Socio-hydrology and vulnerability of levee systems along the lower Illinois River

ABSTRACT The maintenance of flood mitigation levees in the U.S. and elsewhere, has often found to be insufficient. Recent U.S. levee safety inspections have discovered substantial deficiencies in many federally monitored levees resulting in some of them being assigned an ‘unacceptable’ safety rating. Most levees in the U.S. were constructed using federal money but then were turned over to local government entities to maintain, thus socioeconomic characteristics of these communities may impact their ability to maintain their levee(s). We used dasymetric mapping to assign socioeconomic parameters to levee protected areas along the lower Illinois River to explore differences in these characteristics between communities with an unacceptable levee safety rating to those with an acceptable rating. Principal components analysis was used to determine which socioeconomic parameters explained the majority of variance between levee protected communities and the Mann-Whitney-Wilcoxon U-Test for significance testing. These analyses revealed that differences in total population , race, average per-capita income, and number of residential homes were influential indicators for explaining differences between comunities with levees with acceptable versus unacceptable ratings. These results suggest populations who inhabit levee systems with unacceptable safety ratings are white and relatively wealthier than communities located within levee systems with acceptable ratings. This finding is counter to research that shows that the poor and poorer minorities live in areas with higher flood risk. This is perhaps wealthier floodplain property owners inhabitating unacceptable levee systems may be foregoing necessary levee maintenance, inastead relying on other government programs to pay for flood damages.


Introduction
The development of communities along rivers and their associated efforts to maximize riverine services while reducing hydrologic risks such as flooding is a classic example of socio-hydrology, the coupled interactions and associated feedback between humans and their supporting hydrologic systems (Di Baldassarre et al. 2013, 2015. Levees have been used around the world to mitigate flood risk because they are generally viewed as politically expedient because of their effectiveness and relatively low construction costs. This is despite potential increasing flood losses over the long term by encouraging floodplain development because levees are generally constructed to withstand floods of a limited magnitude (generally 0.2 to 10.0% chance annual floods) and require continual maintenance to maintain their level of protection (Tobin 1995;Li et al 2019;Montz and Tobin 2008). This long experience with the construction of levees and their political expediency has led to the construction of more than 64,000 km of levees throughout the U.S. (National Research Council NRC 2013). Further, a recent U.S. wide study found that the national levee database contains only 20% of the total levee length constructed throughout the country (Knox, Wohl, and Morrison 2022).
In the U.S., the maintenance of levee systems has been uneven. For instance, levee safety inspections performed by the U.S. Army Corps of Engineers (USACE) have resulted in a substantial number of levee systems along the Mississippi and its major tributaries, the Illinois and Missouri rivers being assigned an unacceptable safety rating primarily due to maintenance issues (>35% of the systems). This suggests a substantial number of communities are not keeping up with their levee maintenance (Remo 2016). Given most of these levee systems were constructed or enhanced to similar design standards in the mid-20 th century (Remo 2016) suggests socioeconomic characteristics of levee system populations are substantial factors in potentially predicting a levee-protected community's ability to adequately maintain their levee system. To the authors' knowledge, there has been little, if any, research on the capacity of levee-protected communities to maintain their levees.
This study seeks to determine if there are socioeconomic characteristics that can be used to determine which communities can adequately maintain their levee system and consequently preserve these communities' defences against flooding. This effort is an exploratory first step towards developing a predictive model for assessing which communities may have the political, social, and economic capital plus other potentially relevant socioeconomic characteristics to adequately maintain their levee system. This research not only has implication for U.S. flood mitigation policy but, the types of data and methods employed plus the findings of this study can provide novel insights for assessing which levee-protected communities around the world have the necessary capital and other social characteristics to sustain their levee systems.

