Nutrient Retention in Ecologically Functional Floodplains: A Review

: Nutrient loads in fresh and coastal waters continue to lead to harmful algal blooms across the globe. Historically, floodplains—low-lying areas adjacent to streams and rivers that become inundated during high-flow events—would have been nutrient deposition and/or removal sites within riparian corridors, but many floodplains have been developed and/or disconnected. This review synthesizes literature and data available from field studies quantifying nitrogen (N) and phosphorus (P) removal within floodplains across North America and Europe to determine how effective floodplain restoration is at removing nutrients. The mean removal of nitrate-N (NO 3 − -N), the primary form of N in floodplain studies, was 200 (SD = 198) kg-N ha − 1 yr − 1 , and of total or particulate P was 21.0 (SD = 31.4) kg-P ha − 1 yr − 1 . Based on the literature, more effective designs of restored floodplains should include optimal hydraulic load, permanent wetlands, geomorphic diversity, and dense vegetation. Floodplain restorations along waterways with higher nutrient concentrations could lead to a more effective investment for nutrient removal. Overall, restoring and reconnecting floodplains throughout watersheds is a viable and effective means of removing nutrients while also restoring the many other benefits that floodplains provide.


Introduction
Despite global efforts to reduce nutrient loads, harmful algal blooms (HABs) and hypoxic zones are occurring more often, in new places, for longer durations, and at different toxicities [1]. Most are caused by nutrient pollution [1]. The impacts of these algal blooms can be devastating for human and animal health, wildlife habitat, and economics. Nutrient loads have been increasing, especially from nitrogen (N) and phosphorus (P) fertilizer use, river alterations, field and bank erosion, human and animal waste, and expanding aquaculture practices [1][2][3][4][5][6][7][8][9]. Common forms of nutrients in waterways are nitrite (NO2 − ), nitrate (NO3 − ), ammonium (NH4 + ), organic N, particulate P, and soluble reactive phosphorus (SRP). However, the proportions of each form of these nutrients are in constant flux, and the occurrence of each algal bloom depends on the algal species, environmental conditions, other organisms creating those conditions, and the nutrient concentrations and forms present [1]. Each form of N and different ratios of N:P lead to the growth of different algal species and varying levels of toxicity from some species [10][11][12][13][14]. In addition to toxicity from HABs, high concentrations of NO3 − in drinking water are a risk to human health and NO3 − has been prioritized for nutrient removal efforts in many areas, especially in the Mississippi River Basin [4,[15][16][17].
Efforts have been made to reduce nutrient loads from point sources across the globe but some of those sources of nutrients have increased, for example, where populations have increased but sewage treatment has not improved [1]. Furthermore, there is still a significant need for reducing non-point sources of nutrients. Current attempts to reduce non-point nutrient pollution include practices such as riparian buffers, cover crops, contour farming, fertilizer management, wetlands, woodchip bioreactors, and others [18][19][20][21][22][23][24][25]. However, there need to be more options in the riparian corridor that are highly effective at removing nutrients while providing other needed benefits such as water storage, carbon sequestration, fish and wildlife habitat, and recreational benefits. [26].

