Technical Report

INTRODUCTION

Streams are dynamic systems that change through time. Stream-channel change in meandering rivers of Illinois occurs both through the natural operation of fluvial processes under relatively constant environmental conditions (e.g. migration of meanders) and in response to changes in environmental conditions, especially those produced by human alteration of watersheds (e.g. urbanization) and stream channels (e.g. channelization). Knowledge of the magnitudes, rates and mechanisms of stream-channel change is critical for assessing the extent to which human effects have contributed to this change.

Many streams and rivers in Illinois have been directly altered by people to some degree, and most have experienced hydrological changes over the past century. The prevailing perception is that most streams are eroding excessively and need to be actively managed to reduce rates of erosion. In the Illinois River basin, much concern centers on the process of streambank erosion in major tributaries, which is viewed as a source of sediment that contributes to sedimentation in the Illinois River. Reducing the load of sediment entering the main stem is a primary objective of management initiatives for the Illinois River watershed. When streams erode their banks they also change their form or position on the landscape, which often results in the perception that the stream is “unstable”. Although recommendation 10 of the Integrated Management Plan for the Illinois River Watershed (State of Illinois 1997) calls for stabilization of unstable streams in rural and urban areas, this recommendation raises the issue of what constitutes an “unstable” stream.

From a geomorphological perspective, an unstable stream is one that is exhibits abrupt, episodic or progressive changes in location, geometry, gradient, or pattern because of natural or human-induced changes that occur over a period of years or decades and that reflect local or catchment-wide imbalances between sediment inputs and outputs or the proximity of the system to a geomorphic threshold (Rhoads, 1995). The occurrence of bank erosion does not imply that a stream is unstable in the sense that it is actively adjusting to changes in environmental conditions. Bank erosion and lateral movement are part of the natural dynamics of meandering streams and provide mechanisms for remobilization of sediment stored within floodplains. Determinations of instability require information on changes in stream channels over time. In particular, instability implies that rates of changes are increasing over time. Although such determinations could be made on the basis of long-term monitoring, few, if any, studies of this type have been conducted in the past and current programs will have to be sustained well into the future before they yield long-term information on channel change.

The purpose of this pilot study is to demonstrate the value of GIS-based analysis of historical aerial photography for characterizing temporal changes in the location and pattern of meandering rivers in Illinois and for associating this change with human manipulation of streams. This analysis provides a valuable tool for assessing human-induced changes in stream channels and the relation of these changes to background rates of change associated with lateral migration of meanders. It yields relatively rapid, reasonably accurate assessments appropriate for reconnaissance projects such as described in the U.S. Army Corps of Engineers Restoration Needs Assessment (in prep.).

The analytical approach used in this study adopts protocols developed and tested in previous work (Rhoads and Urban 1997; Urban 2000), and draws upon an extensive archive of historical aerial photographs taken by the U.S. Department of Agriculture and the U.S. Geological Survey between the late 1930's and the present. The method is appropriate for assessing planform changes over time, but does not directly evaluate channel changes associated with stream incision. To determine the value of the method for regional- to subwatershed-scale assessment of stream dynamics and to predict resources required for more comprehensive work, we selected for analysis 1.6 km reaches along ten different streams throughout the state. Each of these streams has an associated hydrological record that extends over several decades. The ten stream reaches collectively represent all of the major physiographic divisions (Leighton et al. 1948; Phillips et al., in progress) within the Illinois River watershed. By examining streams from across the basin we provide an overview of watershed-scale variability in stream dynamics. Outcomes of this pilot study, including a tool to automate quantification of planform changes within a GIS framework, will enhance the capacity for rapid assessment of channel dynamics over entire stream networks within subwatersheds of the Illinois River basin.

METHODS

Site Selection

The primary criteria for selection of sites were to sample streams that represent a broad range of geographic conditions and have an available record of hydrologic data. Streams of intermediate size were selected because the largest, navigable rivers in Illinois tend to be maintained, so no planform change is expected, whereas the smallest streams are not readily distinguishable on most air photos. The 1:100,000 scale digital line graph coverage of perennial streams used for the selection process adequately included streams of appropriate size. The focus on gauged streams allows use of stream power in interpretation of observed planform changes (Rhoads 1995). Only gauges with at least several decades of hydrologic data were considered.

We tried to attain geographic representativeness by selecting two sites within each physiographic division in the Illinois River basin (Fig. 1; Table I). Physiographic divisions are distinguished by differences in topography, mean elevation, bedrock geology, and surficial geology, especially the style and timing of glacial sculpting (Leighton et al. 1948 and Phillips et al., in progress). There is some presumed correlation between these parameters and intrinsic stream behaviors. The initial selection used provinces defined by Leighton et al. (1948). During the course of the study, several of the provincial boundaries were revised by Phillips et al. (in progress), placing some reaches in different provinces. Further, there are insufficient numbers of streams with adequate hydrologic data (below) to achieve equal coverage. In the end, four reaches in the Bloomington Ridged Plain, two in the Kankakee Plain, and one in each of the other provinces were selected (Fig. 1; Table I).

Specific selection of study reaches was a compromise between locations near gauging stations, image quality and image availability. The criteria included:

· Most of the reach without heavy forest cover

· Reach not near a tributary to avoid backwater effects

· Reach wide enough for stream banks to be visible in photos

· Reach not on the edge of photos to minimize georeferencing error

· The quality of the entire suite of photos sufficiently good for interpretation

McKee Creek was selected despite the fact that it is an ungauged stream because there is considerable interest from state and federal agencies in assessing the dynamics of this stream. The lack of hydrologic data is typical of streams in the Illinois River basin and the inclusion of McKee Creek illustrates the effectiveness of the method for ungauged streams. Flow conditions for this reach were estimated by comparisons with other gauged streams in the region (see Flow, below). The study reach was identified by random selection of a river mile upstream of the Brown County line.

Airphoto Database

Sources. – Most of the aerial photography for the pilot study was acquired from the Map and Geography Library at the University of Illinois, Urbana-Champaign. The imagery is available in the form of photographic contact sheets that were developed from periodic flights contracted through the U.S. Department of Agriculture (Appx. I). The last image in each series were 3.75 minute digital orthophotoquarterquadrangles (DOQQ) from the U.S. Geological Survey. The DOQQs are available from the Illinois Natural Resources Geospatial Data Clearinghouse (http://www.isgs.uiuc.edu/nsdihome/webdocs/doqs/graphic.html ) as MrSID compressed image files.

Scanning and Georeferencing. – The aerial photographs were digitally captured using a large-format photoscanner. The scanner was operated using Epson Silverfast software, which facilitated the importation of the scanned imagery into Adobe Photoshop 5.0 Limited Edition. From the Silverfast interface, the scans were performed at 720 dpi, 12-bit grayscale. The 720 dpi rate has been demonstrated to provide the maximum value when scanning paper prints (D. Luman, pers. com. 2002). After importation into Photoshop, the images were exported and archived as raw tag image file format (TIFF) files onto CD-ROM

The TIFF images range in file size from 40 to 50 megabytes. Approximately 2 to 3 photographs were scanned for each year of coverage for a given stream reach, especially if the reach lay near the edge of a photo. Typically, one of the overlapping photos was most suited to georeferencing, i.e., it had lower RMSE than the others, but the other photos might be clearer for qualitative analysis. On average, seven complete photosets were obtained for each reach. Thus, scanned photosets typically occupied 600 to 800 megabytes in total.

Photograph scale varied across each time series. Photographs taken before 1970 have a scale of 1:20,000, photographs taken after 1970 have a scale of 1:40,000 (Appx. I). DOQQ were originally obtained at 1:40,000, but processed up to meet National Map Accuracy Standards of 1:12,000. With a 720 dpi scan rate, the 1:20,000 images have a ground resolution (pixel size) of 0.7 m, whereas the 1:40,000 images have a ground resolution of 1.4 m. These are very close to the 1 m resolution of the DOQQ.

Most of the scanned images were georeferenced to DOQQs in Universal Transverse Mercator Zone 16, North American Datum of 1983. There was no DOQQ available for the Spoon R. or Farm Ck. reaches during summer 2001, so those images were georeferenced to the London Mills 1:24,000 topographic map in digital raster graphic format. Note that the Spoon R. and McKee Creek quadrangles (London Mills and Verailles, respectively) have native UTM zones of 15. All data for those sites were georeferenced or reprojected to Zone 16 to conform to the other sites.

Georeferencing was performed using ESRI’s ArcView® GIS software with the ERDAS Imagine® software extension, Image Analyst® v1.0 (see the Procedures highlights for basic system architecture and rectification procedures using the ArcView GUI). The TIFF images were brought into ArcView and referenced off of the DOQQ base layer with the on-screen identification of ground control points (GCPs). A minimum of 10 GCPs were identified, and the resulting planimetric root-mean-squared errors (RMSE) were kept below 15.0 m. The average RMSE of all photos was ± 6 m (Appx. I). GCPs were first established nearest the center of the images where optical distortions are lowest, and then added radially about the center-point. No points were established outside of the circle that would subtend the square image plate. Optical distortions in the regions outside of such a radius are typically large and will contribute to exceptionally high RMSE. Georeferencing attributes for each photograph are given in Appendix I.

The imagery was rectified to the GCP using a six-parameter affine transformation, which adjusts the imagery with respect to the base layer in terms of scale, rotation and translation. Once rectified, the new image was written in the Imagine format (.img). The associated GCPs were saved as attribute point data that included RMSE and other spatial information. The rectification data was written to CD-ROM and archived.

Site Attributes

Slope. – Channel slopes for the ten reaches were derived from topographic maps. Derivation of slopes from topographic maps is more generalized than using a surveyed local channel slope because contours that cross the stream on the map are in some instances several stream miles removed from the study reach. Thus, measures of slope in this study are probably more representative of valley slope than local bed slope. This may or may not be a problematic because local slope is sensitive to discharge variations, as well as local bed conditions (Bledsoe and Watson 2001).

Flow. – Average daily discharge, average monthly discharge, and peak yearly discharge for nine of the ten sites were obtained from US Geological Survey gauging data (USGS 2001). The sites have 4-6 decades of flow data (Table II; Appx. II. There are large gaps in the data for Farm Ck. after WY1984, and data have been collected only periodically since WY1991).

For each gauged site, cross-sectional stream power was calculated over the interval between subsequent aerial photographs using the average monthly flow discharge. Cross-sectional stream power is defined as:

Ω = γQS_e (W∙m^-1)

where γ is the specific weight of water, Q is discharge, and S_e is the energy gradient, which in this study is approximated by the channel slope (Rhoads 1995). Cross-sectional stream power is the average rate of energy expended by the flow per unit length of the reach. The bankfull channel width of a reach can be used as a standardization parameter for comparing different reaches (Rhoads 1995), and so defines mean power, ω = Ω ∙w^-1, or the amount of power applied per unit area of the bed.

For McKee Ck., the only study reach without a gauging station, stream power was calculated using discharge data obtained from the La Moine R. at Colmar gauging station (#05584500; Fig. 1). This station was chosen because it is the nearest long-term gauge record, has a similar watershed size, has comparable land cover and land usage, and has similar flood recurrence intervals to the McKee Ck. reach. The flood recurrence interval for McKee Ck. was determined using regional regression formulas (Curtis, 1987). The McKee to La Moine discharge ratio was approximately 80% for all recurrence intervals. Therefore, the La Moine discharge data was multiplied by 0.80 to attain values more consistent with expected values for McKee Ck. and then used in power calculations.

Other Site Attributes. – Five measurements of channel width at approximately 0.2 km intervals along each reach were taken from the DOQQ at 1:1,000 scale. The channel banks were distinguished by the transition from either bright reflections representing active channel deposition, or, when flow was high, flat reflections representing water, and more variable grays representing floodplain vegetation. Width was also measured at the gauging station for use in mean power calculations.

Unpublished site information was acquired from the USGS for each of the nine gauging stations. Included were descriptions of reach geomorphology and features such as dams, bed substrate, and channel maintenance near the station. For most reaches the records date back to the establishment of the station, and were reexamined or updated sporadically through to 2001. Floodplain soil information was taken from county soil surveys, when available. Surficial geology maps on scales from 1:24,000 to 1:500,000 were used to describe the regional geology. Cultural and engineering histories, and anecdotal evidence of reach behavior were obtained from local authorities for some of the reaches. Watershed information, including basin size and land cover were obtained from the Illinois Watershed Management Clearinghouse (UIUC ACES, 2002), which serves data from several sources.

