The ups and downs of levees: GPS–based change detection, Middle Mississippi River, USA Andrew Flor*, Nicholas Pinter, and Jonathan W.F. Remo Department of Geology, Southern Illinois University Carbondale, Carbondale, Illinois 62901, USA ABSTRACT Changes in levee-crest elevations were measured along 328 km of Mississippi River levees between St. Louis, Missouri, and Cairo, Illinois, in 1998 and 2007. We also compared 1998 and 2007 survey data with 50 yr flood elevation profiles to assess levels of levee protection. Levee heights, stability, and safety are nationwide concerns, especially since Hurricane Katrina and floodwall failures in New Orleans, Louisiana, in 2005. The majority of surveyed levees were stable at the centimeter to decimeter level during the 9 yr measurement interval. Change was measured in other areas, including increases of as much as 1.49 m and decreases of as much as 1.26 m between 1998 and 2007. The increases corresponded to local crevasse repairs, smallscale road maintenance, and larger levee-raising projects. Decreases in levee elevations are interpreted as small-scale surface erosion and larger compaction or subsidence of levee and/or foundation materials. Levee crests in 2007 were locally as much as 2.0 m below the 50 yr floodgrade elevations. Levee degradation reduces protection levels for floodplain residents, often without easily visible symptoms. Intentional levee raising without necessary studies, engineering, or permitting has a range of potential negative impacts including both a false sense of security and the potential for exporting flood risk to neighboring levee communities. Regional surveying of levee elevations and change over time can provide a preliminary tool for assessing levee conditions, stability, and compliance with levee regulations. INTRODUCTION Levee failures, including overtopping of levees and structural failures, have occurred along the Mississippi River corridor throughout the ~300 yr history of levees on the system. The catastrophic failure of levees on the Lower Mississippi River during the 1927 flood reshaped the social fabric of the United States (Barry, 1997) and the national strategy for flood-hazard protection (Morgan, 1971). The 1993 flood also caused widespread levee failures. More recently, levee failures along the Upper Mississippi River in 2008 (Rogers, 2009) and breaches in the New Orleans, Louisiana, flood-protection system in 2005 have made levee safety a renewed national priority. In 2006, one year after Hurricane Katrina, the U.S. Army Corps of Engineers initiated the National Levee Safety Program, compiling a national database that includes levee locations, ownerships, and inspection ratings (U.S. Army Corps of Engineers, 2006b). The risk assessment of these levees, and political responses to these assessments, are ongoing. Levee Degradation Degradation of a levee can be caused by dewatering, oxidation, or compaction of organic-rich soils (Dunbar and Britsch, 2008; Rogers, 2009). These processes can affect the fill composing the levee and the substrate on which a levee is *Current address: Florida Geological Survey, Gunter Building MS 720, 903 W. Tennessee Street, Tallahassee, Florida 32304-7700; E-mail: andrewflor @gmail.com.

built. Mississippi River levees constructed during the nineteenth and early twentieth centuries often incorporated unconsolidated fill and sometimes organic debris (e.g., tree stumps or trunks) that led to volume loss over time, reduced strength, and failure when saturated (Mills, 1978). In the Sacramento–San Joaquin Delta of California, land drainage and oxidation of wetland soils have caused subsidence of 3–7 cm/yr, and some land now is 6–8 m below adjacent river levels and at risk of catastrophic flooding should levees fail (Mount and Twiss, 2005). Regional land subsidence combined with localized subsurface degradation contributed to the failures that inundated New Orleans during and after Hurricane Katrina in 2005 (Dunbar and Britsch, 2008). In this study we set out to directly measure the extent of levee degradation and subsidence as well as increases in levee heights along a portion of the Mississippi River over a 9 yr time span. These 328 km of levees are mostly agricultural and mostly built to the 50 yr level (the 2% annual probability event), but also protect a number of smaller towns ranging in population from 9 (Kaskaskia, the smallest incorporated community in Illinois) to 3632 (Cairo, Illinois). METHODS The Middle Mississippi River is the portion of the Mississippi from the Missouri River downstream to the Ohio River at Cairo, Illinois (Fig. 1). We collected data from just south of St. Louis, Missouri, to Cairo. Data sources here included (1) a high-precision digital elevation

