Urban Floodplain Management

Sept. 1, 2011

Flood events cause more than 3.5 billion dollars, or 90%, of property damage in the United States every year (FEMA 2009). Since 2004, more than 40% of annual natural disasters in the US have been related to flooding. With increasing local, state, and federal expenditures for flood disaster relief, flood prediction and control has become of critical importance. Given the apparent floodplain management crisis, it is somewhat ironic that most contemporary floodplain management strategies remain entrenched in historical floodplain land-use philosophies and economic obstacles. Studies are therefore urgently needed that provide information leading to science-based decisions to improve contemporary floodplain management practices.

In Midwestern states, most floodplain bottomland hardwood forests (BHF) were removed in the 19th and 20th centuries to cultivate rich underlying soils. In many cases, this required installation of drainage and flood control structures, including drainage tiles, ditches, levees, and dams. The channels of many streams and rivers were straightened and enlarged to further reduce flooding. Drainage and flood control structures and channel alterations, coupled with disturbance to vegetation communities and soils, drastically changed the hydrology of streams, floodplains, and remnant BHF (Carter and Biagas 2007). Following World War II, the US underwent rapid human population growth and urbanization. As a consequence of steadily increasing development and imperviousness, more developed areas became prone to flooding, and structural (i.e., engineered) approaches to flood control came under increasing scrutiny for not adequately solving the flood problem. In response, nonstructural methods were investigated to be used in conjunction with engineered alternatives (Bedient and Huber 2002). Nonstructural methods included land-use restrictions (i.e., best management practices), flood warning systems, drainage maintenance programs, and public awareness programs.

A key cornerstone in floodplain management was enactment of the National Flood Insurance Program (NFIP) promulgated by the US government in 1968 through the adoption of the National Flood Insurance Act. The Flood Disaster Protection Act of 1973 and the National Flood Insurance Reform Act of 1994 further defined the NFIP, which was ultimately designed to create a structure of rules that would aid in minimizing losses due to flooding and place responsibility of liability on the floodplain property owner rather than the taxpayer (Mays 2001). With the advent of the environmental movement, increasing emphasis was placed on floodplain management to minimize flood damage. Given the high costs associated with flood damage, it is somewhat surprising that floodplain management has evolved so slowly. Current floodplain management practices are designed to take advantage of natural floodplain processes while not compromising natural and human floodplain resources. Unfortunately, many floodplain management practices applied to retain human sociopolitical and economic desires also directly increase the probability of severe floods (Bedient and Huber 2002) and are therefore often obstacles to effective floodplain management.

Abernathy and Turner estimated in 1987 that less than 25% of pre-European-development BHF remains in many areas of the US. The other 75% was largely lost to cultivation, development, and urbanization (e.g., Owen 1999, Ehrenfeld et al. 2003). It is likely safe to assume that 25 years after the Abernathy and Turner report (1987), the amount of BHF remaining is even less, because human population continues to grow and move to urban centers. Urbanization often leads to conversion of floodplain land from agriculture to other uses that can have positive or more negative impacts on floodplain processes (Zedler and Leach 1998). For example, restored natural areas can improve the hydrologic functioning of floodplains. However, other urban land uses, such as commercial and residential development, can further degrade hydrologic functioning.

Floodplain restoration efforts in the Midwest have been mainly focused on wide alluvial floodplains of large rivers, especially in the lower Mississippi Alluvial Valley (Stanturf et al. 2001). However, riparian and floodplain BHF along lower-order tributary streams also provide critical hydrologic, biogeochemical, and ecological functions. For example, floodplain capacity to attenuate or ameliorate the export of sediment and nutrients from source watersheds in the central US is critical for restoring the ecological health of the Missouri and Mississippi rivers. This holds important implications for the hypoxic zone of the Mississippi delta in the Gulf of Mexico. It is reasonable to expect that urban riparian and floodplain forests of lower-order tributaries play particularly vital roles in absorption, attenuation, and treatment of storm flows due to coupling to urban development. However, a quantifiable assessment of those roles remains largely unknown in urban floodplain environments.

