About Us | Help Videos | Contact Us | Subscriptions

Journal of Environmental Quality - Environmental Models, Modules, and Datasets

Agricultural Conservation Planning Framework: 2. Classification of Riparian Buffer Design Types with Application to Assess and Map Stream Corridors


This article in JEQ

  1. Vol. 44 No. 3, p. 768-779
    unlockOPEN ACCESS
    Received: Sept 15, 2014
    Accepted: Feb 13, 2015
    Published: April 10, 2015

    * Corresponding author(s): mark.tomer@ars.usda.gov
Request Permissions

  1. M. D. Tomer *a,
  2. K. M. B. Boomerb,
  3. S. A. Portera,
  4. B. K. Gelderc,
  5. D. E. Jamesa and
  6. E. McLelland
  1. a USDA–ARS, National Laboratory for Agriculture and the Environment, 2110 University Blvd., Ames, IA 50010
    b The Nature Conservancy, 5410 Grosvenor Lane, Ste. 100, Bethesda, MD 20814
    c Iowa State Univ., Dep. of Agricultural and Biosystems Engineering, 1340 Elings Hall, Ames, IA 50011
    d Environmental Defense Fund, 1875 Connecticut Ave. Ste 600, Washington, DC 20009


A watershed’s riparian corridor presents opportunities to stabilize streambanks, intercept runoff, and influence shallow groundwater with riparian buffers. This paper presents a system to classify these riparian opportunities and apply them toward riparian management planning in hydrologic unit code 12 watersheds. In two headwater watersheds from each of three landform regions found in Iowa and Illinois, high-resolution (3-m grid) digital elevation models were analyzed to identify spatial distributions of surface runoff contributions and zones with shallow water tables (SWTs) (within 1.5 m of the channel elevation) along the riparian corridors. Results were tabulated, and a cross classification was applied. Classes of buffers include those primarily placed to (i) trap runoff and sediment, (ii) influence shallow groundwater, (iii) address both runoff and shallow groundwater, and (iv) maintain/improve stream bank stability. Riparian buffers occupying about 2.5% of these six watersheds could effectively intercept runoff contributions from 81 to 94% of the watersheds’ contributing areas. However, extents of riparian zones where a narrow buffer (<10 m wide) would adequately intercept runoff but where >25 m width of buffer vegetation could root to a SWT varied according to landform region (p < 0.10). Yet, these wide-SWT riparian zones were widespread and occupied 23 to 53% of the lengths of stream banks among the six watersheds. The wide-SWT setting provides opportunities to reduce dissolved nutrients (particularly NO3–N) carried via groundwater. This riparian classification and mapping system is part of a ArcGIS toolbox and could provide a consistent basis to identify riparian management opportunities in Midwestern headwater catchments wherever high-resolution elevation data are available.


    CZ, critical zone; DEM, digital elevation model; DRV, deep-rooted vegetation; HUC, hydrologic unit code; LiDAR, light detection and ranging; MSB, multi-species buffer; RAP, riparian assessment polygon; SSG, stiff-stemmed grasses; SBS, stream bank stabilization; SWT, shallow water table

Riparian buffers can provide a range of environmental, ecological, and social benefits (Schultz et al., 2009). Private landowners and government agencies have recognized this and have planted riparian zones with perennial vegetation in many watersheds under programs, such as the USDA Conservation Reserve Program (Lovell and Sullivan, 2006). Research conducted on riparian buffers has been published extensively. However, management of riparian zones and river corridors continues to be a challenging and costly endeavor, and few studies have shown how riparian corridor management can be improved from the watershed perspective (Bernhardt et al., 2005; Wohl et al., 2005). Yet the application of precision conservation technologies (Delgado and Berry, 2008) along riparian corridors could help link research, implementation, and evaluation of riparian practices. Digital elevation models (DEMs) obtained from LiDAR (light detection and ranging) surveys are a new data resource and are becoming increasingly available (USGS, 2013), which can help us map and evaluate riparian zones and their potential to provide a range of ecosystem services across watersheds. These sources of data may allow the development of precision conservation tools for site-specific design of riparian practice to account for the variety of stream-side settings found throughout a watershed’s stream network.

