Mussel Species Turnover Along the River Continuum · MUSSEL SPECIES TURNOVER ALONG THE RIVER...
Transcript of Mussel Species Turnover Along the River Continuum · MUSSEL SPECIES TURNOVER ALONG THE RIVER...
MUSSEL SPECIES TURNOVER ALONG THE RIVER CONTINUUM
Carla L. Atkinson University of Oklahoma &
Oklahoma Biological Survey
Organization of Fish and Wildlife Information Managers Meeting October 2011
Why mussels? • Dominated biomass in rivers
• Ecosystem services
BACKGROUND IMPLICATIONS RESULTS METHODS
Why mussels? • Dominated biomass in rivers
• Ecosystem services
• ~300 species in North America
BACKGROUND IMPLICATIONS RESULTS METHODS
BACKGROUND IMPLICATIONS RESULTS METHODS
Why mussels? • Dominated biomass in rivers
• Ecosystem services
• ~300 species in North America
• Highly imperiled
BACKGROUND IMPLICATIONS RESULTS METHODS
Freshwater mussels
Very threatened as a group
3 federally listed species in the Kiamichi and Little Rivers
River Continuum Concept in a
Nutshell
Functional feeding groups vary in a downstream direction There are also differences within functional groups!
BACKGROUND IMPLICATIONS RESULTS METHODS
Vannote et al. 1980
Study Area -3 rivers
-Kiamichi (18 species) -Little (16 species) -Mt. Fork (18 species)
-Mussels -28 species included in analysis -High densities (up to 100 mussels/m2)
- Landuse -Primarily forest (70-80%) -Human use (water extraction, agriculture)
BACKGROUND IMPLICATIONS RESULTS METHODS
Rivers of southeastern Oklahoma
Relatively pristine water
High biodiversity, but imperiled
Source: The Nature Conservancy
BACKGROUND IMPLICATIONS RESULTS METHODS
Methods
• Mussel data – 16 sites (collections in 1994, 2006, 2010)
– 28 species – how to describe the community?
– Bray-Curtis Ordination (Bray & Curtis 1954, Ecology) • 1st – a dissimilarity matrix is computed among sites
• 2nd – selection of 2 sites as poles (find the community separated by the greatest distance)
• Ordination is projected (along a gradient)
BACKGROUND IMPLICATIONS RESULTS METHODS
Predictable Species Shifts
-Bray-Curtis value is indicative of community composition
-Communities that are the same distance from the headwaters are more similar
Explain
• What are the factors that lead to shifts in community composition?
– Physiography?
– Land use?
GIS Data
• DEM – extracted watersheds for each site
• NLCD – all landuses (%forest, %urban, etc.)
• SSURGO - CaCO3, % frequently flooded
• DEM and NHD combined - stream gradient, 100 m buffers
• NAIP aerial photographs - verification
BACKGROUND IMPLICATIONS RESULTS METHODS
Multiple Spatial Scales
• Allan (2004) suggested the use of 3 spatial scales
– Watershed (upstream of the site)
– Buffer (100 m around all channels)
– Site (buffer scale 1 km upstream from site)
BACKGROUND IMPLICATIONS RESULTS METHODS
Example
• Created 100-m buffer using spatial analyst
• Measured 1 km upstream and selected the buffer for that length
3 Scales • For each scale I extracted:
– Landuse
– Information on soils
• Gradient was calculated different ways for each scale – Watershed – total change in elevation/stream
length
– Buffer – change in elevation 10 km upstream/10km
– Site – change in elevation 1km/1km
BACKGROUND IMPLICATIONS RESULTS METHODS
Watershed Scale
River Site
Mainstem
Distance
Downstream
(km)
Watershed
Land Area
(km2)
Gradient*
(m/km)
% of
Watershed
that is
Frequently
Flooded**
% Open
Water
%
Urban
%
Barren
%
Forest
%
Grassland/S
hrubs
%
Agriculture
%
Wetland
Kia
mic
hi
KM1 61.88 443.87 4.85 8.84 0.07 2.67 0.00 87.14 1.77 8.04 0.30
KM2 81.50 815.19 3.77 8.47 0.29 3.08 0.01 75.94 4.85 14.84 1.00
KM3 109.17 1040.14 2.91 8.29 0.28 3.14 0.01 73.51 6.26 15.44 1.35
KM4 117.39 1860.24 2.80 8.16 3.11 2.64 0.01 70.91 6.83 15.23 1.26
KM5 131.63 1946.57 2.60 8.33 3.00 2.70 0.01 71.10 6.78 15.15 1.27
KM6 152.01 2043.85 2.30 8.31 2.89 2.68 0.01 71.25 6.82 15.06 1.28
Litt
le
LM1 18.49 73.55 15.08 7.94 0.00 1.56 0.01 84.64 12.10 1.23 0.45
LM2 51.96 362.87 6.90 9.83 0.21 2.66 0.01 80.21 14.36 2.05 0.52
LM3 59.94 414.48 6.18 9.98 0.20 2.68 0.01 79.18 14.41 3.00 0.53
LM4 67.10 717.24 5.36 10.20 0.15 3.04 0.00 77.53 15.63 3.22 0.44
LM5 93.23 1055.20 4.30 11.38 0.18 3.25 0.00 77.86 15.96 2.34 0.41
Mt.