Little guidance is available for restoration of forested urban floodplains. What information does exist suggests that methodologies for floodplain assessment and restoration to forested conditions are of limited value in urban environments, mainly due to the magnitude and variety of disturbances over time (i.e., legacy effects) in urban environments (Ehrenfeld 2000, Simmons et al. 2007, Ravit et al. 2008). Urban stream water-quality studies documented strong correlations between imperviousness of a drainage basin and the health of receiving streams (Arnold and Gibbons 1996). This observation has lead in part to a movement of water-quality control by means of volume-based flow reductions (Hubbart et al. 2010). Volume-based flow reductions exemplify the types of land-use practices that have often been applied with unknown consequences, often to the detriment of floodplain flow attenuation efficacy. Ultimately, there is little thought to legacy effects of land use in general (Kochendorfer and Hubbart 2010) and floodplain management in particular when considering the most appropriate future management strategies. Considering compounding effects of various management practices over the decades on floodplain lands could hold important implications for how floodplains are best managed now and in the future. Floodplains along urbanized streams in the lower Midwest represent unique and challenging opportunities for BHF preservation and restoration activities. Fortunately, because of the tendency toward public ownership, floodplains in urban areas are ideal for restoration. In addition, urban floodplain location in or near population centers make them ideally situated for public education and involvement during restoration activities. Urban-floodplain-associated wetlands can also provide nearby opportunities for recreation and aesthetic enjoyment, along with rare urban wildlife habitat.

Given the preceding discussion, it is clear there is a great deal that must be learned about floodplain processes in urban environments to achieve their proper management. A logical starting point is to quantify the physical differences between remnant BHF and an agricultural floodplain in an urban setting in terms of the most obvious differences (i.e., vegetation and soils). The following project was undertaken to provide quantifiable baseline information about floodplain vegetation and soil processes in lower Hinkson Creek (HUC 103001020907), a 303(d)-listed impaired stream located in Columbia, MO. Specific goals were to: a) quantify the canopy density of a section of old growth urban BHF, b) quantify the difference in soil infiltration capacity between BHF and agricultural sites, and c) estimate soil water-holding capacity by soil characteristic analysis. Conceivably, these baseline indices could provide information to estimate the potential role of BHF in flood attenuation by means of increased infiltration and consumption of water by transpiration.

Study Site: The Hinkson Creek Watershed
This research is focused on an approximately 6-mile reach of the lower Hinkson Creek. The reach is located between two permanent hydroclimate gauging sites on the main channel that are part of an urban watershed study implemented in 2008 (Hubbart and Freeman 2010, Hubbart and Gebo 2010, Hubbart et al. 2010). The contributing watershed to the floodplain project reach contains a large portion of the most intensively developed land in the city of Columbia (Figure 1).

Figure 1. Locations of five permanent gauging stations (right), and nested floodplain study sites (left) in the Hinkson Creek watershed

Figure 2. Aerial photographs of Hinkson Creek in 1939 and 1992 flowing through the current floodplain study reaches in central Missouri

Hinkson Creek watershed (HCW) is located within the Lower Missouri-Moreau River Basin (LMMRB) in central Missouri. The HCW was instrumented with a nested-scale experimental watershed study design in the fall of 2008 to study contemporary land-use and urbanization effects on hydrologic processes, water quality, and localized effects on climate and biological community health (Hubbart and Freeman 2010, Hubbart and Gebo 2010, Hubbart et al. 2010). Nested watershed study designs use a series of sub-basins inside a larger watershed to examine environmental variables. Sub-basins are determined based on dominant land use and characteristics of the hydrologic system (Capel et al. 2008). A nested watershed study design enables researchers to quantifiably ascertain the influencing patterns and processes observed at each location (Pickett et al. 1997, Hubbart et al. 2007, Hubbart et al. 2010). Hinkson Creek flows through a catchment basin of approximately 231 square kilometers (81.2 square miles). The creek flows approximately 42 kilometers (26 miles) in a southwesterly direction, ultimately flowing into the Missouri River approximately 8 miles away. Elevation ranges from 177 meters (580 feet) to 274 meters (890 feet) above mean sea level in the headwaters. Hinkson Creek is classified as a Missouri Ozark border stream located in the transitional zone between Glaciated Plains and Ozark Natural Divisions (Thom and Wilson 1980). Average annual temperature and precipitation (30-year record) is approximately 14ºC (57ºF) and 980 millimeters (38 inches), respectively. Soil types range from loamy till with a well-developed clay pan in the uplands (Chapman et al. 2002) to thin cherty clay and silty to sandy clay in lower reaches. Urban areas are primarily residential (approximately 100,000 residents) with progressive commercial expansion. Land use in the watershed is approximately 34% forest, 38% pasture or cropland, and 25% urban area; the remaining land area is wetland, open, or shrub/grassland areas. Instrumentation in the HCW is complimented by a US Geological Survey gauging station (USGS-06910230) that has collected stage data intermittently since 1966 (Figure 1, site 4).

