the tiles have no data values. The results tab rendered an image with the desired parameters. Band 4 (red) and Band 5 (near infrared) .tif files were downloaded for each image, as these are the necessary bands for normalized difference vegetation index computation. To calculate NDVI from the downloaded Landsat imagery, a Python notebook within ArcGIS Pro was created. Images taken from the same date had the same prefix; thus, a script was needed to loop through the files and conduct the NDVI computation on bands 4 and 5 from the same date of collection. The output of this script was NDVI .tif files for each date of collection.  NDVI files were also re-projected to a consistent coordinate reference system. Another Python notebook was used to alter the NDVI raster set so that any negative values were considered as no data or null. Using “Create Random Points” in geoprocessing tools for ArcGIS Pro, a feature class of 100 points was randomly placed within each site polygon. Next, the tool “Extract Multi Values to Points” was used to calculate the daily NDVI for each point. Finally, using the “Summary Statistics” tool in the geoprocessing toolkit for ArcGIS Pro, the average daily NDVI was extracted for each site from the initial point sample values. For each site and time frame, there were on average 98.9 points of the 100 used to derive the mean NDVI. This tool skipped null values in the computation of the mean. The exclusion of null values ensured that open water, which can have negative values, was not used in the computation of the mean NDVI. Each image has a distinct date of collection; thus, the date was employed to distinguish data headings. This process was run using another Python notebook. The join field tool, as a batch, was used to join the mean NDVI data to the “Sites” feature class attribute table. The input table was the “Sites” feature class, and the join field was the “ObjectID” column. The summary statistic table served as the join table, and the join table field was “CID.” Other parameters were kept on the default settings, and the tool ran successfully. The delete field tool was then used to clean up the “Sites” attribute table and remove irrelevant columns. Fields which were kept included ObjectID, Type, Location, UniqueID, AREA, and the statistic that followed the format MEAN_{date}.