**2. Methods**

#### **2.1 Geomorphology**

Riverbed slope, channel geometry, and substrates are well known for engineering purposes. System-wide topographic mapping and channel surveys undertaken for each significant channel improvement plan were completed in 1890 and 1930. Surveys are much more

pollution, support navigation, and enhance habitat. The outcomes of these changes differ

Alterations to hydrology, geomorphic structure, and direct impacts from historical land use change have substantially altered the form and function of ecological communities and processes in the UMRS. The flow of energy is a critical function in ecosystems and alterations to energy pathways can cascade through ecosystems in many ways (Welcomme, 1979; Vannote et al., 1980, Ward and Stanford, 1983; Junk et al., 1989; Ward et al., 1989). Early formal models for stream ecosystem energetics emphasized linear pathways transporting and utilizing metabolic energy differently along a river continuum (Vannote et al., 1983). The early stream ecosystem conceptual models were then tailored to account for nutrient cycling (Newbold et al., 1982), anthropogenic disturbances (Ward et al., 1999), different types of rivers (Junk et al., 1989; Wiley and Osborne, 1990), internal processes (Thorp et al., 1994), watershed influence (Benda et al., 2004 ), and geomorphic structure (Thoms, 2003). We developed system scale data to focus on the relationships expressed in the hydrogeomorphic methodology (Brinson, 1993; Klimas et al., 2009) and the River Ecosystem Synthesis (Thorp et al., 2006, 2008). Land cover, aquatic area, hydrology, and geomorphology data were derived for the entire UMRS for historic, contemporary, and simulated conditions. They can be compared among functional units such as Functional Process Zones or Hydrogeomorphic Patches defined by Thorp et al. (2008) or reference conditions (Nestler et al., 2010; Theiling and Nestler, 2010; SER, 2004) to create simulations

depending on location in the river system (Theiling and Nestler, 2010).

of potential vegetation communities under alternative management scenarios.

**2. Methods** 

**2.1 Geomorphology** 

Ecosystem restoration initiatives require estimates of the natural resource benefits that may be achieved by alternative project plans or project features to ensure accountability and success in Federal projects (USACE, 2000). Recent guidance also calls for the use of adaptive management in Federal water resource planning (Section 2039 of U.S. Water Resources Development Act 2007; Council on Environmental Quality, 2009). The models described here are important elements of adaptive management because they can estimate anticipated outcomes for comparison during monitoring and evaluation stages of the adaptive management cycle (Christianson et al., 1996; Walters, 1997; Williams, 2009). Many restoration plans include plant community or habitat models that estimate community response to physical forces (U.S. Water Resources Council, 1983; USACE, 2000; Council on Environmental Quality, 2009). Predicted plant and habitat response can then be used to support species or community habitat suitability models (USFWS, 1980). Dynamic physical forces are well known ecological drivers in large rivers (Doyle et al. 2005). Methods and data presented here can help estimate physical-ecological cascades resulting from hydrologic and geomorphic alteration of large rivers. We have made great progress developing data needed for potential vegetation models for the entire system. We also discuss the need for a rigorous landscape analysis that includes forest composition in the pre-settlement land cover data.

Riverbed slope, channel geometry, and substrates are well known for engineering purposes. System-wide topographic mapping and channel surveys undertaken for each significant channel improvement plan were completed in 1890 and 1930. Surveys are much more frequent in the modern era. The Valley's floodplain has been mapped to document the relative age of geomorphic surfaces and associated deposits to help manage cultural resources (Bettis et al., 1996). The studies developed Landform Sediment Assemblages (LSA) which are mappable landforms and their underlying deposits that occur with predictable characteristics (Figure 3; Hajic, 2000). U.S. Department of Agriculture (USDA) soil maps are widely available, but generally lack detail in frequently flooded parts of the floodplain.

