Covariate Spatial Table

One of the great strengths of the AKN is the ability to attach a wealth of spatial characteristics to any given point where a bird has been observed. We can provide a fully populated table to any interested data provider.

Over 1300 of these spatial covariates can be ascribed to any given point on the landscape and allow us to use what we know of the bird abundance (or presence/absence) at a given point to try to make predictions about other points on the landscape that share similar characteristics. We have used this to good advantage to generate predictive maps describing the occurrence of species such as Northern Cardinal, Yellow Warbler, and others. Below is an explanation of what goes on behind the scenes to generate this spatial information.

Much of the back-end work at the AKN has gone towards the development and maintenance of our spatial covariates. This information comes from several widely-available GIS layers that we use to assign landscape characteristics to a given point. The lion’s share of these layers stem from climatic maps from the Climate Atlas of the United States (CAUS) and land cover maps from the National Land Cover Data (NLCD) for 1992 and 2001. These two datasets provide most of our spatial information, totaling some 1323 discrete variables for every single point on the landscape! For the NLCD layers, although the original layers provide the covariates that describe the specific characteristics of the point, we also have generated information for the surrounding landscape. Thus, for a given point we describe not only the NLCD covariate for the point, but also the percentage of each NLCD covariate that occurs in the neighborhood.

We are continually searching for more layers to integrate that will improve our ability to describe the landscape in which AKN bird observations have taken place—many such layers are already in the process of being integrated. As more spatial information is represented in integrated GIS layers, more opportunities will arise for using these layers in our analysis and this process is ongoing within the AKN.

One weakness we are acutely aware of is that the habitat descriptions in the NLCD data is quite coarse. For example, “evergreen forest” does not distinguish between Pitch Pine woods with nesting Eastern Towhees and Prairie Warblers, White Pine woods with nesting Pine Warblers and Blue Jays, or spruce forest with nesting Blackpoll Warblers and Yellow-bellied Flycatchers. One of our long-term goals is to identify and incorporate habitat layers that provide more granularity to habitat classifications and thus are more relevant to avian distribution.

Below we describe these variables in more detail and if you’d like to see the complete list, you can download an excel file of the descriptors and their definitions here.

The Climatic Atlas of the United States (CAUS), developed by NOAA’s National Climatic Data Center, provides GIS maps that show the spatial distribution of major climatic elements, including temperature, precipitation, snowfall, wind, pressure, and more. We have incorporated most of these maps, including all those that have obvious relevance to bird distribution and abundance patterns. You can read the full descriptions of the elements here or download our complete list of Covariates, including CAUS data, here.

National Land Cover Data (NLCD)
These layers characterize landscape features based on a 30x30m grid. There are two iterations of these data: 1992 and 2001. The greatest difference between the 1992 iteration and the 2001 version, is that while the 1992 layer included only land cover information, the 2001 version also includes two additional metrics: imperviousness and canopy density. Some categories from 1992 were retired (e.g., Quarries/Strip Mines/Gravel Pits and Orchards/Vineyards/Other), while the 2001 version includes four new habitat classes for Alaska and nine new classes for coastal areas. Further, the more recent version revised the algorithms for calculating the land cover boundaries, improving the precision in the more recent version. Finally, NLCD 2001 was expanded from the 48 conterminous states to now include Alaska, Hawaii, and Puerto Rico. You can read the full descriptions of the 1992 elements here, the 2001 elements here, or download our complete list of Covariates, including NLCD 1992 and NLCD 2001 data, here. Below is an explanation of the main categories.

Canopy density (NLCD 2001 only): Each 30x30m grid cell that is ranked as forest (i.e., deciduous, evergreen, mixed, or woody wetland) within the land cover data is scored with a value of percent canopy cover (0-100%, thus 101 possible values).

Impervious surface (NLCD 2001 only): As for canopy cover, each 30x30m grid cell is classified with 101 possible values to indicate the percentage of impervious surface (i.e., pavement). This gives a quick assessment of the percentage of a given area that is developed. Parking lots and buildings have 100% impervious surface while natural areas with grass or earth have 0% impervious surface.

Land cover: Each 30x30m grid cell is characterized with one of 16 classification schemes (habitats). An additional four categories are unique to Alaska. These are fairly general with respect to bird habitat (i.e., evergreen forest does not differentiate longleaf pine forest of the Southeast from boreal spruce-fir forest of Alaska and northern Maine) but do provide a quick view of bird habitats.

Miscellanious additional layers

In addition to the CAUS and NLCD data discussed above, the Covariate Spatial Table also includes a selection of other layers that give additional information.

Subnational1 (or State):  This layer provides geopolitical information at the first level below country. For the USA, Mexico, and several others countries, this is a “state” layer, while in certain other countries it shows the governmental equivalent (e.g., provinces in Canada and Panama and departments in Guatemala).

Subnational2 (or County): For regions that have additional subdivisions below the level of state, province, or department (or equivalent), this layer provides that information. For most of the United States these are counties, although in Louisiana they are parishes.

Elevation: Provides elevation (in meters). We actually use two different layers to provide elevation. For the Lower 48 States, the Spring 2005 version of the National Elevation Dataset (NED48) gives elevation at a fine scale (about 30x30m, or 60x60m in Alaska). On a Global scale, the December 2004 GTOPO30 layer provides elevation throughout the world, but uses a wider grid of about 1 square kilometer.

Population: Human population per square mile based on ESRI data from the 2004 US Census. We also have a layer defining the Block Group from the ESRI 2004 Census layer.

Bird Conservation Region: Bird Conservation Regions (BCRs), delineated in 1998 by a mapping team under the guidance of the North American Bird Conservation Initiative (NABCI), are ecologically distinct regions in North America with similar bird communities, habitats, and resource management issues. See http://www.nabci-us.org/bcrs.html for more information.

Lat/Long Grids: It is sometimes useful to group data by cells of latitude and longitude (latilongs). We have two layers devoted to assigning latilongs: one for a one degree cell and one for a four degree cell.

If any data provider is interested in receiving a file showing their datapoint locations populated with this additional covariate information, simply contact the NADC coordinator Marshall Iliff for assistance.

Table 1. The National Land Cover Data (NLCD) layers for 1992 and 2001 are listed and compared. Roughly equivalent layers are listed on the same line, while those on different lines highlight differences between the 1992 and 2001 iterations. Layers from 2001 marked with a single asterisk (*) indicate layers that are unique to Alaska, while those marked with a double asterisk (**) indicate layers that are strictly coastal.