model (DEM) generated based on ground-based survey measurements in 1998, (2) global positioning system (GPS)−based levee-crest surveys completed in 2007, and (3) flood profiles of the Mississippi River through the study reach. A high-precision DEM (4.6 m horizontal and ~0.3 m vertical accuracy) was constructed for the St. Louis District of the U.S. Army Corps of Engineers (EarthData International, 2001). GPS points for this survey were collected in the spring of 1998. The elevation data used to construct this DEM were obtained from the U.S. Army Corps of Engineers in ASCII XYZ format, and the original 1998 GPS points measured along levee crests were filtered from this data set. Elevations were specified relative to the National Geodetic Vertical Datum (NGVD) of 1929. Our 2007 levee-crest elevations were derived from a kinematic post-processed survey using dual-frequency (geodetic grade) GPS receivers. We used a pair of 12-channel receivers, one as a base station at a benchmark of known position, and the other as a kinematic receiver. Precise kinematic surveying is possible so long as the roving receiver first obtains fixed-integer solutions for each satellite and then maintains a continuous lock on those signals. We utilized

Figure 1. Levees in study area, from south of St. Louis, Missouri, to Ohio River confluence at Cairo, Illinois. Detailed survey lines are shown in Figure 2 and in Data Repository (see footnote 1).

© 2011 Geological Society of America. For permission to copy, contact Copyright Permissions, GSA, or [email protected]. GEOLOGY, January 2011 Geology, January 2011; v. 39; no. 1; p. 55–58; doi: 10.1130/G31493.1; 3 figures; 1 table; Data Repository item 2011034.

55

six separate base-station locations to maximize measurement accuracy, and the study area was subdivided into 11 survey units. With a maximum baseline distance of 27.67 km, formal survey uncertainties were 3.8 cm or less (Trimble Navigation, 1998), with additional error derived from the nature of the kinematic survey (see following). Levee-crest elevations were measured kinematically with the rover receiver mounted on our field vehicle. Following initialization, the vehicle was driven on the centerline of all the main levees in each survey unit at 20–35 km/h, the receivers collecting data every 1 s (every 5–10 m). We collected >53,000 levee-crest positions through the study area. Antenna height included the height of the field vehicle, although random variation of several centimeters due to vehicle motion was visible in survey elevations. Following collection, survey data were post-processed, with erroneous or extraneous points manually removed. Poor vector solutions were identified and were reprocessed by filtering the number of satellites, satellite signals, or other parameters. Little manual processing of the data was needed. GPS-based elevations were calculated using NAD (North American Datum) 83 benchmark ellipsoid heights, transformed to elevations using the NAVD 88 benchmark elevations, and then converted to NGVD 29 elevations using the National Geodetic Survey’s VERTCON conversion methodology. We also evaluated 50 yr flood profiles for the study region (U.S. Army Corps of Engineers, 2004). Flood-profile elevations (specified in NGVD 29) were interpolated to each 2007 levee elevation point. A freeboard (required safety margin) height of 90 cm (3 ft) was added to the profiles. Comparison of measured levee heights with these flood profiles shows the level of protection afforded by each levee. The survey coordinates from each of our 11 survey units were imported in to ArcMap. Next, the 1998 levee points were filtered from the ASCII XYZ data file. Since the 2007 and 1998 points did not precisely coincide, a 2 m buffer was used to spatially join the two data sets. Finally, total distance along each levee unit was calculated as the cumulative sum of horizontal distances between measured points, and elevations were plotted versus cumulative distance. RESULTS Detailed survey maps and longitudinal elevation plots are provided for the 11 survey units composing the study area (Table 1). A total of 59.1 km of measured levee showed increases in elevation of >0.25 m, and 54.9 km of levees in the study area decreased by >0.25 m. Using a more conservative filter (change >0.35 m on ≥2 adjacent segments, with ≥3 measured GPS and DEM

56

TABLE 1. SUMMARY OF LEVEE ELEVATION CHANGES, 1998–2007 Elevation change

Length of changes

(0.1 km segment averages)

Valmeyer N Valmeyer S Modoc Kaskaskia Grand Tower N Grand Tower S Chester E Chester W Wolf Lake Commerce Cairo Total

Filter 1

Filter 1

Filter 2

Filter 2

µ (m)