Geomorphic mapping in the Valley generally followed the protocol defined by Bettis et al. (1996) with slight variations. U.S. Geological Survey topographic quadrangle maps, aerial photos, soils maps, boring records, and literature were used to construct geomorphic maps. Geomorphic classifications were done at several different scales which allows for more detailed site-specific analysis than reported here. Mapping under modern aquatic areas was not possible and most of the low elevation features (active floodplain and some paleofloodplain) were inundated in the lower ends of navigation pools 2 through 13 between Minneapolis, Minnesota and Clinton, Iowa. We unioned four separate LSA data sets (Bettis et al., 1996; Madigan and Schirmer, 1998; and Hajic, 2000) and reclassified them using a common classification scheme in GIS. The data were clipped to the bluff to bluff floodplain extent (Laustrup and Lowenburg, 1994). LSAs were summarized using a river mile segmentation floodplain overlay. River mile segments are unequal because the width of the floodplain varies and there are curves in the river that create wedge-shaped polygons. These results are a first approximation and open to further interpretation. Higher resolution mapping and analysis will be required for site-specific studies (Heitmeyer, 2010), but this generalized classification matches flood inundation mapping, historic land cover mapping, and regional habitat assessments (Theiling et al., 2000) quite well.

Our LSA geomorphic classification has nine classes described below. Characteristics were derived from Bettis et al. (1996), Madigan and Schirmer (1998), and Hajic (2000) and mapped as follows:


Hydro-Geomorphic Classification

(Heitmeyer, 2008).

**2.2 Hydrology** 

**2.2.1 Topographic data** 

and 500-yr expected recurrence interval flood).

and Potential Vegetation Mapping for Upper Mississippi River Bottomland Restoration 169

• **Paleo-Floodplain – Well Drained** is infrequently flooded mid elevation floodplain of lateral accretion origin that includes inactive scrolls, bars, meander belts, and splays. Soils are variable with fine silt, loams, and clays overlying sand and pebbly sand. Paleo-Floodplain is mapped mostly in the Rock Island and St. Louis Districts. In the Rock Island District it is an association with early and mid Holocene surfaces that define the wet areas and paleo-channels that derive the dry areas. In the St. Louis District this LSA comprises large meander scrolls that occupy a major proportion of the more elevated floodplain area. There is almost no paleo-floodplain in the St. Paul District because Holocene channel incision has isolated older surfaces as infrequently flooded terraces.

• **Natural Levees** are slightly elevated, well-drained areas that parallel relatively stable channel reaches. Levees may also occur at crevasse splays that extend from channels cut into the natural levee and spreading into adjacent low-lying wet paleo-floodplain. Deposits of this LSA are stratified loam, sand, silt, clay, and sand. Levees are discontinuous linear areas that appear most abundant on the Illinois River because the Illinois River mapping was done at a smaller scale (higher resolution; Hajic, 1990). Several large levee areas are mapped in the Rock Island District and smaller levee areas are common along the channel in the St. Paul District where they are not submerged. • **Alluvial/Colluvial Aprons** are elevated, bluff-base areas underlain by a variety of sediments derived from adjacent slopes and small tributary valleys. This LSA typically is quite messic and is rarely inundated. The most notable abundance of this LSA occurs

• **Sandy Terraces** occur throughout the river and were formed during the last glacial period (Knox and Schumm in West Consultants, Inc., 2000). They are most abundant in the Illinois, Minnesota, Chippewa, Maquoketa, and Iowa River reaches. Downstream of

High resolution topographic data and updated river stage-discharge relationships were developed following the "Great flood of 1993" when there was a comprehensive review of floodplain management (Interagency Floodplain Management Review Committee, 1994). Photogrammetric methods were used to create a high accuracy digital elevation model for the entire Upper Mississippi floodplain for use in hydrologic modeling to re-define the river stage frequency rating curves. We created GIS overlays of the water surface elevation profiles corresponding to the rating curves, superimposed on the high resolution topography to map potential flood inundation patterns (Figure 4) for 8 annual exceedance probability floods: 50, 20, 10, 4, 2, 1, 0.5 and 0.2 percent (i.e., 2-, 5-, 10-, 25-, 50-, 100-, 200-,

The U.S. Geological Survey National Elevation Database available through the National Map Seamless Server provided online access to digital elevation data in an easily accessible and well documented format. Upper Mississippi, Illinois, and Missouri Rivers floodplain

in Illinois near Quincy where there are other high floodplain features.

the Iowa River Reach this LSA merges with the paleofloodplain LSA.

