3. PREDICTED ANIMAL DISTRIBUTIONS
AND SPECIES RICHNESS
Limitations and Discussion
All species range maps are predictions about the occurrence of
those species within a particular area (Csuti 1994). Traditionally,
the predicted occurrences of most species begin with samples from
collections made at individual point locations. Most species range
maps are small-scale (e.g., >1:10,000,000) and derived primarily
from point data for field guides. The purpose of the Gap analysis
vertebrate species maps is to provide more precise predictions of
the current distribution of individual native species within their
general ranges. With this information, better estimates can be made
about the actual amounts of habitat area and the nature of its configuration.
Gap analysis maps are produced at 1:100,000 and are intended for
applications at the landscape or "gamma" scale (heterogeneous areas
generally covering 1,000 to 1,000,000 hectares and made up of more
than one kind of natural community). Applications of these data
to site- or stand-level analyses (site - a microhabitat, generally
10 to 100 square meters; stand - a single habitat type, generally
0.1 to 1,000 ha; Whittaker 1977, see also Stoms and Estes 1993)
are likely to be compromised by the finer-grained patterns of environmental
heterogeneity that are resolved at those levels.
Gap analysis uses the predicted distributions of native vertebrate
species to evaluate their conservation status relative to existing
land management (Scott et al. 1993). However, the maps of vertebrate
species distributions may be used to answer a wide variety of management,
planning, and research questions relating to individual species
or groups of species.
Previous to this effort there were no maps available, digital or
otherwise, showing the likely present-day distribution of species
by habitat type across their ranges. Because of this, ordinary species
(i.e., those not threatened with extinction or not managed as game
animals) are generally not given sufficient consideration in land-use
decisions in the context of large geographic regions or in relation
to their actual habitats. Their decline because of incremental habitat
loss can, and does, result in one threatened or endangered species
"surprise" after another. Frequently, the records that do exist
for an ordinary species are truncated by state boundaries. Simply
creating a consistent spatial framework for storing, retrieving,
manipulating, analyzing, and updating the totality of our knowledge
about the status of each vertebrate species is one of the most necessary
and basic elements for preventing further erosion of biological
The species nomenclature used in this report generally follows
the standards of The Nature Conservancy's (TNC) Central Scientific
Databases for animals. Where this standard is simply a name change,
we have used the new TNC name along with the WHR code number. In
a few cases, the reclassification of species has led to a split
or aggregation of species in the WHR database. It was often not
possible to reasonably model the distribution of such a taxon, and
in such cases we have retained the original taxonomy as defined
for the WHR database.
Specific taxonomic issues were resolved as follows:
- The Yellow-bellied
sapsucker, as currently classified, does not occur in California.
Populations formerly assigned to Sphyrapicus varius are
now called S. nuchalis, the Red-Naped Sapsucker.
- The Western
flycatcher (Empidonax difficilis) has been split into two
distinct species, the Pacific-slope flycatcher (which retains
the original scientific name) and the Cordilleran flycatcher (E.
occidentalis). Because the two species occupy distinct geographic
ranges in California, we were able to associate the portions of
the overall range with the appropriate species and predict their
habitats in this gap analysis. E. occidentalis was assigned
a new WHR code for this project (B990).
- A similar
situation occurred with the Black-tailed gnatcatcher. Polioptila
melanura was split into a desert-dwelling Black-tailed species
which retained the scientific name, and the California gnatcatcher
(P. californica) in coastal sage habitats. The latter was
assigned a WHR code of B991, and the two species were modeled
independently in this analysis.
- Shrews along
the west coast have been revised. The Pacific shrew (Sorex
pacificus) no longer occurs in California. The populations
formerly called S. pacificus are now assigned to S.
sonomae, the Fog Shrew.
- The Pacific
giant salamander (Dicamptodon ensatus) was recently split
into three species, two of which occur in California (D. ensatus
and D. tenebrosus). We were unable to distinguish the individual
ranges of these two species and chose to leave these two species
as one for this analysis. Conclusions about the management status
of either species may not be reflected accurately in the results
for the combination of the two.
- The Olympic
salamander was split into two species. Currently, Rhyacotrition
olympicus does not occur in California, but the range is occupied
by R. variegatus, the Southern Torrent Salamander.
- The slender
salamanders are undergoing major revisions in taxonomy. Because
of the volatility of the nomenclature and uncertainties about
the ranges of new species, we kept the California slender salamander
(Batrachoseps attenuatus) and the Pacific slender salamander
(B. pacificus) as portrayed in the California Wildlife
Habitat Relationships System.