Floodplain Processes Driving Nutrient Cycling
Ecologically functional floodplains are connected to the river without barriers that prevent flooding, and they are capable of supporting plant and animal communities native to that region. They can either be naturally occurring or restored via human action to reconnect the floodplain and restore the natural ecosystem features. Ecologically functional floodplains provide multiple processes that improve nutrient retention ( Figure 1). Many of these processes are similar to those driving nutrient cycling in treatment wetlands [27]. . This is a depiction of a disconnected and non-functional versus an ecologically functional floodplain. Ecologically functional floodplains flood regularly without constructed barriers or incised main channels that prevent water from entering the floodplain during high-flow events. They are also capable of supporting plant and animal communities native to that region.
The best permanent removal mechanism for N is usually denitrification if N is in the form of NO3 − or first converted to it [28]. Denitrification is a microbially facilitated reduction of dissolved NO3 − to a gaseous form of N usually under anaerobic conditions. Most denitrifying microbes use NO3 − as an electron acceptor for respiration when oxygen is depleted, although some microbes prefer NO3 − over oxygen [27]. When soil is saturated for extended periods of time, the oxygen becomes depleted and microbes switch to using NO3 − instead of oxygen. Several other conditions are necessary for denitrification including abundant carbon, warm temperatures, readily available NO3 − , and a sufficient population of microbes [29]. If water levels fluctuate to the extent that oxygen is replenished, then most microbes will switch back to oxygen and could even increase the NO3 − concentration by converting other forms of N to NO3 − [27]. Because particulate P is the primary form of P in large rivers including the Mississippi River [4], sedimentation, soil accretion, and burial are the primary means of removing P throughout those basins [30]. Floodplains are thought to be net sinks for sediment [31]. However, in basins where SRP is the primary form, biological uptake is likely the primary means of removing P. Generally, the dissolved form of SRP is the form that is most available for uptake by plants and algae and contributes disproportionally to algal blooms [32]. It can exchange between the dissolved or particulate phases depending on the conditions and through processes such as decomposition, biological uptake, redoximorphic release, or sedimentation and accretion [18]. Dissolved forms are often released from soil during anaerobic conditions; changes to pH can also drive release of P from soils that fluctuate between wet and dry conditions. Soil properties that favor retention of P through iron and other chemical complexes strongly influence the ability of floodplain soils to retain P [33].

Limitations to the Nutrient Cycle in Floodplains
With extensive alterations to riparian corridors associated with human settlement along rivers, many limitations have impeded the cycling of nutrients in floodplains. Floodplains have historically been some of the most nutrient-rich soil for agriculture and provide land with easy access to the riverfront. Rivers have undergone drastic changes to accommodate transportation, urban development, and agriculture. Some of the earliest activities responsible for changes to the floodplain include timber harvesting, agriculture, and river impoundment. For example, surveys from the Mississippi River above its confluence with the Illinois River indicate that 56% of the floodplain was forested and 41% was prairie before European settlement. The floodplains were reduced to 35% forest cover by 1938 and 6% prairie by 1994 [34]. Llewellyn [35] estimates that only 20,000 km 2 of the formerly 85,000 km 2 of bottomland hardwood forests along the Mississippi River remain. By the 1990s, an estimated 90% of the entire Mississippi River floodplain had been disconnected from the main channel due to levee construction [36] and up to 90% of floodplains across North America are cultivated and thus have lost most, if not all, ecological functionality [37]. In the Upper Mississippi River Basin alone, there are over 8000 miles of known levees disconnecting floodplains from their river channels [38].
Floodplains typically become wider and provide more flood storage moving downstream as streamflow and river size increase. Lower river reaches tend to have a longer floodwater retention, promoting nutrient removal. Floodplain soils are dynamic, as sand, silt and clay are deposited, or resuspended, by the river. Floodplain soil is often layered as the sediment is deposited at different rates that depend on grain size during each flood event. Larger grain sizes (typically sands or gravel) deposit first as swift moving water slows. Silt and clay drop out of the water column later, as velocities are reduced, typically further away from the main channel boundary [30]. Phosphorus retention is highest in floodplain soils that have more silt and clay because P can bond with the chemically reactive surfaces of clay and silt-sized organic matter particles. Soils with abundant iron, aluminum and manganese often form insoluble phosphate compounds that remove P from more mobile, soluble forms. Depending on pH and oxygen levels, P may dissolve and become available again, as SRP, particularly during anaerobic conditions [39][40][41].