Older surfaces in the St. Paul District occur as terraces.

formed backwater lakes and sloughs prior to significant floodplain drainage

Fig. 3. Landform Sediment Assemblage maps characterize surficial and underlying characteristics that help define local edaphic factors. This map depicts parts of Pools 16 to 20 between Muscatine, Iowa and the junction of the Des Moines River.


formed backwater lakes and sloughs prior to significant floodplain drainage (Heitmeyer, 2008).


## **2.2 Hydrology**

168 Studies on Environmental and Applied Geomorphology

Fig. 3. Landform Sediment Assemblage maps characterize surficial and underlying

between Muscatine, Iowa and the junction of the Des Moines River.

characteristics that help define local edaphic factors. This map depicts parts of Pools 16 to 20

• **Active Floodplain – Well Drained** is frequently flooded low elevation floodplain of lateral accretion origin. It is underlain by less than 1.5 meters of fine-grained alluvium that buries sand and pebbly sand. Despite high frequency inundation, it does not retain water. Dry active floodplain may also be associated with alluvial fans and deltas. Dry active floodplain is common on the Illinois River and occurs in patches in the St. Paul

District. This class was not mapped in the Rock Island and St. Louis Districts. • **Paleo-Floodplain – Poorly Drained** is infrequently flooded mid elevation floodplain of vertical accretion origin. These floodplain areas contain former channel and lake features that have transitioned to terrestrial area. Deposits and soils are variable with fine silt, loams, and clays overlying pebbly sand. They function as overflow channels on the rising and receding flood or as ponded groundwater at high river stage. They High resolution topographic data and updated river stage-discharge relationships were developed following the "Great flood of 1993" when there was a comprehensive review of floodplain management (Interagency Floodplain Management Review Committee, 1994). Photogrammetric methods were used to create a high accuracy digital elevation model for the entire Upper Mississippi floodplain for use in hydrologic modeling to re-define the river stage frequency rating curves. We created GIS overlays of the water surface elevation profiles corresponding to the rating curves, superimposed on the high resolution topography to map potential flood inundation patterns (Figure 4) for 8 annual exceedance probability floods: 50, 20, 10, 4, 2, 1, 0.5 and 0.2 percent (i.e., 2-, 5-, 10-, 25-, 50-, 100-, 200-, and 500-yr expected recurrence interval flood).

#### **2.2.1 Topographic data**

The U.S. Geological Survey National Elevation Database available through the National Map Seamless Server provided online access to digital elevation data in an easily accessible and well documented format. Upper Mississippi, Illinois, and Missouri Rivers floodplain

Hydro-Geomorphic Classification

**2.2.3 Floodplain inundation** 

area.

**2.3 Land cover** 

**2.3.1 Presettlement land cover** 

and Potential Vegetation Mapping for Upper Mississippi River Bottomland Restoration 171

Triangulated Irregular Network (TIN) files were created from the cross section feature lines for each separate flood stage frequency (Figure 4). Each flood stage TIN: 50%, 20%, 10%, 4%, 2%, 1%, 0.5%, and 0.2% annual exceedance probability (i.e., 2-yr, 5-yr, 10-yr, 25-yr, 50-yr, 100-yr, 200-yr, 500-yr expected recurrence interval flood) was overlayed in a cut-fill analysis on the high resolution floodplain topography for each navigation pool or reach. The area represented as inundated by the cut-fill procedure for each flood stage was separated out as a conditional GRID analysis that selected areas with volume > 0 and output a single GRID with a count of the 20X20 m cells below the elevation of the water surface elevation (Figure 4). This value was exported to a spreadsheet where grid counts were converted to area estimates (acres) for the navigation pool scale at which they were created. The resulting GRID was converted to a shapefile to merge with other layers to create system-scale layers