U.S. levee systems and their inspection
Levee systems are the earthen embankments and the associated supporting infrastructure used to operate the levee to mitigate flood risk and include components such as drainage pumps, culverts, ditches, relief wells, etc. After the Hurricane Katrina Disaster in 2005 caused a breach of floodwalls and levees resulting in the inundation of the city of New Orleans (Mossa 2013), the USACE established the Levee Safety Program to work with local sponsors (communities, levee districts, and private entities) to help understand, prioritize, and manage flood risk related to levee systems (USACE 2018a).
Levees systems within the USACE's levee safety program are divided into three categories base on whether the system is a federal project (congressionally authorized) and who retains the primary responsibility for operation and maintenance. These categories are: 1) federal levee, levee sponsor operated and maintained; 2) federal levee, USACE operated and maintained; 3) nonfederal levee, locally operated and maintained. All levees within the levee safety program are eligible for the USACE's Rehabilitation Program authorized under Public Law 84-99 which allows the USACE to restore a levee system to its pre-disaster status at no cost to the levee owner (categories 1 and 2), and at 20% cost to the eligible non-Federal system owner (category 3). In addition, costs of routine maintenance and necessary upgrades for non-USACE operated and maintained levee system are born by the sponsor (generally local government entities; categories 2 and 3) and the USACE operated and maintained levees systems are largely maintained using congressionally authorized funds (i.e. federal funds; category 1).
The National Levee Database holds information on the locations, ownerships, construction characteristics, associated features, and inspection status of USACE inspected levee systems. USACE, as the portion of its levee safety program, performs routine and periodic levee inspections. The annual routine inspection consists of a visual examination by civil engineers and others to validate and rate a given levee district's operation and maintenance of its levee under the USACE levee safety program. Periodic inspections are completed every five years and have three key components: 1) collection of existing data on varying aspects of the levee including levee operation, its maintenance history, inspection record, emergency action plans, and flood-fighting records; 2) the field inspection which is performed via a multidisciplinary crew overseen by a professional engineer; and 3) an inspection report with a rating attribute for the operation and maintenance of the levee system plus additional recommendations, identified deficiencies, or aspects requiring assessment (USACE 2017).
At the time this study was undertaken, inspection ratings were given for routine and periodic inspections for the maintenance and operation of the levee systems. Each of the levee systems has an inspection rating: unacceptable, minimally acceptable, or acceptable (USACE 2017). The USACE (2017) defines 'An acceptable levee as all inspection items are rated as acceptable'. A minimally acceptable levee has 'one or more inspection items are rated as minimally acceptable, or one or more items are rated as unacceptable, and an engineering determination concludes that the unacceptable inspection items would not prevent the segment/system from performing as intended during the next flood event'. An unacceptable levee is defined as one or more inspection items are rated as unacceptable and would prevent the system from performing as intended, or a serious deficiency noted in past inspections has not been corrected within the established timeframe". In this study, we group levee systems with minimally acceptable or acceptable levee rating together as an 'acceptable levee safety rating' for ease of communication.
Recent studies by the USACE have found that >95% of the inspected levees systems need substantial maintenance and upgrades to continue their current levels of protection to an adequate margin of safety (USACE 2018a). Federal, state, and local government agencies currently struggle to finance important repairs and system upgrades (American Society of Civil Engineers [ASCE] 2017). The USACE estimates that $21 billion is needed to improve and maintain the moderate to highrisk levees in its portfolio, which represents only about 15% of the levee systems monitored by the USACE under their levee safety program. The maintenance needs of the remaining 85% are currently under evaluation or unknown (USACE 2018a).

Hazard insurance
In the U.S. there are two main types of federally backed insurance programs intended to offset flood losses for private property owners. These programs are the National Flood Insurance Program (NFIP) and the Federal Crop Insurance Program (FCIP). Federal flood insurance from the NFIP is available in participating communities to help individuals and small businesses recover from flood damages (Congressional Research Service CRS 2011). Portions of this program are subsidized to encourage participation (Congressional Research Service CRS 2021b). The FCIP provides farmers the opportunity to purchase insurance coverage against financial losses caused by a wide variety of risks, including losses from flooding. The federal government subsidizes the premiums that farmers pay for these insurance policies to encourage farmer participation. Federal crop insurance premium subsidies vary based on the coverage level and unit type the producer selects however, subsidies range from 38% up to 100% for catastrophic losses (Congressional Research Service CRS 2021a).