Stream Threads and Buffering

Stream channel centerlines (stream threads) were digitized on-screen in an ArcView interface from the rectified images zoomed in to 3-times hardcopy resolution. At such a zoom level, in-channel features such as active bars and banks were usually observable and used to identify the channel centerline (Figs. 2, 3). Line vertices were generated every few meters, with a higher density over areas of greater curvature to yield a reasonably precise arc representing the center of the channel. Vegetative overgrowth occasionally obscured or covered the channel. In those instances, centerlines were inferred using multiple photographs from the same flight year. The image color ramp on some images was adjusted to highlight the channel features. Error caused by misidentification of the channel centerline during digitization is a concern, particularly along wider channels. It could possibly be overcome by constructing stream polygons through digitization of both channel banks. Quality control experiments by B. Rhoads have shown, however, that digitization error is far outweighed by georeferencing error. In cases where digitization error is significant, i.e., results in potential over-quantification of planform change (see below), the error can be inferred in the interpretation phase.

Quantifying Change: Automated Intersection Tool

Buffers were applied to each stream thread using the respective image RMSE to represent the degree of lateral uncertainty that exists in the digitization. Thus, the buffer encompasses the RMSE distance on either side of the channel; any channel migration beyond the limits of the centerline buffers for subsequent years then represents a significant change in stream planform. Areas between significantly separated channel traces were generated in the GIS as polygonal thematic data, enabling spatial-temporal analysis of stream channel change.

The areal change polygons were created by intersecting pairs of RMSE-buffered polygons and running a custom script on the merged data to extract out the interstitial polygons, or “donut-holes”. The donut-holes are regions of statistically significant planform change (Fig. 4). The script was developed using ESRI’s proprietary Avenue object-oriented programming code. The script is executed through the ArcView Graphical User Interface (GUI) as a button tool (arrow, Fig. 4). The tool is portable to ArcView and ArcGIS sessions as an .avx script, or extension (calculate_enclosed_areas.avx).

When activated, the extension operates on active polygonal features within the current frame given that encapsulated voids exist in the dataset. It creates a new polygon theme and writes Shape, ID and Area fields into the new attribute table using the View’s current unit settings (Fig. 5). When the script finishes executing, it closes the editing session and outputs a new Shapefile into the active View frame. Polygons are not automatically created where the buffered polygons diverge at the head or foot of the reach. These polygons are closed manually after the new shapefile has been created.

Classification and Interpretation Rules

Change polygons were divided into 5 dynamic behavior classes (Table III), lateral or downstream migration, avulsion, channelization, post- channelization, or chute development. Lateral or downstream migration (hereafter, “migration”) is the typical planform movement of stream meanders as they erode sediment from their outside banks and deposit sediment on their inside banks. It is recognized by progressive evolution of channel planform. By this mechanism, all material between progressive channel positions is reworked down to the channel depth. Avulsion is the tendency of a channel to abruptly change positions. Material between positions is not reworked. Avulsion occurs when a chute developed on the floodplain during high flow is rapidly incised and ultimately captures all flow from the main channel. Some chutes may also be ephemeral features and abandoned before they become fully developed. Channelization is usually recognized by an abrupt change in channel planform. In the sense that floodplain material between channel positions remains unworked, channelization is not unlike avulsion. However, the channelized reach is typically quite straight, and airphotos and topographic maps may show levees and other evidence of the construction activities. Where an avulsion, chute formation, or migration occurred along a subreach that had been previously classified as channelized, the polygon was classed as Post-Channelization. The classification does not necessarily imply the evolving planform was caused by channelization, but rather indicates that the planform does continue to evolve.

Some polygons created by the automated intersection tool were rejected as caused by digitization error. Digitization error was most likely when stream channels were relatively wide making determination of the channel centerline imprecise, photo contrast was either too high or too low, the image was blurry, the subreach was forested, or when the RMSE varied greatly between images. One test for the last was to re-buffer the stream thread from the lowest RMSE image, e.g., the DOQQ, to the higher RMSE of the second image, e.g., almost any other photograph. If a gap between the buffers disappeared after this “buffer test”, the polygon was attributed to digitization error and rejected as a change polygon; if a gap between the buffers still existed, the polygon was accepted as a change polygon and assigned to the appropriate dynamic class. Potential digitization error along wide reaches can be circumvented by digitizing both sides of the channel rather than just the centerline, although the method was not followed here for consistency.

RESULTS

Blackberry Creek, Kendall County

Setting. – The Blackberry Ck. study reach lies in a predominantly rural area in northeastern Illinois, about 6 km upstream of its confluence with the Fox R. and 3 km upstream of a gauging station (Fig. 1; Table I). The watershed above the gauging station is 70.2 mi² (181.8 km²). The creek lies within the Bloomington Ridged Drift physiographic province, an area dominated by landforms of the Wisconsin glaciation (Fig. 1). The study reach cuts across silty clay loam diamicton ground moraine and glacifluvial sand and gravels (Willman 1971). The stream bed is comprised of “shifting sand and silt” (USGS, written com.).

Five aerial photographs taken between 1939 and 1998 comprise the base data set (Table II; Appx I). The reach is highly meandering and is bisected by a railroad (Fig. 6). The active floodplain near the study reach, interpreted as the area that appears saturated (low reflectance) in some photos, is about 150 m wide and lies about 5 m below the surrounding interfleuve. Meander and chute scars on the floodplain and in-stream bars attest to a dynamic fluvial regime along the reach. The reach was not extensively channelized through the period, although some channel maintenance may have occurred at the railroad bridge and in association with small bridges that were emplaced in the 1980’s. A low slope of 0.0008 is median for the reach set (Table IV).

Land use throughout the area decreased from 100% to approximately 75 % agriculture between 1939 and 1998 (Fig. 6), as residential developments grew and a golf course was constructed. Directly surrounding the creek, the upstream ¾ of the reach remained in agriculture throughout the period, while the southern ¼ changed from row crops to a golf course between the 1967 and 1988. In the 1939 photo, fields were farmed to the edge of the creek. The active floodplain did not appear to be cropped in 1967, and by 1998 a narrow zone of riparian vegetation, 30-180 m wide, had grown along both sides of the creek.

Flow data from the gauge at Yorkville are available since WY1960 (Appx. II), and thus overlap the last 4 photographs of the time series (1967-1999). Records are thought to be generally good though poor in winter because of ice effects (USGS writ. com. 2002). The average discharge for the entire record is 1.5 m³∙s^-1, with flows typically varying 10-fold between winter baseflow (Oct.-Feb.) and summer peak floods (Mar.-Sep.) (Appx. II). Mean stream power had relatively low average value and variability of our study set (Fig. 7).

Reach Evolution. – Over the 59 years of aerial photography, the most frequent type of change that occurred was attributable to lateral and downstream migration (Fig. 8; Table V; Table VI). Between 1939 and 1998, for example, a meander at the head of the reach became more arcuate – tending towards neck-cutoff – and then moved downstream (red arrow, 1939, Fig. 9). Lateral and downstream migration accounted for slightly less than half of the total planform change over the time series, but was spread evenly along the reach. Slightly more than half of the total change was attributed to avulsion and chute development. Although these events caused a relatively large amount of planform change, they comprise only 4 instances involving relatively short sub-reaches. Two significant channel bifurcations developed over 1967-1988 (Fig. 10) and 1993-1998 (Fig. 9, 1998), when floodplain chutes were enlarged and deepened, creating anabranching channels. These bifurcations could be precursors to avulsion. Potential avulsion of the earlier (and upstream) bifurcation was averted, however, by construction of a small bridge across the reach between 1988 and 1993 within the golf course. The bridge may have restricted flow, backing up water upstream and causing channel widening. Dredging or damming of the creek, or both, may have accompanied bridge construction to make a pleasant pond for viewing from the course restaurant. The downstream bifurcation occurred over the last study period, so further evolution is not known. A second bridge was constructed in the golf course about 275 m downstream of the first. Some channel widening, channelization and possibly hardening of the banks can be seen there, but the amount of planform change was too small to form a change polygon.

There was no correlation between planform change and mean stream power (Fig. 11).

Du Page River, Will County

Setting. – The Du Page R. study reach lies in the Kankakee Plain, a province of predominantly low relief and low slope between Wisconsin Episode moraines (Fig. 1; Table 1). The basin was significantly modified by lacustrine sedimentation and sculpting by large glacifluvial floods during the late Pleistocene. The surficial geology varies from thick morainal deposits, small and thin lacustrine deposits, fluvial terraces and bars, to bedrock exposed in some stream valleys. The Du Page R. near the study reach flows southwards through a low area between fluvially-sculpted portions of the Minooka and Rockdale moraines (Willman 1971). This geomorphology, particularly the relief, is not typical of the Kankakee Plain (Phillips et al., in progress). The surrounding low areas are underlain by diamicton with a veneer of glacilacustrine clay, silt, and sand in depressions. The study reach appears to be one of the few aggradational reaches with significant accumulation of recent alluvial silt, sand and some gravel (Larsen 1976). The substrate near the stream gauge about 350 m downstream of the study reach comprises bars of sand and silt (USGS, writ. com. 2002). Dolomitic bedrock crops out or is near surface immediately upstream, downstream, and along the east bank of the study reach (Willman 1971).

Six aerial photographs from 1939 to 1998 were available for study (Table II). Image quality except for 1993 is generally good. The study reach is an anomalous section of the river (Fig. 12). While most of the river is straight to slightly meandering, the middle of the study reach is anastomosing, with two main channels separating and reconnecting several times. Floodplain features through the anastomosed section include splay fans, abandoned chutes and meander scars. The reach is the widest of the study at about 40 m, and has a moderately high slope of 0.0015 (Table IV). A low dam occurs about 300 m downstream (USGS, writ. com. 2002).

Row crops were farmed across the area of the 1939 photograph. Upstream of the study reach and along the downstream left (eastern) bank, row crops extended up to the river’s edge (Fig. 12). Significant development after 1954 throughout the area included residential areas and a four-lane expressway which follows the Du Page on part of its course. Active agriculture immediately adjacent the reach appeared to have ceased by the 1967, although these tracts of land were not incorporated into the Hammel Woods Forest Preserve until 1978 (D. Robson, Will County Forest Preserve, writ. com. 2002). After incorporation the land was left to grow wild. In 1967, the downstream right bank of the study reach was forested, and rest of the riparian area was grassy field except for a narrow field of row crop along the upstream left bank. The forested zone grew wider and denser with time along the study reach, with some woody growth eventually becoming established on the islands. There was no direct human modification of the study reach, although the low dam downstream of the gauge and visible in airphotos affected low flow gauge records (USGS, writ. com., 2002).

Continuous flow data are available since WY1940 (Appx. II) from a gauge 350 m downstream of the study reach. The records are thought to be generally good except when ice jams occur during spring break-up (USGS, writ. com., 2002). The average monthly discharge for the entire record is 8.1 m³∙s^-1 with a standard deviation of 5.63 m³∙s^-1. Baseflow appears to have increased steadily since the early 1960’s (Appx. II), possibly from either wetter climate or increased runoff associated with increased impervious landcover. Du Page has the highest median mean power of the study, as well as the greatest variability in power magnitude (Fig. 7).

Reach Evolution. – The general shape of the study reach remained constant throughout the 59 years of photographic record (Fig. 13). During wet periods, the river floods old chutes, emphasizing the anastomosing planform. Lateral and downstream migration was the dominant mode of planform change (Table V). It occurred at cutbanks along the entire study reach throughout the time series, and at a greatly accelerated pace over the last period, from 1993 to 1998 (Fig. 14). About 15% of the planform change was attributed to the formation and abandonment of chutes or secondary channels. These changes are depicted in Figure 15 where formation of an emergent bar between 1939 and 1967 creates a chute, and a meander becomes increasingly necked as downstream migration of the upstream pointbar/cutbank occur. There is no obvious correlation between stream power and planform change in Figure 14, although both increased over the time series.

Farm Creek, Tazewell County

Setting. – The Farm Ck. study reach flows east to west through an intermorainic basin of the Bloomington Ridged Drift near the westernmost extent of the Wisconsinan Episode glaciation (Fig. 1; Table 1). Upstream of the study reach, the stream has dissected the Washington and Bloomington moraines, features of relatively high relief comprised of clay loam to loam diamicton (till) with lenses of stratified silt, sand, and gravel and covered with a thick blanket of silt to silt loam (loess; Johnstone, in review). Downstream from the moraines, the stream has incised through thick outwash sands and gravels that now comprise the bluffs of the modern valley, and into Illinois Episode till and glacifluvial sediment (Hunter 1966). Although fluvial dissection has probably reduced the slope of the alluvial valley significantly, the slope at the study reach, 0.006, is the highest of this study (Table IV). Sand loam, loam, and silt loam greater than 2 m thick have been deposited on the modern floodplain (Teater 1996; Johnstone, in review).