Floodplain inundation classes (i.e., 50, 20, 10, 5, and 1 percent annual exceedance probability floods) were summarized by river mile and compared by geomorphic reach. Leveed areas were then extracted from inundation layers to assess changes in flood distribution attributable to levees. These data were also summarized by river mile and geomorphic reach. The inundation classes in leveed areas were subtracted from the maximum simulated inundation surface in each geomorphic reach (i.e., 1 percent or 0.02 percent annual exceedance probability flood) and data were normalized as percent of maximum inundation

Land cover databases are the foundation of our vision of UMRS landscapes and habitats over multiple reference conditions. Early explorers described interesting new landscapes, vast abundances of strange new animals, and drew crude maps as they moved through North America (Carlander, 1954). As settlers followed explorers, the Public Land Survey (PLS) mapped and characterized the mostly unsettled Louisiana Territories to sell land to the westward-expanding population of the United States (Sickley and Mladenoff, 2007). The PLS methods first divided the region into 36 square mile townships and then subdivided each one into 36 one mile square sections. Along the township and section lines, the surveyors set posts every half mile at locations called ½ section corners (where section lines intersected) and quarter section corners (midway between the section corners). Between two and four bearing trees were marked near each post and recorded in their notebooks by species, diameter, and compass bearing and distance from the post. The surveyors recorded other features that they encountered along the survey lines in the notebooks as well, including water features, individual trees located between the survey posts, boundaries between the ecosystems through which they were traveling, boundaries of natural and anthropogenic disturbances, and cultural features such as houses, cultivated fields, roads, and towns. Initial pilot studies reconstructing PLS surveys in the UMRS (Nelson et al., 1996) proved to be very valuable, so The Nature Conservancy's Great River Partnership contracted the University of Wisconsin Forest Ecology Lab to complete a comprehensive interpretation in a GIS for the entire UMRS (Sickley and Mladenoff, 2007). PLS data extend beyond the bluff into upland habitats, but the data were clipped to the bluff to bluff extent

of the potential water distribution at each flood stage (Figure 4).

elevation data were updated in 1998 using high resolution stereographic techniques (Interagency Floodplain Management Review Committee, 1994; Scientific Assessment and Strategy Team (SAST), David Greenlee, USGS EROS Data Center, Sioux Falls, South Dakota, personal communication). The Mississippi River floodplain ("bluff-to-bluff") digital terrain model data was designed and compiled so that spot elevations on well-defined features would be within 0.67 feet (vertical) of the true position (as determined by a higher order method of measurement) 67% of the time. It is approximately 1/6th of a contour interval (4 foot contours; U.S. Army Corps of Engineers, 2003, 2004b). High river stages when photography was acquired limited their utility to visualize and model low river stages in mid reaches of the Mississippi River and prevented their use for this project on the Lower Illinois River. The NED2003 floodplain elevation data were used for the Illinois River floodplain inundation mapping. Issues regarding vertical datum conversions were evaluated and determined to be insignificant at the scale and intended application for this study (Theiling, 2010).

Data can be accessed at several levels of resolution, we used the default 1 arc second download format to conserve data processing requirements over large geographic regions and because subsequent hydrologic modeling analyses were completed at similar resolution. Rectangular tiles covering about 100 miles each were downloaded and data extracted by a mask of the floodplain as represented by the prior defined floodplain extent for each pool (Laustrup and Lowenburg, 1994). We combined the pool scale DEMs into a DEM for the entire floodplain using default mosaic procedures in ArcGIS. Metric elevations were converted (i.e., times 3.281 in Raster math) to English units to match river stage in feet and discharge in cubic feet per second (cfs) which is the vernacular of the Flow Frequency Study.

#### **2.2.2 Flow frequency study**

Hydrologic analyses were accomplished with 100 years of record from 1898 to 1998 using the log-Pearson Type III distribution for unregulated flows at gages. Mainstem flows between gages were determined by interpolation of the mean and the standard deviation for the annual flow distribution based on drainage area in conjunction with a regional skew. Flood control reservoir project impacts were defined by developing regulated versus nonregulated relationships for discharges, extreme events were determined by factoring up major historic events, and the UNET unsteady flow program was used to address hydraulic impacts. The result of the hydrologic aspects of the study was a discharge and related frequency of occurrence for stations or given cross sections located along the Mississippi and Illinois Rivers (Figure 4; USACE, 2004b).