Lower Illinois river
The lower Illinois River extends 235 km from the confluence with the Mackinaw River, just downstream of Peoria, IL, to the Mississippi River confluence (Guida, Remo, and Secchi 2016). Along this river segment, a total of 29 levee systems mitigate inundation on ~98,000 ha of land (US Army Corps of Engineers USACE 2018b). The flood magnitudes in which these levee systems mitigate flood inundation is variable and ranges from the 0.5% to the 10% annual-chance flood (Scientific Assessment and Strategy Team SAST1995); Figure 1).
The State of Illinois in 1879 passed the Levee Act which allowed for the formation of levee districts which are quasi-local-government entities authorized to create special tax districts to fund and maintain the lower Illinois River and other levees across the state of Illinois (Hannah 1960). However, given the limited coordination between and oversite of these early levee districts along the Illinois River, the levees they constructed ranged substantially in their level of protection, engineering standards, and the amount of area which they mitigated flood risk (Thompson 2002).
Large floods which occurred in 1922 and 1926-27 caused substantial damage to many of the ILR levees. Falling agricultural commodity prices during this period resulted in reduced tax revenue taken in by the levee districts. The flood damage coupled with a lack of funds to repair the levees resulted in many lower Illinois River levees falling into disrepair. However, the Great Flood of 1927 along the Mississippi River and its major tributaries including the Illinois River, which resulted in several levee breaches and failures, led congress to pass the 1928 Flood Control Act (FCA). Congress, through the 1928 FCA, tasked the USACE with the authority to design and oversee the construction, repair, and then improvement of the existing levees along the Illinois, the Mississippi, and its other major tributaries. The 1928 FCA and several subsequent FCAs passed in the 1930s through the 1960s led to a substantial amount of federal funding to strengthen and complete nearly all lower Illinois River levees to federal standards (25 out of 29; Thompson 2002;(USACE 2018b).
In terms of the USACE's levee safety program, 25 of the lower Illinois River levees systems were federally constructed levees that are sponsor operated and maintained (i.e. levee districts which are local government entities) and the remaining 4 levees are non-federal levees, that are solely operated and maintained by local government entities or private individuals (USACE 2018a). However, 3 of the 25 federally constructed levees have been removed from the USACE levee safety program (USACE 2018a) because these levee systems are currently managed by non-governmental organizations or government agencies attempting to naturalize floodplain areas along the lower Illinois River (Guida, Remo, and Secchi 2016).

Study objectives
In this study, we explore the differences in the socioeconomic characteristics of lower Illinois River levees systems rated by the USACE as acceptable versus those systems rated as unacceptable. The purpose of this assessment is to identify the socioeconomic characteristics that can potentially indicate which levee systems are likely to be sustainable (i.e. capable of maintaining a levee system's integrity at the authorized flood mitigation level achieving an acceptable safety rating). The three research questions we seek to inform include: (1) What are the socioeconomic characteristics which differentiate the communities with levee systems with acceptable verse unacceptable levee safety ratings? (2) Do relatively poorer communities have levee systems that are rated as unacceptable where relatively wealthier communities are protected by acceptable levee systems? (3) Do levee systems that have acceptable ratings protect agricultural areas which possess higher agricultural profits?
Answering these questions will help identify if there are socioeconomic characteristics that can be used to determine which communities can adequately maintain their levee system and consequently maintain the community's protection against flooding.

Data sources
In this study we evaluated 29 lower Illinois River levee systems. Three of these levee systems are undergoing naturalization where their land use is being converted from agricultural to floodplain forest and/or wetland. These include Kelly Lake, West Matanzas, and Meredosa Lake Drainage and Levee Systems ( Figure 2). We chose to retain the three levee systems undergoing naturalization in our analysis because portions of these systems still contained substantial agricultural or residential land uses in which the levees still mitigate some flood risk. For the 29 levee systems, socioeconomic data, political boundaries, hydrological features, land cover, building inventory, and levee information were compiled for the analyses undertaken in this study. The socioeconomic and building inventory data were compiled from the Federal Emergency Agencies (FEMA) Hazus-MH loss estimation software. The data contained within this software was compiled from the 2010 US Census data and other national-level sources (FEMA 2017). Land cover data and hydrologic features were compiled from U.S. Geological Survey (USGS 2014). Levee information was compiled from the USACE's National and Scientific Assessment and Strategy Team (SAST 1995) Upper Mississippi River levee databases. Floodplain land agricultural profit was obtained from Guida, Remo, and Secchi (2016).  Figure 2. Levee systems, their protection level, and safety rating. * Estimated flood magnitude return period in years which reference the maximum water-surface elevation in which the levee was constructed to withstand. Italics indicated levee system undergoing conversation from crop land to floodplain forest and wetlands.