Selection Criteria for Literature Search
With the extent of floodplain acres currently disconnected from the Mississippi River and other rivers across the globe, nutrient removal by floodplains has been greatly reduced. However, attempts to quantify and project these potential reductions are still limited. Restored and constructed wetlands have been thoroughly reviewed several times already [42][43][44][45] and a small collection of floodplain studies have been reviewed [18]. This paper synthesizes the literature and data available from field studies quantifying N and P removal within floodplains across the globe. Floodplains from rivers and streams of all sizes were included. Laboratory experiments, mesocosm studies, and modeling projects were not included.
The review started with a systematic search using literature database search engines (Google Scholar, Worldcat, Web of Science, etc.) and university library databases to find peer-reviewed publications and technical papers as well as reference lists within other papers. The searches used terms including "phosphorus," "nitrogen," "nutrients," and each nutrient form combined with terms including "floodplain," "riparian," "wetland," and synonyms of these terms. The initial list included over 200 peer-reviewed articles, technical papers, and graduate student papers. Of those sources, 40 provided new empirical data from in-field measurements ( Figure 2). Those 40 sources measured NO3 − removal in 29 floodplains and total or particulate P in 42 floodplains-10 of which were particulate P only. The definition of a floodplain varied throughout the literature reviewed, but in this review an ecologically functional and reconnected floodplain is defined as a low-lying area adjacent to a river that becomes inundated regularly during high flows or floods, and it is capable of supporting plant and animal communities native to that region. High areas adjacent to rivers that are not flooded by the river in its current flow regime, known as terraces, were not included in this analysis. Connected floodplains containing unnatural surfaces void of vegetation, constructed treatment wetlands receiving no stream flow, and two-stage ditches were a few systems that were also excluded from this review since their connectivity was not fully restored or they were very small in areal extent. Restored oxbow wetlands and floodplains reconnected using large, riparian water pumps were included in the review.

Evaluating Nutrient Removal from Previous Studies
Nutrient removal studies occurring specifically in floodplains varied in the metrics and methods they used, thus complicating cross-study comparisons. The original goal was to collect data from floodplain studies for comparisons using the plug-flow area-based first-order model [46,47], a model that has become widely used to represent removal kinetics, but too few studies provided all the variables needed for the equation. The review therefore used simpler evaluations of the data. The studies most commonly expressed results of removal effectiveness as changes in concentration, mass balances, mass accumulations, percent removal, or denitrification measurements. The review of these results focused primarily on studies that summarized mass balances within floodplains (i.e., mass in versus mass out or mass removed). Studies measuring changes in concentration from upstream to downstream points in the main channel were excluded due to nutrient spiraling being complex and the difficulty in assuming reductions took place due to proximity to the adjacent floodplain. Some studies measured nutrient cycling in vegetation and the impact of plant communities on nutrient removal [18,[48][49][50][51][52]. From each study, removal values and potential controlling factors were collected for statistical analyses.
Using the Spearman rank correlation test, this review compared the loads and removal rates of N and P among wetland sizes, nutrient loads, and hydraulic load. Statistical analyses included ANOVA (normal distribution) or the Mann-Whitney U test (non-normalized distribution) to compare categories of floodplains and their removal rates based on flooding frequency. When analyzing flooding frequencies, the floodplains were grouped into categories of permanent inundation or seasonal inundation. Permanent inundation did not mean the entire floodplain was impounded, but it meant there was standing water somewhere in the floodplain year-round. Seasonal inundation meant the floodplain dried at some time each year. Analyses were performed in Rstudio ® and Microsoft ® Excel [53].

Seasonal Variations and Climate
Most floodplains are inundated seasonally rather than the entire year. In line with this hydrological variation, studies varied in the time frame over which they presented their nutrient removal rates. Many results were presented as mass removed per day, but this does not explain how much total mass those floodplains could remove annually. Therefore, the daily reductions from a study were converted to annual reductions by multiplying the daily reductions by the average number of days per year of inundation reported in that respective study. If a study did not mention the number of days of inundation per year, it was not included in these comparisons. This conversion to annual retention also aided in accounting for varying climates. All studies included in the statistical evaluation were from the northern hemisphere, but some of the floodplains were located far enough south to remain unfrozen year-round. In these cases, the floodplains were usually inundated for more days each year, but the mass retentions would still be multiplied by the number of days of inundation per year reported in the study.