Fo
rk
MF1 43.19 437.99 2.63 5.10 0.31 4.06 0.03 76.14 4.01 15.31 0.14
MF2 53.01 565.94 2.34 5.59 0.27 3.75 0.02 78.77 4.28 12.76 0.15
MF3 62.19 673.38 2.23 5.87 0.26 3.85 0.02 78.09 4.94 12.67 0.17
MF4 71.98 831.49 2.04 5.60 0.24 3.83 0.02 78.09 6.36 11.30 0.17
MF5 76.66 1091.94 1.97 5.92 0.22 3.86 0.01 79.67 7.09 8.82 0.32
*Gradient considering the elevation and length of the entire mainstem channel
**The percentage of land area in which that the chance of flooding is more than 50 percent in any year but is less than 50 percent in
Further Analysis
• Used Akaike’s Information Criterion (AIC) approach to select model to predict community composition*
• Several multiple linear regressions generated
• Provides the best compromise between predictive power and model complexity
* See Burnham and Anderson 2001 for more information
BACKGROUND IMPLICATIONS RESULTS METHODS
AIC results
Watershed and buffer scale variables best at predicting mussel community composition
Scale Parameters in Model K F-value R2 AIC Δi wm
Watershed Gradient, %Open Water, %Urban 3 55.79 0.933 -84.224 0.000 0.169
Gradient, % Open Water, %Urban, %Grassland/Shrub 4 38.85 0.934 -82.416 1.809 0.068
Gradient, %Open Water, %Urban, %Frequently Flooded 3 38.76 0.934 -82.381 1.843 0.067
Buffer Gradient, %Open Water, %Grassland/Shrub 3 56.76 0.934 -84.483 0.000 0.146
Gradient, %Open Water, %Agriculture 3 51.19 0.928 -82.942 1.541 0.068
Gradient, %Open Water, %Frequently Flooded, %Agriculture 4 36.11 0.927 -82.924 1.559 0.067
Site %Urban, %Forest, %Agriculture 3 3.8 0.488 -51.647 0.000 0.042
Gradient, %Agriculture, %Wetland 3 3.75 0.484 -51.540 0.107 0.040
%Frequently Flooded, %Urban, %Forest, %Agriculture 4 3.12 0.531 -51.073 0.574 0.031
BACKGROUND IMPLICATIONS RESULTS METHODS
2 Most Predictive Scales – Watershed & Buffer
• Gradient best predictor in the model – geomorphic control (corroborates with Arbuckle and Downing 2002)
• %Open Water another good predictor (likely driven by Sardis Lake)
BACKGROUND IMPLICATIONS RESULTS METHODS
Sardis Lake Dams Jackfork Creek
Management Implications -- Water Management
-- Watershed Management
Other drivers
• Watershed Scale
– %Urban in top 3 models
• Buffer Scale
– Very variable
• Site Scale
– %Agriculture in top models
• Not included
– Fish
BACKGROUND IMPLICATIONS RESULTS METHODS
Why does this matter?
• Changes in hydrology and land use are influencing community composition
• Protecting site ≠ protecting the community
• Need to consider the buffer scale, and the whole watershed
BACKGROUND IMPLICATIONS RESULTS METHODS
New Biogeochemistry Data Corroborates
Biogeochemical signature is indicative of the watershed.
Extinction Debt?
• Mussels are long-lived and slow-growing
• Time lag between disturbance and species extinctions
BACKGROUND IMPLICATIONS RESULTS METHODS
2011 Drought – Little River 2011 Drought – Kiamichi River
Acknowledgements
• Vaughn lab / OK Biological Survey
• Robert S. Kerr Lab – Environmental Protection Agency
• Landowners
• Funding Sources:
– Sigma Xi
– OU College of Arts and Sciences
– OU Graduate Student Senate
– OU Zoology Dept.
“Mussels are not dismissible, even by those who have little interest in the natural world.
Their presence is a signature of healthy aquatic ecosystems, to which they contribute as living water filters.”
- E.O. Wilson
ANY QUESTIONS?