As with all of our working papers, this is a preliminary analysis. The results described here should not be used to claim that critical habitat designation does or does not work: this a starting point only.

Background

Critical habitat occupies a contentious position in endangered species policy (see, e.g., James and Ward 2016). Although much has been written about the intent, scope, and effectiveness of critical habitat designations (Gibbs and Currie 2012; Mullen, Peterson and Todd 2013; Nelson et al. 2015; Taylor, Suckling and Rachlinski 2005), very little is known about its current ecological condition. This knowledge gap makes it difficult, if not impossible, to understand whether the U.S. Fish and Wildlife Service (FWS) is adequately protecting critical habitat or the extent to which critical habitat is furthering recovery.

In this working document, we test the hypothesis that remotely-sensed data–particularly the National Land Cover Database (NLCD)–can help close this pervasive knowledge gap by providing estimates of the extent of habitat changes over a 10-year period within designated critical habitat. Using a set of 42 ESA-listed species with critical habitat designated before 2000, we calculate the change in acreage of potentially suitable land cover types during the period 2001-2011. The focus on potentially suitable land cover types is important because not all areas within critical habitat polygons is regulated as critical habitat: the “physical and biological elements” a species needs must be present. We expect that high rates of land cover change and declining species status may signal inadequate protection of critical habitat, and may reveal instances of “destruction or adverse modification.” Conversely, we expect that low rates of land cover conversion and high species recovery indicate appropriate levels of protection. Our preliminary results indicate that most of the 42 listed species have not witnessed significant critical habitat losses during the study period.

Methods

To test our hypothesis, we first created a list of all ESA-listed species with critical habitat designated before 2001. Because of the complexities of linking land cover changes to aquatic critical habitat, we eliminated fishes from the list. Last, we removed species for which habitat disturbance was not the primary threat to their recovery, e.g., wolves (threatened by hunting) and condors (threatened by lead). The filtered list of 42 species with critical habitat designated before 2000 included plants, invertebrates, mammals, birds, and herps.

We used only publicly available data for our analyses, including:

  • critical habitat, provided by the Fish and Wildlife Service through its Environmental Conservation Online System, ECOS); and

  • the National Land Cover Database (NLCD) provided by the Multi-Resolution Land Characteristics Consortium, MRLC).

Within the NLCD, we used several products, including:

  • 2001 Land Cover (2011 edition),

  • 2001 Percent Developed Imperviousness (2011 editon),

  • 2011 Land Cover,

  • 2011 Percent Developed Imperviousness, and

  • 2001-2011 Land Cover From To Change Index.

We made a geodatabase of critical habitat with a seperate feature class for each of our 42 case study species. Using ArcGIS Model Builder, we clipped each of the NLCD products to each of the feature classes (i.e., 5 rasters for each of our case study species). For the land cover products, we used arcpy scripts to calculate the percent cover (2001 and 2011), the total acreage change, and the percent acreage change of each land cover type for each feature class. For the percent developed imperviousness, we used arcpy scripts to calculate the total number of acres with > 50% developed imperviousness for 2001 and 2011 for each of our 42 species. Those numbers were used to calculate total and percent acreage change in imperviousness.

Results

To provide context, we first present two examples of the NLCD layers within polygons of designated critical habitat:

Ex. 1. Mount Graham red squirrel habitats in 2001 (top) and in 2011 (bottom) after the large wildfires of 2004.

Green is evergreen forest; tan and whitish are shrub/scrub and grassland.



Ex. 2. Least Bell’s Vireo habitat in CH in 2001 (top), 2011 (middle), and the change in habitats from 2001-2011 (bottom).

Reds are classified as development of varying intensity; shades of brown and gray are ‘natural’ habitat.





Species

The percentage of critical habitat that changed from one NLCD class to another ranged from 0.1 to 53.7 per species. The average amount of change was 5.1% (median = 2%).

Note

The habitat changes for the three species with the highest percentage change are mostly attributable to fires. Some of the changes within the Amargosa vole (#4, by percent) habitat are beneficial, e.g., shrub/scrub to emergent wetland vegetation.

The proportions of habitat transitions varies by species. For example, the Mt. Graham red squirrel had the largest proportion of its habitat change, primarily because of large fire in 2004:

Whooping crane, which saw 3%, or >11,000 acres, of habitat change within critical habitat:

We include the from:to habitat transitions for all 42 species evaluated in an Appendix.

Habitats

We next asked if there were systematic habitat transitions across species. Note that the percentages are calculated over the area within CH designations, not the total extent of the habitat type in the US.

To see which habitat was converted to which habitat, we can create a heatmap. Note that the acres are log10-transformed because the area of shrub/scrub is so large relative to other areas:

We leave this working paper here - without a Discussion - because the results are incomplete. We will update this document as we collect and analyze more data in 2017.

Appendix

To facilitate species-by-species evaluation of habitat changes, we provide habitat ‘from:to’ heatmaps for each of the 42 species in the dataset. When examining these heatmaps, remember that our primary interest is with changes from and to habitat types that the species requires. (Because not all areas within critical habitat polygons is regulated as critical habitat: the “physical and biological elements” a species needs must be present.) The diagonal of each heatmap is the “no-change” amount for each habitat type within designated CH. We are therefore most interested in the off-diagonal blocks, which are read as the acreage from the habitat type on the y-axis to the type on the x-axis. Ideally, we want to see larger changes from types that are not suitable for a species to types that are suitable. Forthcoming work will focus on calculating the amount of ‘good’ and ‘bad’ changes for each species.