σ (m)

Max. (m)

Min. (m)

Up (km)

Down (km)

Up (km)

Down (km)

–0.12 0.1 –0.01 –0.05 –0.48 0.29 0.03 0.18 –0.08 0.59 0.25

0.29 0.31 0.39 0.27 0.34 0.45 0.32 0.28 0.32 0.34 0.34

0.87 0.79 1.49 0.69 0.53 1.49 0.79 0.78 0.84 1.33 1.19 1.49

–1.21 –0.63 –1.26 –0.65 –1.26 –0.4 –0.73 –0.68 –1.02 –0.46 –0.44 –1.26

0.9 8.1 8.3 3.4 0.2 4.7 4.7 9.2 1.7 14.5 3.4 59.1

5.2 4 9.5 6 13.7 0.9 5.4 2.1 7.6 0.2 0.3 54.9

0.2 4.7 4.3 1.1 0 3.5 1.3 4 0.9 11.5 3 34.5

2.6 1.1 4.1 1.4 5.7 0 1.6 0.3 3.1 0 0 19.9

Note: µ—mean; σ—standard deviation; Max.—maximum; Min.—minimum. Filter 1—Elevation change >0.25 m for average of 0.1 km survey segments. Filter 2—Elevation change >0.35 m, ≥3 measured points/0.1 km, adjacent segment also ≥0.35 m change.

points per segment), 34.5 km of levees increased and 19.9 km of levees decreased in elevation between 1998 and 2007. The maximum upward change in levee-crest elevation was 1.49 m, in the Grand Tower S survey unit; the maximum decrease was 1.26 m, in the Modoc survey area. Detailed results for two sample areas are illustrated here, for the Kaskaskia (Fig. 2A) and the Grand Tower S (Fig. 2B) units; all other maps and plots are provided in the GSA Data Repository1 (Figs. DR1−DR18). The 2007 survey elevations show centimeter- to low decimeter-scale variability from point to point, reflecting some combination of position uncertainties, irregularity in levee roads, and motion of the vehiclemounted GPS receiver. Much of this random variability was reduced by averaging individual survey points within 0.1 km of levee survey segments (see Table 1). In addition, many profiles contain single-point spikes up to ~1 m above the mean of nearby elevations, coinciding with features noted in the field (e.g., pipeline crossings, pump stations, and transitions in road construction type). Road construction types varied between paved, gravel, and unimproved. The plotted profiles also show several gaps in both the 1998 and the 2007 profiles where levees could not be surveyed because they were gated off or otherwise impassable. The Kaskaskia survey unit includes the ringshaped levee around the town of Kaskaskia, Illinois, and the portion of the surrounding floodplain known as Kaskaskia Island (Fig. 2A, inset). This so-called island is an area of Illinois now located on the west side of the Mississippi 1 GSA Data Repository item 2011034, detailed survey results, is available online at www.geosociety. org/pubs/ft2011.htm, or on request from editing@ geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.

River channel that was cut off from the rest of the state by a channel avulsion during the 1881 flood. The Kaskaskia levee has failed throughout its history, including in 1973 and in 1993. The 1973 failure occurred approximately at point b, and the 1993 failure at point e (Fig. 2). Comparing the 1998 and 2007 levee-crest elevations (light and dark lines in Fig. 2), the Kaskaskia levee includes areas where elevations increased (e.g., between points c and d, averaging +42 cm) and other portions of the levee crest where elevations decreased (e.g., between points b and c, average drop of 38 cm). Comparison with the flood profile shows that the Kaskaskia levee protects to about the 50 yr level or less. The Grand Tower S survey unit (Fig. 2B) includes 12.4 km of levees just to the north and to the south of the town of Grand Tower, Illinois. Several bedrock outliers are present in the floodplain here, and the levee segments tie laterally into those promontories. The Grand Tower levee did not fail during the 1993 flood, although it was locally damaged and required repair. Comparison of the 1998 and 2007 levee-crest profiles shows broad increases in elevation during that period of time, as much as 1.49 m (between points c and d). The areas of the greatest elevation increases were areas that were the lowest in 1998, generally well below the 50 yr flood profile. Inferred explanations for changes in these and the other surveyed areas are discussed below. DISCUSSION Comparison of the 1998 and 2007 levee-crest elevations through the study area documents 214−274 km of surveyed levees where no discernable change occurred, but 35−59 km of levees (11%–18% of the total) where the 2007 survey elevations were distinguishably higher, and 20−55 km of levee crests (6%−17%) where the elevations were lower than in 1998.

