relating riparian vegetation to different landforms and flow regimes

advertisement
1
Electronic Supplementary Materials
2
A. Modeling stream flow and flood extents
3
Streamflow estimates for the Little Carp River were developed using the U.S. Army Corps of
4
Engineers Hydrologic Engineering Center (HEC) Hydrologic Modeling System (HMS)(US
5
Army Corps of Engineers 2000). The HMS simulates rainfall-runoff processes in a watershed or
6
region and models surface hydrology by arranging hydrologic elements in a dendritic network,
7
providing many different methods for defining the precipitation inputs of a watershed,
8
determining losses, and routing water through the stream network.
9
HEC-GeoHMS (Hydrologic Engineering Center 2000) was used to extract hydrologic,
10
topographic and topologic information from digital spatial data of a Little Carp River watershed.
11
The HEC-GeoHMS system for hydrologic modeling performs the following basic operations,
12
including, 1) raster-based terrain analysis and network definition, (2) vectorization of the
13
hydrologic elements, (3) parameterization of the hydrologic elements, and (4) topologic analysis.
14
During the first step, topographic analyses were used to define the hydrologic system based on a
15
digital elevation model (DEM) derived from a 1:24,000 USGS digital topographic map and GPS
16
coordinates of numerous stream cross sections. Using several GIS functions and the DEM, a
17
single downstream cell (in the direction of the steepest descent) was defined for each terrain cell,
18
so that a unique path from each cell to the basin outlet was determined. This process produced a
19
dendritic cell-network of 6 sub-basins within the Little Carp watershed that represented the paths
20
of the watershed flow system. Once these sub-basins were generated, the average SCS curve
21
number (an index the represents the integration of a hydrologic soil group, land cover, and
22
condition of the land cover) of each sub-basin was calculated as the average of the curve number
23
values within the sub-basin polygon. This curve number grid was developed using vegetation
1
1
cover data obtained from the Michigan Resource Information System (MIRIS), the percentage of
2
each hydrologic soil group (A, B, C and D) obtained from the U.S. Department of Agriculture’s
3
Natural Resources Conservation Service (USDA NRCS) Sate Soil Geographic (STATSGO)
4
database, and a look-up table that relates land use and soil group with curve numbers. Finally,
5
the sub-basin lag-time was calculated with the SCS formula where tp (minutes) is the sub-basin
6
lag-time (measured from the centroid of the hyetograph to the peak time of the hydrograph), Lw
7
(feet) is the length of the longest flow-path, S (%) is the slope of the longest flow-path, and CN
8
is the average curve number in the sub-basin. For a more detailed description of our hydrologic
9
model development see Goebel (2001).
10
The performance of the hydrologic model was iteratively evaluated by examining the mean
11
deviation, mean square deviation, and correlation between estimated and observed discharge
12
measurements collected over a 3 year period at the downstream end of each study reach
13
(Reynolds 1984). Specifically, we developed rating curves for each stream reach by collecting
14
streamflow data bi-monthly over a 3 year period when we could access the study sites (mid-April
15
to early December) and during various storm events (at least 3 storms form each study reach).
16
Rating curves were developed by measuring streamflow at the bottom of each reach using
17
standard techniques (Gore 1996). Additionally, we collected stream level data continuously over
18
the same 3 year period using pressure transducers located at the bottom of each study reach.
19
From these data (rating curves and stream level), we developed continuous flow estimates for
20
each reach that were used to help parameterize and verify the HEC-GeoHMS hydrologic model.
21
To examine the performance of the hydrologic model, we calculated the mean deviation
22
23
(MD), a measure of bias where large values indicate a poor fit, using the following equation:
MD = Σ (Ej – Oj)/P.
(1)
2
1
where Ej is the estimated discharge and Oj is the observed discharge for day j, and P is the total
2
number of observations. Positive MD values result if values of discharge are overestimated,
3
while negative values suggest the model underestimates discharge. We also calculated the mean
4
square deviation (MSD) which measures the precision of the surface hydrology model to predict
5
the observed or real discharge. MSD was calculated using the following equation:
MSD = Σ (Ej – Oj)2/P.
6
(2)
7
Estimates of discharge with small MSDs are considered more precise than those with large
8
MSDs (Palmer 1990). Finally, the Pearson correlation coefficient between the estimated and
9
observed values is reported to determine the adequacy of the model to estimate discharge at each
10
11
stream cross section.
Although the hydrologic model has a positive bias for each of the cross sections (MDs are all
12
positive), the model tends to predict discharge well on average, as indicated by the low MD
13
values (a measure of poorness of fit), especially for the clay lake plain reach (Table A1).