Although riparian buffers provide a variety of ecosystem services, this paper focuses on water quality improvement and the use of riparian vegetation to mitigate contaminant losses to streams along three pathways, namely direct losses from bank erosion, surface runoff, and subsurface (groundwater) flow. Buffers can be designed to stabilize soils and reduce stream bank erosion, to slow the velocity of runoff to trap sediment and enhance infiltration of runoff water, and to increase the uptake and transformation of dissolved nutrients in subsurface waters (Schultz et al., 2009). These benefits are made possible by a variety of physical, biological, and geochemical processes, as reviewed by Dosskey et al. (2010), which vary according to the soil–landscape setting and may be enhanced using different types of buffers. In this paper we summarize how riparian vegetation mitigates each pathway of nutrient/sediment loss and how varying buffer widths and vegetation types can optimize benefits to water quality. We then outline how opportunities to influence each of these three pathways for water quality benefit can be identified and mapped within a stream network based on the landscape setting. We next propose a strategy to map the functional potential for water quality improvement along a riparian corridor and demonstrate the results in a set of six small watersheds in Iowa and Illinois. We finally compare results among watersheds and validate the interpretability of these maps using field observations and photographs. We aim to demonstrate how geographic information system technologies can be used with high-resolution DEMs to map riparian corridors and allow conservation plans to vary the design of riparian buffers based on the range of site-specific opportunities for water quality improvement found throughout a watershed’s riparian corridors. These opportunities depend on specific attributes of landscape and soil type that can be defined and mapped at watershed scales, provided that the required data are available. The scope of our study comprises headwater watersheds in glaciated areas of the midwestern United States.

Mitigating Three Pollutant Source Pathways with Riparian Vegetation

Direct Stream Bank Losses

Stream banks comprise the major source of sediment loads in agricultural watersheds across much of the central United States (Simon and Klimetz, 2008; Wilson et al., 2008). Riparian vegetation can stabilize stream banks and can thereby control bank sources of sediment and phosphorus (P). Conservation cover using trees and grasses has been shown to decrease bank erosion compared with grazed or cropped streambanks (Zaimes et al., 2008). Trees have been reported to provide better bank stability than grasses (Zaimes et al., 2008; Langendoen et al., 2009), but this may depend on soil type, bank height, amount of vegetative cover, and the types and rooting-depth distributions of grasses and trees being compared. Control of bank erosion can be achieved by buffers that are little wider than the bank height unless the channel is actively incising or widening, in which case a buffer width of three times the bank height is recommended (Schultz et al., 2009). Therefore, if the bank height is 3 m or less, then buffers of 6 to 9 m width will suffice to limit bank erosion. Banks higher than 3 m will frequently exceed the rooting depth of the buffer vegetation, which increases the likelihood of bank undercutting by the stream. Mitigation of bank undercutting may require an intervention to stabilize soils at the toe of the bank. Flood-tolerant shrubs are often used for bank-toe stabilization; options are detailed by Schultz et al. (2009). Streambank erosion is normally found along the outside banks of meander bends; therefore, efforts to improve bank stability should be of greatest benefit for reducing stream-sediment loads wherever evidence of active bank erosion dominates beyond the outside banks of meander bends (Bjorkland et al., 2001). Our observation is that sod-forming prairie grasses, if grazing is restricted, can limit erosion in many small Midwestern streams if banks are not subject to undercutting.