General approach
The geospatial analyses performed for this study are diagramed in Figure 3. The first task in the analytical framework applied in this study was to export the geospatial data layers of socioeconomic and building inventory parameters from FEMA's Hazus-MH loss estimation software for the Lower Illinois River Valley at the U.S. Census Block Level. Next, these socio-economic geospatial layers were attributed to a specific levee system using a dysametric mapping approach utilizing USGS land cover data and then the count-based parameters were normalized by estimated population for each levee system. Then levee system information and agricultural profit information were attributed to each of the 29 evaluated levee systems ( Figure 2). Next, we assessed the compiled parameters using a principal components analysis (PCA) to see which of the compiled variables explained the variance between the levee systems to reduce the number variables assessed for significance  testing (Table 1). Finally, we performed a two-sample Mann-Whitney-Wilcoxon U-test to assess for statistically significant differences in the socioeconomic parameters between levees that ranked acceptable versus unacceptable to see which socioeconomic characteristics could be affecting a community's ability to adequately maintain their levee system.

Compiling of the socioeconomic data and building inventory data
Census block level socioeconomic and building inventory data for the lower Illinois River Valley were extracted from the Hazus-MH (v 4.0; FEMA 2017) database and processed using the tools within ArcMap (v.10.3). Since many census blocks extended beyond the boundary of a given levee system, we used a dasymetric mapping approach to assign population and building inventory information based on the developed land cover as depicted in the USGS's 2011 National Land Cover Database (NLCD; USGS, 2014). Developed areas were defined here as low, medium, and highly developed area land cover types in the 2011 NLCD. The overall goal of using dasymetric methods in this study is the creation of a map that implements a depiction of the spatial distribution of socioeconomic and building inventory characteristics for each levee system as truthfully as possible. Our dasymetric modelling approach is similar to the methods described in Zandergen and Ignizio (2010).
Using the zonal statistics tools within ArcMap, the developed areas were summed for each census block. Next, the sum of the developed area within a given census block and levee system were also calculated. Then, the percentage of the developed land cover for each census block contained within the levee system was calculated. This percentage of developed land cover was then used as a weighting factor to appropriate the portion of the population counts, socioeconomic data, and the amount of building inventory to the fraction of the census block with developed land cover contained within the levee system (Alruzuq, 2018). For each levee system, the proportionally weighted socioeconomic and building inventory data from each census block were summed to produce an estimate of each parameter using the spatial join tool in ArcMap.
After the socioeconomic parameters were compiled, the five races contained within the census data were consolidated into two parameters (white and nonwhite) and the seven income parameters into two annual income parameters <$40,000 and > The income per capita < $10,000 Count US Census (2010) Inc30000to40000 The income per capita between $30,000 to $40,000 Count US Census (2010) Inc40000to50000 The income per capita between $40,000 to $50,000 Count US Census (2010) Inc50000to60000 The income per capita between $50,000 to $60,000 Count US Census (2010) Inc60000to75000 The income per capita between $60,000 to $70,000 Count US Census (2010) Inc75000to100000 The income per capita between $70,000 to $100,000 Count US Census (2010) Incover100000 The income per capita > $10,0000 Count US Census (2010) (2010)  Less75000 The average annual income per capita under $75,000 Dollar US Census (2010)  More75000 The average annual income per capita over $75,000 Dollar US Census (2010)  Less40000 The average annual income per capita under $40,000 Dollar US Census (2010)  More4000 The average annual income per capita over $40,000 Count US Census (2010)  The safety level of the levee system Nominal USACE (2018b) Acres The area within the levee system Acres USACE (2018a) $40,000 per year. These parameters were consolidated because many LIR levee systems had zero values for certain categories of race and income, and PCA analysis cannot be performed on parameters that have a value of zero. The $40,000 annual income was selected to consolidate the seven U.S. Census into two income bins because this is the value commonly used to distinguish between poverty and non-poverty level incomes for most family sizes in Illinois .