The channel substrate at the stream gauge, which lies approximately 1200 m downstream, is sand and gravel (USGS, writ. com., 2002). Although the study reach and gauge are separated by a dam, the brightness of channel bars on aerial photographs suggest that they, too, are predominantly sandy. A major landscape feature that affects the modern stream is the flood control dam erected in 1951. The spillway at an elevation of 188 m and flow restriction by a 2 m diameter outlet pipe places the Farm Ck. study reach within the 100 yr floodplain (ISWS 1996). The study reach is joined by several low order tributaries.

Seven aerial photographs from1939 to 1998 comprise the base data set (Table II). The Farm Ck. study reach displays many changes over the course of the study due to the extensive human manipulation of the creekbed and floodplain (Fig 16). In the 1939 photo the highly meandering reach featuring several large loops and curves appears to be relatively unmodified. A railroad ran abreast of the creek at portions, crossing the creek twice along the study reach. Where the channel abutted the railroad, the banks were likely rip-rapped. In advance of dam construction, more than 8 km of the railroad was realigned to south of the creek. By 1969 almost the entire study reach was channelized, cutting off the three large loops. Channelization or channel maintenance continued through 1998.

The landscape surrounding Farm Ck. underwent substantial change, as well. In 1939, the small town of Farmdale north of the stream was the only development, comprising less than 5% of land over the entire area of the photograph (Fig. 16). The right (north) floodplain was in row crop with little riparian buffer. The left floodplain was partly agricultural, but also included stretches of light forest and field. By 1998, the city of East Peoria (Fig. 1) had expanded over the uplands to cover more than 50% of the DOQQ of that year, though all of the floodplain was forested. A sewage treatment plant constructed in or before 1963 is permitted to discharge into Farm Ck. near the head of the study reach (sewage lagoons are dark, oblong patches, Fig. 16; EPA Envirofacts).

Flow data at the gauge downstream of the dam (Table I) was collected continuously from WY1948 through WY1984, but was not collected for water years 1985-1990 and 1994-1997, and collected only during the wet season for 1991-1993 and 1998 (Appx. II). Active bars made records poor (USGS, writ. com. 2002), and dam impoundment at medium to high flows probably both substantially reduces flow magnitudes and attenuates flood events. For this analysis, the data gaps were ignored and annual average values from existing data were assumed applicable for the entire period. High mean stream power values for 1988-1994 are probably biased by the inclusion of data from the relatively wet year of 1993 (three floods exceeded the 2-yr recurrence interval) and the exclusion of all potential low flow values (compare Appx. II to Fig. 17). On the other hand, positive biasing might partly offset negative biasing caused by flow attenuation at the dam. Farm Ck. had the lowest mean annual discharge (0.581 m³∙s^-1) of all the study reaches, and is very flashy (USGS, writ. com., 2002). Mean stream power, however, was moderate in magnitude and variability (Fig. 7).

Reach Evolution. – Almost the entire study reach was channelized at some point between 1939 and 1998, but the planform continued to evolve after channelization (Fig. 18; Table VI). The cutting off of 2 large meanders during railroad realignment before 1951, a third before 1969, and several smaller modifications caused about 95% of the planform change. Along non-channelized reaches, lateral and downstream migration of meanders was the main dynamic mode, although accounted for only 6 % of the gross change.

Continued evolution of channel planforms after channelization was attributed to both avulsion and meander migration (Table VI). An avulsion observed between the 1939 and 1951 photos was attributed to post-channelization change because the very straight section immediately upstream in the 1939 photo was interpreted to have been modified to protect the railroad embankment (Fig. 19). Figure 20 shows downstream migration redeveloping over 25 years after channelization. Note the marked increase in planform sinuosity between 1969 and 1994. The rate of change of downstream migration in the channelized reaches is of similar magnitude to downstream migration in the unchannelized reaches (Table VI).

There was no simple correlation between mean stream power and planform change (Fig. 17). Stream power varied 3-fold over the time series, and planform change was dominated by channelization.

Ferson Creek, Kane County

Setting. – The Ferson Ck. study reach occurs just upstream of the confluence with the Fox R. in the northeastern portion of the Bloomington Ridged Plain (Fig. 1.; Table I). The site was selected at the beginning of the study as representative of the Wheaton Morainal Province as defined by Leighton et al. (1948), but that provincial boundary was moved east in the reevaluation by Phillips et al. (in progress). The reclassification was partly based upon stratigraphic reclassification of the underlying till (Lemont Formation) into an older episode of glaciation than formed the morainal complex to the east. The provincial topography is dominated by subparallel morainal ridges, kamic moraines, outwash deposits in the valleys. Ferson Ck. flows northeasterly, cutting across the St. Charles and Minooka moraines. Post-glacial valley deposits in lower Ferson Ck. include silt and sand with some gravel derived from glacifluvial sand and gravel deposits and loam to clay loam diamicton in ice-marginal deposits (Grimley and Curry, in press; Curry et al. 1999). The stream bed is comprised of sand to gravel with cobbles (USGS, writ. com. 2002). The study reach channel has one of the highest slopes of the study, 0.002, but is relatively narrow (Table IV).

Seven aerial photographs from 1939 to 1999 comprise the base data set (Table II). Photos were obtained at both high and low flow stages (Appx. II). Images from 1975-1994 were dropped from the analysis because image quality was fair to very poor including very high RMSE in 1988. Ferson Ck. is moderately to highly sinuous, with some straight reaches. The study reach is less sinuous with lower amplitude meanders than is the creek upstream (Fig. 21). Abandoned channel tracks, possibly formed by avulsion because no scroll bar and related features can be distinguished, can be seen on the left (north) floodplain in both photos.

Several human modifications affected the reach over the photo series. Between 1939 and 1961 a large bridge was constructed across the lower third of the study reach (Fig. 21,1999). A pond just upstream of the bridge appeared in 1961. The straight edges of the pond suggest it was dredged; it may have been a borrow pit used in bridge construction. In 1961 the main channel bypassed the pond to the south, whereas the pond had a spillway which flowed into the main channel. In 1969 the main channel flowed through the pond (then a pool?) and occupied the spillway. The bridge was widened once in 1984, and reconstructed in 1990.

Two small bridges or dams or both were constructed across the upstream end of the study reach prior to 1939 (light linear features across creek, Fig. 21; prominent in 1998). A small rock dam 240 m downstream of the gauging station was noted in station data (USGS, writ. com. 2002). This dam may support the upstream-most of two ponds in the downstream 200 m of the study reach. The ponds fluctuated in size throughout the study series, possibly because of periodic neglect and repair of the dams. Pond formation caused some widening of the creek immediately upstream by backwater affects, and significant channel realignment may have been associated with dam history.

In 1939, the floodplain was open field with light forest cover (Fig. 21). Croplands and lightly to heavily forested patches covered the uplands. Residential areas grew rapidly after 1975 at the expense of agricultural land, but forest cover around the stream grew increasingly dense.

A gauging station was installed at the downstream bridge crossing in 1960. Records are generally good except in winter, possibly because of ice jams (USGS, writ. com., 2002). Average annual discharge was the third lowest of the study, but mean stream power was about average in magnitude and variability (Fig. 7). The mean annual stream power for WY1960 was applied to the 1939-1961 photograph interval. This value is probably low because high flows that year were lower than flows during the rest of the record.

Reach Evolution. – Almost all change observed on the study reach was attributed to lateral or downstream migration (Fig. 22; Table VI). One small change polygon was attributed to channelization where the channel appeared to be deflected south by creation of the pond/borrow pit during bridge construction (Fig. 23). Although other channel changes directly associated with this feature could be observed through the time series, they were smaller than the buffer width. A large medial bar upstream of this pond formed before 1967 (anabranching in Fig. 22; Fig. 23, 1967). This was contemporaneous with bridge construction and could possibly be related to backwater effects. That connection cannot be clearly made, however, so associated planform changes were not classed as “channelization” or “post-channelized”. The bar eventually became stabilized by vegetation, creating an anastomosed channel (Fig. 22). This bar, as well as several unstable medial and lateral bars observed in the photos, are evidence of an aggradational regime. Migration of the channel at mid-reach was caused by delta formation between 1939 and 1961 where a small tributary enters the reach (Fig. 23, 1961). The high sediment loading to the tributary cannot be attributed to any landuse practices interpreted from the airphotos, and the tributary appeared inactive in several of the photos. It is possible the tributary is an abandoned channel scar, and was only active during flood.

Both the rate of areal planform change and mean stream power increased dramatically between 1961 and 1975 (Fig. 24). After 1975 the rate of planform change plummeted, whereas mean stream power decreased only slightly. In general, the 1975-1994 photos were rather poor with either high RMSE or blurriness, so changes in planform may have been not well-characterized over the intervals. The biggest planform changes from 1967-1975 involved two narrow polygons downstream of the stream gauge. This section was relatively straight, and although the channel thread could be reasonably identified, the 1975 image was very blurry so these polygons may have been created by digitization error. By comparing the 1967 to 1994 photos, by contrast (dashed line, Fig. 24), 28 m² yr^-1 of planform change can be attributed to development of the anastomosed pattern and downstream meander migration.

Kickapoo Creek, DeWitt County

Setting. – Unlike several of the other reaches in this study, the geomorphological setting of the Kickapoo Ck. study reach is uniform for several kilometers up- and downstream. Kickapoo Ck. occupies a former glacial valley which was incised into till plain between the Bloomington morainal complex to the northeast through the Shelbyville moraine southwest and downstream of the study reach, and is included in the Bloomington Ridged Drift physiographic province (Fig. 1; Table I; Leighton et al. 1948). The stream wanders across the entire glacial valley floor with both gently meandering and straight reaches. Much of the study reach flows near the northern valley wall. One tributary enters about 300 m downstream of the head of the study reach (Fig. 25).

Uplands around the study reach comprise a blanket of Peoria Silt (loess) over loamy till of the Tiskilwa Formation, Delevan Member (till) (Hunt and Kempton 1977; Hansel and Johnson 1996). Tiskilwa Formation as well as Glasford Formation (silty till with sand and gravel lenses deposited during the Illinois Episode) crop out in the glacial valley walls (Hunt and Kempton 1977). Modern silty to sandy fluvial deposits in the floodplain are 2 to 6 m thick and overlie sands and gravels of Henry Formation (Hunt and Kempton 1977). The floodplain surrounding the study reach is silt loam to silty clay loam (Windhorn 1991); the modern stream could be incised locally into Henry Formation. At the USGS stream gauge (Table II) about 4.25 km downstream of the study reach, the stream bed comprises silt, sand, and fine gravel (USGS, writ. com. 2002). The low channel slope of 0.006 is moderate among streams in the study; the channel width of about 16 m is also moderate (Table IV).

Five aerial photographs from 1941 –1998 comprise the base data set (Table II). They were obtained at both base flow (1966) and high stages (1998; Appx. II). Image quality was excellent to fair. The floodplain south of the upper 2/3 of the study reach features prominent swales (Fig. 25). Because there are no scrollbars, these channels were probably either from crevasse splays or were abandoned by avulsion. Indeed, a new crevasse splay occurred between 1994 and 1998. Evidence of active sedimentation includes extensive sandy point bars (bright lensoid areas in 1941 photo), a small delta at the confluence of the intermittent tributary, and small crevasse splays.

The land surrounding the study reach was primarily in row-crop agriculture with no significant land use changes over the entire 57 years photographed (Fig. 25). Nearly all of the reach was farmed to the stream’s edge at some point since 1941, but the density of a narrow band of forest along the channel margin increased steadily until 1988, and then remained fairly constant. A drainage ditch constructed between 1966 and 1988 joined lower ¼ of the reach (“ditch”, 1998, Fig. 25). The intermittent tributary was straightened between 1966 and 1988.

Flow data began to be collected at Waynesville in February 1948, 7 years after the first photograph in the series was taken (Table II; Appx. II). This gap was ignored when interpreting the influence of stream power on planform change (Fig. 26). Assuming that flow during that period was average, stream power estimates for the interval 1941-1966 are about 30% too low. Flow records are thought to be good (USGS writ. com., 2002). Average annual discharge was below average among streams of the study, but average mean stream power and its variability were quite typical (Fig. 7).