Nutrient Forms
Nitrogen and P both appear in multiple forms in the environment. However, NO3 − was the most prevalent form of N in removal studies, and TP and particulate P were the most common forms of P reported. Although the other forms of these nutrients are important to study as part of their cycles in floodplains (i.e., NH4 + , organic N, and SRP), they were not evaluated in this review because there were too few studies on those nutrients. While TP includes SRP, TP and particulate P were analyzed together in this review due to the majority of TP being particulate and removal being through deposition.

Size
There were large variations in sizes of floodplains studied. Larger rivers typically have larger floodplains [54]. Although floodplain-to-watershed area ratios would have been helpful for assessing nutrient removal efficiency, the watershed area was seldom mentioned. In order to reconcile the variation in floodplain size, the mass removal rates were divided by the floodplain area.

Nutrient Removal Results
Mass reductions in NO3 − -N in floodplains ranged from 2.35 to 962 kg-N ha −1 yr −1 , with a mean reduction of 200 (SD = 198) kg-N ha −1 yr −1 . Total P reductions ranged from a net release of 14.6 to net retention of 130 kg-P ha −1 yr −1 , with a mean retention of 21.0 (SD = 31.4) kg-P ha −1 yr −1 . Only 1 of the 41 floodplains in the studies had a net release of P rather than net retention ( Table 1). The average concentration of NO3-N entering floodplains in 14 studies was 2.3 mg L −1 (SD = 1.7).
As the hydraulic load increased, the mass removal of NO3 − also increased (y = 5.6568x + 79.61, ρ = 0.79, p = 0.004, Figure 3) but the percent removal of NO3 − decreased, i.e., they were inversely related (y = −8.255ln(x) + 84.542, ρ = −0.76, p = 0.007, Figure 4). For TP, there was a positive relationship between the hydraulic load (m yr −1 ) and TP mass removal (kg ha −1 yr −1 ; y = 5.8913ln(x) + 4.8823, ρ = 0.75, p = 0.003, Figure 5).    Although there was no significant relationship between the inflow concentration and retention in this review, mass removal increased as the NO3-N load into the floodplain increased (y = 0.2946x + 115.67, ρ = 0.70, p = 0.004; Figure 7). The mean percent reduction in NO3 − -N in floodplains in this review was 64.2% (SD = 19.7%), and the mean percent reduction in TP in floodplains was 26.5% (SD = 24.8%; Table 2). There was a negative relationship between NO3 − -N load and percent removal. As the NO3 − -N load increased, the percent removal decreased exponentially (y = 68.75 × , ρ = −0.66, p = 0.009; Figure 8). For TP, there was no correlation between the TP input load and percent removal, but the mass removed did increase as the input load increased (y = 0.0326x + 11.936, ρ = 0.63, p = 0.01; Figure 9).   Vegetation harvest was not a common practice for removing N from floodplains because it removed such a small portion of the N load. There were, however, some studies about harvesting P in floodplain vegetation. From the 8 studies that included vegetation harvest, the mean P removal was 8.8 kg-P ha −1 yr −1 in the floodplain (Table 3). With the mean TP load in the studies reviewed being 15.8 kg-P ha −1 yr −1 (SD = 16.1, 2 outliers removed), this could be a significant portion of the P load, but vegetation has a limit for how much it assimilates into tissue. There are many other biomass studies in constructed or highly managed wetlands, and two of those measured P removal exceeding 15 kg-P ha −1 yr −1 [55] and 20-60 kg-P ha −1 yr −1 [56] by harvesting Typha sp. stands. The site with the greatest P assimilation in floodplains included alfalfa, switchgrass, and cottonwood trees that were harvested every four years [48]. Table 3. Mean phosphorus mass removed from floodplains through vegetation harvest. Ranges reflect the range of means from multiple floodplains in the study listed.