Floodplain study grids were located within large remnant BHF and abandoned agricultural sections of the reach in the spring of 2010. The BHF site of this study is dominated by the woody species: Acer saccharinum (Silver Maple), Acer negundo (Boxelder), Ulmus americana (American Elm), Populus deltoides (Eastern Cottonwood), and Juglans nigra (Eastern Black Walnut). The stand was established in approximately 1945 on alluvium soils (Figure 1). Older Silver Maples remain in the northern half of the BHF study site dating to 1917 (established by tree core dating). Historical photos of the study sites indicate that Hinkson Creek was much more meandering in the early 1900s relative to recent photos of 1992 (Figure 2) and current conditions (Figure 1). Study grid dimensions measured 120 (east-west) by 110 (north-south) square meters. The southern 120-meter side was located 10 meters from, and parallel to, Hinkson Creek. Floodplain study grids were divided in to 20-square-meter sub-grids in the summer of 2010 (July and August), from which 42 sampling locations (120 meters by 110 meters, n=42) were established to quantify canopy cover and infiltration, and 25 sampling locations were located in the center of the larger grid (80 meters square, n=25) to collect soil samples to estimate equivalent depth of soil between the BHF and abandoned agriculture site.

Photo: Edward Bulliner
Figure 3. Hemispherical photo (left) of Hinkson Creek floodplain forest canopy and photo gap analyzer analysis for LAI (right)

Leaf Area Index. Leaf area index (LAI) is the ratio of total upper leaf surface of vegetation divided by the surface area of the land on which the vegetation grows and is a dimensionless value with typical range of zero for bare ground to six for dense forest (Campbell and Norman 1998). LAI is useful to predict primary production, which is strongly related to plant transpiration and therefore soil water status. Two methods were used to collect LAI data in the floodplain: the ceptometer method and the hemispherical method. The ceptometer method required the use of a Decagon Devices LP-80, which measures photosynthetically active radiation (PAR) along an array of 80 sensors mounted on a 1-meter light bar. The amount of PAR that is transmitted through a vegetative canopy is a direct function of canopy structure and density, characterized by the canopy’s LAI. One ceptometer was placed on a tripod approximately 1.6 meters aboveground within a clearing to log reference PAR every minute. Measurements of PAR within the clearing were compared to measurements taken beneath the canopy to calculate the ratio between the two measurements. Ceptometer measurements were collected in four cardinal directions at all 42 sampling locations of the BHF site, for a total of four samples per location (n=168). LAI was calculated as per Decagon Devices (2006).

LAI was also estimated using hemispherical photography (Figure 3) using a Nikon D60 digital SLR camera with a Sigma 4.5-millimeter F2.8 EX DC circular fisheye lens. Photos were taken at the same locations and same time as the PAR data. Photographs were taken with the lens pointing vertically upward and camera base mounted 1.3 meters above ground. An aperture of f5.6 was used for all photographs. Hemispherical photographs were analyzed using Gap Light Analyzer software (Frazer et al. 1999). The algorithms in Gap Light Analyzer software relate gap fraction (percent open sky seen at a point versus obstructions such as leaves) to leaf area index, and rely on an inversion of Beer’s law with the assumption that leaves in the canopy are randomly distributed. Since the penetration of direct light through a forest canopy is a function of leaf density through which light travels, LAI (a function of leaf density) can be calculated by quantifying the amount of light penetrating the canopy as seen from a given location (e.g., Stenberg et al. 1994).

Infiltration Capacity. Devices used to measure the rate of water percolation in to soils are called infiltrometers. Double-ring infiltrometers use two rings to measure infiltration, an inner and outer ring, thus greatly increasing accuracy of infiltration estimates. The purpose of the double ring assemblage is to create one-dimensional flow of water from the inner ring. An inner ring is driven into the soil and a second larger-diameter ring around that helps control the flow of water through the first ring. Water is supplied either with a constant or falling head condition, and the operator records how much water infiltrates from the inner ring into the soil over a given time period. This arrangement accounts for lateral movement of water around the infiltration ring blades and also creates a seal. Therefore, the inner ring will supply a much more accurate estimate of infiltration rates. The infiltration rings can also be used at other depths of soil to investigate infiltration changes with depth (Ahuja et al. 1976, Bodhinayake et al. 2004). For more information on the method the reader is referred to the American Society for Testing and Materials D3385-09 Standard Test Method for Infiltration Rate of Soils in Field Using Double-Ring Infiltrometer
(www.astm.org/Standards/D3385.htm).