Reach Evolution. – The Kickapoo Ck. study reach experienced continuous planform changes from meander migration, avulsion, and channelization (Fig. 27; Table VI). The largest single polygon was attributed to relocation of a 460 m meander in the lower half of the study reach between 1941 and 1966. In the 1941 image (Fig. 25) a north-south oriented linear feature is prominent. Its southern end turns abruptly towards the existing stream channel. This feature is interpreted as a new channel under construction. By 1966, the realigned reach had already begun to redevelop meanders. There was considerable change of features within this portion of the channel through the rest of the photo series. Some of this evolution was extensive enough to be characterized quantitatively as migration-postchannelization (Table VI). Several small avulsions and meander migrations immediately downstream of the confluence of the realigned section with the old channel were also attributed to post-channelization change as that portion of the channel responded to the new configuration. Planform evolution along the rest of the study reach continued primarily by downstream and lateral migration of meanders (Figs. 26, 27). Meander migration was equally distributed throughout the study reach, but occurrences were larger outside the channelized reach than inside it (Table VI).

McKee Creek, Brown County

Setting. – Leighton et al. (1948) included the McKee Ck. drainage basin in the Galesburg Plain, an upland covered by pre-Wisconsin Episode glacial sediments and deeply incised by modern (Fig. 1) tributaries to the Illinois R., which enter that valley through a steep bluff line at the eastern edge of the province. Phillips et al. (in progress), subdivided an additional province including McKee Ck., the Griggsville Plain, where uplands are less extensive, drainages more deeply incised, and drainage patterns more complex than on the rest of the Galesburg Plain. These features could be explained by relatively thin and thus less active ice cover during the Illinois Episode.

McKee Ck. drains west to east from its headwaters in Adams County. Low order tributary valleys are much steeper than the trunk valley; the channel slope at the study reach is only 0.0005, one of the shallowest of the study (Table IV). The valley alternately widens and narrows, possibly because the modern drainage cuts across the grain of bedrock valleys (Herzog et al. 1994). The study reach occurs where the valley opens abruptly from a narrow constriction. The reach meanders along the south valley wall, with abundant scroll bars and channel scars indicating a history of active channel migration (Fig. 28). Average channel width is 36 m, second widest of the study (Table IV). Two tributaries, Avery Ck. and an unnamed creek, join the study reach from the north valley wall. Their confluence with the main channel is at a high to oblique angle (Fig. 28).

The geology is known only from small-scale maps and as extrapolated from a quadrangle mapping project in Adams County. Upland drift is relatively thin and bedrock crops out on valley walls (M. Barnhardt, ISGS, pers. com. 2002; Piskin and Bergstrom 1975). The upland drift comprises Wisconsin Episode silt (loess) over Illinois and pre-Illinois Episode loamy diamicton (till) over shale, limestone, and sandstone bedrock (Berg and Kempton 1988). Valley fill sediments are less than 20 m thick, and may contain glacial outwash at the base of the pile. Floodplain soils adjacent the study reach are mainly silt loam over sandy loam (Berning 1988).

Eight aerial photographs for McKee Ck. were taken between 1938 and 1998. Image quality was generally good, but georeferencing error was the highest of the study (Appx. I). The high RMSE can be attributed to high relief and the occurrence of sparse yet clustered GCPs mainly in the uplands, not the floodplain. Over this sixty year period, the land remained completely in field crop with scattered farmsteads on the uplands (Fig. 28). Small (two-lane) roads skirt the valley walls to the west and north. There was periodic straightening and maintenance of the two tributaries.

McKee Ck. is ungauged. Flow was estimated from records of the La Moine R. at Colmar, modified by relative flood magnitude as determined from regional regression (see Methods). The estimated flow record for McKee Ck. extends from 1944 to present. Average annual discharge was applied to the period 1938-1944 for stream power calculation. Average annual discharge was unremarkable, and mean stream power was below average in magnitude and variability among streams in the study (Fig. 7). Airphotos were taken at both low and moderate flood (1969) stages (Appx. II).

Reach Evolution. – The study reach evolved continuously through both meander migration and avulsion (Fig. 29). No portion of the main channel reach was modified directly by people. In 1938, the study reach featured prominent meanders separated by straight sections (Fig. 28). Qualitative evidence of active sedimentation throughout the series includes large sandy point bars, bright areas on photographs indicating recent overbank flooding, and abundant and complex scroll bars on the floodplain. By 1998, only one of the original meanders remained; the other two had become relatively straight through avulsion and extensive lateral migration.

Most (60%) of the planform change on the McKee Ck. reach was caused by avulsion, although lateral and downstream migration of meanders accounted for about 40% (Table VI). Two large avulsive events occurred between 1938-1950 and 1979-1988 where the unnamed tributary feeds into McKee Ck. (arrows, Figs. 29 A and B, respectively). The first of these events left an oxbow lake that persisted through perhaps 1969. That area was then cleared and the tributary straightened by 1979. The second avulsion exploited a chute that had been repeatedly occupied and modified during flood conditions since before 1938. The remaining quantifiable avulsion was confined to 2 small polygons, which accounted for <1% of the class. The three polygons over the time series classified as chutes were also small contributors to the total areal change (Table VI). The two large avulsions occurred during periods of highest mean stream power, but no avulsion occurred during a similar mean power regime in 1957-1963 (Fig. 30).

Although lateral and downstream migration accounted for less planform change than avulsion, quantifiable evolution occurred more frequently (38 polygons vs. 4 for avulsion), and along the entire study reach (Table VI). Meander migration does not appear to be simply correlated with mean stream power, for similar migration rates occurred over a range of power conditions (Fig. 30). Indeed, high power between 1957 and 1963 was associated with minimum overall planform change.

Two tributaries enter McKee Ck. at obtuse angles along the study reach. The obtuse confluence – tributaries in gradually evolving systems typically join trunk streams at acute angles pointing downstream -- is further evidence of rapid planform change in the main channel. In addition, evolution of these tributaries attests to high dynamics in the entire river system. The lower 0.75 mi of the Avery Branch was straightened between 1950 and 1957, but by 1963 showed considerable growth of new meanders through lateral and downstream migration, and continued to evolve thereafter. The eastern, downstream unnamed tributary was also straightened between 1950 and 1957. Subsequent photographs showed cycles of redevelopment of a wandering planform, followed by restraightening.

South Fork Sangamon River, Sangamon County

Setting. – The study reach of the South Fork Sangamon R. (hereafter, “South Fork”) flows SW-NE in a largely rural area just east of Lake Springfield and the City of Springfield. The reach occurs in the Springfield Plain, a province of very low upland relief dissected by shallow valleys. The land surface is mainly underlain by glacial deposits of the Illinois Episode. The South Fork lies over a bedrock valley that has been filled with mainly glacifluvial and glacilacustrine sediments. In a boring 1.6 km downstream of the study reach were described 0.6 m of post-glacial alluvial silt over 4 m of Wisconsin Episode silty lacustrine and glacifluvial sediments (Bergstrom et al. 1976). Uplands are underlain by silty glacilacustrine sediments from the Illinois Episode with a blanket of silty loess (L. Follmer, ISGS, pers. com. 2002[1]). The bedrock surface is highly variable and bedrock crops out both up- and downstream of the study reach (Bergstrom et al. 1976).

The South Fork valley is irregular in width and may have been constructed by an older fluvial system with much wider meanders than the modern channel, which wanders gently across a narrow meander belt by contrast. Although this portion of the stream is mainly straight to wandering, the study reach comprises a single necking meander (Fig. 31). The confluence with the trunk Sangamon R. about 8 km downstream. The reach has the lowest slope of the study (0.0001), but a moderate width (19 m; Table IV).

Eight aerial photographs taken from 1939 to 1998 comprise the base data set (Table II). Image quality was generally good. Consistent and moderate RMSEs are partly attributable to occurrence of the reach near the middle of most images (Appx. I). Photographs were obtained during low to moderate flow conditions (Appx. II).

In 1939, the entire photographed region surrounding the study reach was undeveloped and primarily covered by row crop. By 1998, the region was still mainly agricultural, but areas to the east and west of the reach had become highly developed, as well as areas upstream and closer to Lake Springfield in the southwest. The city of Springfield to the west of the reach crept eastward with residential developments along the highway which crosses the South Fork just downstream of the study reach. Rochester to the west also exhibited growth towards the river. There was little evidence of channelization upstream or downstream of the study reach over the course of the aerial photography, although some channel maintenance was likely associated with a highway and rail crossing immediately downstream. Clearing and temporary road construction on the outside of the meander bend in 1968. Sewage lagoons were constructed on the right (eastern) floodplain before 1978.

Forest cover along the study reach fluctuated significantly. In 1938, dense forest on both sides of the reach was 20-200 m wide (Fig. 31). There were major clearings in 1950 and 1968 of tree stands on inside bends of the river. By 1998, however, a riparian border at least 20 m wide had reestablished along the entire length. Consistently overhanging canopy made width measurements difficult.

Flow data were collected at a station near Rochester since 1949 (Appx. II; Table. II). Flow records are considered fair because of ice jams in winter (USGS, writ. com. 2002). Average annual discharge was second highest of the study, but mean power was lowest (Fig. 7). Flow data collection began one decade after the first photograph of the series (Table II). Similar to other sites with flow record gaps, the average annual mean stream power for 1949-1950 was thus applied to the entire 1939-1950 photo permutation.

Reach Evolution. – This portion of the South Fork remained relatively unchanged in its overall shape over the 59 year study period (Fig. 31). The total planform change was one of the lowest of the study (Table VI). All change occurred along cutbanks and was classed as lateral and downstream migration (Fig. 32; Table VI). Although change polygons tended to be very narrow (in the cross-channel sense), they were accepted as significant because gaps remained after the buffer test (see Methods). The widest change polygon was at the location that was completely cleared with roads to the stream edge in 1968, however. That activity could have contributed to widening. There was no clear correlation between average annual planform change and mean stream power (Fig. 33).

Spoon River, Fulton County

Setting. – The Spoon R. is a major drainage of the Galesburg Plain, although the headwaters fall from the western (down-glacial) flank of the terminal Wisconsin Episode moraine of the Bloomington Ridged Plain (Fig. 1). The extensive, gently undulating surface of the uplands resulted from deposition of a relatively thin (generally less than 7.6 m though up to 30 m) blanket of glacial sediments during the Illinois and pre-Illinois glacial episodes (CTAP 1998a). Where streams have dissected the surface, valley walls are steep.

The main valley of the Spoon R. is typically wide and flat with steep walls, although it is somewhat constricted just upstream of the study reach. At the study reach the channel lies close to the western valley wall. The river has both tightly-looped and straight reaches. The study reach flows NE-SW with straight and meandering reaches, though it evolved considerably with time (below). The channel has a very low slope (0.0002) and is relatively wide (30 m; Table IV). Tributary valleys tend to be narrow and steep. Tributaries enter the study reach along the right (western) bank at the middle and foot (Cedar Ck.).

Uplands are comprised of diamictons (predominantly till) with lenses of sand and gravel, overlain by a loess blanket up to 3 m thick. Tills crop out in valley walls of the well-incised drainages (CTAP 1998a). The underlying bedrock surface is complex; a large pre-glacial bedrock valley underlies the modern Spoon R. (Horberg 1950). The valley is filled with glacifluvial and modern sediments. No soil survey exists for Fulton County, but in Knox County less than 1 Km upstream of the study reach, the floodplain consists of silty to sandy sediments deposited by the modern river (Windhorn 1986).

Nine aerial photographs from 1940 to 1999 comprise the base data set (Table II). Photos were taken during minimal to low flow conditions (Appx. II). Image quality was generally good.

The head of the study reach starts on the outskirts of the town of London Mills. The urban area did not change significantly between 1940 and 1999 (Fig. 34). Two bridges crossed the middle of the study reach. The upstream-most, a railroad crossing, was removed between 1969 and 1978. Improvements to the highway bridge (I116) appear to have begun by 1988. The upper half of the study reach down to just below the railroad bridge was channelized prior to 1940 as evident by its straightness, levees, and a kink in the planform where the channelized reach stops (Fig. 34). A small road anchors the right (western) bank in this channelized reach. There was only a thin forested riparian border along most of the banks, although patches of dense forest occurred along the right bank. Over time use of the floodplain varied from cropland to recreational. There are several prominent drainage ditches across the floodplain. A levee was constructed on the left (east) floodplain south of the highway between 1957 and 1963.

