1. Introduction
Water quality impairment and inland flooding are multi-billion-dollar problems that are expected to worsen in response to climate change. The World Resources Institute estimated that the number of people worldwide impacted per year by riverine flooding events will more than double from 2010 (65 million) to 2030 (132 million [
1]). Likewise, the annual costs of flooding in riverine urban areas are expected to more than triple, from USD 157 billion in 2020 to USD 535 billion by 2050 [
1]. A separate study estimated that using natural infrastructure to protect against climate change threats, such as flooding, could save USD 248 billion at a cost of only half that of equivalent grey infrastructure [
2]. Meanwhile, agriculture globally is the largest contributor to water pollution, causing damage to ecosystems and human health and costing billions of dollars annually [
3]. In the U.S.A. alone, nitrogen losses from agriculture to surface and groundwater are in excess of USD 150 billion per year [
4].
In this paper, we review the opportunity for the use of natural infrastructure in rural landscapes to provide water quality and flood mitigation benefits for downstream communities. We define “natural infrastructure” as:
Durable structural and/or native perennial vegetative measures embedded in a landscape or riverscape that are inspired and supported by nature, restore ecological processes, and deliver multiple environmental benefits to downstream communities.
We use the term “natural infrastructure” rather than “nature-based solutions” or “natural flood management” because our primary focus is on ecosystem services—water quality and flow regulation—that have traditionally been delivered through heavily engineered “grey infrastructure,” which attempts to control, rather than work with nature. We expand the definition of natural infrastructure beyond structural measures in order to capture the multiple environmental benefits to downstream communities of converting annual cropland to native perennial vegetation. Our emphasis on perennial vegetation, with its implications of longevity, excludes annual measures such as cover crops that are part of a farmer’s crop rotation.
While some of the natural infrastructure measures that we describe share similarities in terms of ecological processes with plot- and local-scale “green infrastructure” measures used in urban landscapes [
5], our context—the rural landscape and the ditches, streams and rivers that flow through it—is quite different. In addition, where much urban “green infrastructure” has necessarily been developed at the site scale in response to parcel-by-parcel urban development, we emphasize using natural infrastructure as part of a strategic and systemic approach to planning at the watershed scale. Watershed-scale design is critical to respond to the interests of multiple stakeholders and to account for both synergies and trade-offs across multiple ecosystem services [
6]. Likewise, while there is a growing body of literature describing the potential role of natural infrastructure in mitigating storm surge and sea level rise in coastal areas (e.g., [
7,
8]), there has been much less discussion of its role in mitigating pluvial and overbank flooding in inland areas, which is our focus.
Our emphasis is on natural infrastructure (NI) measures within rural landscapes that promote both the storage and slow release of water and the physical, chemical and biological processes that remove or transform waterborne pollutants. From a flood mitigation perspective, NI includes a range of interventions that: (i) reduce runoff generation, (ii) increase water storage, and/or (iii) attenuate flow from hillslopes to small streams and within the larger hydrologic network (e.g., by disconnecting runoff channels or by increasing channel roughness). These same physical modifications to hydrologic flows are likely to be effective in mitigating particulate pollutants such as phosphorus and sediment which largely follow surface flow paths. For removing dissolved pollutants such as nitrate, however, additional biochemical processes such as denitrification are required. Where dissolved pollutants are challenging, NI measures may only be effective for water quality improvement if they create the appropriate biogeochemical environment—for nitrate, this typically means multi-day residence times under reducing conditions in the presence of a carbon-rich substrate. Thus, NI measures suitable for flood mitigation may or may not provide water quality benefits. Likewise, NI measures that create an appropriate biogeochemical environment for pollution treatment may or may not provide flood mitigation benefit, depending on whether they also store water and delay its release downstream. For our review of water quality impacts, we chose to focus on nitrogen because of its impacts on both ecosystem and human health, for example, the degradation of drinking water. Furthermore, N is more important for ocean estuary degradation, while P is more important for freshwater body degradation [
9].