Surface Runoff

Buffer vegetation can slow runoff velocity to enhance infiltration of runoff water and trap sediment (and other pollutants) originating from uplands. Buffers that are solely designed to accomplish this function are often called filter strips (Dosskey et al., 2011). Warm-season, stiff-stemmed grasses (e.g., switchgrass [Panicum virgatum L.]) will often be included in filter strips and in riparian buffers that are designed to capture runoff and sediment. It is possible for narrow strips (<1 m) of densely planted stiff-stemmed grasses to impede runoff and encourage sediment deposition (Dabney et al., 2012). In a meta-analysis of research results, Liu and Zhang (2008) found that sediment reductions of 95% can be accomplished with buffer widths of 10 m or less, unless the slope of the buffer is >10%. This information supports the idea that relatively narrow buffers are often sufficient for control of surface runoff. However, the ratio of buffer to contributing area may need to be considered in buffer design; this ratio is often used to model sediment trapping efficiency, and National Resources Conservation Service technical guides for buffer design suggest this ratio should be 0.02 or greater, depending on the rainfall factor in the Revised Universal Soil Loss Equation (2nd version). See Dosskey et al. (2011) for details on the relationship of buffer-area ratios to runoff and sediment trapping efficiency.

Groundwater Flow

Shallow riparian groundwater may be subject to a number of processes that reduce contaminant concentrations, including sorption on mineral or organic substrates and biological uptake and conversion. One key biological conversion is denitrification, a microbially mediated process that reduces NO3–N to N2 gas under anaerobic conditions (Seitzinger et al., 2006). Deep-rooted and phreatophytic (i.e., saturation-tolerant) vegetation can, at least in concept, increase cycling of soil carbon at depth to encourage denitrification in or near the saturated zone. The actual efficiency of denitrification in a buffer will largely depend on soil and hydrogeologic conditions that encourage slow water movement and depletion of dissolved oxygen concentrations (Simpkins et al., 2002; Mayer et al., 2007). Where these conditions are met, decreases in nitrate concentrations beneath riparian buffers have been reported as high as 90% in the Midwest (Osborne and Kovacic, 1993; Simpkins et al., 2002). A meta-analysis of N removal in buffers (Mayer et al., 2007) found a mean N-removal effectiveness of 77%, with buffers <25 m wide averaging 58% N removal and buffers >50 m wide averaging 85% N removal. Herein, we assume that riparian buffers can be designed to enhance denitrification in shallow groundwater wherever roots can interact with a shallow water table (SWT) (i.e., <1.5 m deep) and that wide buffers (i.e., >25 m) are needed to ensure substantial denitrification rates are achieved in groundwater naturally flowing beneath the buffers. Although plant uptake can also contribute to nitrate removal from shallow groundwater in buffers, harvest of plant biomass is needed to ensure this N pathway can be considered a long-term removal.

A Framework To Match Buffer Design to the Landscape Setting

Our aim here is to provide guidance for designing buffers to match actual opportunities for water quality improvement on a site-specific basis using geographic information system technologies. We assume there is a minimum goal to limit bank instability with riparian vegetation along all streambanks given that a buffer width of only 6 m may be needed to accomplish this. In addition, in many watersheds it will be useful to map locations where bank heights exceed 3 m (i.e., where toe stabilization may be needed to control bank erosion). Beyond ensuring bank stability, site-specific design for a riparian buffer depends on whether either or both of the other two benefits (i.e., treatment of surface runoff and/or groundwater flows) can be achieved at a particular riparian location. In other words, how do we identify those riparian zones where buffer vegetation should be planted wider than a minimum 6 to 9 m width (for bank stability) to effectively trap runoff and/or influence groundwater most efficiently throughout a watershed? A two-way classification for riparian sites is proposed to accomplish this mapping goal (Fig. 1). Our approach first identifies where a significant potential for runoff contribution exists to show where buffer vegetation should be widened to at least 10 m and include stiff-stemmed grasses to effectively intercept runoff. We classify riparian sites into three groups according to runoff-contributing areas and the ratio of contributing area to buffer area as follows. These three classes are also depicted on the left side of Fig. 1.

  • High: These riparian sites have the potential to receive overland flow from large upslope areas, typically via the largest ephemeral waterways in the watershed. If surface runoff were to be generated throughout the watershed with every unit of area contributing equally, half the surface runoff from the watershed to the stream would pass through these “High” riparian zones.