GEOLOGY, January 2011

c

2007 GPS

d

1998 DEM

b

Elevation (m amsl NGVD29)

50-year (≤2% flood profile + freeboard)

e

121

f

a/g 0

1500

3000 m

120

a

119

d

c

b

0

5000

10,000

e

g

f 20,000

15,000

a b

2007 GPS

117

Elevation (m amsl NGVD29)

1998 DEM

c

50-year (≤2% flood profile + freeboard)

d

116 e

115 f 0

2000

g

4000 m

114

113

a 0

c

b 2000

d 4000

e 6000

8000

g

f 10,000

12,000

Distance along levee (m)

Figure 2. Elevation profiles for 1998 levee crest (light blue) and 2007 (dark blue). A: Kaskaskia survey unit. B: Grand Tower S survey unit. Locations (and points a–g) are shown in insets and in Figure 1. Approximate profiles of 50 yr flood profile plus freeboard (90 cm safety margin) are shown in red for reference. GPS—global positioning system; DEM—digital elevation model; NGVD29—National Geodetic Vertical Datum of 1929; AMSL—above mean sea level.

Levee Degradation At least 19.9 km of levees in the study area decreased in elevation by >0.35 m over broad and coherent areas (200 m to several kilometers) between 1998 and 2007. Other more localized areas of subsidence as well as smaller scale elevation decreases (<0.25 m) are suggested, but confirmation would require more precise and focused surveying. Even localized subsidence is of concern, however, as levee safety is only as great as the lowest or weakest point on a given levee. Many of the surveyed levees were capped by unimproved or gravel roads, and many of these areas show consistent but small-scale elevation decreases (e.g., b–c in Kaskaskia, Fig. 2A; b–c in Chester E, Figs. DR9 and DR10; a–b in Chester W, Figs. DR11 and DR12). These elevation changes were generally 0.1–0.2 m (i.e., at or below the resolution of the surveying here) but consistent over broad areas with unpaved road surfaces. In contrast, levee crests with paved roads generally showed greater correlation between the 1998 and 2007 survey elevations, but also included areas of distinguishable and coherent decreases (e.g., b–c in Valmeyer N, Figs. DR1 and DR2; a–b, c–d, e–f in Grand

GEOLOGY, January 2011

Tower N, Figs. DR7 and DR8). We suggest that these paved areas document subsidence due to degradation of either the levee fill or the underlying floodplain alluvium. Some areas of concentrated levee subsidence coincided with recent crevasses (levee breaches). During the 1993 Mississippi River flood, levee failures occurred in the Valmeyer N, Kaskaskia, Chester E, and Cairo survey units. The 1993 Valmeyer levee failure caused widespread inundation of the Mississippi floodplain in Illinois and led to Federal Emergency Management Agency buyouts and relocation of the town of Valmeyer. Two crevasses, located adjacent to the end of the survey at point g (Figs. DR1 and DR2), breached the Valmeyer levee. The 2007 survey was as much as 1.22 m lower than the 1998 levee crest near the location of these crevasses. We interpret this pattern as crevasse repairs after the 1993 flood but before the 1998 survey, followed by consolidation and settling in the repair area between 1998 and 2007. Elevation Increases A minimum of 34.5 km of the levees surveyed in this study area showed clear increases