- The Spotted
frog (Rana pretiosa) may not have occurred in California as specimens
may have been mis-identified. We have retained this species in
the predictive modeling, but readers are warned that this may
- The Black-collared
lizard (Crotaphytus insularis) has been split into C.
insularis and C. bicinctores, the Mojave black-collared
lizard. We were unable to distinguish their individual ranges
from the CWHR range maps and have modeled them as a single species.
of the Western aquatic garter snake (Thamnophis couchii)
were recently elevated to full species status. Because range data
is not yet provided in CWHR for the Santa Cruz garter snake (T.
atratus) or the Giant garter snake (T. gigas), our
modeling and analysis are for T. couchii as originally
Distribution of vertebrates were modeled based on a six-step procedure
that takes advantage of the known habitat preferences of species.
First, criteria were developed to select the species to be included
in gap analysis. Second, the distributional limits of each selected
species was determined from available range maps. Third, the existing
California Wildlife-Habitat Relationships (CWHR) database (Airola
1988) was used to assign suitability ratings to habitat types. Fourth,
the land-cover map (Chapter 2)
was reclassified as wildlife habitat types. Fifth, the range data
and the CWHR database were used in a GIS-modeling process that assigned
species to mapped habitat polygons (Chapter
2). Distributions were also summed for 7.5 minute quadrangles
to create maps of species richness for each taxonomic group. Sixth,
the modeling was compared to observed species lists for many parks
and with the Breeding Bird Survey data.
There are 650 terrestrial vertebrate species included in the CWHR
database. Some of these are introduced. Some do not breed in the
state or are rare visitors. Some are marine mammals and pelagic
birds that would not be appropriately modeled with GAP data on terrestrial
environments. Therefore we shortened the list by the following criteria:
- Exclude introduced
species such as the bullfrog (Rana catesbeiana).
- Exclude species
not related to habitats of the terrestrial surface of the state,
i.e., marine mammals, pelagic birds, waterfowl, bats.
- Exclude species
that do not breed in the state, such as migratory birds that only
winter in California.
screening step, 455 species remained, including 205 birds, 134 mammals,
45 amphibians, and 71 reptiles.
range of each species was represented as presence or absence in
7.5 minute quadrangles based on digital maps of range limits that
were produced by the California Department of Fish and Game for
the three-volume series on California wildlife (Zeiner et al. 1990).
These maps were originally prepared at 1:3,500,000 scale based on
available scientific literature, museum records, and major physical
and/or vegetational features. Some of these maps have been more
recently revised at a 1:1,000,000 scale by Fish and Game. The range
maps were summarized as a look-up table by Fish and Game as a list
of quadrangle-species pairs. The November, 1997, version of the
ranges by quadrangle look-up table was used for our predictive modeling.
The exception to this was for birds, where the quadrangle lists
did not distinguish breeding, summer range from winter range for
migratory birds. For the birds, therefore, we compiled new lists
by quadrangle from the latest set of range maps.
ratings by habitat type for each species were obtained from Version
5.3/6.0 of the California Wildlife-Habitat Relationships database.
This database had been compiled and revised by an interagency team
of wildlife biologists to contain all available information on habitat
requirements or terrestrial vertebrates (Airola 1988). CWHR ranks
each habitat type as High, Medium, Low, or Unsuitable for breeding,
feeding, and cover. For gap analysis, we only employed the suitability
rating for breeding.
2, we described the development of the CA-GAP land-cover map.
In addition to classifying landscape units into plant community
types, we also classified them into CWHR habitat types. Barry Garrison,
the Fish and Game manager of the WHR Program, assisted us in cross-walking
the combination of dominant species in each polygon into the habitat
types described in the CWHR system (Mayer and Laudenslayer 1988,
Schultze 1994). As each polygon in the vegetation map is considered
to be a landscape mosaic of several habitat types, three major CWHR
vegetation types can be assigned to each polygon as well as several
different CWHR wetland/riparian types, the latter being coded as
attributes of each polygon during the original mapping. The cross-walking
procedure used a scoring system that rated the propensities of plant
species in the CA-GAP land-cover database to be associated with
various CWHR habitat types. The habitat type with the highest combined
score for the combination of plant species was assigned to the polygon.