14
Furthermore, the modeled discharge estimates for each of the stream cross sections appears to be
15
relatively precise based upon the MSD values, and the relationships between the predicted and
16
observed discharge following storm events are strong (Table A1). As we are most interested in
17
the hydrologic models ability to predict streamflow associated with flood events, the strong
18
relationship between the predicted and observed discharge following storm events suggests the
19
hydrologic model is acceptable for our analyses (Table A1).
20
Once the HMS framework was in place and the model accuracy determined, we used the
21
HMS model to develop a long-term hydrologic record and flood frequency curves for each study
22
reach. To accomplish this, we used daily precipitation, snow depth, and temperature data from
23
1969 to 1999 obtained from the closest weather station maintained by the National
3
1
Oceanographic and Atmospheric Association (NOAA) located at Bergland Dam, Michigan
2
(46°35'N / 89°33'W, approximately 30 km to the south of the Little Carp River watershed).
3
Flood-frequency curves were generated for each stream cross section using the highest simulated
4
daily discharge values estimated by the HMS for each water year (November 1 – October 31)
5
over a 31-year simulation. Estimates of discharge for various flood intervals (e.g., 2, 10, 25, and
6
50-year flood events) were then calculated for each stream cross section using log-normal
7
frequency analyses (McCuen 1998). Modeled streamflow estimates for different flood
8
recurrence intervals of each study reach are listed in Table A2.
9
Next, we used the U.S Army Corps of Engineers Hydrologic Engineering Center River
10
Analysis System (HEC-RAS) to predict flood surface elevations for each stream cross section.
11
HEC-RAS is an integrated package of hydraulic analysis programs that performs steady flow
12
water surface profile calculations (Hydrologic Engineering Center 1998). The HEC-RAS utilizes
13
geometric data, which consists of connectivity information for the stream system and valley
14
cross-section data. Additionally, the model incorporates an estimate of channel roughness
15
(Manning’s n) for each stream cross section and adjacent banks, allowing us to incorporate the
16
influence of adjacent riparian vegetation. Once the boundary conditions (e.g., the starting water
17
surface elevations or bankfull conditions which were visually estimated in the field) of the model
18
are set and the geometric data entered into the HEC-RAS, the model requires only the discharge
19
(m3·sec-1) of the desired flood event to be entered in order to calculate flood surface elevations
20
(Hydrologic Engineering Center 1998).
21
After entering the stream and valley cross section data for each transect into the HEC-RAS,
22
we estimated values of Manning’s n using standard methods of field observation (Gore 1996).
23
Water surface elevations at each cross section in each of the three study reaches were then
4
1
estimated using predicted discharge values for frequent flood events (recurrence intervals of 2
2
and 10 years) and infrequent flood events (recurrence interval of 25 and 50 years) as modeled
3
using the HMS (Table A2).
4
5
References
6
Goebel PC (2001) Hydrogeomorphic controls on riparian areas of the northern Lake States.
7
8
Ph.D. Dissertation. Michigan Technological University
Gore JA (1996) Discharge measurements and streamflow analysis. In: Hauer FR, Lamberti GA
9
(ed) Methods in Stream Ecology. Academic Press, San Diego, CA, pp 53-75
10
Hydrological Engineering Center (1998) River analysis system HEC-RAS 2.0. User’s guide.
11
U.S. Army Corps of Engineers, Hydrologic Engineering Center, Davis, CA.
12
Hydrologic Engineering Center (2000). Geospatial hydrologic modeling system HEC-GeoHMS.
13
User’s guide. Version 1. U.S. Army Corps of Engineers, Hydrologic Engineering Center,
14
Davis, CA.
15
16
McCuen RH (1998) Hydrologic analysis and design, 2nd Edition. Prentice Hall, Upper Saddle
River, NJ
17
Reynolds, Jr.MR (1984) Estimating the error in model predictions. Forest Science 30:454-468
18
U.S. Army Corps of Engineers (2000) Hydrologic modeling system HEC-HMS. User’s Guide.
19
Hydrologic Engineering Center, Davis, CA
20
21
5
Table A1. Performance of hydrologic model to predict discharge (m3·sec-1).
Reach
Percent
r2
r2
MD
MSD
overestimates
Jun - Aug
Storms
Upper
0.12
0.06
77.53
0.26
0.88
Middle
0.24
0.37
77.18
0.14
0.71
Lower
0.07
0.25
58.24
0.19
0.68
1
2
6
1
Table A2. Modeled discharge (m3•sec-1) and flood surface elevations (m) for three study
reaches of the Little Carp River, Upper Michigan.