A hydraulic analysis was required to establish the water surface elevation associated with each frequency of discharge at each location or cross section along the river reach. The main procedures were to use the UNET unsteady flow numerical modeling tool with recent channel hydrographic surveys (routinely obtained for navigation channel maintenance), and floodplain digital terrain data collected in 1995 and 1998. Levee overtopping was established at the top of existing levee grade based on an upstream and a downstream point. Using these station rating curves and the station frequency flows developed during the hydrology phase, frequency elevation points were obtained for each cross section location. Connecting the corresponding points resulted in flood frequency elevation profiles (USACE, 2003).

### **2.2.3 Floodplain inundation**

170 Studies on Environmental and Applied Geomorphology

elevation data were updated in 1998 using high resolution stereographic techniques (Interagency Floodplain Management Review Committee, 1994; Scientific Assessment and Strategy Team (SAST), David Greenlee, USGS EROS Data Center, Sioux Falls, South Dakota, personal communication). The Mississippi River floodplain ("bluff-to-bluff") digital terrain model data was designed and compiled so that spot elevations on well-defined features would be within 0.67 feet (vertical) of the true position (as determined by a higher order method of measurement) 67% of the time. It is approximately 1/6th of a contour interval (4 foot contours; U.S. Army Corps of Engineers, 2003, 2004b). High river stages when photography was acquired limited their utility to visualize and model low river stages in mid reaches of the Mississippi River and prevented their use for this project on the Lower Illinois River. The NED2003 floodplain elevation data were used for the Illinois River floodplain inundation mapping. Issues regarding vertical datum conversions were evaluated and determined to be insignificant at the scale and intended application for this

Data can be accessed at several levels of resolution, we used the default 1 arc second download format to conserve data processing requirements over large geographic regions and because subsequent hydrologic modeling analyses were completed at similar resolution. Rectangular tiles covering about 100 miles each were downloaded and data extracted by a mask of the floodplain as represented by the prior defined floodplain extent for each pool (Laustrup and Lowenburg, 1994). We combined the pool scale DEMs into a DEM for the entire floodplain using default mosaic procedures in ArcGIS. Metric elevations were converted (i.e., times 3.281 in Raster math) to English units to match river stage in feet and discharge in cubic feet per second (cfs) which is the vernacular of the Flow Frequency

Hydrologic analyses were accomplished with 100 years of record from 1898 to 1998 using the log-Pearson Type III distribution for unregulated flows at gages. Mainstem flows between gages were determined by interpolation of the mean and the standard deviation for the annual flow distribution based on drainage area in conjunction with a regional skew. Flood control reservoir project impacts were defined by developing regulated versus nonregulated relationships for discharges, extreme events were determined by factoring up major historic events, and the UNET unsteady flow program was used to address hydraulic impacts. The result of the hydrologic aspects of the study was a discharge and related frequency of occurrence for stations or given cross sections located along the Mississippi

A hydraulic analysis was required to establish the water surface elevation associated with each frequency of discharge at each location or cross section along the river reach. The main procedures were to use the UNET unsteady flow numerical modeling tool with recent channel hydrographic surveys (routinely obtained for navigation channel maintenance), and floodplain digital terrain data collected in 1995 and 1998. Levee overtopping was established at the top of existing levee grade based on an upstream and a downstream point. Using these station rating curves and the station frequency flows developed during the hydrology phase, frequency elevation points were obtained for each cross section location. Connecting the corresponding points resulted in flood frequency elevation profiles (USACE, 2003).

study (Theiling, 2010).

**2.2.2 Flow frequency study** 

and Illinois Rivers (Figure 4; USACE, 2004b).

Study.