A species can
then be predicted to be present or absent in a landscape map unit
through the following GIS-based modeling process. The map of habitat
types is overlaid with quadrangles. For each polygon, the model
refers to a look-up table to determine if the polygon falls within
the species' range as represented by the quadrangle maps. If so,
the model then checks another look-up table of habitat suitability
rankings for each of the three habitat types recorded in the polygon.
Then the areal extent of each suitability rank in every habitat
polygon is summed, accounting for the relative proportion of the
primary, secondary, and tertiary habitat types present. Thus a habitat
polygon might be modeled as having 30% High suitability habitat,
10% as Medium, 20% as Low and the remaining 40% as unsuitable. The
model may also indicate that one or more of the wetland/riparian
habitats originally recorded as polygon attributes may also be suitable
(or in fact, only the wetland/riparian habitat may be suitable).
Rather than simple presence/absence results, therefore, the CA-GAP
wildlife modeling produced a vector of proportions of suitability
levels for each polygon.
process can create problems at the edge of a species range limits.
If any part of the habitat polygon overlaps a quadrangle within
the range, one option would be to model the entire habitat polygon
as potentially suitable habitat. For large habitat polygons spanning
many quadrangles, this alternative solution could predict species
presence far beyond its actual range limit. Alternatively, the predicted
distribution could be arbitrarily truncated at the quadrangle boundary,
creating an unnatural, rectilinear edge. Our solution was to check
the proportion of each polygon within the range limits. If less
than a majority of the habitat polygon was within the range limits
as summarized by quadrangle, the entire habitat polygon was considered
these complex, multivariate data, we categorized the output into
a combination of area and suitability to give a single class per
polygon that could be displayed and analyzed further (Table 3-1).
Thus a polygon that contained a large proportion of the best habitat
could be distinguished from one with only a small amount or one
where an arbitrary threshold for habitat suitability is used to
include or exclude polygons. We believe this categorization is more
meaningful than simple presence/absence data in which all the suitability
levels and their extent is ignored and avoids making arbitrary decisions
in the predictive modeling phase about what suitability level or
areal proportion should be used as a thresholds to include or exclude
polygons from the predicted distribution. Note that the categories
are assigned in descending order, so that a polygon is assigned
to category 4 only if it does not also qualify in category 5. Category
1 is primarily restricted to those habitat polygons in which none
of the three major habitat components of the landscape mosaic (i.e.,
primary, secondary, tertiary) are considered suitable for a species,
but where one or more of the wetland/riparian types is suitable.
These types were only recorded as being present in a polygon where
their areal extent was too small to include them as one of the dominant
types. This category is also used in some cases where Joshua trees
are known to be present. If their density was unknown, as described
in Chapter 2, we could not confidently
assign the polygon to Joshua Tree Woodland habitat. Instead, the
polygon was rated as Category 1. Consequently there is no associated
areal extent for these minor habitat types and no way to determine
in which higher suitability category they might be included. Category
1 may in fact be very critical habitat for the species, but we can
not use that information in spatial analysis. If a wetland or riparian
type was large enough to be one of the three major types in a polygon,
however, that would be a factor in assigning the polygon to categories
2-5. For mapping species richness (below) and for gap analysis (Chapter
6), we only include suitability ranks of 4 and 5 and omit the
lower suitability levels.
>50% Medium or High Medium or High Suitability FONT>
Low, Medium or High Suitability
Low, Medium or High Suitability but >0%
habitat in wetland/riparian types only (no areal estimate)
3-1. Categories of predicted habitat quality based on suitability
and areal extent.
Once a list
of species is ascribed to each polygon, the predicted distributions
were summed into species richness maps for the four major taxonomic
groups: birds, mammals, amphibians, and reptiles. This involved
resampling the species distributions mapped on habitat polygons
to a uniform grid system of equal-area units (Stoms 1994). We selected
the 7.5 minute USGS quadrangles and counted every species present
in any polygon within the quadrangle with a suitability category
of 4 or 5. That is, only species with the best habitat quality polygons
were counted. This count would be similar to what could be calculated
directly from the quadrangle by species look-up tables used as input
to the GIS-based modeling. The difference here is that habitat distribution
and suitability has been used to refine the coarser level representation
of species range.