Flow data has been collected at a gauging station 500 m upstream of the head of the study reach since 1942 (Appx. II). Records are considered good (USGS, writ. com. 2002) and are complete. Spoon R. has the highest discharge of the study, but mean stream power is moderate (Fig. 7).

Reach Evolution. – In 1940, the study reach was straight to slightly sinuous. The upper third of the study reach channel was straightened prior to 1940 (Fig. 34). A recently active chute is evident at meander A in Figure 34, and an older large bar immediately downstream of the Il 116 bridge is covered by vegetation.

The planform of the channelized subreach did not change over the study period (Fig. 34). Several polygons were created by the automated intersection tool, however. These were attributed to digitization error and rejected by application of the “buffer test”. The width of the channel is sufficiently great that identification the channel centerline was difficult. This may be a situation where digitization of both banks rather than the stream thread may be valuable.

By contrast, the unchannelized reach exhibited considerable movement including formation and erosion of bars and meander migration (Fig. 35). There is a dramatic contrast between the straight planform of the river in 1940 and the sinuous planform of 1999 (Fig. 35). The net planform change due to downstream meander migration was 27,590 m², the second highest change by that mechanism in the study (Table VI). Between 1940-1950, the channel cut a new path beneath Il-116 and into the old bar. Meander A in Figures 34 and 35 became increasingly and steadily necked over the period 1940-1978 as the cutbank migrated downstream while the downstream limb remained stationary. Between 1978-1988, the reach downstream of meander A suddenly evolved from straight to sinuous and thus opening the meander. Downstream migration of the meanders continued through 1999.

There was an apparently simple correlation between mean stream power and planform change over the period 1940-1988 (Fig. 36). Both rates of changes decreased gradually from the beginning of the series to a minimum in 1963-1969. After 1969 mean stream power jumped to a higher, relatively steady level. Planform change also increased over 1969-1978, but then again fell gradually to a second minimum in 1988-1994.

Other dynamic behavior was not quantified in the analysis because it was below the photo resolution, but could be inferred from evolution of geomorphological features. The tributaries contributed significant sediment loads to the Spoon throughout this series as evident by episodic delta formation. Floodplain chutes both added sediment to the main channel by scouring the floodplain as well as eroded the banks at reentrant channels. The channel below the foot of the study reach appears to be even more active.

Sugar Creek, Iroquois County

Setting. – Sugar Ck. near Milford, IL, lies in the Kankakee Plain, a province Wisconsin Episode moraines and till plains that were modified by glacilacustrine sedimentation and large magnitude glacifluvial floods (Fig. 1; Leighton et al. 1948). The surface is gently undulatory with morainic islands, large fluvial bars and terraces, and lake plain. Headwaters of many Sugar Ck. tributaries begin on the up-ice (northern) flank of moraines of the Bloomington Ridged Plain to the south. The Sugar Ck. valley is an outwash channel headed in western Indiana (in which about half of the watershed lies) that dissects a low morainal bank at the study reach. The modern alluvial valley comprises silty clay loam to silt loam deposits between 6-9 m thick (Kiefer 1982; CTAP 1998b). These are incised into outwash sands and gravels of Wisconsin Episode less than 6 m thick, which in turn overlie diamictons, predominantly till, of Wisconsin and Illinois Episodes (Berg and Kempton 1988, CTAP 1998b). The valley fill may be thin; field notes describe bedrock cropping out or near the surface within the study reach (USGS, writ. com, 2002). Surrounding uplands are comprised of morainal diamictons and thin deposits of glacilacustrine silts and clays (Stiff 2000). A veneer of silty loess less than 2 m thick map cap the uplands (CTAP 1998b).

The study reach flows south to north along the left (western) wall of the outwash valley (Fig. 37). The channel substrate is sand and gravel with mud (USGS, writ. com, 2002). The reach slope of 0.0003 is the second lowest of the study. The channel width at the gauge is 36 m, but average width along the study reach was 20 m (Table IV). Prior to channelization, the reach was widely meandering (Fig. 37). Abandoned meander and chute scars were visible within and adjacent to the modern meander belt.

Seven aerial photographs for Sugar Ck. taken from 1940-1998 comprise the base data set (Table II). Taken at varying times of the year, these photos show the creek during a wide range of flows, from very dry in August, to flood stage in April (Appx. II). Land use change surrounding the Sugar Ck. study reach was low during the study period (Fig. 37). The area remained in extensive agricultural production across the floodplain and beyond. Most of the land made contiguous by channelization was put into crop production. Farming continued to the very edges of the stream channel, although a few patches remained forested throughout the study period. Dense canopy within the forested patches obscured the channel in several of the photographs, making it necessary to infer the channel centerline during digitization. Flow was monitored at a stream gauge at the head of the reach since 1948 (Table II). The records are continuous and considered of good quality (Appx. II; USGS, writ. com. 2002). Average annual discharge was moderate among streams in the study. By contrast, mean stream power was lowest in magnitude and variability (Fig. 7).

Reach Evolution. – The planform of the Sugar Ck. study reach was significantly modified by channelization throughout the study (Fig. 37; Table VI). By contrast, the creek directly up and downstream of the study reach was not channelized over the study period. In 1940, Sugar Ck. featured several wide meanders, with evidence of chutes and meander scars located sparsely along the floodplain. By 1973, Sugar Ck. was significantly channelized, with most of the work occurring between 1954 and 1967 (Fig. 38). Migration or remeandering of channelized threads is relatively small, and on the order of observed evolution along unchannelized reaches (Table VI).

There was no correlation between areal change and stream power (Fig. 39). Average annual mean stream power reached a minimum in 1957-1963, but was otherwise steady to slightly increasing. Extensive planform change between 1940-1973 was attributed to channelization. Meander migration after 1973 occurred at very low rates, though it did occur (Table VI).

West Branch DuPage River, DuPage County

Setting. – The West Branch DuPage R. (hereafter, “West Branch”) flows in a glacial outwash valley that dissects the Valparaiso Morainic System of the Wheaton Morainal Country physiographic province (Fig. 1; Leighton et al. 1948; Willman 1971). The Valparaiso Morainic System is a complex of overlapping end moraines deposited during repeated minor ice front fluctuations. The complex is onlapped by younger moraines further to the east. The province is rugged with relatively high relief (Phillips et al., in progress). Rivers in the province are highly constrained by the morainal and proglacial landforms. The outwash valley in which the West Branch lies was constructed by drainage from a younger ice front position than the one which constructed the Valparaiso system. Uplands are pebbly silty clay diamicton (till) with lenses of silt and sand (Hansel and Johnson 1996). The West Branch valley is filled with stratified glacifluvial sands and gravels. Modern floodplain deposits are mainly silt loam to silty clay loam (Mapes 1979). Channel substrate near the USGS gauging station is silt and gravel with sparse boulders (USGS, writ. com. 2002).

Five photographs from 1939-1998 comprise the base data set (Table II). The study reach occurred along the edge of the 1961 photo, however, causing unacceptable image shifts after georeferencing. It was thus not used for quantitative analysis. The image is valuable, however, for it shows the reach in flood conditions (daily mean discharge for 9/26/1961 was ~2.8 m³∙s^-1, c.f. monthly mean in Appx. II).

The study reach flows from the east turning to the southwest through the Hawk Hollow County Forest Preserve, which was established in 1976 (Fig. 40; D. Labrose, Du Page County Forest Preserves, pers. com. 2002). The modern alluvial valley is in the middle of the outwash valley. The channel planform is tightly meandering to straight. Chutes and abandoned meanders, including oxbow lakes, are apparent on the floodplain. Two small order tributaries draining the right (north) valley wall join the lower half of the study reach. The channel slope (0.001) is moderate among streams in the study (Table IV). In 1998, average reach width was 12 m, although that average is strongly increased by one short subreach 23 m wide.

In 1938, the area was almost completely rural (Fig. 40). Row crop farming occurred in the uplands, but the floodplain appeared to be open field with few trees. Farm roads crossed about 500 m from the head (“Stearns Road access”) and at the foot of the study reach. The bridge for the former was removed in the 1980’s (D. Labrose, DPCFP, pers. com. 2002). Cul de sac construction in 1967 heralded a wave of residential development that crested between 1988 and 1998. Establishment of the Forest Preserve kept development out of most the valley bottom, although a two-lane bridge connecting residential neighborhoods was constructed between 1993 and 1998. Tree cover was sparse until after 1988.

The USGS stream gauge is about 6 km downstream (Table II). Discharge from a sewage treatment plant constructed approximately 0.5 Km upstream of the reach between 1967 and 1988 has probably affected stream discharge magnitude and variability, but it is not known how much (USGS, writ. com. 2002). Certainly, base flow (based on minimum average monthly discharge) increased steadily from 1965 onwards (Appx. II).

Reach Evolution. – Although change in the shape and position of stream threads could be discerned between each photograph, most change was below the RMSE (Fig. 41). However, significant downstream meander migration and changes in channel sinuosity are evident in comparison of the 1939 to 1998 photos (Fig. 40; Table VI; Fig. 42). Meander loops were cutoff by avulsion between 1961-1967 and 1967-1988 (loops A and B, respectively, Figs. 40 and 41). A small path or drainage ditch confluence crossed meander A before and after the avulsion. We cannot determine if the feature had a role in the event. The event was classed as a natural avulsion and not channelization based upon the smoothness of the planform and consideration that relocating that subreach would have little benefit (Table VI).

The strangest changes occurred at the Stearns Road access crossing. In 1939, there was an obvious kink where the channel intersected the road, then flowed south for a few meters before passing through a culvert or similar structure (Fig. 40). During flood conditions in 1961, overbank flow funneled through a chute. The chute eventually became the primary channel between 1988-1998 contemporaneous with bridge removal, although the earlier channel continued to carry water. Even though there was no length of the reach that was channelized sensu stricto, the anabranching was classed as “chute/post-channelization” because of the probable influence of the bridge or bridge removal (Table VI), and may be a precursor to avulsion. Not quantified was significant widening of the subreach immediately upstream of the crossing from about 10 m to 25 m at the same time. A linear feature trending SW-NE across the floodplain and through the widening is apparent in the 1998 photo. Its origin could not be determined through conversation with local authority (D. Labrose, DPCFP, pers. com. 2002). Possibly it was a construction access road related to bridge removal or borrowing of sediment from the stream channel.

DISCUSSION

Representativeness of Study Reaches

Each of the stream reaches selected for study appeared to be anomalous in some way, but perhaps that is actually typical: all streams have unique features depending upon local geology and human practices. The 1.6 km reaches selected are not necessarily representative of the entire watershed. For example, several of the sites (e.g. Fig. 13; 41) featured anabranching reaches that are neither typical of the rest of the stream nor of streams in the Illinois River basin in general. At McKee Ck. high, quasi-continuous rates of planform change (Fig. 29; Table VI) contrast with relatively little planform change along reaches immediately up and downstream.

Landuse practices and landcover adjacent the study reach may differ significantly from conditions elsewhere in the watershed. At Farm Ck., South Fork, and West Branch study reaches, sewage treatment plants near the head of the reach discharge into the streams. The discharges affect baseflow and flow fluctuations by unknown amounts, but certainly contribute water load without a concomitant increase in sediment load. At other sites, e.g. DuPage R. and West Branch, floodplains were covered by wide expanses of riparian vegetation or open field, but immediately upstream row crops were farmed to the stream banks or dense residential developments grew. The direct effects of these factors are not known and interpretation in this study is partly limited by the study reach length and site selection, but local effects on sediment load, water load, bank stability, or all three are possible. The effects of these conditions are more likely to be discernable by increasing the area covered in the study to span a sub-watershed.

Farm Ck. has a distinctive setting in that it lies upstream of a flood control dam, within the 100 yr floodplain caused by dam impoundment, and has been highlymodified. There are, however, many stream reaches in the Illinois River basin that lie within dam impoundment areas, though generally smaller, or have been highly modified. For this reason it remains instructive to investigate the Farm Ck. reach.

Singularity in site configuration as described above should be expected. Useful conclusions about stream dynamics can be drawn from each reach despite the variability, however. The analysis of longer reaches would give greater context to the variability and may help identify effects.

If the assessment method used here were expanded across the Illinois River basin, estimation of flow as was performed for the McKee Ck. reach would be the rule rather than the exception because very few of the streams have been gauged. The methods we used to estimate flow were not very sophisticated but probably sufficient for our needs. In other situations, generation of a synthetic flow curve may be appropriate (c.f., Knapp and Myers 2001). The difficulty in using estimates rather than actual gauge data is that local effects such as landuse variability will be harder to assess in the flow data.