Principal Component Analysis and the Mann-Whitney-Wilcoxon Test
PCA is a common tool for data reduction (Jolliffe and Cadima 2016). In this study, PCA was used to identify the socioeconomic variables that explained the largest amount of variance to assist in the reduction of variables assessed during the socioeconomic parameter significance testing between levee systems with acceptable versus unacceptable levee safety ratings. For significance testing, the Mann-Whitney-Wilcoxon test (U-test) was employed in this study. This test analyzes if the dependent variable distribution is equivalent for both groups and consequently from the same population. This U-test is a non-parametric corollary of the independent sample t-test. This means that the test does not anticipate any qualities in relation to the dependent variable distribution in the analysis. Compared to the t-test, the U-test does not analyse average scores but rather the median score of two samples. Because of this, the Mann-Whitney does not need a specific distribution of the dependent variable in the calculation; unlike the 2-sample t-test. The U-test does not require a normal distribution or that the variances of the two populations be equal. The only assumptions for carrying out a U-test are that the two groups must be independent and that the dependent variable is ordinal. This makes U-test the appropriate analysis to use when the measured variable does not have a normal distribution and of an ordinal scale (Hollander et al., 2014). The purpose of using the U-test in this study is to compare the median from the dependent variables between the two acceptable versus the unacceptable levee system.

Socioeconomic conditions within the lower Illinois river valley levee systems
As of the 2010 census, ~10,300 people were inhabiting flood-prone areas protected by the 29 LIR levee systems evaluated in this study. The overwhelming majority (97%) of the population lived within levees with an acceptable levee safety rating. The majority (59%) of these leveeprotected floodplain inhabitants are between the ages of 16 and 64, they are nearly evenly split between males (51%) and females (49%), and the overwhelming majority (74%) are white. Nearly all of the non-white population (98%) lived within Sid Simpson Levee System that protects the region's largest economic and population centre, the City of Beardstown, which had an acceptable levee safety rating. The levee system with the highest average per-capita income was Clear Lake Special and Hanger Slough Levee System ($83,500). The lowest per-capita income was Valley City Drainage and Levee District ($48,500). Nearly all (99%) of the household with an income below the poverty line in Illinois (<$40,000) resided in levee systems with an acceptable levee rating. However, the per capita income was ~6% higher in levee systems with an unacceptable levee safety rating (Table 2; Figures 4, 5, 6, and 7). Land cover within these levee protected floodplain areas is 83.3% agriculture, 6.5% open water or wetlands, 5.2% developed, and 5.0% forested. The total building exposure within these levee systems is estimated to be nearly $1.0 billion, which includes ~5200 residential structures. The average estimated building value ranges from $49,000     up to $200,000, with an average of $100,000 (Figure 7). The Sid Simpson Levee system contains the most development; it has the highest average building value and contains most of the building exposure (53%) within the areas assessed. The overwhelming majority (~90%) of the residential structures are located within levee systems that have an acceptable levee safety rating ( Table 2). The agricultural profit estimated by Guida, Remo, and Secchi (2016) for each of the lower Illinois levee systems ranges from $74,000 to $3,370,000 per year, with an average of $1,277,000 per year (Figure 8). The agricultural profit within the levee system with an acceptable safety rating was 26% lower than the levee systems with an unacceptable rating ($1.31 versus $1.65 million). On average, the levee-protected agricultural lands along the lower Illinois River had a profit of $141 per acre. The agricultural profit per acre for a levee with an acceptable was ~8% lower than the levee systems with an unacceptable levee safety rating ($132 versus $164; Table 2).