Triangulated Irregular Network (TIN) files were created from the cross section feature lines for each separate flood stage frequency (Figure 4). Each flood stage TIN: 50%, 20%, 10%, 4%, 2%, 1%, 0.5%, and 0.2% annual exceedance probability (i.e., 2-yr, 5-yr, 10-yr, 25-yr, 50-yr, 100-yr, 200-yr, 500-yr expected recurrence interval flood) was overlayed in a cut-fill analysis on the high resolution floodplain topography for each navigation pool or reach. The area represented as inundated by the cut-fill procedure for each flood stage was separated out as a conditional GRID analysis that selected areas with volume > 0 and output a single GRID with a count of the 20X20 m cells below the elevation of the water surface elevation (Figure 4). This value was exported to a spreadsheet where grid counts were converted to area estimates (acres) for the navigation pool scale at which they were created. The resulting GRID was converted to a shapefile to merge with other layers to create system-scale layers of the potential water distribution at each flood stage (Figure 4).

Floodplain inundation classes (i.e., 50, 20, 10, 5, and 1 percent annual exceedance probability floods) were summarized by river mile and compared by geomorphic reach. Leveed areas were then extracted from inundation layers to assess changes in flood distribution attributable to levees. These data were also summarized by river mile and geomorphic reach. The inundation classes in leveed areas were subtracted from the maximum simulated inundation surface in each geomorphic reach (i.e., 1 percent or 0.02 percent annual exceedance probability flood) and data were normalized as percent of maximum inundation area.

#### **2.3 Land cover**

#### **2.3.1 Presettlement land cover**

Land cover databases are the foundation of our vision of UMRS landscapes and habitats over multiple reference conditions. Early explorers described interesting new landscapes, vast abundances of strange new animals, and drew crude maps as they moved through North America (Carlander, 1954). As settlers followed explorers, the Public Land Survey (PLS) mapped and characterized the mostly unsettled Louisiana Territories to sell land to the westward-expanding population of the United States (Sickley and Mladenoff, 2007). The PLS methods first divided the region into 36 square mile townships and then subdivided each one into 36 one mile square sections. Along the township and section lines, the surveyors set posts every half mile at locations called ½ section corners (where section lines intersected) and quarter section corners (midway between the section corners). Between two and four bearing trees were marked near each post and recorded in their notebooks by species, diameter, and compass bearing and distance from the post. The surveyors recorded other features that they encountered along the survey lines in the notebooks as well, including water features, individual trees located between the survey posts, boundaries between the ecosystems through which they were traveling, boundaries of natural and anthropogenic disturbances, and cultural features such as houses, cultivated fields, roads, and towns. Initial pilot studies reconstructing PLS surveys in the UMRS (Nelson et al., 1996) proved to be very valuable, so The Nature Conservancy's Great River Partnership contracted the University of Wisconsin Forest Ecology Lab to complete a comprehensive interpretation in a GIS for the entire UMRS (Sickley and Mladenoff, 2007). PLS data extend beyond the bluff into upland habitats, but the data were clipped to the bluff to bluff extent

Hydro-Geomorphic Classification

**2.3.2 Contemporary land cover** 

scheme of the PLS on the contemporary data.

**2.3.3 Land cover classes** 

information.

**2.4 Data analysis** 

(Theiling, 2010).

and Potential Vegetation Mapping for Upper Mississippi River Bottomland Restoration 173

Environmental Management Program Long Term Resource Monitoring (LTRM) has compiled several system-wide land cover data sets. The 2000 land cover data extent was used to define the floodplain area for other GIS coverages. LTRMP Land cover data were interpreted from 1:15,000 scale infra-red aerial photography with a minimum map unit of one acre. Several land cover classifications schemes have been used, but National spatial data standards have helped optimize and standardize the scheme. The current classification scheme includes 31 classes that are ecologically or socially relevant. The scheme can be lumped or split as necessary to match other data sets. The HNA-18 land cover classification was reclassified to the general ecosystem classes compatible with the PLS data (Theiling, 2010). LTRM land cover data were combined in a spatial join to replicate the point sampling