in elevation between 1998 and 2007. Summing all incremental measured changes, there was a net upward shift of 8500 m2 [Σ (δz × δl)], where z is elevation, and l is incremental distance. Activities that may add height to levees include routine road maintenance, crevasse repairs, other permitted levee reconstruction projects, and unpermitted levee raises. Field observations during our survey suggest areas of road maintenance and addition of fresh gravel to some portions of the surveyed levees. In the Chester E survey unit, for example, the levee crest between points d–f and g–i were gravel roads that increased in elevation 1998–2007. These increases were generally small (<0.25 m; near the resolution limit here), but extensive and broadly consistent. Areas of small increases like this, probably associated with routine road maintenance, are not tabulated among the 34.5 km of levee-elevation increases cited here (increases >0.35 m). Larger increases in levee-crest elevations, to a maximum of 1.49 m, are broadly distributed across the study area. Increases >0.35 m (column 8 in Table 1) were documented in all of our survey units except Grand Tower N, although the extent of these areas varied substantially. The largest elevation increases, including 3.6 km of levees that increased >1.0 m, are concentrated in discrete portions of four survey units, Modoc, Grand Tower S, Cairo, and Commerce. The Modoc levee included at least 1.1 km where the crest increased >1.0 m (f–g in Figs. DR5 and DR6), filling a pronounced low in the 1998 profile and bringing the levee up to the ~50 yr protection level in that area. In the Cairo survey area, the Len Small levee was breached during the 1993 at point d (Figs. DR17 and DR18). Between 1998 and 2007, this portion of the levee increased in elevation by an average of 0.80 m, consistent with levee repairs undertaken after 1998. Other areas of levee increases, for example in the Grand Tower S survey unit, do not appear to have been planned, sanctioned, or carried out by the U.S. Army Corps of Engineers or other federal agencies. The town of Grand Tower, Illinois (population ~600), is typical of many rural floodplain communities. With minimal economic infrastructure at risk, the town cannot attract federal funds to upgrade its levees, a step that local residents believe necessary to grow and expand its financial base. Other increases in levee-crest elevations, generally in the 0.35–1.0 m range, are documented through much of the study area. Although the original intent of these increases cannot be ascertained, many apparently occurred under the guise of road repair. Field observations documented fresh gravel and other evidence of resurfacing on many of these levees. In one case, an earlier reconnaissance documented several

57

decimeters of new gravel over a preexisting and seemingly well maintained paved road surface (Fig. 3A). We question whether 0.35–1.0 m or more of added height is consistent with routine maintenance of these small levee-crest roads. Field observations also documented other types of levee raising. Along several of the areas surveyed, sandbags were exposed in the roadbed (e.g., in the Valmeyer N levee, points e and f; Figs. DR1 and DR2). This area showed no increase in elevation from 1998 to 2007, as expected if the bags were emplaced during an earlier flood (1993 or 1973). Sandbags are employed for flood fighting during large floods when levee overtopping is feared, but regulations require that all such additions to levee height be fully and promptly removed after the water recedes (U.S. Army Corps of Engineers, 2006a). Anecdotal evidence, however, suggests that such flood-fighting fill often is not removed, but merges with the levee and becomes undetectable until exposed via erosion or by a subsequent levee failure (Fig. 3B). CONCLUSIONS Comparison of 1998 and 2007 levee-crest elevations documents 214–274 km of surveyed levees where no discernable change occurred, but 35–59 km were distinguishably higher (to 1.49 m higher) and 20–55 km were distinguishably lower (maximum decrease of 1.26 m) in 2007 compared to 1998. Given the heightened national awareness of levee safety following the 2005 New Orleans disaster, both increases and decreases of these magnitudes suggest lessons for river, floodplain, and levee management in the U.S. Decreases in levee-crest elevations may result from (1) compaction or degradation of the levee fill, (2) subsidence of the underlying substrate, and/or (3) erosion of the levee-crest surface over

time. Distinguishing such contributing mechanisms generally requires detailed geotechnical and/or geodetic work. However, the results here document surprising elevation changes in just a 9 yr span, and the distributions of these declines (including drops over recent crevasse repairs and on paved roads) suggest significant compaction or subsidence in a number of locations. Levee measurements here document a net upward change of 8500 m2 (cumulative elevation change times incremental distance), which requires large-scale human intervention. These increases in levee heights parallel increases in the extent of Mississippi levees over time (Pinter, 2005; Pinter et al., 2008), both of which trends pose challenges, even threats, to residents behind the levees as well as to their floodplain neighbors. If a new or raised levee excludes floodwater, nearby floodplain properties may experience deeper and more frequent inundation (Pinter et al., 2008, 2010). For residents behind a levee thus raised, the unengineered nature of some of these projects suggests that any added security may be illusory. Our work has documented levee crests raised to the 50 or 100 yr protection level, while other portions of the same levees remain much lower. Field work also documented the addition of decimeters (to >1 m) of gravel-rich (and probably permeable) aggregate to levee crests, without corresponding increases in levee width. High-precision, GPS-based kinematic surveying represents a valuable reconnaissance tool for assessing levee safety and stability and for monitoring compliance with regulations. Local degradation and subsidence is a threat to many levees that has contributed to catastrophic failures in the past. Regular surveys would permit up-to-date monitoring of levee protection levels and target corrective actions for areas of concern prior to failure. Similarly, ad hoc levee elevation increases