Principal Components Analysis
PCA was used to identify the differences in the socioeconomic conditions between the lower Illinois River levee system evaluated in this study. PCA analysis revealed five components consisting of 19 variables explained 85% of the variance in these parameters between the evaluated levee systems. The five components were labelled based on the variables that comprised each component. The five components were exposure characteristics, population characteristics, wealth characteristics, development characteristics, and levee protection, which explained 26.3%, 24.2%, 18.2%, 8.9%, and 7.2% of the variance, respectively. The exposure characteristics component was comprised of the levee-protected area, population, average residential building value, number of residential homes, and total exposure variables. The population characteristics component was comprised of the percentage of the population between 16 and 64, percentage of the population younger than 16 and older than 64, percentage male, percentage female, percentage white, and percentage of the population making more than $40,000 a year. The wealth characteristics component was comprised of agricultural profit, agricultural profit per acre, average residential building value, percentage of the population making less than $40,000 a year, and average per capita income variables. The development component was comprised solely of the percent development variable, and the levee-protection component was composed of the levee-protection level variable (Table 3).

Comparison of socioeconomic variables between acceptable and unacceptable levee systems
The two-sample Mann-Whitney-Wilcoxon U-test was used to calculate the differences in the socioeconomic variables between levees that ranked acceptable versus unacceptable in USACE's levee safety inspections conducted prior to 2017. The distribution shape between each of the socioeconomic variables is only significantly different at the α ≤ 0.05 confidence interval for the parameter percentage of the population that identifies as white. However, using a confidence interval of α ≤ 0.2as a guide in selecting the socioeconomic parameters which may be important indicators of a levee system's safety rating indicates total population within a levee system, average per-capita income, number of residential homes, and total population might also be crucial socioeconomic characteristics related to a levee system's safety rating (Table 4). All other assessed variables had α > 0.2, indicating that they are likely not important socioeconomic characteristics related to a levee system's safety rating along the lower Illinois River.

Assessing differences in the socioeconomic conditions within lower Illinois river levee systems
Five components were identified from the PCA analysis that helped categorize the socioeconomic parameters, which can be used to differentiate differences in social and economic characteristics between lower Illinois River levees systems. These components included exposure characteristics, population characteristics, wealth characteristics, development characteristics, and levee protection. Overall, these five components were comprised of 19 socioeconomic variables explaining 85% of the variance in these parameters between the lower Illinois River levee systems. The exposure characteristics, population characteristics, wealth characteristics, development characteristics, and levee protection explained 26.3%, 24.3%, 18.2%, 8.9%, and 7.2% of the variance, respectively (Table 3).
Most of the socioeconomic parameters which make up these components have been identified as important characteristics that describe social and/or flood vulnerability of a given population of people (Cutter, Mitchell, and Scott 2000;Finch, Emrich, and Cutter 2010;Cutter et al. 2013;Remo et al. 2016). However, PCA analysis performed in this study also identified agricultural profit and agricultural profit per acre as important potential parameters in defining the differences between the studied levee systems. These parameters have largely not been explored in the flood vulnerability literature and warrant further investigation to determine their importance in both flood vulnerability assessment and levee system sustainability. These parameters are likely important because agricultural profit coupled with value of homes and buildings within levee-protected areas may serve as a proxy to assess the robustness of the tax base to support the maintenance and consequently the sustainability of a given levee system. The assumption being if there is not a sufficient tax base to maintain the levee system, the flood risk reduction the system provides will not be able to be sustained. Given that most the operation and maintenance of nearly all levee systems along the lower Illinois River and most throughout the US are funded by local government entities (i.e. levee districts, city, or county government), a robust local tax base is critical for the financial sustainability of these levee systems.