Discussion
Results from this review have major implications for floodplain reconnection initiatives, nutrient reduction strategies, and floodplain restoration designs. Restored and reconnected floodplains can remove significant masses of nutrients from rivers and streams. When properly restored, these systems also provide habitat for fish and wildlife, floodwater storage, recreation, and other benefits. The NO3 − removal results were much lower in this review than those found by Dee et al. [59], but P removal estimates in this review were similar. Some of the key findings from this review can be described in the following categories.

Hydrology
Reconnecting floodplains to rivers and designing them with the correct flow regime is key for properly restoring the function of floodplains. Hydraulic load and water depth play significant roles in nutrient removal effectiveness in wetlands and floodplains, and there are flow rate recommendations for effective treatment wetlands [27,60]. Because nutrient removal increases as hydraulic load increases, restoring floodplains with greater connectivity to the river could prove a more effective investment for floodplain reconnection projects. However, that effectiveness decreases as the hydraulic load becomes too great. Optimal hydraulic load from floodplains in this review was roughly less than 20 m yr −1 for NO3 − removal and less than 50 m yr −1 for P removal. While the volume of water entering the floodplain was an important factor for nutrient removal, the duration of inundation was also important.
Floodplains with year-round inundation somewhere on the floodplain (i.e., restored wetlands within the floodplain) removed significantly more N each year than floodplains that would dry completely for any period of time each year. This follows the expectations that flow regimes will impact nutrient removal effectiveness, but the important implication is that flooding lawns, crop fields, or other areas as temporary floodways would not be as effective at removing nutrients as permanently restoring and reconnecting floodplains. Restoring ecological functionality of floodplains would be more effective at nutrient removal, but restoring some permanently inundated wetlands would provide even more benefit. These results could be related to permanent wetlands containing sustained denitrifying microbe populations in the wet areas, organic-rich soils, or fringes of the wetlands [61][62][63][64][65][66]. Some have raised concerns about connecting floodplains on small-order streams because flood duration is shorter on smaller streams and therefore may not provide enough time for impactful nutrient removal or may even lead to a greater release of nutrients [67]. However, the floodplains on many small-order streams may be lacking a permanent wetland due to channelization and land use change along stream buffers. Floodplain reconnections and restorations along these smaller streams may require a permanently inundated or saturated wetland restoration component for the floodplain to effectively remove nutrients.

Microbes
In research investigating the role of microbes in floodplains, Argiroff et al. [61] measured a greater genetic potential for denitrification in floodplains that were more frequently flooded than those that were only occasionally flooded. Hernandez and Mitsch [68] measured greater denitrification in permanently flooded marshes than in occasionally flooded areas. Tomasek, Staley, et al. [62] concluded that more frequent inundation is likely to lead to greater denitrification and greater denitrifying microbe abundance. However, a higher population of denitrifying bacteria does not always lead to a correlatively higher denitrification rate, and sometimes the greatest denitrifier gene abundance is in the littoral zone or periodically inundated area rather than in the main stream channel or permanently inundated sediments where the greatest denitrifying potential was measured [63][64][65]. It is possible that these fringe habitats with fluctuating wet and dry conditions could be denitrification hot spots where denitrifying microbes are most active after a flood pulse [66]. Restoring more areas of permanent inundation in floodplains would create more fringe areas in return, especially if there is a high ratio of fringe to pooled areas.
More topographic and geomorphic diversity in the floodplain could capture the greater denitrification potential found in permanently inundated areas and enhance populations of denitrifying microbes in more fringe areas. It is important, however, to avoid creating permanent pools that have limited fringe zones and are too deep. These deep pools can limit vegetative growth, the organic matter provided by the vegetation, and the bacteria dependent on organic matter [61]. Furthermore, areas of inundation should have slow flow through them to avoid flushing out too much organic carbon and microbes attached to the sediment [69]. Residence times in the floodplain should be at least five days to maximize denitrification [27]. Restoring floodplains with swale wetlands in them that sit below the water table coupled with ridges that remain above the water table may therefore help to maximize denitrification.
It is also important to design floodplains so the majority of sedimentation occurring in the 50 m closest to the river does not fill swales that were meant to be permanent denitrification wetlands [70,71]. However, creating ridges that will slow down the flow and allow for better sedimentation will provide better P retention. While vegetation can remove a large portion of P from the soil, sedimentation and the subsequent burial of particulate P is likely the best means of permanently retaining large loads of P in the floodplain [72].