The data on
species by quadrangle also allowed us to examine the turnover of
species between sites, that is the similarity of species in one
quadrangle to the set of species in every other quadrangle. This
information is complementary to species richness because it can
illustrate that two equally rich sites may, in fact, have distinctly
different species composition and one site could not be substituted
for another solely on the basis of richness. Our measure of similarity
is the Jaccard similarity coefficient, which is simply the percentage
of species common to both sites in relation to their combined set
of species. An index of 100% would indicate identical species composition
and therefore the sites would be interchangeable in any conservation
plan based on this biological information alone. A low index occurs
when the two sites share very few species. The index is calculated
as a "distance" between every pair of sites, and so a
map of the similarity index must always be in reference to a specific
In this section,
we present a sample of representative maps of predicted species
distributions showing the relative suitability rank of each habitat
polygon. (The full set of distributions can be viewed in the CD-ROM
and GIS database products.) The complete table of total mapped distribution
(suitability ranks 4 and 5 only) and percent area of the entire
state of California, is included in the gap analysis table in Appendix
in California tend to have relatively wide distributions. The ext
ent of high and moderate suitability habitat predicted for most
species was between 10-50%. Very few species had more than 50%,
e.g., turkey vulture. The Black-headed grosbeak (Figure 3-1) had
high and moderate suitability habitat predicted for 31% of California's
land area. The highest quality habitats are found in the foothills
surrounding the Great Central Valley, portions of the north coast
and urban areas such as Sacramento and Los Angeles. (See the Limitations
and Discussion section at the end of this chapter about the suitability
of urban lands as species habitat).
3-1. Predicted distribution and suitability categories of the Black-headed
grosbeak (Pheucticus melanocephalus).
the mammals were predicted to have high and moderate quality habitat
over much of the land area of California. These species include
the Black-tailed jack rabbit, Botta's pocket gopher, Western harvest
mouse, deer mouse, coyote, american badger, and bobcat, all with
distributions greater than 70% of the state. The Western gray squirrel
is widespread (26%) through the montane areas of the state (Figure
3-2). Its highest suitability habitat, however, occurs in the area
surrounding the Great Central Valley. Habitat suitability tends
to be low on the central and southern coastal regions.
3-2. Predicted distribution and suitability categories of the Western
gray squirrel (Sciurus griseus).
3-3. Predicted distribution and suitability categories of Western
spadefoot (Scaphiopus hammondii).
Many of the
amphibians have very narrow ranges in California, generally less
than 10%. The widest predicted occurrence for any amphibian was
for the Pacific chorus frog (Pseudaris regilla, WHR code
# A039) with highly suitable habitat in nearly half the land area
of the state. Several species were predicted to occur in no quadrangles
with a suitability rank of at least 4. These species all had very
small ranges, such as the Black toad (Bufo exsul) which only
occurs in a single wetland in Deep Springs Valley in Inyo County.
When the only suitable habitat for the species was wetland or riparian
and this habitat was below the mapping resolution of CA-GAP, our
method predicted a suitability rank of 1, which was not included
in calculating species richness or in the gap analysis of predicted
species distributions. The predicted distribution of the Western
spadefoot is shown in Figure 3-3 which includes all suitability
ranks. Although its range extends across the Great Central Valley
and the central and southern coastal regions of the state, its most
suitable habitat is relatively limited (roughly 11% of the state)
to a narrow band around the valley.
to have wider distributions than the amphibians. Many species' distributions
were predicted to exceed 20% of the state. With the exception of
the gopher snake and the common kingsnake, however, no reptiles
had highly suitable habitat over more the 50% of the state. The
predicted distribution shown in Figure 3-4 is for the California
mountain kingsnake. Although it is relatively widespread, there
is virtually no high quality habitat (rank = 5) for this species.
3-4. Predicted distribution and suitability categories of California
mountain kingsnake (Lampropeltis zonata).
determined by tallying all vertebrates with at least one polygon
with a suitability rank of 4 or 5 in a quadrangle. Thus the entire
quadrangle should not be assumed to be high suitability, only that
some Medium or High quality is present. Similarly, richness is based
on predicted habitat suitability, not on known occurrences of species.