Equivalent Depth of Soil Water. Equivalent depth of soil water (EDSW) is calculated based on soil characteristics (e.g., volumetric water content, porosity, bulk density), which are often determined using the soil core method (Hillel 2003). This is a simple function of soil volumetric water content (VWC) calculated as follows:

VWC=Vw/Vt

Where Vw is the volume of water in a soil sample often quantified by drying a soil sample of known volume at 105°C (221ºF) for 24 hours and subtracting the dry sample weight from the wet sample weight, and where Vt is the total volume of the sample. VWC is then distributed by soil depth to compute EDSW. Ideally, the researcher will have retrieved multiple soil samples from multiple soil depths to be able to provide a depth integrated average EDSW. This is important, because soil type and thus water-holding capacity and EDSW will vary spatially (i.e., vertically and horizontally) in soils (Hillel 2003).

Figure 4. Average leaf area index in a bottomland hardwood forest of lower Hinkson Creek
Figure 5. Infiltration capacity comparisons between a bottomland hardwood forest and agricultural floodplain site in the lower reaches of Hinkson Creek

Results and Discussion
Climate over the period of study was typical for central Missouri; 372 millimeters (14.6 inches) of precipitation fell between July 1 and August 31, 2010, in the city of Columbia, which received more than 1,346 millimeters (53 inches) of precipitation in 2010. Intuitively, we should expect that BHF should improve infiltration and subsurface flow of water by virtue of large, extensive root systems. LAI is assumed highly correlated to higher transpiration (i.e., more leaf area=higher rates of transpiration); therefore, EDSW should also be higher in BHF, indicating more flood protection by increased percolation/infiltration and additional removal (i.e., consumption) of water by transpiration. The notion of consumptive water use is important in this context because consumptive use of water removes the water from the watershed. This is opposed to non-consumptive uses that use the water and then return some or all of the water to the source. An example of non-consumptive use is excess irrigation water pumped from a stream that flows overland or laterally through the subsurface back to the stream. Plant transpiration is a good example of consumptive water use, which reduces water to vapor that is then generally removed from the watershed by local turbulence (i.e., local climates). Figure 4 shows average estimated LAI using both methods discussed earlier. Average LAI was 2.98 (SD=0.70) and 3.06 (SD=0.65) using the ceptometer and hemispherical methods respectively, indicating close agreement between methods. Figure 4 shows distribution of LAI spatially throughout the BHF study plot, where darker green colors indicate higher LAI.

Results of infiltration tests (n=42) using the double-ring infiltrometer indicated average infiltration capacity (maximum steady-state infiltration under saturated conditions) of 23 (SD=21.0) and 38 (SD=29.0) centimeters per hour in the agricultural and BHF sites respectively (Table 1 and Figure 5). Minimum infiltration rates between the sites were 0.1 and 3.0 millimeter per hour for the agricultural and BHF sites, respectively. Maximum infiltration capacity was 69.0 and 126.0 centimeters per hour for the agricultural and BHF sites, respectively. Obviously, maximum infiltration rates measured in the BHF created a dramatic difference in mean infiltration between the two sites. High maximum infiltration rates were associated with locations of root systems of larger trees corresponding to higher LAIs (Figure 4).

Estimations of EDSW were calculated using the soil core method (Hillel 2003) extracted at depths of 15, 30, 50, 75, and 100 centimeters every 20 meters within the 80-square-meter study grid in the center of each 120- by 110-meter study plot (n=25) for a total sample size of n=150 from each floodplain study site (Figure 6). When integrated over 100 centimeters (1 meter) depth, average EDSW in the agricultural site was calculated to be 33.3 centimeters per meter (SD=2.24 cm/m), or approximately 33% of soil depth to 1 meter was water. EDSW in the BHF site was estimated to be 36.9 centimeters per meter (SD=2.68 cm/m), or approximately 37% of the soil depth to 1 meter was water. Figure 7 illustrates the quantifiable differences between the agricultural and BHF sites. Notably, a two-sample t-test was conducted, indicating that EDSW is statistically different between the sites at the 99% confidence level.