Land cover data from historic and contemporary periods were generalized to a common 12 class scheme (Theiling, 2010). The classification scheme combined several forest classes from the contemporary classification and two from the historic classification. The savanna class combined 11 classes from the PLS surveys, but none from the modern surveys because the habitat is only rarely present in the modern landscape. A "bottom" class was evident in the historic data but not clear in the contemporary data which were lumped as "forest." Similar to forests, the historic data allowed separation of several prairie classes: prairie, bottom prairie, and wet prairie which were not separable in the modern data. The historic classification identified forested wetlands as swamps, but that distinction is not made in the contemporary data where forested wetlands were not identified. Shrubs were represented in both data sets. Water was classified as several aquatic area types in the historic data, but in the modern data distinctions among aquatic classes depended on the presence of vegetation. Agriculture and developed classes were not common in the historic data, but they were very important in the modern data. PLS data have been criticized for inaccurate and inconsistent identifications and naming conventions. Their use at the general landscape level here is to provide a broad view of the system without consideration of species and precise locational

We overlayed the river reach segmentation schemes on land cover layers to provide proportional estimates for each land cover class to show plant community composition change along the river. A GIS extension was built to complete point counts for each land cover class at each river mile (Tim Fox, U.S. Geological Survey, Upper Midwest Environmental Sciences Center, La Crosse, Wisconsin). We also summed point counts by geomorphic class and hydraulic inundation frequency. Data were normalized as a proportion of total points within each segment (i.e., river mile, pool, reach, etc.) to assess the relative importance of each class in each area. The normalized data were plotted by river mile here and also used in multivariate statistical analyses examining the distribution of geomorphic, hydrologic, and land cover characteristics among river reaches at several scales

for this initial analysis. The Nature Conservancy dataset, and recently available statewide PLS plat map GIS coverages, provide a snapshot to speculate on ecological community associations in the undeveloped landscape.

Scale and resolution are important issues to consider when using PLS data. The quarter section and ½ section corners are a half mile apart and are generally marked by two to four trees each. A single section is commonly bounded by eight corners, which means that a square mile in the data would contain information on about only 16 to 32 trees. This is too sparse to be used at a stand or site level in anything other than the most qualitative sense. It is recommended to use the data at broad spatial extents (tens to thousands of square miles) and at resolutions of no less than a square mile (Schulte and Mladenoff, 2001; Theiling, 2010).

Fig. 4. Images depicting examples of elevation data, hydraulic model cross-sections, derived TINs, cut/fill interpolation, and grid and shapefile products.

#### **2.3.2 Contemporary land cover**

172 Studies on Environmental and Applied Geomorphology

for this initial analysis. The Nature Conservancy dataset, and recently available statewide PLS plat map GIS coverages, provide a snapshot to speculate on ecological community

Scale and resolution are important issues to consider when using PLS data. The quarter section and ½ section corners are a half mile apart and are generally marked by two to four trees each. A single section is commonly bounded by eight corners, which means that a square mile in the data would contain information on about only 16 to 32 trees. This is too sparse to be used at a stand or site level in anything other than the most qualitative sense. It is recommended to use the data at broad spatial extents (tens to thousands of square miles) and at resolutions of no less than a square mile (Schulte and Mladenoff, 2001; Theiling, 2010).

Fig. 4. Images depicting examples of elevation data, hydraulic model cross-sections, derived

TINs, cut/fill interpolation, and grid and shapefile products.

associations in the undeveloped landscape.

Environmental Management Program Long Term Resource Monitoring (LTRM) has compiled several system-wide land cover data sets. The 2000 land cover data extent was used to define the floodplain area for other GIS coverages. LTRMP Land cover data were interpreted from 1:15,000 scale infra-red aerial photography with a minimum map unit of one acre. Several land cover classifications schemes have been used, but National spatial data standards have helped optimize and standardize the scheme. The current classification scheme includes 31 classes that are ecologically or socially relevant. The scheme can be lumped or split as necessary to match other data sets. The HNA-18 land cover classification was reclassified to the general ecosystem classes compatible with the PLS data (Theiling, 2010). LTRM land cover data were combined in a spatial join to replicate the point sampling scheme of the PLS on the contemporary data.