While some types of NI, such as wetlands, have been identified as potential solutions to water quality problems in agricultural landscapes (e.g., [
10]), there has been less discussion of the potential role of such measures in flood mitigation. While there has been growing interest in “natural flood management” in the U.K. and Europe, little of this work has addressed the associated water quality benefits (see [
11,
12] for examples wherein both concerns are addressed). As a result, downstream communities—which often experience both water quality and flooding problems—lack information on how NI can be used to address both issues. Our paper is intended to help fill this information gap.
Our review was guided by our experience in the Mississippi River Basin (MRB) of the central U.S.A., where downstream communities are impacted by both water quality degradation and more frequent and severe flooding. Agricultural intensification (including changes in land use, artificial drainage and nutrient additions) as well as climate change [
13,
14,
15] have contributed to increased flooding [
16] and increased harmful algal blooms [
17]. Agriculture is extremely important in the MRB, with over two thirds of the region’s land identified as farmland [
18], and the region’s economy contributes to 18% of the U.S. gross domestic product [
16]. While these water quality and flooding problems are not unique to the MRB, the region’s role as a major global exporter of corn, soybeans and wheat creates the potential for tension over land use, with agricultural producers concerned about any loss of productive cropland. An as-yet unanswered question is how much cropland would need to be converted to natural infrastructure to achieve regional goals related to water quality improvement and flood risk reduction. With this in mind, our review places particular emphasis on identifying those NI measures that could deliver the greatest environmental benefit with little or no cropland conversion.
We reviewed scientific literature describing the use of NI in agricultural landscapes in North America, Europe and the U.K., compiling data on (i) the effectiveness of these measures in improving water quality and/or mitigating floods and (ii) the strength of the evidence for such benefits. Based on our review, we identify a suite of NI measures that, if implemented in a strategic and systemic approach, can deliver both water quality and flood mitigation benefits while minimizing the loss of productive agricultural land.
2. Materials and Methods
We conducted a search of peer-reviewed papers in Google Scholar dating from 2000 to 2020 using the following search terms, individually and in combination: [agricultural land use conversion, agricultural landscapes, alluvial forest restoration, backwater reconnection, bottomland hardwoods, constructed wetlands, depressional wetlands, detention ponds, engineered wetlands, farm ponds, flood attenuation, flood mitigation, floodplain, floodplain forest restoration, floodplain restoration, flood risk reduction, flooding, leaky barriers, levee removal, levee setback, natural flood management, natural flood management measures, natural infrastructure, nitrogen removal, nutrient retention, offline ponds, overland flow interception, oxbow reconnection/restoration, perennial vegetation, prairie pot-hole wetlands, re-meandering the river, retention ponds, riparian buffers, riparian wetlands, riparian forested wetlands, river restoration, runoff attenuation features, saturated buffers, two-stage ditches/channels, vegetated ditches, water quality, wetland restoration]. Occasionally, we found fewer than 2 papers for a specific NI measure in the specified time frame. In those cases (riparian forest buffers and farm ponds), we expanded the search to include any time period. In addition, we used a snowball technique of manually tracing references and citations from recent papers. Our search retrieved a total of 145 papers for further review.
Subsequently, we screened these papers to identify those reporting on NI measures in temperate climates (North America, Europe and the U.K.), as this is where most of the work on NI measures has been carried out, and to ensure that results could be more easily compared across studies. For some NI measures, there is little-to-no research in North America, and the European/U.K. literature was more robust, or vice versa. For water quality, we selected only those studies reporting actual field measurements, and further narrowed our selection to those papers reporting results on nitrogen (N) because N impacts downstream communities in a variety of ways (e.g., contaminated drinking water and toxic algal blooms). In contrast, since the flood mitigation literature is dominated by studies reporting model simulations, we included model-based studies as well as those reporting results.
Scientists have used a wide variety of metrics to report the benefits of NI measures (see
Table 1). Given our interest in minimizing the area footprint of new NI, we looked for studies reporting NI performance using area-based units. For water quality, we therefore selected studies reporting the reduction of nitrogen load per area (kg N ha
−1) for a given time period. This metric was used in 38 studies covering 15 different NI measures. Unfortunately, there is no comparable area-based metric that is commonly used to report on flood risk reduction. We chose the percentage reduction in flood peak because it appeared to be the most commonly used metric in the literature and would therefore enable us to compare a wide variety of NI measures. Recognizing that flood mitigation performance data were drawn from studies covering a wide range of storm and watershed sizes, both of which might be expected to influence performance, we also compiled data on these factors where available.