The net change over the period of record can be determined by comparison of only the earliest and latest photographs, typically about 60 years apart. Indeed, for Blackberry Ck., Spoon R. and West Branch the net change seemed to better represent stream dynamics than did the shorter permutations because planform change over the shorter intervals was less than the combined error of the photograph pairs. Net change analysis is not necessarily sufficient, however, because competing processes can cause similar planform evolution. Some of the processes can be identified by analyzing photos at higher temporal resolution, even if the analysis is qualitative.

Effects of Setting on Stream Power

One goal of this pilot project was to characterize the variability of stream dynamic regimes across the basin. Physiographic regions after Leighton et al. (1948) were used as a primary basis for stratification of the sites under the premise that each region is distinct topographically, which in turn is the summation of the geological materials and landform development (Leighton et al. 1948; Phillips et al., in progress). That is, there is some similarity within each region of watershed relief, configuration, valley slope, sediment type, and bedrock effects. These properties have a presumed fundamental effect on stream dynamics.

Geological aspects of physiographic region delineation mainly reflect distinctions in the uplands rather than the valleys (Leighton et al. 1948; Phillips et al., in progress). Thus, despite the initial study reach stratification, all reaches have a similar valley setting. The modern valley of each of the reaches lies in a former glacial outwash valley or outwash plain, and coarse glacifluvial sediment underlies varying thicknesses of post-glacial deposits. One expected effect is that abundant sand and gravel are potentially available for transport by the streams by contrast to the sediment that comprises most of the uplands. This may be coincidence of the decisions made in siting stream gauges and our need to use reaches with gauge data, and thus may not be characteristic of all streams in the Illinois River basin. Bedrock occurs near the surface or crops out near the Du Page R., South Fork, and Sugar C. valleys and may have some effects on stream dynamics as discussed incidentally below.

Climate, vegetation, and landcover are fundamental parameters not distinguished by physiographic province but they are captured in the flow data. Our use of average monthly discharge as a statistic smoothed out much of that variability, however, leaving mainly climatic effects and perhaps fundamental landscape changes. Such smoothing may be unavoidable in general application of the method because most watersheds are ungauged, so flow must be estimated as we did for McKee Ck. Regional regression after Curtis (1987) is based on regional climatic characteristics and was used to modify the flow data borrowed from a neighboring watershed. It would be inappropriate to assume that details in the original flow data were generally applicable to McKee Ck.

More northern sites tend to have more observations of ice effects in winter at gauging stations (USGS, writ. com. 2002). Although the occurrence of ice is noted the actual effects of ice on flow records or stream dynamics are not well known at these sites. Ice damming may cause overbank flooding and high stream power unleashed at breakup may be locally important for bank erosion and sediment transport.

The effects of competing parameters that control the energy available to move sediment are summed by determination of stream power (Rhoads 1995). Mean stream power, total power standardized by cross-sectional width, allows comparison between streams of variable size (discharge). Mean stream power varies across the Illinois River basin, but is not simply correlated to physiographic region (Fig. 7). For example, streams with both the very highest and very lowest mean power lie within the Kankakee Plain. Rather, there appear to be two energy regimes, the first with lower median mean power and lower variability, the second with higher median mean power and higher variability.

Median mean power may not be the most meaningful statistic because it mainly reflects regional climate, as mentioned above. After seminal work by Wolman and Miller (1960), many have assumed that channel morphology is controlled by bankfull flows that occur statistically at about 1- to 2- year intervals. These flows, the “effective discharge”, are sufficiently large that substantial volumes of sediment are moved, and sufficiently frequent that there is not enough time for channels to reduce through deposition to the size of the more frequent but smaller and less powerful flows that occur in the channel. That is, the effective discharge is that flow magnitude which transports most sediment over the long term. Nash (1994) criticized simple application of this model by demonstrated that a wide range of effective discharges exist as functions of regional and local controls. For the sake of argument, however, mean power equivalent to 1.05 and 2 yr flood discharges are plotted in Fig. 43. Similar to Fig. 7, there appear to be two clusters of data, one with low “effective mean power”, ω_e,and a second with much higher ω_e. These clusters, again, cross provincial boundaries.

What then most influences stream power? Mean power and its components of discharge, slope, and width are plotted in Fig. 44. There is no correlation between ω and any of its components. Even when Farm Ck. is dropped from the power vs. slope plot as an outlier, correlation remains low. The lack of any direct relationship clearly demonstrates that the interaction between the components of stream power is independent, and makes the important point that the geomorphology of whole watersheds must be studied to understand the dynamics of any given stream.

At seven of the reaches, mean stream power was higher at the end of the study than at the beginning (compare Figs. 11, 14, 17, 24, 36, 39, 42). The trends were not steady, but highly variable. Of these, mean power increased the most (by a factor of ~2) at the reaches which became the most developed, Blackberry Ck., DuPage R., Ferson Ck., Farm Ck., and West Branch. The South Fork watershed also experienced significant development but mean power decreased with time. By contrast, mean power was steady if variable at Kickapoo and McKee creeks, two reaches that showed very little change in their dominantly agricultural land use. Thus increasing mean stream power with time may possibly be explained by increases in surface runoff associated with increased imperviousness and point source discharges.

The inverse exponential relationship between slope and discharge in Fig. 44 is well-recognized in alluvial systems although its origin is complex. There is a significant configurational effect because discharge increases with drainage basin size, basin size grows with position downstream, and slope decreases downstream. However, bed particle size, an indicator of channel roughness and which also tends to decrease downstream has an important role (Hack 1957). Schumm and Kahn (1972) and Bledsoe and Watson (2001) showed bed particle size to be a predictor of channel pattern.

Variability in Dynamic Regime

We determined changes in stream planform between temporally adjacent photographs (short permutations) as well as by comparing the first and last photos in the series (net permutation). Each photograph suite included different numbers of photographs taken at different intervals (Table II). Not all photographs were useable because of poor image quality or very high georeferencing error. Therefore, we cannot compare change between any two streams at a particular time interval directly, but instead must compare trends over the entire period. In addition, some change was better captured by a long interval than by a shorter one.

Table VI shows the sum of quantified change for all the short permutations (“Aggregate Change”) as well as the net change. The aggregate change can be broken down into change unaffected by human activity (“non-channelized”) and change that occurred after some type of stream modification occurred (“post-channelized”). Planform evolution after stream modification cannot be identified in the net change because the time of modification cannot be known a priori. The net change does show, however, incremental changes that were too small, i.e., below the total image error, to be discerned in the short permutations.

Thus if the aggregate change is greater than the net change in Table VI, planform evolution was occurring within a narrow belt and old features were reworked into new ones. By contrast, where net change was greater than aggregate change – West Branch, Blackberry Ck., and Spoon R. are the prominent examples – much of the incremental change was below the photo RMSE of the short permutations.

Are some streams more dynamic than others? Unequivocally, yes, but the controls on the behavior are less clear. Total areal change excluding channelization suggests that three dynamic regimes exist (ΔA vs. ω, Fig. 44). The planform of McKee Ck. is highly dynamic, with moderate meander migration but extensive avulsion (Fig. 45). Moderately dynamic are Spoon R. and DuPage R., which both change mainly by meander migration (Fig. 44). All other streams exhibit a low dynamic regime. These classes can be discerned in the rankings of total areal change excluding channelization in Table VII. This analysis assumes, however, that areal change is uniform, when in fact we know that avulsion creates much greater planform change than the power required for channel creation, and meander migration may occur at different rates along any given subreach. Consider instead the number of instances of planform change (Table VI). The rankings by polygon frequency after lumping polygons from both unchannelized and post-channelized categories and from all dynamic classes are also given in Table VII. Instances from the net permutation were used for West Branch because that analysis characterized changes along that stream better (i.e., captured more change) than the short permutations. Three regimes can again be separated as high (McKee and DuPage), medium (Kickapoo, Blackberry, West Branch), and low (all others). Several study reaches changed position between the rankings, in particular Spoon R. and West Branch. Areal change and frequency of change have different implications for stream management, and it is important to note that channel planform evolution occurs everywhere.

Planform Evolution and Stream Power

Channel planform evolution was observed in the form of chute development, avulsion, downstream and lateral migration of meanders, and channelization. As a single class, channelization caused most of the total planform change followed by migration, avulsion, then chute development (Table VIII). Stream power is not a strong predictor of dynamic behavior in the Illinois River basin (Fig. 44). Correlation between planform change by avulsion and ω should not be expected because, unlike meander migration, the material between successive channel positions is not reworked. The only geomorphic work expended is in the cutting of the new channel. In this sense the relationship between geomorphic work and total planform change is similar to that for channelization, which Rhoads (1995) described as effective geomorphic work. That is, it is the amount of work that would have been expended had the same planform change evolved through migration. On the other hand, if ecosystems or human structures are to be considered, the areal planform change does quantify the potential impact of an avulsion avulsion in terms of the magnitude of spatial change in channel position on the landscape.

By contrast, correlation between planform evolution by migration and ω might be expected because all the material between successive channel positions is moved. Such a relationship is not obvious in Fig. 45, however. This may be partly attributed to the use of average monthly values. As discussed above, that statistic mainly reflects regional climate. Further research using dominant or effective discharge (Wolman and Miller 1960; Nash 1994) which should more accurately reflect subwatershed forcings may be fruitful.

With McKee Ck. dropped as an outlier, planform change by migration is replotted against the components of stream power in Fig. 46. Planform change is not correlated to width – and the correlation would be cryptic if it were – nor with slope. There is, however, a statistically significant linear correlation between ΔA_m and Q_AV (Pearson’s r² = 0.772). The importance of the correlation is not clear, for it can be largely attributed to the fact that larger streams have higher discharges and thus have greater erosion rates than small streams. Nonetheless the correlation suggests it may be possible to predict by Q_AV the dynamic regime of individual streams where migration is the dominant mode of change. This is contingent, however, upon developing a better understanding of why migration on McKee Ck. is uncorrelated with discharge, and whether similar behavior occurs in other streams.

The dominantly avulsive regime at McKee Ck. was unexpected, and accounted for more planform change there than did migration. Such a regime may be unique in the Illinois River basin, but might be exist in other creeks that drain the Griggsville Plain (Phillips et al., in progress). Smaller avulsions causing neck cutoff occurred along other meandering reaches and should be expected as one of the typical mechanisms of reach evolution in meandering regimes. Chutes may be precursors to or failed neck cutoffs.

Avulsive behavior is typical of systems with high sediment loads relative to stream power, unvegetated banks, or subsiding basins (rising base level). The stream systems may have high width to depth ratios, i.e., are relatively shallow and thus prone to out-of-bank flooding. It is possible that high sediment loads are most responsible for the active planform change characteristic of this reach. This prediction can be examined by investigating texture of surficial materials, landuse and landcover evolution upstream, channel dynamics upstream, as well as instream investigations. Richards et al. (1993) suggested that avulsing rivers cannot be understood solely by examining in-channel geomorphology, but must be considered as systems. They do not fit into the straight-meandering-braided continuum of planform developed by Leopold and Wolman (1957), which can be well explained available stream power and sediment resistance to transport. This condition emphasizes the need for studying the geomorphology of entire basins rather than focusing on channel conditions in geomorphic assessments for stream restoration.

There is no correlation between areal change and stream power in McKee Ck. (Fig. 30) because avulsion was the dominant dynamic process. Valley expansion at the McKee study reach is interpreted as the immediate cause of the avulsive regime because flow has freedom to expand and drop bedload. Examination of a few airphotos, surface, and bedrock topographic maps from the region suggest this is because there are alternating constrictions and expansions of the valley, possibly by bedrock control. Avulsive regimes similar to the study reach may also occur along valley expansion elsewhere in the watershed.

The stream reaches studied were dominantly single-thread with straight (or wandering) to meandering planforms. As mentioned, however, part of the DuPage R. reach was anabranching with a stable form (Fig. 12), and although there are other anabranching subreaches of similar extent elsewhere on the river, these features are rare. The origin of the pattern is not known. The study reach appears to be in an aggradational portion of the DuPage river based upon mapping of Cahokia Formation (Willman and 1971). The elevation range of that subreach is exactly the same as flood elevations when a small dam downstream of the stream gauge is overtopped, suggesting temporary base-level fluctuation create a depositional regime. However, the pattern was shown on the 15 minute Joliet topographic quadrangle of 1923. The dam presumably post-dates that map because there were no cultural features noted in the area. Local aggradation could be attributed to the stream being free to drop sediment load after passing over a reach where bedrock crops out in or is near the channel bottom (Willman and Linebeck 1970). Alternatively, the reach could have been the site of a beaver dam, whose backwater effects persisted through time (c.f. Prince 1997).