  • Medium: A buffer that must be wider than 10 m to provide a buffer-contributing area ratio of 0.02 (and thereby meet National Resources Conservation Service technical guidance).

  • Low: A narrow buffer (10-m wide or less) provides the minimum recommended contributing area ratio of 0.02 (i.e., the buffer widths required for bank stabilization and for runoff interception are similar).

Fig. 1.
Fig. 1.

General approach to identify types of riparian settings, which are used to map where riparian buffers can improve water quality by pathway-specific processes on a stream reach basis (adapted from Tomer et al. [2013b]). H, high; L, low; M, medium.


Next, riparian sites are identified where there are opportunities to influence shallow groundwater with a widened buffer including deep-rooted vegetation (DRV) with species tolerant of saturated soils. These plantings would be appropriate where the water table is expected to be <1.5 m deep (Jencso et al., 2009; Murphy et al., 2009), but these SWT zones need to be relatively wide to consistently influence groundwater nitrate concentrations (Mayer et al., 2007), leading to the following proposed classification of riparian sites. These three classes are represented along the top row of the table in Fig. 1 (i.e., as >50 m, 25–50 m, and <25 m, respectively):

  • Wide: A shallow water table (<1.5 m depth) extends on average to distances >50 m from the stream.

  • Moderate: A shallow water table (<1.5 m depth) extends on average to distances between 25 and 50 m from the stream.

  • Narrow/absent: A shallow water table (<1.5 m depth) on average extends to distances <25 m from the stream.

Buffer Design Types Based on Dominant Transport Pathways

We identify five types of riparian settings (Fig. 1), each carrying a suggested range of buffer widths and suggested types of vegetation, which are listed as follows.

Locations where opportunities to influence both surface and subsurface pathways are colocated should be prioritized for multispecies buffers (MSBs) (Fig. 1), which would be designed along the lines specified by Schultz et al. (2009). Multispecies buffers are frequently designed with three vegetation zones to intercept runoff furthest from the stream, stabilize banks closest to the stream, and provide other ecosystem services in the intermediate zone, which could include mitigation of groundwater nutrients if deep rooted vegetation is included.

If particularly large volumes of runoff contributions can be anticipated to enter the riparian zone where a wide zone of SWT occurs, this setting should be considered a critical zone (CZ) buffer type (Fig. 1). A CZ buffer is an environmentally sensitive area with intensified hydrology where a wide MSB should be planted to stabilize soils and manage varied rates of overland and subsurface flows. The need for permanent vegetation may extend above the riparian zone along pathways of runoff contribution and concentrated flow.

Where there is the opportunity to influence hydrologic flows that are dominantly along surface runoff pathways, then a buffer designed to filter runoff and trap sediment, often by including stiff-stemmed grasses (SSG) as a key vegetative component (SSG-type buffer) (Fig. 1), would be the preferred buffer design option.

If the opportunities to influence hydrologic flows are dominantly along subsurface pathways due to the presence of a wide SWT zone, then a relatively wide buffer that includes a deep-rooted vegetation (DRV)-type buffer (Fig. 1) should offer the greatest water quality benefits.

Where neither surface nor subsurface waters can be readily influenced, then a narrow buffer designed to protect or stabilize the streambank (SBS-type buffer) (Fig. 1) can be the preferred option.

This proposed approach is simple in concept but requires detailed topographic data to provide consistent results. The increasing availability of high-resolution, LiDAR-based topographic data, obtained by aerial survey, makes it possible to examine riparian landscapes in greater detail than has been possible in the past. Tomer et al. (2013a) demonstrated that these topographic data could be used to identify locations suitable for nutrient-removal wetlands. Here we will evaluate the potential to use LiDAR-based, 3-m resolution DEMs to map conservation opportunities throughout the riparian zone of a watershed. What we are describing is one component of a precision conservation framework for agricultural watershed planning described by Tomer et al. (2013b). Our objective is to apply a classification scheme to identify conservation opportunities throughout a riparian network to six Midwestern hydrologic unit code (HUC)12 watersheds and compare the results among watersheds.