Vegetation
While sedimentation and accretion are likely the best means for retaining P in floodplains, vegetation will provide the organic matter in the soil to drive denitrification and assimilate some of the nutrients into plant tissue. Vegetation type varied among the studies, so this review was not able to determine the best community for nutrient removal. Therefore, the best plant community for floodplain restorations should depend on location and historic communities more than optimal nutrient removal at this time. However, as explained more thoroughly by Dee et al. [59], each community type may establish more or less successfully based on the flood regime and be more or less successful at removing nutrients [72]. Herbaceous communities with dense ground cover, like grasses and rushes, are better at slowing flow velocity, increasing retention time, promoting sedimentation, and burying P [19,73]. Forested floodplains are likely to be better at assimilating nutrients more permanently, establishing growth better in less-frequently flooded areas, tolerating larger floods and sediment loads, and removing more NO3 − from groundwater [58,74,75].
Harvesting vegetation can have a positive impact on removing P from the floodplain, but there are still unknowns about its impact on the plant community and if there can be a market for selling harvested vegetation. If the plant community is already overwhelmed by invasive species, such as Typha sp., a harvesting regime could improve diversity [76,77]. Although highly productive species such as Typha sp. were highly effective at assimilating P, they are usually invasive in North America and would be a detriment to the native biodiversity if planted specifically for nutrient removal. If vegetation is harvested, there still needs to be an investment in the proper harvesting equipment, someone who wants to purchase the biomass, and good timing to be able to access the floodplain when it is not flooded. If vegetation is not harvested, some of it can decompose and release the nutrients it assimilated the previous season. However, the organic material is oftentimes accreted into the soil layers where more sediment deposits on top and buries it into storage [72].

Nutrient Loading
Some of the most commonly referenced and most effective NO3 − removal wetlands in the Midwest United States are the Iowa Conservation Reserve Enhancement Program (CREP) wetlands [78][79][80]. When calculating the mass of N removed for each hectare converted to wetland pooled area and buffer of 79 CREP wetlands, they removed an average of 450-500 kg-N ha −1 yr −1 through 2016 [78]. Of the 16 sites consistently monitored, they were significantly more effective than all the floodplains in this study (mean = 224 kg-N ha −1 yr −1 , p = 0.002), but they were not significantly different than floodplains that had permanent inundation (mean = 312 kg-N ha −1 yr −1 , p = 0.1). Some smaller NO3 − treatment wetlands in Illinois and Minnesota ranged from an average of 166 to 619 kg-N ha −1 yr −1 [81,82]. This difference between load removals is possibly related to the incoming NO3 − -N concentration and load to each wetland or floodplain. The average concentration entering the Iowa CREP wetlands was 14 mg L −1 , the Illinois wetlands was 8.5-13.0 mg L −1 , and the Minnesota wetland was 15.6 mg L −1 . Concentrations of NO3 − -N in tile drainage discharge can be very high, oftentimes exceeding 20 mg-N L −1 [24,83,84] or even 40 mg-N L −1 in some areas [27]. The mean concentration entering the floodplains in this review was 2.3 mg L −1 (SD = 1.7). The lack of a correlation between inflowing concentration in floodplains and retention may have been due to a limited range of and difference among inflow concentrations, but there was a linear correlation between NO3-N load and NO3 − -N removal effectiveness. As the NO3 − -N load increased, the mass removed increased, but the percent removal decreased. Similarly for TP, as the load increased, the mass removal increased. In the same way, as the hydraulic load increased, so did the masses of N and P removal increase while percent removal decreased. Therefore, prioritizing floodplain reconnections along rivers or streams with higher concentrations may lead to greater masses of N and P removed without sacrificing percent removal effectiveness. Although P removal increased as load increased, floodplains had a maximum capacity for P removal, but floodplains did not lose nutrients as the flow increased.