2
Stream Reach Type
Flood recurrence
Low-gradient
High-gradient
Low-gradient
interval
poorly entrenched
moderately entrenched
deeply entrenched
Discharge (m3•sec-1)
4.7
9.4
9.4
Elevation (m)
0.53
0.61
0.63
Discharge (m3•sec-1)
9.9
15.3
19.0
Elevation (m)
0.76
0.80
0.86
Discharge (m3•sec-1)
14.1
22.0
29.2
Elevation (m)
0.90
0.98
1.04
Discharge (m3•sec-1)
17.5
28.4
38.1
Elevation (m)
1.00
1.13
1.17
1.4 yearsa
10 years
25 years
50 years
a
Roughly equivalent to bankfull discharge.
7
1
2
B.B. List of common species (occurring on > 5% of sample plots) encountered across riparian
ecotones of the Little Carp River watershed.
Functional
Functional
Guild
Species
Guild
Species
Graminoids
Brachyelytrum erectrum
Pteridophytes
Athyrium fillix-femina
Forbs
Carex arctata
Dryopteris intermedia
Carex stricta
Gymnocarpium dryopteris
Cinna latifolia
Lycopodium clavatum
Leersia oryzoides
Onoclea sensibilis
Oryzopis asperifolia
Thelyptris phegopteris
Anemonella thalictroides
Woody
Abies balsamea
Aralia nudicaulis
Acer rubrum
Clintonia borealis
Acer saccharum
Eupatorium maculatum
Acer spicatum
Galium triflorum
Alnus incana
Hydrocotlye americana
Betula alleghaniensis
Maianthemum canadense
Fraxinus nigra
Polygantaum biflora
Lonicera canadensis
Scutellaria laterifolia
Ostrya virginiana
Smilicina racemosa
Prunus serotina
Solidago gaegantea
Rubus parviflorus
Streptopus roseus
Rubus pubescens
Trientalis borealis
Thuja occidentalis
Urtica diocia
Tsuga canadensis
Urtica procera
Viola cuculata
Viola pubescens
8
1
C. Ordination diagrams
2
500
Low Gradient
Poorly Entrenched
DCA Axis 2
400
TsCa
ArNu
LyCl
StRo
BeAl
DrIn
TrBo
AbBa
MaCa
300
200
LoCa
SmRa
ClBo
RuPa
AtFi
PoBi BrEr
AcRu UrDi
ViPu
CaAr
OnSe
CiLa
OrAs
GaTr
AnTh
AcSa
100
SoGa
RuPu
HyAm
LeOr
CaSt
ViCu
0
EuMa
ScLa
UrPr
-100
-200
-100
0
100
200
300
upland
400
500
600
stream
DCA Axis 1
Fig. C1. DCA ordination of ground-flora species by valley type. Species codes are first two
letters of genus and first two letters of specific epitaph (e.g., TsCa – Tsuga canadensis).
9
1
2
upland
250
DrIn
200
High Gradient
Moderately Entrenched
DCA Axis 2
150
100
AcSa
TsCa
50
0
MaCa
ClBo
ThPh
BeAl
stream
-50
AcSp
-100
-150
0
50
100
150
200
upland
250
300
stream
DCA Axis 1
Fig. C1. Continued.
10
1
2
200
stream
3
150
High Gradient
Deeply Entrenched
RuPa
CaAr
DCA Axis 2
100
ThPh
GyDr
AtFi
50
ArNu
RuPu
DrIn
0
TrBo
LyCl
AcSp
LoCa
-50
MaCa
AcRu
upland
CaSt
ClBo TsCa
BeAl
-100
AcSa
-150
-50
0
50
100
150
200
upland
250
300
stream
DCA Axis 1
Fig. C1. Continued.
11
1
400
2
Low Gradient
Poorly Entrenched
FrNi
DCA Axis 2
300
TsCa
BeAl
200
OsVi
AlRu
100
PrSe
AcSa
0
AbBa
-100
-50
0
50
100
150
200
upland
250
300
350
stream
DCA Axis 1
Fig. C2. DCA ordination of overstory species by valley type. Species codes are first two
letters of genus and first two letters of specific epitaph (e.g., TsCa – Tsuga canadensis).
12
1
upland
2
350
High Gradient
Moderately Entrenched
300
BeAl
DCA Axis 2
250
200
AcSa
150
OsVi
100
AcRu
stream
50
0
TsCa
-50
-100
-50
0
50
100
150
upland
200
250
300
stream
DCA Axis 1
Fig. C2. Continued.
13
1
2
400
Low Gradient
Poorly Entrenched
FrNi
DCA Axis 2
300
TsCa
BeAl
200
OsVi
AlRu
100
PrSe
AcSa
0
AbBa
-100
-50
0
50
100
150
200
upland
250
300
350
stream
DCA Axis 1
Fig. C2. Continued.
14
Download