Assessing differences in socioeconomic parameters between levee systems with acceptable and unacceptable ratings
The Mann-Whitney-Wilcoxon U-test was used to determine differences in the socioeconomic parameters between levee systems, which were graded as at least minimally acceptable, versus levees that were unacceptable by USACE's levee safety inspections. This statistical test identified one parameter, the percentage of the population that was white was significantly different (α ≤0.05) between levee systems which were rated acceptable versus unacceptable. Other notable differences in the parameter medians (α <0.2) between levee systems with at acceptable verse unacceptable ratings for the total population within a levee system, race, average per-capita income, average value of residential homes, and the total population living within a given levee system. These results show the approach applied in Table 3. Summary of principal components analysis of socioeconomic, building exposure, and agricultural profit data for lower Illinois River levee systems.  this study can be used to determine important socioeconomic characteristics related to levee safety ratings. The spatial analyses performed in this study show most of the population and associated building infrastructure protected by lower Illinois River levees are located within levee systems, which have an acceptable safety rating. The analyses also showed that the population residing within levees with an unacceptable rating are nearly exclusively whites with incomes larger, on average. Agricultural profit per acre was also higher within levee systems which had an unacceptable rating (Table 2). These findings are counter to research that has shown more vulnerable populations (the poor and minorities) are generally found within areas with a higher risk in relation to flooding and other natural hazards (e.g. Cutter, Mitchell, and Scott 2000;Finch, Emrich, and Cutter 2010;Cutter et al. 2013;Remo et al. 2016). It was also interesting to note that average residential building values within levee systems that were unacceptable were higher on average (Table 2). This too was an unusual finding because buildings located within unprotected or less protected floodplains are typically discounted in value because of their increased flood risk (Kousky and Walls 2014).
While the results of this study's analyses were not designed to attribute causality to why a white, relatively wealthier population with more profitable farmland are found within lower Illinois River levee systems with unacceptable safety rating, future research should attempt to evaluate the robustness of this finding in similar levee-protected regions. Previous flood vulnerability research in Illinois has shown populations that inhabit rural floodplains are predominantly white. However, this study did not compare the relative wealth between different floodplain inhabitance . The results may suggest wealthier floodplain property owners within unacceptable levee systems are foregoing necessary levee maintenance and instead are relying on insurance and other government programs to pay for potential future flood damages. In the case of insurance, given the high subsidization rates for crop and property insurance through the FCIP and NFIP, it is possible the inhabitance of unacceptable levee systems are foregoing necessary levee maintenance and relying on insurance to protect against flood damages because it is more economical than taxing themselves at a sufficient rate to maintain their levees system to the federal safety standards.
Another potential explanation for this observation may be attributed to the idea of the 'charity hazard'. Browne and Hoyt (2000) define the charity hazard as the tendency of an individual at risk not to procure insurance or other risk financing because of a reliance on expected charity from others, such as government disaster relief programs, non-governmental organizations, and others. As applied to the situation along the lower Illinois River levee systems, the charity hazard would be not maintaining a levee system to a satisfactory condition to meet the requirements of receiving an acceptable levee safety rating.
Given recent substantial disaster and other government assistance experienced after large flood events along the lower Illinois River in 2001, 2013, and 2019(FEMA 2022 and the governmental obligations to repair levee systems under the Public Law 84-99 after such events at little to no cost to the levee maintaining entity (i.e. levee districts, local municipalities, and county governments; USACE 2018a), communities with an unacceptable levee system may be anticipating substantial future government assistance to recover from a flood event which exceeds the capacity of their deficient levee system. While the charity hazard has been explored in relation to purchasing of flood insurance (e.g. Browne and Hoyt 2000;Hung 2009;Robinson et. al 2021), it has not been explored in relation to the maintenance of levees and other structural flood mitigation systems. Understanding the interplay between government programs to support levee system maintenance, flood insurance, and crop insurance needs to be further studied to understand their impact on the economic sustainability of levee systems along the lower Illinois River and elsewhere where these systems are employed to protect people and property from flood risk.