Short anabranching subreaches also evolved at other sites by chute development (e.g. Kickapoo Ck., Fig. 27), stabilization of medial bars (Ferson Ck., Fig. 22), and possible local increases in sediment load by bank instability (West Branch, Fig. 41). Human activity, including ditching and bridge construction, may have been directly or indirectly responsible in several of these cases, but, except for West Branch, cannot be unequivocally linked.

Portions of four of the study reaches were channelized during the period of photographic record (Fig. 45). At Farm, Kickapoo, and Sugar creeks more effective geomorphic work was expended by channelization than by the other mechanisms. After channelization, each of the channelized subreaches continued to evolve (upper right, Fig. 45; Table VI). At Sugar Ck. there was an order of magnitude more migration along the channelized subreaches than along unchannelized subreaches, although the change values do not account for the relative subreach lengths and the amount of change was small relative to the other study sites. The quantified post-channelization change along the channelized reach at Kickapoo Ck. was also small, but there was an obvious increase in channel sinuosity with time that was not captured by the method, although higher quality images and better image georeferencing may have helped. Evolving sinuosity can be seen by comparing the 1998 channel thread with the 1941 channel excavation in the left image of Figure 25. As well, braiding and avulsion immediately downstream of the foot of the channelized reach was likely caused by a combination of hydrological effects of the channelization and discharge from a newly-constructed drainage ditch. Channelization of portions of Blackberry Ck. occurred during the last photograph permutation (1993-1998), so no post-channelization evolution could be identified. Chute development as a precursor to avulsion and width changes at West Branch were clearly associated with bridge removal (Fig. 40). Although the total areal change was small – about one half of the net change of the study reach by other mechanisms, the response to stream modification was immediate and dramatic (Table VI; Fig. 45).

The upper subreach at Spoon R. was channelized and unchanging through the series (Fig. 35). It is not known how much of the steadiness can be attributed to maintenance. Removal of a bridge across the reach may have been associated with medial bar formation, i.e., local increase in bedload transport. By contrast, the lower subreach exhibited considerable change. Local increases in stream power from increased slope caused by the channelization may play a role. This hypothesis could be tested by analyzing for continuity of the regime or waning of the regime with distance downstream.

Planform Change and Incision

Incision is predicted in models of the evolution of “disturbed” stream channels, such as Simon (1989). Channel widening is a consequence. Although incision cannot be interpreted directly from photos, there is potential that trends in stream width could be. Measurements of channel width for this study show no consistent trends, though the measurement error is quite large relative to the width. It is likely, however, that the reaches showing avulsion are not incising because avulsion occurs in aggrading systems. The study reach at McKee Ck. and similar floodplain widenings throughout its watershed might be considered as a local base-levels, although that remains to be proven with further research. That reach has been aggrading, not incising, through the period of record. Likewise, evidence of active bedload movement and crevasse splay at Kickapoo Ck. also suggest aggradation there.

Potential and Limitations of The Method

The method gives a quantitative understanding of stream change over the past 60 years with limited investment of resources. The GIS database for 16 km of reach was compiled and digitized, including calculation of change polygons less than 20 person-weeks. Most of that time was consumed by the scanning and georeferencing. With suggested method improvements, especially the use of Orthobase to orthorectify images (see below), we can expect to considerably reduce this base investment. Analysis of the geological setting and interpretation of change is dependent upon data availability, planform complexity, and the amount of change. The greatest value of method for initial watershed assessments is that it quantifies how a given stream changes in a historical perspective. The analysis identifies dominant processes and geological targets for more intensive field study, reveals the variability of stream planform dynamics, and demonstrates that total geomorphology of the system needs to be evaluated to understand stream behavior.

This pilot study focused on how far the analysis could go with available data. The geological setting was developed only generally because of limited data. In all cases except for Ferson Ck., geologic maps were only available at scales of 1:100,000 or smaller. Soil surveys typically give reasonably detailed assessments (~1:16,000) of floodplain materials and their properties, but additional interpretation is required to assess the geological history of the floodplain. As well, only small scale soil surveys were available at several sites. The only bed substrate information available was from stream gauge records (USGS, writ. com.) and was mainly anecdotal. Most needed are geological maps at the 1:24,000 scale for establishing the geologic setting, especially the thickness of post-glacial valley fill and depths to older sediments or bedrock. Such maps should be supplemented by focused higher resolution field studies of floodplain and channel sedimentology and river geomorphology.

Channel incision cannot be directly assessed from airphotos. Trends of increasing channel width with time could possibly be surrogate for assessing incision following channel evolution models (Simon 1989), however. We found no such trends, but georeferencing error was quite high relative to channel width for many of the images in this study. Width analysis may be more definitive with expected error reduction through use of crisper source images and georeferencing methods (below).

Improvement of Methods

Our predominantly manual methods worked sufficiently well for a small study such as this one. To examine an entire river or subwatershed would require compiling many more georeferenced digital images. Increased production would lower the cost of developing the GIS database. Although our georeferencing method proved adequate for quantification of dominant evolutionary behaviors, more accurate quantification of change and improvement of interpretations could be achieved with lower georeferencing error RMSE and crisper images. These steps are recommended:

· Scans of airphotos taken by USDA and contractors since 1955 can be obtained directly from that agency for $10/image, although lead times are approximately six months at the time of writing (USDA, 2002). The scans are of the original negatives and so should have highest possible image quality and the potential for minimum georeferencing error. A database of scans of the earliest aerial photography (~1938-1939) is being compiled by the Historical Airphoto Archive (Luman 2002) on a county by county basis. These images are made with a high resolution scanner of original prints in good condition.

· Images from the mid-1940’s to early 1950’s still would have to be scanned from photograph archives. Although the scanner that we used was sufficient for the purposes of the study (see methods), a professional high-end scanner would yield crisper images as has been the experience with the HAA (D. Luman, writ. com. 2002). As well, it would probably be more efficient for a large project to use an outside source.

· The use of ERDAS Orthobase software to produce orthorectified images would substantially increase image accuracy over the polynomial georeferencing method used here. Flight data is available for some series of airphotos obtained after 1973 (USDA, 2002). Georeferencing error can be kept below 1 m for images with flight data (D. Szafoni, INHS, pers. com. 2002). Relatively low error can also be obtained even without flight data, however. Experience in compiling the Historical Airphoto Archive has shown that errors less than 3 m are typical (Lund 2001). The software may have the additional advantage of increasing production.

· It would be valuable to recast the automated intersection tool to preserve topology. This would permit analysis of reach length involved in any event.

· Our use of average monthly discharge and consequent use of average monthly stream power reduced the sensitivity of the analysis to local flow effects. When sufficient flow and sediment monitoring data are available to determine effective discharge (Nash 1994), parameters such as the flow that exceeds effective discharge may be useful. We recommend research into effective discharge and its variability across the Illinois River basin, and whether or not a stronger correlation than we determined exists between effective discharge and meander migration.

CONCLUSIONS

· The airphoto analysis method is suitable for baseline assessments of stream dynamics in watersheds over the period of record.

· The method gives essential historical perspective on channel evolution.

· The method helps focus subsequent assessment by identifying dominant dynamical behaviors and variability in rates of planform change along reaches.

· Several specific recommendations were made that should significantly improve the productivity and accurateness of the method.

· Five of the study reaches exhibited change in a low dynamic range; four were moderately dynamic; and one was extremely dynamic.

· Most planform change occurred by channelization followed by meander migration. Avulsion also occurred and was the dominant mode at McKee Ck.

· Planform evolution continues after channelization, and some changes may be caused by channelization or other human activities. Planform change after channelization was observed to both decelerate and accelerate.

· Although several of the reaches exhibited the progressive increase in rate of change expected for unstable streams,

· There is no single factor controlling dynamic regime. Analysis of geomorphology of entire watersheds, not just the channel, is essential for fully understanding stream dynamics.

REFERENCES

Bledsoe, B.P, and Watson, C.C. 2001. Logistic analysis of channel pattern thresholds: meandering, braiding, and incising. Geomorphology 38: 281-300.

Berg, R.C., and Kempton, J.P. 1988. Stack-unit mapping of geologic materials in Illinois to a depth of 15 meters. C542, Illinois State Geological Survey, Champaign, IL, 23 p.

Bergstrom, R.E., Piskin, K., and Follmer, L.R. 1976. Geology for Planning in the Springfield-Decatur Region, Illinois. C497, Illinois State Geological Survey, Champaign, IL, 7 maps.

Berning, G.V. 1988. Soil Survey of Brown County, Illinois. Illinois Agricultural Experiment Station. 163 p, 45 maps.

CTAP. 1998a. Spoon River Area Assessment, Vol. 1 (Geology). Critical Trends Assessment Project, Illinois Department of Natural Resources, Springfield, IL, 92 p.

CTAP. 1998b. Kankakee River Area Assessment, Vol. 2 (Geology). Critical Trends Assessment Project, Illinois Department of Natural Resources, Springfield, IL, 83 p.

Curry, B.B., Grimley, D.A., and Stravers, J.A. 1999. Guidebook K28, Illinois State Geological Survey, Champaign, Illinois, 40 p.

Curtis, G.W. 1987. Technique for estimating flood-peak discharges and frequencies on rural streams in Illinois. U.S. Geological Survey, Water-Resources Investigations Report 87-4207, 79 p.

EPA. May 2002. Envirofacts Database Warehouse, Environmental Protection Agency. http://www.epa.gov/enviro/

Grimley, D.A., Curry, B.B. in press. Surficial Geology of the Geneva Quadrangle (map). Illinois State Geological Survey, Champaign, IL.

Hack, J.T. 1957. Studies of longitudinal stream profiles in Virginia and Maryland. U.S. Geological Survey Prof. Paper 29-B: 45-97.

Hansel, A.K., and Johnson, W.H. 1996. Wedron and Mason Groups: Lithostratigraphic reclassification of deposits on the Wisconsin Episode, Lake Michigan Lobe area. Illinois State Geological Survey Bull. 104, Champaign, IL, 166p.

Herzog, B.L., Wilson, S.D., Larson, D.R., Smith, E.C., Larson, T.H., and Greenslate, M.L. 1994. Hydrogeology and Groundwater Availability in Southwest McLean and Southeast Tazewell Counties. Part 1: Aquifer Characterization. Coop. Groundwater Rept 17, ISGS, Champaign, IL, 70p.

Horberg, L. 1950. Preglacial Erosion Surfaces in Illinois. Journal of Geology, v. 54, no. 3, p. 179-192.

Hunt, C.S., and Kempton, John, P. 1977. Geology for Planning in De Witt County. EGN 83, Illinois State Geological Survey, Champaign, IL, 42 p, maps.

Hunter, R.E. 1966. Sand and gravel resources of Tazewell County, Illinois. C399, Illinois State Geological Survey, Champaign, 22 p.

ISWS. 1996, One-Hundred and Five-Hundred Year Floodzones for Unincorporated Areas in Illinois by County (map). Illinois State Water Survey, Champaign, IL.

Johnstone, P. in review. Surficial Geology of Tazewell County (map). Illinois State Geological Survey, Champaign, IL.

Kiefer, Linus M. 1982. Soil Survey of Iroquois County, Illinois. Illinois Agricultural Experiment Station. 184 p, 162 maps.

Knapp, H. Vernon, and Michael W. Myers. 2001, Streamflow Assessment Model for the Little Wabash River Watershed : Hydrologic Analysis. ISWS CR 2001-14, Illinois State Water Survey, Champaign, IL, 106 p.

Larsen, Jean I. 1976. Geology for Planning in Will County (map). OFS 1976-6, Illinois State Geological Survey, Champaign, IL.

Leighton, M.M., Ekblaw, E., and Horberg, Leland. 1948. Physiographic Divisions of Illinois. ISGS Report of Investigations No. 129, Illinois State Geological Survey, Champaign, IL, 19 p, 1 map.

Leopold, L.B., and Wolman, M.G. 1957. River Channel patterns; braided, meandering and straight. U.S.G.S. Prof. Paper 282-B.

Luman, D.E. 2002. Preserving the Past: The Development of a Digital Archive of Historical Aerial Photography, paper presented at the ACSM-ASPRS Annual Conf., April 24, 2002, Washington, D.C.