## **2.3.3 Land cover classes**

Land cover data from historic and contemporary periods were generalized to a common 12 class scheme (Theiling, 2010). The classification scheme combined several forest classes from the contemporary classification and two from the historic classification. The savanna class combined 11 classes from the PLS surveys, but none from the modern surveys because the habitat is only rarely present in the modern landscape. A "bottom" class was evident in the historic data but not clear in the contemporary data which were lumped as "forest." Similar to forests, the historic data allowed separation of several prairie classes: prairie, bottom prairie, and wet prairie which were not separable in the modern data. The historic classification identified forested wetlands as swamps, but that distinction is not made in the contemporary data where forested wetlands were not identified. Shrubs were represented in both data sets. Water was classified as several aquatic area types in the historic data, but in the modern data distinctions among aquatic classes depended on the presence of vegetation. Agriculture and developed classes were not common in the historic data, but they were very important in the modern data. PLS data have been criticized for inaccurate and inconsistent identifications and naming conventions. Their use at the general landscape level here is to provide a broad view of the system without consideration of species and precise locational information.

#### **2.4 Data analysis**

We overlayed the river reach segmentation schemes on land cover layers to provide proportional estimates for each land cover class to show plant community composition change along the river. A GIS extension was built to complete point counts for each land cover class at each river mile (Tim Fox, U.S. Geological Survey, Upper Midwest Environmental Sciences Center, La Crosse, Wisconsin). We also summed point counts by geomorphic class and hydraulic inundation frequency. Data were normalized as a proportion of total points within each segment (i.e., river mile, pool, reach, etc.) to assess the relative importance of each class in each area. The normalized data were plotted by river mile here and also used in multivariate statistical analyses examining the distribution of geomorphic, hydrologic, and land cover characteristics among river reaches at several scales (Theiling, 2010).

Hydro-Geomorphic Classification

Open Water Active river channels,

Riverfront Forest Bar-and-chute and

Swales Point bar swales and

Slope Forest Alluvial fans, colluvial

Savanna Alluvial fans, colluvial

Floodplain Forest

Bottomland hardwood forest

Bottomland Prairie

**3. Results** 

**3.1 Gemorphology** 

side channels

braided bar

depressions

Wet Point bar and terrace swales and depressions

Bottomland Lake Abandoned channels Clay, silt-clay with

tributary riparian zones

Mesic Prairie Point bar edges and terraces Sandy loam, silt

Missouri) in relation to geomorphic surface, soils, and flood frequency.

Table 1. Hydrogeomorphic matrix of historic distribution of major vegetation

Backswamp, larger point bar swales and floodplain

and Potential Vegetation Mapping for Upper Mississippi River Bottomland Restoration 175

**Habitat Type Geomorphic Surface Soil Type Flood Frequency** 

Ridges Point bar ridge Loam, sandy loam 2 – 5 year

Intermediate Point bar ridges Silt loam >5 year

communities/habitat types in the American Bottoms geomorphic reach (near St. Louis,

Land Sediment Assemblage abundance plotted by rive mile illustrates the distribution of each class and the relative width of the floodplain (Figure 5, top). Geomorphic reach overlays helped identify characteristics that separated reaches in a multivariate analysis (Theiling, 2010). The Chippewa River Reach (RM 650 – 750) is separated downstream by the narrower Wisconsin River Reach (RM605-650) which runs through resistant dolomite valley walls (Knox, 2007). The floodplain widens again through erosive shale in the Maquoketa

Riverine Riverine

sand/loam plugs

Sand, sandy loam and silt loam in

Silt loam, slit clay

veneer 1 – 2 year

Silt loam, silty clay 2 – 5 year

Clay, silt clay 2 – 5 year

loam >20 year

Abandoned channels Clay, silt-clay Permanent-

swales

aprons, terrace edges Mixed erosional >20 year

aprons, terrace interface Silt loam 10 – 20 year

Permanent Permanentseasonally dry

seasonally dry

Permanent to semi-permanent

1 – 2 year