So richness itself is only an estimate of the actual number of species
present. Because richness patterns are seldom the same among taxonomic
groups (Prendergast et al. 1993), we computed it separately for
the four orders of terrestrial vertebrates (Figures 3-5 through
of species richness at the scale of 7.5 minute quadrangles are quite
distinct among the four taxonomic groups. Land birds (summer ranges
of breeding birds only) are most diverse in the lower elevations
of the Sierra Nevada and Cascade ranges and in the mountains of
southern California. The maximum number of species in a quadrangle
is 126, or 61% of all species of land birds. Mammals have a similar
pattern except that the richest quadrangles tend to be at higher
elevations than birds. They also tend to have greater turnover in
composition than birds, with a maximum of only 49% (66 species)
of all mammals in the richest quadrangle. Amphibians are most diverse
in the cooler, wetter parts of California, primarily on the northern
and central Coast Ranges and the central Sierran foothills. Of the
45 native amphibians in the state, as many as 15 (33%) are predicted
with highly suitable habitat in the richest sites , while many quadrangles
in the deserts contain none. Reptiles, in contrast, thrive in the
hot, arid deserts of southeastern California. Because of their concentration
in just a portion of the state, the richest quadrangle for reptiles
contains 73% (52 species) of all reptiles. Clearly, it would be
impractical to base a conservation strategy solely on protecting
the richest sites for one of the se groups, since it would seldom
be a site that was rich for other taxa. If all terrestrial vertebrates
are combined, the number of birds dominates the totals that the
pattern is essentially the same as seen in Figure 3-5. As a result,
the richest sites for all vertebrates, in the northeastern part
of the state would capture few amphibians and reptiles.
3-5. Species richness of land birds (summer ranges of breeding birds
only) by 7.5 minute quadrangle. (Only species with at least category
4 or 5 suitability are counted).
3-6. Species richness of mammals by 7.5 minute quadrangle. (Only
species with at least category 4 or 5 suitability are counted).
3-7. Species richness of amphibians by 7.5 minute quadrangle. (Only
species with at least category 4 or 5 suitability are counted).
3-8. Species richness of reptiles by 7.5 minute quadrangle. (Only
species with at least category 4 or 5 suitability are counted).
can be a misleading indicator of biodiversity. In Figures 3-9 through
3-11, we illustrate one of the dilemmas of relying on species richness
as a conservation criterion. Each of these 3 maps depicts the similarity
in species composition of mammals of all quadrangles in relation
to a reference quadrangle. For this set of maps, we selected 3 reference
quadrangles that each contain 52 mammals (top 5% of all quadrangles
in mammal richness), Taylorsville in the northern end of the Sierra
Nevada range, Cascadel Point in the low- to mid-elevations of the
southern Sierra, and Lake Arrowhead in the San Bernardino Mountains
in southern California. All 3 quadrangles are characterized by a
mix of montane conifer and hardwood forest habitats with some shrub
types. Yet they show little similarity in composition. The two Sierran
sites only share 2/3 of their combined species. The Taylorsville
and Lake Arrowhead quadrangles share only 1/3. In fact, it is striking
how rapidly the similarity of species composition declines as one
moves away from the reference quadrangle. The average similarity
of all other quadrangles with these 3 ranged from only 36-42% Selecting
one site to represent biodiversity, simply because the site has
high species richness, would tend to miss many species. This result
emphasizes the importance of the complementarity of sites,
since each contains species not represented by the others.
3-9. Jaccard similarity index of mammals by 7.5 minute quadrangle
in reference to the Taylorsville quadrangle in the northern Sierra
Nevada region. (Only species with at least category 4 or 5 suitability
3-10. Jaccard similarity index of mammals by 7.5 minute quadrangle
in reference to the Cascadel Point quadrangle in the southern Sierra
Nevada region. (Only species with at least category 4 or 5 suitability
3-11. Jaccard similarity index of mammals by 7.5 minute quadrangle
in reference to the Lake Arrowhead quadrangle in the Southwestern
California region. (Only species with at least category 4 or 5 suitability
When this report
was written, was written, the accuracy assessment/validation of
the predicted species distribution modeling had not been completed.
Two approaches are being taken concurrently: 1) comparison of predicted
species lists using the CA-GAP data and methodology with lists of
observed species from state parks and other manages areas (in collaboration
with Dr. James Quinn at University of California, Davis), and 2)
comparison of predicted species with the Breeding Bird Survey (BBS)
data. Dr. Quinn's group at Davis has assembled and edited species
lists from a large number of managed areas. From the comparisons
with his data, we hope to be able to estimated the overall quality
of the CA-GAP vertebrate data and identify which groups of species
tend to be predicted with greater or lesser accuracy. Because the
managed areas are various sizes, we should be able to estimate the
spatial resolution at which the predictions are of satisfactory
accuracy. The expectation is that the predictions will be less accurate
for small sites than for large ones, where many of the small local
errors are overcome as one encounters more habitat types. For instance,
predicted lists for the entire state should be virtually perfect.