Study limitations
The 2010 US Census socioeconomic data contained within FEMA's Hazus-MH were used for the comparison of socioeconomic variables between levee systems along the lower Illinois River Valley. Two primary issues result from using these data to address the research questions posed in this study. First, more than ten years have passed since the data was collected. Changes in housing and population have occurred over this time. However, this area is not a growing area, and a comparison of the 2010 to 2020 population total shows a slight decrease in population (~3.0% decrease). Also, the comparison of regional income and housing parameters are similar between the 2010 and 2020 U.S. Census data (U.S. Census 2020). Given these likely small changes, the conclusions presented in this work still provide essential and relevant insights into the research questions addressed in this study. The second issue with the socioeconomic data is that the U.S. Census Bureau has acknowledged that persistent data collection error happens because of overcounts and undercounts (U.S. Census 2012). The undercounts have for a long time been prevalent with ethnic and racial minorities, renters, migrant farmworkers, undocumented immigrants, and children (Hannah 2001;Passel 2005). While these issues likely added uncertainty to the analyses performed in this study, the U.S. Census data and associated Hazus-MH building inventory represent the best, widely accessible data available for the Lower Illinois River valley and should not substantially impact the findings from the study.
The agricultural profit data from Guida, Remo, and Secchi (2016) should be considered a minimum estimate for each levee district and the Lower Illinois River study segment. This is due to this study not accounting for secondary economic impacts and potential other revenue sources from agriculture, such as conservation payments or crop payments. Some of these secondary sources of income might substantially increase agricultural profits from levee-protected agricultural lands along the lower Illinois River. However, these agricultural profit values provide a useful comparison of the relative differences in agricultural profits between the evaluated levee systems.
While the number of levee systems evaluated in this study was sufficient to perform a two-sample Mann-Whitney-Wilcoxon U-test, the results from this test were not robust (i.e. the differences in the median of most of the parameters were not significant at the confidence interval commonly used for significance testing [e.g. >95% confidence interval]). Future studies should seek to evaluate a larger sample of levee systems to increase the statistical power of the analysis and to test the robustness of this study's findings. As for developing a predictive model using socioeconomic parameters to determine which levee system may or may not be able to sustain the maintenance of their levee system, while this study suggests socioeconomic parameters may be useful for such an endeavour, a much larger dataset, likely >100 levee systems will need to be developed to construct and validate such an empirical model.

Conclusion
The overarching goal of this study was to develop a better understanding of community's socioeconomic characteristics, which allows for the sustainment of levees systems for the mitigation of their flood risk. To achieve this goal, we sought to answer three research questions. The first was, what are the socioeconomic characteristics which differentiate the communities with levee systems with acceptable versus unacceptable safety ratings along the lower Illinois River? The analyses revealed that the percentage of the population that was white was significantly different (α ≤0.05) between levee systems which were rated acceptable versus unacceptable. Our analyses also revealed substantial differences (α <0.2) in average per-capita income, the average value of residential homes, the total population within a levee system, and the percentage of the population with income below the poverty line (<$40,000) were also noted. The results suggest that there are notable differences in the socioeconomic characteristics between the communities protected by a levee system with acceptable verse unacceptable levee safety rating, and the approach applied in this study can identify them.
The second question we attempted to answer is, do levee systems that protect poorer communities have systems that are rated as unacceptable where relatively wealthier populations are protected by acceptable levee systems? Overall, the majority (97.0%) of the populations located within the lower Illinois Levee systems are located within a system with an acceptable rating. In addition, our results show that nearly all (99.0%) of those with an income <$40,000, which is below the poverty line for this region of Illinois, and the nonwhite population (99.9%) were located within levee systems with an acceptable levee rating. Interestingly, the portion of the population which resided in levee systems with an unacceptable rating were white and had higher average incomes than those populations that resided in a levee system with an unacceptable rating.
The third question we assessed was, do levee systems that have acceptable ratings protect agricultural areas which possess higher agricultural profits? Average agricultural profit for a given levee system was found to be 26% higher in systems with an unacceptable levee safety rating versus an acceptable rating.
The observations that white, relatively wealthier, with more profitable agricultural lands were located within levees with an unacceptable safety rating were counterintuitive findings because they are counter to research that has shown more vulnerable populations (the poor and minorities) are generally found within areas with a higher risk in relation to flooding and other natural hazards (e.g. Cutter, Mitchell, and Scott 2000;Finch, Emrich, and Cutter 2010;Cutter et al. 2013;Remo et al. 2016). In addition, those with higher levels of income are generally more likely to have higher levels of personal financial capital and a greater likely hood of access to external governmental resources (Kunreuther and Pauly 2004;Finch, Emrich, and Cutter 2010;Cutter et al. 2013;Rufat and Botzen 2022) to support the maintenance and consequently the sustainability of their levee systems. These findings may suggest wealthier floodplain property owners within unacceptable levee systems are foregoing necessary levee maintenance and instead are relying on insurance and/or other government programs to pay for potential future flood damages.