home

Status Report: The Missouri Aquatic GAP Pilot Project
SCOTT P. SOWA
Missouri Resource Assessment Partnership, Columbia, Missouri

Over the last year, significant progress has been made on all of our
major objectives. We are on schedule to complete our pilot project
by the contract deadline of September 2001. This project is by far
the most difficult but most rewarding project I have worked on. It
has forced all of us working on this project to think outside the
typical stream resource management “box,” to educate ourselves in
areas unfamiliar to most stream resource professionals, to put new
meaning into the words “agency cooperation,” and to overcome tech-
nical challenges that have opened the doors to a new age of aquatic
resource conservation. Below is a brief update on the progress we
have made on four major tasks of the project.

Defining and Mapping Assessment Units
Measures of species diversity must be expressed relative to biogeo-
graphic units of a determined spatial scale if they are to be mean-
ingful (Levin 1981). We are currently working with The Nature
Conservancy to generate a national map of biogeographic units that
circumscribe relatively distinct freshwater communities. Empiri-
cal evidence has shown that both major drainages and ecoregions
describe a significant amount of the variation in aquatic communi-
ties (Pflieger 1989; Angermeier and Winston 1999; Rabeni and
Doisy, in press). Consequently, these biogeographic units will con-
sist of major drainages (yet to be delineated) stratified by ecoregions
(likely Bailey’s Ecological Sections; Bailey 1995). We are assum-
ing that these biogeographic units serve as “ecological and evolu-
tionary units” with distinct genetic histories and thus serve as mean-
ingful units for partitioning environmental, life history, and genetic
variation in populations and species that have yet to be discovered
through intense phylogenetic, life history, and habitat requirement
research. This national map will hopefully serve as a standardized
base coverage of freshwater assessment and planning units for the
aquatic component of GAP, which can be refined by individual states
through expert review.

To build this coverage we will be using a national fish database
(and possibly other taxa) to calculate similarity coefficients for each
pair of 4, 6, and/or 8-digit Hydrologic Units (HUs) and also to ex-
amine levels of endemism within each of these units. While trying
to complete this same task for Missouri, we quickly realized that
we needed to place our similarity coefficients in a broader context.
For instance, using fish data alone, Jacaard similarity coefficients
among the 66 8-digit HUs that cover Missouri ranged from around
0.2 to 1.0. Obviously, HUs that contain all of the same species (i.e.,
similarity coefficient = 1.0) should be combined into a single unit.
However, what about HUs that have similarity coefficients of 0.9,
0.8, 0.7, or even 0.6? To answer this question we need to under-
stand the range of similarities and the average similarity among
HUs across the nation. This will allow us to generate statistical
criteria for identifying HUs with relatively similar aquatic commu-
nities and ultimately produce a national coverage of major drain-
ages that harbor relatively distinct communities. We will then in-
tersect this “major drainage” coverage with the national coverage
of Ecological Sections to generate a final coverage of freshwater
biogeographic units.

We hope to have the national coverage completed by January 2001.
However, before we can begin generating a national coverage of
freshwater biogeographic units, the fish database has to go through
an expert review, which could take six months or more.

Defining and Mapping Valley Segment Types
Valley segment types are distinct hydrogeomorphic units. Each
valley segment type will be relatively homogeneous with respect to
energy and nutrient dynamics, flow regime, physical habitat, and
position within the drainage network. These valley segment types
are the lotic counterparts of wetland types and are used in the same
manner that the vegetation classes are used to carry out assessments
in the terrestrial component of GAP. Our effort to classify stream
networks into distinct valley segment types largely follows The
Nature Conservancy’s Aquatic Community Classification Frame-
work (Lammert et al. 1997).

To date we have identified the list of classification variables that
will be used to delineate valley segment types and have obtained
the necessary digital data layers that will be used in the classifica-
tion process. In December 1999 we finally worked out the many
technical details of developing this data layer and have completed
the classification process for two 8-digit HUs, the Meramec and
Current River basins. It is truly exciting to see the end product of
this classification process. The final digital maps provide us with a
view of these watersheds unlike anyone else has ever seen (Figure
1); a continuous stream network broken down into distinct
patches much like viewing a digital image of vegetation alliances
for an entire ecoregion. We anticipate having a statewide 1:100,000
valley segment coverage completed in the next 12 months.