All species known to occur would be predicted to occur somewhere
in the state. But in very small managed areas of 100 ha or less,
generally the habitat suitability method tends to greatly overestimate
the number of species (errors of commission). At the same time,
the generalization of the CA-GAP land-cover map will often miss
individual patches of unique habitat and therefore fail to predict
species at the site (errors of omission). The difficulty in making
these comparisons is standardizing the predicted and observed lists
to include the same types of species. For instance, park species
lists often include every species ever seen in the managed area,
not just the breeding species modeled in gap analysis.
The BBS is a
long-term program of the U. S. Fish and Wildlife Service to collect
data on the trends in birds across the nation. Each spring, 25 mile
long sections of a sample of roads are surveyed, with stops at 0.5
mile intervals. Birds that are seen or heard are recorded, both
the species and the number of individuals, in a brief stay at each
stop. Comparisons of the predicted bird lists and the BBS will allow
us both to test the CA-GAP wildlife predictions with actual field
data over a large area as well as to evaluate the significance of
management status of the transects in predicting the trends in bird
abundance. If the assumptions of gap analysis are valid, we should
expect to see declines in the birds that are least well-represented
in managed areas.
Limitations and Discussion
The methods followed by CA-GAP made use of the best available information
to create meso-scale maps of predicted species distribution, but
did not involve the collection of new specimens or field observations.
Because locational records of collection and observation of species
are grossly incomplete, CA-GAP relied strongly on the use of expert
opinion in the development of range limits, habitat association,
and review of predicted distributions. The resulting maps then are
testable hypotheses that we encourage field biologists to assess
whenever conducting field studies.
These procedures work best for species with habitat preferences
that can be described in terms of land cover and other digitally
mapped features or characteristics. It provides less accurate predictions
of the presence of species that are highly variable in spatial and/or
temporal occurrence (Krohn 1996), and it will work for habitat specialists
only if their specific habitat requirements are available as mapped
features or are well associated with other mapped characteristics
such as land cover types.
An additional caution is that species with very restricted distributions
(occurring in one or a few locations of a size below the GAP MMU)
cannot be predicted to occur in seemingly appropriate habitat within
their distributional limits. Because of their rarity, these species
are often the subject of special attention from state and federal
resource agencies and are ranked G1-G2 by the Natural Heritage Central
Databases (see ranks and their description at
http://www.heritage.tnc.org). The specific locations where they
are known to occur are usually tracked by NHPs, CDCs, and the USFWS.
GAP makes use of the data from NHPs and CDCs to report the presence
of populations of such species within a geographic unit. For security
purposes, the exact locations of these populations should be distributed
only by the data owner.
It was noted earlier that urban habitats were rated in CWHR as
High or Medium suitability for some species, such as the Black-headed
grosbeak. CWHR does not distinguish between land use types within
urban areas as it does for agricultural types and so the ratings
are applied equally to all habitats from low density housing to
fully developed industrial areas. Neither the CA-GAP land-cover
nor the CWHR suitability databases have sufficient detail to discriminate
between these different levels of use. We examined the CWHR habitat
suitability ratings to determine the extent this would misrepresent
both the distribution of some species but also their management
status. Many of the species with this habitat rating are non-natives,
which had been eliminated from our analysis. Of the remaining native
species, perhaps 10% had this type of rating. Most of these were
common birds that one would expect to be found at least in suburban,
residential areas. In general, the species potentially affected
by this simplification are not considered vulnerable.
Successful assessment of the management status of vertebrate species
through gap analysis requires accurate mapping of their distributions.
As of the date of this report, the accuracy assessment in still
in progress, so quantitative measures of accuracy are not yet available.
There have been many concerns expressed about species richness
criteria in conservation assessment and planning. Our purpose is
not to identify conservation priorities as to portray relatively
fine-grain biogeographic information that has not been available
until compiled by CA-GAP. Species richness patterns are highly dependent
on the size of the sampling unit (Stoms 1994). Our portrayal of
species richness by 7.5 minute quadrangle is just one of many possible
representations at other scales. We also constrained the count of
species to be those with the highest categories of suitable habitat
(i.e., 4 or 5). Consequently, the richness maps ignore species with
small areas of highly suitable habitat or larger areas of low suitability
habitat that may actually contain such species.