Figure 6. Equivalent soil water depth comparisons at five soil depths between a bottomland hardwood forest and agricultural floodplain site in the lower reaches of Hinkson Creek

Figure 7. Box plot showing descriptive summaries of equivalent depth of soil water between an agricultural and a bottomland hardwood forest site in Columbia, MO

Even though significant at the 99% confidence interval, an approximate 3.5 centimeters per meter difference may not seem like a lot. However, it equates to an approximate 11% difference in soil water between the sites, which can be substantial in terms of flood control (note: 3.5 cm=1.4 inches). As expected, the BHF site had higher soil water holding capacity. If we assume that BHF trees transpire 2 to 3 millimeters of water per day during summer months (likely a conservative estimate), then it would take between 12 and 18 days to transpire the 3.5 centimeters of additional water in the BHF floodplain. Over the course of a summer, this could amount to a great deal of water consumed, especially in larger expanses of BHF. Notably, this estimation is only in the top meter of soil. There are a variety of implied assumptions with this estimate, including the assumption of a homogenous forest canopy in terms of density and age distribution (among other assumptions). Exacting differences in this estimation supplies impetus for ongoing efforts to understand plant water use and floodplain processes. It is worth further recognizing that many BHF root systems will penetrate deeper soils (greater than 1 meter), thus removing deeper soil water, potentially into the saturated zone. Regardless of the assumptions of this estimate, there is little argument that tree transpiration consumes water and creates soil pore space, which creates a concentration gradient for more water, thus improving floodplain attenuation capacity and water consumption.

This work shows quantifiably the potential benefit of reestablishing floodplain forests to enhance storage capacity and improve attenuation of water, thus reducing flooding and mitigating development and urbanization. Ongoing research in the floodplain study plots used in this work will include information gleaned from automated piezometric networks and soil volumetric water content sensor profiles. Ultimately, there may be many benefits to reestablishing BHF in urban floodplains besides floodwater attenuation, including improving water quality, improving aquatic ecosystem health, creating inner city parks, improving human health, and also carbon sequestration.

It is understood that forests play a significant role in carbon sequestration in aboveground woody biomass accumulation and to an even greater extent in forest soils (Cason et al. 2006, Londo et al. 1999). Worldwide, forest ecosystems store approximately 1,240 PgC (1.24 x 1015 grams carbon). This accounts for approximately two-thirds of the terrestrial carbon on the planet, excluding rock and sediment (Sedjo 2001, Dixon et al. 1994). At least one-third of total forest carbon is contained in wetland and floodplain soils (Trettin and Jengursen 2003, Dixon et al. 1994). One of the largest sources of CO2 is conversion of forests to agricultural lands (Hendrickson 2004, House et al. 2002), which consequently contain lower soil organic carbon stocks than their potential capacity. Clearly, afforestation of converted floodplain lands could not only dramatically improve flood safety and losses, but also significantly increase soil carbon sequestration (Trettin and Jurgensen 2003,
Wigginton et al. 2000. Turner et al. 1995, Lal 2005), and perhaps combat to some extent local and regional anthropogenic climate change phenomena and potential increased flooding.

Based on this work, it is arguable that perhaps Mother Nature got it right in the first place. Growing trees on flood prone land increases infiltration and transpiration, and also directly sequesters carbon both above and belowground. Characteristics of former BHF agricultural sites include poor soil quality and competing vegetation (Stanturf et al. 2004). To successfully reforest urban floodplains, there is much to learn. However, reestablishment of floodplain forests has the potential to reduce flooding and to improve soil and water quality, among a broad range of other beneficial uses.

Conclusions
In urbanizing watersheds such as Hinkson Creek, comprehensive management approaches are imperative that examine not only the volume of water causing impairments and the variable-use landscape, but also the pollution load being transported. For the HCW and other similarly affected watersheds across America, the work presented here is timely given legal mandates to provide quantifiable estimates of total maximum daily loads (TMDLs), improve water quality, reduce stormwater runoff, and decrease flood risk. Given the relative size of the HCW and the scope of land uses in the watershed, the HCW serves as common ground encouraging cooperation, trust, and innovation among watershed stakeholders to reach a common goal to improve and sustain water quality, and as a model urban watershed for similar studies.

Initial results from this ongoing urban watershed project indicate that floodplain forests are good not only for natural resources but also for human health. The current work shows that with a single two-month snapshot, more than a 10% improved floodplain attenuation capacity is possible in a bottomland hardwood forest floodplain. Continued work will seek to quantify annual BHF transpiration and interception rates and to establish multiannual vadose and saturated zone water flux data to show quantifiably the benefits of reestablishing bottomland hardwood forests in the floodplains of the American Midwest.  

About the Author

Jason Hubbart

Jason A. Hubbart is an assistant professor of forest hydrology and water quality at the University of Missouri and director of the university’s Interdisciplinary Hydrology Laboratory in the Department of Forestry.