Figure 3. A: Fresh road aggregate placed on levee in Alexander County, Illinois; material overlies paved road surface (image taken 28 May 2002; courtesy of R. Heine). B: Levee breach along Sangamon River, Illinois, exposed old sandbags within interior of levee; new sandbags emplaced during flood that breached levee are visible on ground surface (image courtesy of P. Osman).

58

serve neither local residents nor the national interest in sound floodplain management. ACKNOWLEDGMENTS This project was funded by National Science Foundation award BCS0552364 and by a research grant to Flor from the Geological Society of America. We thank Charles Theiling and an anonymous reviewer for very helpful suggestions. REFERENCES CITED Barry, J.M., 1997, Rising tide: The great Mississippi flood of 1927 and how it changed America: New York, Simon & Schuster, 528 p. Dunbar, J.B., and Britsch, L.D., III, 2008, Geology of the New Orleans area and the canal levee failures: Journal of Geotechnical and Geoenvironmental Engineering, v. 134, p. 566–582, doi: 10.1061/(ASCE)1090-0241(2008)134:5(566). EarthData International, 2001, Mississippi River DEM/DTM, Project Metadata: Gaithersburg, Maryland. Mills, G.B., 1978, Of men & rivers: The story of the Vicksburg District: Vicksburg, Mississippi, U.S. Army Corps of Engineers, 244 p. Morgan, A.E., 1971, Dams and other disasters: A century of the Army Corps of Engineers in civil works: Boston, P. Sargent, 422 p. Mount, J.F., and Twiss, R., 2005, Subsidence, sea level rise, and seismicity in the Sacramento– San Joaquin Delta: San Francisco Estuary and Watershed Science, v. 3, 18 p. Pinter, N., 2005, Policy forum: One step forward, two steps back on U.S. floodplains: Science, v. 308, p. 207–208, doi: 10.1126/science.1108411. Pinter, N., Jemberie, A.A., Remo, J.W.F., Heine, R.A., and Ickes, B.S., 2008, Flood trends and river engineering on the Mississippi River system: Geophysical Research Letters, v. 35, L23404, doi: 10.1029/2008GL035987. Pinter, N., Jemberie, A.A., Remo, J.W.F., Heine, R.A., and Ickes, B.S., 2010, Cumulative impacts of river engineering, Mississippi and Lower Missouri rivers: River Research and Applications, v. 26, p. 546–571. Rogers, J.D., 2009, Overview of post-flood surveys of the Upper Mississippi River valley in the summer of 2008, in Criss, R.E., and Kusky, T.M., eds., Finding the balance between floods, flood protection, and river navigation: St. Louis, Missouri, St. Louis University Center for Environmental Sciences, p. 41–46. Trimble Navigation, 1998, 4800 operation manual: Sunnyvale, California, Trimble Navigation Ltd, 52 p. United States Army Corps of Engineers, 2004, Upper Mississippi River system flow frequency study: Rock Island, Illinois, United States Army Corps of Engineers, http://www.mvr. usace.army.mil/pdw/pdf/FlowFrequency/Documents/FinalReport/default.asp. United States Army Corps of Engineers, 2006a, Levee owner’s manual for non-federal flood control works: The Rehabilitation and Inspection Program, Public Law 84-99, 176 p: http:// www.nfrmp.us/docs/USACE_NonFed Levee Owner’s Manual_Mar06.pdf. United States Army Corps of Engineers, 2006b, National Levee Safety Program, fact sheet: Washington, D.C., United States Army Corps of Engineers, 2 p. Manuscript received 21 June 2010 Revised manuscript received 5 August 2010 Manuscript accepted 10 August 2010 Printed in USA

GEOLOGY, January 2011