One major unresolved issue is that at the scale of 1:100,000, head-
water valley segment types are grossly underrepresented. Currently,
we are working with two 1:24,000 digital stream networks. While
working with these higher-resolution networks, it has become ap-
parent that there is no way to effectively deal with this problem
except for actually developing the valley segment layer at 1:24,000
instead of the coarser 1:100,000. Unfortunately, 1:24,000 digital
hydrography for Missouri will not be completed for another two
years, and for most other states the waiting period will be even
longer. The ramifications of this problem are that a) very unique
headwater valley segment types will not even be identified, and b)
very common headwater valley segment types will be identified as
rare and underrepresented in our current network of conservation
lands. At a scale of 1:100,000, the only way around these problems
is to remain conscious of their existence and somehow factor this
understanding into your conservation assessment.

Mapping Known Distributions
The community-sampling databases used in generating statewide
known distributions are nearing completion. The fish, mussel, and
crayfish databases have been populated, and each sampling loca-
tion has been linked to the Environmental Protection Agency’s River
Reach Files. We also identified the ecological section, subsection,
land type association, and the 8-, 11-, and 14-digit HUs in which
each sample was located. This leaves the snail database as the only
one yet to be completed.

The fish database contains 3,719 community fish samples that were
collected from 2,484 different stream reaches for a total of 63,166
species occurrence records. The mussel database contains 1,156
community mussel samples that were collected from 814 different
stream reaches for a total of 12,604 species occurrence records.
The crayfish database contains 949 community crayfish samples
that were collected from 793 different stream reaches for a total of
1,855 species occurrence records.

We are using the USGS/Natural Resources Conservation Service
(NRCS) Hydrologic Units to generate our statewide distribution
maps. To minimize errors of commission and omission we are spe-
cifically using 11-digit HUs for widely distributed species and the
smaller 14-digit HUs for narrowly distributed species. We have
finished developing the preliminary statewide distribution maps for
each fish, mussel, and crayfish species (Figure 2 .) These maps
are then sent out for professional review. Professional review of
the fish distribution maps has been completed, and we are currently
incorporating the revisions into our sampling database and our fi-
nal distribution maps. The mussel and crayfish maps are still out
for professional review.

It appears that adequate statewide sampling data exist to generate
detailed (i.e., using 11- or 14-digit HUs) statewide distribution maps
for both fish and crayfish. Unfortunately, sampling data for mus-
sels and snails are much more clumped, with numerous samples in
certain 11-digit HUs and only a handful or no samples in others.
Because of this clumped sampling distribution, it is likely that the
final known distribution maps for most mussel and snail species
will have to be represented at a coarser resolution (perhaps 8-digit
HU).

Developing Habitat-Affinity Models
Our habitat-affinity models are used in conjunction with the valley
segment coverage and the statewide known distributions to predict
the potential species composition of each individual valley segment
in the state. This is analogous to predicting all of the species likely
to inhabit each grid cell in a statewide land cover data layer. A
major difference is that we have no way of determining the envi-
ronmental quality of every single valley segment. For instance, a
valley segment could be of the right size, temperature, gradient,
and flow for a particular species yet have serious water quality prob-
lems that make the environment unsuitable for that species. Conse-
quently, our predictions reflect potential species composition un-
der pristine conditions as with GAP’s terrestrial vertebrate distribu-
tions.

To develop our models we are conducting extensive literature re-
views to find habitat-affinity information for each fish, mussel, cray-
fish, and snail species found in Missouri (368 total species). We
have completed the initial literature review for each species in all
four taxonomic groups. More intensive reviews have been com-
pleted for both fish and mussels. We have also completed the first
draft of the habitat-affinity database for fish.

In conducting the literature reviews, two major problems have been
identified. First, there is simply a lack of habitat-affinity informa-
tion for several species. This problem is most prevalent for mus-
sels, crayfish, and snails but also exists for several nongame fish
species. There is an obvious need for increased emphasis on, and
funding for, basic life-history research of freshwater biota. The
second problem is that much of the habitat-affinity information that
does exist is at the “microhabitat” scale. That is, most research on
the habitat requirements of aquatic species has identified the range
of depths, velocities, and substrates utilized by a particular species.
This would be analogous to research that documents if a bird spe-
cies is primarily a ground-dwelling species or utilizes the lower,
middle, or upper forest canopy without documenting the particular
forest types it inhabits. Such microhabitat data are useful for pre-
dicting the distribution of an organism within a particular stream
segment but not the actual stream segments it is likely to inhabit.
What we need is substantially more habitat-affinity research at
broader spatial scales that identifies associations between the pres-
ence of a species and factors such as stream size, temperature, el-
evation, permanence of flow, local and basinwide geology, soils, and gradient.