The selection process resulted in a total of 46 papers suitable for in-depth review and data extraction. To compare the performance of various NI measures, we developed a scoring rubric as follows (see also
Table 2 and
Figure 1): We characterized the strength of the evidence for water quality improvement or flood risk reduction based on the number of peer-reviewed articles we could find where a given NI measure showed positive impacts: One, Few, or Many. “Few” was defined as two articles and “Many” was defined as three or more. To account for the effectiveness of a given NI measure in improving water quality or reducing flood risk, we categorized impacts as Low, Medium and High. For water quality, we characterized those NI measures that delivered nitrogen reductions of less than 500 kg N ha
−1 yr
−1 as “Low” impact, while measures delivering 500–1000 kg N ha
−1 yr
−1 were designated as “Medium” impact, and measures that delivered greater than 1000 kg N ha
−1 yr
−1 were characterized as “High” impact. For flood reductions, impacts were scored as “Low” for peak flow reductions less than 15%, “Medium” for reductions of 15–25%, and “High” for reductions above 25%. As we are not aware of any absolute levels of flood mitigation or nitrogen reduction that would be considered “good” or “beneficial,” we chose to compare the measures relative to each other rather than an external standard. We chose these thresholds based on the range of values in the data for a roughly even distribution of measures in each category.
Figure 1 shows that we considered NI measures categorized as “Many” for evidence and “High” for impact as “High” priorities for implementation, whereas we would consider those categorized as “Few; High” and “Many; Medium” as “Medium” priorities for implementation.
5. Conclusions
Our analysis of published data on the environmental performance of NI measures suggests that a variety of NI interventions in agricultural landscapes could benefit downstream communities experiencing water quality and flooding problems. In particular, the restoration of wetlands and of floodplains are likely to provide dual benefits. A number of interventions associated with ditches and streams (vegetated ditches, two-stage ditches, riparian forest buffers and stream restoration) hold promise for improving water quality, although more work is needed to establish evidence of any flood mitigation benefits. In more upland areas, conversion of cropland to native vegetation (especially forests) and farm ponds designed to hold runoff are likely to reduce downstream flood risks.
Our analysis also highlights the difficulty of comparing the performance of multiple types of NI given the wide variety of metrics used for reporting improvements in water quality and flood risk reduction. In part, this reflects the variety of approaches (direct monitoring, hydrologic modeling and hydraulic modeling) used to assess performance, as well as the location and scale at which performance is assessed. In order to facilitate comparison between measures, and to assist decision-makers who often need to understand the absolute (rather than relative) value of various NI measures in attaining regional water quality or flood reduction goals, we encourage researchers to report the results of their work in absolute (e.g., N removal per unit area of measure) rather than relative (e.g., percentage reduction of flow or contaminant) metrics.
Recognizing that in agricultural landscapes the implementation of NI will require modest conversion out of cropland (though this may be only temporary in some cases, such as runoff attenuation features), we sought to determine which NI measures could be integrated into the landscape with minimal cropland conversion. Again, the limited use of area-based metrics in reporting NI performance made this challenging. Our preliminary assessment suggests that structural measures that provide new water storage capacity, such as created and restored wetlands and restored floodplains, may be more effective than vegetative measures (cropland conversion). This may reflect a combination of factors: the importance of water storage in flow modification, the increase in hydraulic retention time (and increased biogeochemical processing capacity) associated with increased water storage, and the (usually) strategic placement of structural measures to intercept important flow paths.
In summary, we recommend that communities experiencing water quality and flooding problems consider the role that NI measures in upstream agricultural watersheds could play in addressing these problems. Our analysis suggests that there is the potential for NI measures implemented across the landscape from farmers’ fields to river floodplains to provide Medium to High benefits to water quality and flood risk reduction. While additional work is needed to assess the benefits of more recent and innovative measures such as runoff attenuation features and two-stage ditches, there is sufficient evidence on more established measures such as restoration of floodplains and depressional wetlands to support their inclusion in local government watershed management plans.