Lund, D. 2001. Creating Historical Digital Orthophotography from a Print Archive (Abs.). Fall Mtg., Illinois Geog. Inf. Sys. Assoc., Lisle, IL

Mapes, D.R. 1979. Soil Survey of Du Page and part of Cook Counties, Illinois. Illinois Agricultural Experiment Station. 217 p, 126 maps.

Nash, D.B. 1994. Effective sediment-transporting discharge from magnitude-frequency analysis. Journal of Geology 102:79-95.

Phillips, A.C., Smith, L., and Mushrush, M. in progress. Landforms of the Illinois River Basin. Contract Report, Illinois State Geological Survey, Champaign, IL.

Piskin and Bergstrom. 1975. Glacial Drift in Illinois. Circular C490, Illinois State Geological Survey, Champaign, IL. 2 maps.

Prince, H. 1997. Wetlands of the American Midwest: A Historical Geography of Changing Attitudes. The University of Chicago Press, Chicago, 395 p.

Rhoads, B.L. 1995. Stream Power: A Unifying Theme for Urban Fluvial Geomorphology. In Stormwater Runoff and Receiving Systems, CRC Press, Inc.: 65-75.

Rhoads, B.L and Urban, M.A. 1997. A Human-induced Geomorphic Change in Low-energy Agricultural Stream: An Example from East-central Illinois. In Management of Landscapes Disturbed by Channel Incision, Wang, S.S.Y., Langendoen, E.J., and Shields, F.D., (eds.), University of Missippi, Oxford, MS. pp. 968-973.

Richards, K., Shobhit, C., and Friend, P. 1993. Avulsive channel systems: characteristics and examples. In Best, J.L. and Bristow, C.S. (eds.), Braided Rivers, Geological Society Spec. Pub. 75:195-203.

Schumm, S.A., and Khan, H.R. 1972. Experimental study of channel patterns. Geol. Soc. America Bull. 83: 1755-1770.

Simon, A. 1989. A model of channel response in disturbed alluvial channels. Earth Surf. Proc. Land 14:11-26.

State of Illinois 1997. Intergrate Management Plan for the Illinois River Watershed, Technical Report. Office of Lt. Governor, Springfield, IL, 61 p.

Stiff, B.J. 2000. Surficial Deposits of Illinois (map). ISGS, Champaign, IL.

Teater, W.M. 1996. Soil Survey of Tazewell County, IL. Illinois Agricultural Experiment Station. 210 p, 100 maps.

UIUC ACES. February 2002. Illinois Watershed Management Clearinghouse, University of Illinois, College of Agricultural Consumer and Environmental Sciences. http://web.aces.uiuc.edu/watershed/maps.html

Urban, Michael A. 2000. Conceptualizing anthropogenic change on the upper Embarras River. University of Illinois dissertation, Champaign, IL, 558 p.

USDA. June 2002. Aerial Photgraphy Field Office, United States Department of Agriculture. http://www.apfo.usda.gov

USGS. December 2001. NWISWeb Data for Illinois, United States Geological Survey. http://waterdata.usgs.gov/il/nwis/nwis

U.S. Army Corps of Engineers. In preparation. Illinois River Ecosystem Restoration Needs Assessment. U.S. Army Corps of Engineers, Rock Island District, Rock Island, Illinois.

Willman, H.B. 1971. Summary of the Geology of the Chicago Area. C460, Illinois State Geological Survey, C460. Urbana, IL, 77 p, 1 map.

Windhorn, Roger D. 1986. Soil Survey of Knox County, Illinois. Illinois Agricultural Experiment Station. 161 p, 105 maps.

Windhorn, Roger D. 1991. Soil Survey of De Witt County, Illinois. Illinois Agricultural Experiment Station. 133 p, 52 maps.

Wolman, W.R., and Miller, J.P. 1960. Magnitude and frequency of forces in geomorphic process. Jour. Geology 68: 54-74.

TABLE CAPTIONS

Table I. Geographic Setting of Study Reaches

Table II. Base Data Availability Site, years of record for photos and flow

Table III. Rules for Dynamic Classification

Table IV. Study Reach Geometry

Table V. Summary of Reach Change

Table VI. Dynamic Regime of Study Reaches by Rank

Table VII. Total Planform Change Attributed to Each Dynamic Class

FIGURE CAPTIONS

Figure 1. Location of study reaches by physiographic province. Province boundaries after Phillips et al. (in progress).

Figure 2. McKee Creek study reach, 1939, prior to stream thread digitization. Image at original scale. River miles 8 and 9 (upstream of Brown County line) identified in yellow.

Figure 3. Digitization of stream thread. Screen shot of McKee Creek study reach at River Mile 9, Brown County, zoomed to 3 times original scale for digitization. Channel centerline in red.

Figure 4. Paired stream threads prior to intersection. Screen shot of ArcView interface. Polygons A and B are the RMSE-buffered stream threads; polygons C are the non-overlapping donut-hole features that represent significant areal change. Black arrow points to tool button.

Figure 5. Output from Automated Intersection Tool. Output includes polygon theme and attribute table. Yellow indicates areas of statistically significant planform change.

Figure 6. Airphotos of Blackberry Creek study reach, 1939 and 1998.

Figure 7. Box plots of average monthly mean power for each study reach. Study reaches are grouped by physiographic province. Box plots show the distribution of values. The ends of the box (hinges) are at the 1^st and 3^rd quartiles; thus fifty percent of the values, lie within each box or the hinge spread. The median is indicated by a horizontal bar. The whiskers (vertical lines beyond hinges) indicate the range of values within 1.5 hinge spreads of the hinges, the inner fence. Outside values lie within 3 times the hinge spread and beyond the inner fence, and are indicated by asterisks. Far outside values lie beyond the outside values and are indicated by empty circles.

Figure 8. Meander maps, Blackberry Creek, 1939-1998. Stream threads buffered to average stream width at gauge (36 m). The red arrow shows the direction of meander migration.

Figure 9. Channel Evolution at Blackberry Creek. Meander migrates downstream between 1939 (fat) and 1998 (skinny). Note increased necking of subsequent meander. Chute (1998, arrow) existed in earlier photos but increased substantially in width after 1993. Further development may lead to neck cutoff by avulsion. Other meanders have migrated downstream.

Figure 10. Chute development at Blackberry Creek. In 1967, chute scar on floodplain (1967, arrow) was evident by dark-toned path. An embayment in the creek occured at downstream confluence. Anabranching formed by 1988 as chute became primary channel, leaving old channel as secondary.

Figure 11. Average annual mean stream power (ω) and planform change (ΔA), Blackberry Creek, 1939-1998.

Figure 12. Airphotos of Du Page River study reach, 1939 and 1998.

Figure 13. Meander maps, Du Page River, 1939-1998. Stream threads buffered to average stream width at gauge (41 m).

Figure 14. Average annual mean stream power (ω) and planform change (ΔA), Du Page River, 1939-1998.

Figure 15. Channel Evolution on Du Page River. An emergent bar forms a chute between 1939 and 1967 (1967, red arrow). Downstream migration of an adjacent pointbar and cutbank make a meander increasingly necked between 1939 and 1998 (1998, red arrow). Note several crevasse splay features (white arrows) on floodplain to NW. Scale ~ 1:6,200.

Figure 16. The Farm Creek study reach in 1939 and 1998.

Figure 17. Average annual mean stream power (ω) and planform change (ΔA), Farm Creek, 1939-1998.

Figure 18. Meander maps of Farm Creek study reach, 1939-1998. Stream threads buffered to average stream width at gauge (16 m).

Figure 19. Farm Creek, 1939-1951. Avulsion between 1939 and 1951 (arrow) may have been influenced by hardening or other bank maintenance to protect railroad embankment. Note increase in channel width and bright patches in channel suggesting active sediment transport. Flow was probably lower in 1951 than in 1939, revealing more in-channel features. Scale = 1:4,000.

Figure 20. Farm Creek, 1969-1994. Redevelopment of meandering habit after channel had been completely modified before 1963. Note increasing sinuosity of channel thread with time. Large meanders cutoff in 1939-1951 and 1963-1969 remain visible and periodically flooded at high water. Significant lateral migration upstream of right-of-way (arrow 1), and avulsion at downstream end (arrow 2). Scale 1:18,000

Figure 21. Ferson Creek study reach, 1939 and 1999.

Figure 22. Meander map of Ferson Creek, 1939-1998. Stream threads buffered to average stream width at gauge (15 m). Arrows indicate progression of some meanders by downstream migration.

Figure 23. Geomorphic processes at Ferson Creek study reach. Stream deflection by sediment deposition at tributary mouth, upper study reach, 1961 (left). Stabilization of medial bar in 1967 (right). By 1975, an anastomosed planform had developed around the feature. Scale 1:2,500.

Figure 24. Average annual mean stream power (ω) and planform change (ΔA), Ferson Creek, 1939-1998.

Figure 25. Kickapoo Creek study reach, 1941 and 1998. “Relocation” on 1998 image shows realignment of natural channel between 1941 and 1967. “Crevasse splay” formed between 1994 and 1998 joins swale of pre-1941 channel scar.

Figure 26. Average annual mean stream power (ω) and planform change (ΔA), Kickapoo Creek, 1941-1998.

Figure 27. Meander map of Kickapoo Creek, 1941-1998. Stream threads buffered to average stream width at gauge (16 m).

Figure 28. McKee Creek Study Reach, 1938 and 1998. Avery Branch confluence in northwest.

Figure 29. Meander map of McKee Creek, 1938-1998. Stream threads buffered to average stream width at most upstream point of study reach (36 m). Arrows identify major avulsive changes.

Figure 30. Average annual mean stream power (ω) and planform change (ΔA), McKee Creek, 1938-1998.

Figure 31. South Fork Sangamon River Study Reach, 1939 and 1998.

Figure 32. Meander map, South Fork Sangamon River. A. 1939-1978; B. 1979-1998. Stream threads buffered to average stream width at most upstream point of study reach (19 m).

Figure 33. Average annual mean stream power (ω) and planform change (ΔA), South Fork Sangamon River, 1939-1998.

Figure 34. Airphotos of Spoon River study reach, 1940 and 1999. Evolution of Meander A is discussed in text.

Figure 35. Meander map of Spoon River study reach, 1940-1999. Stream threads buffered to average stream width at most upstream point of study reach (30 m). Evolution of Meander A is discussed in text.

Figure 36. Average annual mean stream power (ω) and planform change (ΔA), Spoon River, 1940-1999.

Figure 37. Sugar Creek study reach, 1940 and 1998.

Figure 38. Meander map of Sugar Creek study reach, 1940-1998. Stream threads buffered to average stream width at most upstream point of study reach (36 m).

Figure 39. Average annual mean stream power (ω) and planform change (ΔA), Sugar Creek, 1940-1998.

Figure 40. West Branch Du Page River, 1939 and 1998.

Figure 41. Meander map of West Branch Du Page River study reach, 1939-1998. Stream threads buffered to average stream width at most upstream point of study reach (10 m). A and B indicate meander neck cutoffs referred to in text.

Figure 42. Average annual mean stream power (ω) and planform change (ΔA), West Branch Du Page River, 1939-1998.

Figure 43. Mean power equivalent to discharges for 1.05- and 2 yr recurrence intervals. The 1.05 yr recurrence interval cannot be determined for McKee Creek because discharge was estimated by regional regression (Curtis 1987).

Figure 44. The components of stream power. ΔA_All is net change for all dynamic classes, S is channel slope, W is channel width, Q¯ is average annual discharge, and ω is average annual stream power.

Figure 45. Planform change along unchannelized reaches by dynamic class versus annual mean power. Blackberry Creek (B), Spoon River (Sp), and West Branch (W) values are net change from 1938-1998; others are aggregate change (Table V). “Post-channelization” includes avulsion, chute, and lateral/downstream migration classes. There is a break in scale for the much larger aggregate McKee Creek values; also note different scales on channelization and post-channelization axes.

Figure 46. Planform change by migration, ΔA_m, and the components of mean stream power. The point for McKee Ck. is dropped as an outlier in the lower graph. W is channel width, S is channel slope, and Q¯ is average monthly discharge.

[1] The uplands west of the South Fork valley are mapped as underlain by silty diamicton (Glasford Formation) in Bergstrom et al. (1976) and statewide maps, but L. Follmer (pers. com. 2002) now believes they are underlain by stratified silts and sands (Tenneriffe Silt).