Nutrient Release from Floodplains
Due to the dynamic nature of floodplains, most processes of nutrient storage are not permanent. However, contrary to any concerns that may arise about floodplains releasing nutrients or scouring currents resuspending them during the next flood, this release did not exceed retention in the literature reviewed. Only one TP study measured a net release of P [85], and one study measured an increase in both nutrients in the river channel next to a floodplain during some seasons of the year [86]. However, SRP contributes disproportionally to downstream algal blooms and is an important consideration in some situations [32]. Over long time-scales sediment and nutrients are returned to the river via the process of lateral stream migration and streambank erosion which removes part of the floodplain each year [30,40]. While sediment and nutrient resuspension from the floodplain are a part of the nutrient cycling process, it was not a common observation to see release be greater than retention in these studies.

Future Research
There need to be more studies on the nutrient removal effectiveness of floodplains. This review is necessary to provide initial estimates of floodplain removal effectiveness and better understand some of the trends and factors impacting that effectiveness, but the calculations are based on fewer than 50 studies. With the limited number of studies included in the statistical analyses, some trends could be influenced by outliers or could change if more values are included. Because P removal in floodplains is likely mostly from deposition [30], particulate P and TP were grouped together in this analysis, but they should be separated in future studies in order to better distinguish particulate and soluble P removal. Furthermore, as results from more empirical studies become available about nutrient cycling in floodplains, a review using the areal rate constant would be beneficial. It has become a widely used model to represent nutrient removal kinetics, but not enough studies provided all the variables needed for the equation at this time [46,47]. There also need to be more studies on the retention of other forms of N and P in floodplains. The most common forms studied, by far, were NO3 − and total or particulate P. However, recent studies have revealed that NH4 + and ratios of nutrient forms to each other may play larger roles in influencing which algae are present in HABs and how toxic the bloom may be compared to understandings of HABs in decades past [1,13,14]. Furthermore, studies from only North America and Europe were included in this review due to the difficulty finding studies from elsewhere, written in English and meeting the selection criteria. More research is needed on nutrient removal in other parts of the globe, particularly tropical rivers in the southern hemisphere.

Conclusions
Hypoxic zones, HABs, and accelerated eutrophication continue to be problems in bodies of water across the globe as nutrient loads remain at high levels. While many efforts are underway to reduce those nutrient loads, especially from point sources and agriculture, more options need to become available and be proven effective. Restoring and reconnecting floodplains to rivers and streams throughout watersheds is a viable and cost-effective practice for removing nutrients. The following design considerations could improve removal effectiveness:

•
Engineer the floodplain to optimize hydraulic load. Although more flow across the floodplain could lead to a greater total mass of nutrients removed, the floodplain will lose effectiveness, (the percent removal) as flow rates increase.

•
Incorporate a permanently inundated wetland in the floodplain area to improve NO3 − -N removal.
• Ensure geomorphic diversity across the floodplain to increase both N and P removal due to improved microbial habitat for denitrification and more areas for sedimentation and accretion of P.

•
Restore dense vegetation to improve nutrient removal by providing organic matter for denitrifying microbes and slow water flow for better sedimentation and accretion.

•
Harvest vegetation from floodplains where feasible to aid in P removal, but caution should be taken due to the unknown impact on native plant communities.

•
Restore floodplains along waterways with higher concentrations of nutrients to increase the load of nutrients into the floodplain. Limit flow to maximize nutrient removal.
Reconnecting floodplains to their respective rivers or streams and restoring their geomorphology and vegetation would greatly reduce nutrient loads. Floodplains are dynamic systems that provide the benefits of nutrient retention, water storage, fish and wildlife habitat, aquifer recharge, and recreation. As more empirical studies are conducted on nutrient removal in floodplains, especially on retention of other forms of N and P, more design considerations could be incorporated into these projects to achieve better nutrient removal results across the globe.