Materials and Methods

Trial Watersheds and Input Data

Six watersheds were selected to provide a pair of watersheds in each of three general glacial landform regions found in the upper Midwest (Table 1; Fig. 2 and 3). These settings represent a sequence of older glacial landforms (Illinoian age, about 500,000 yr) with well-developed stream networks and rolling hills (watersheds 1 and 2) to young glacial landforms (Wisconsinan age; about 14,000 yr) that exhibit minimal landscape dissection through stream development (watersheds 5 and 6). Watersheds 3 and 4 exhibit an intermediate extent of landscape dissection; they are also of Illinoian age, but long periods of peri-glacial conditions limited stream incision (Prior, 1991). Each of these six HUC12 watersheds can be characterized as a headwater catchment (i.e., the watershed area is not traversed by a high-order stream), providing a relatively simple and consistent basis to evaluate our mapping concepts and make comparisons among watersheds. More detailed descriptions have been published for four of these six watersheds: Schilling and Spooner (2006) described watersheds 1 and 2, Tomer et al. (2008) described watershed 5, and Tomer et al. (2013a) described watershed 6. The diversity of Midwestern streams is under served by the diversity of names given to them; to avoid confusion, we refer to these watersheds by number, as indicated in all tables and in Fig. 2 and 3.

View Full Table | Close Full ViewTable 1.

Comparative data on soil and slope characteristics for the six study watersheds.

Watershed Dominant soil series and classification Portion of watershed covered by hydric soils Average slope of cropland in watershed
1. Walnut Creek South Tama (Typic Argiudoll) 4.8 7.1
2. Squaw Creek Tama (Typic Argiudoll) 9.1 5.8
3. Beaver Creek Saude (Typic Hapludoll) 9.7 3.7
4. Headwaters Blackhawk Creek Tama (Typic Argiudoll) 15.9 3.1
5. Headwaters Beaver Creek Webster (Typic Endoaquoll) 25.3 3.7
6. Lime Creek Osco (Mollic Hapludalf) 19.6 2.8
Fig. 2.
Fig. 2.

Hydrologic unit code 12 watersheds selected for trial of the riparian buffer design mapping scheme and their placement in Major Land Resource Areas of Iowa and Illinois. Watersheds 1 and 2 (Walnut Creek South and Squaw Creek) are on the oldest glacial landforms and exhibit the greatest extent of stream network development. Watersheds 3 (should be 5) and 6 (Headwaters of Beaver Creek and Lime Creek) are located in areas of Wisconsinan glaciation (12,000–14,000 yr in age) and exhibit the least stream network development.

Fig. 3.
Fig. 3.

Maps of six hydrologic unit code 12 watersheds showing agricultural land use, field boundaries, stream networks, and (shaded relief) elevation models. CB, corn/soybean.


A DEM was developed for each watershed, providing a set of 3-m grid rasters originating from aerial LiDAR survey data, each with grid-cell elevations interpolated to 1 cm resolution. The LiDAR survey data were obtained from the State of Iowa (IDNR, 2013) for watersheds 1 through 5 and as described by Tomer et al. (2013a) for watershed 6. Absolute errors in LiDAR elevation data are typically 0.10 to 0.18 m, depending on slope and survey density, as discussed by Bater and Coops (2009). Relative errors are less than absolute errors but depend on vegetative cover. Effects of vegetative cover were minimized by conducting surveys in late winter or early spring, before regrowth of perennial vegetation. Preprocessing was conducted on each DEM so that overland flow routing could be modeled using terrain analysis software (Wilson and Gallant, 2000). False impoundments (dominantly roadway bridges and culverts) were digitally edited as described by Tomer et al. (2013a), but this editing was conducted under software control (with review and manual editing of results). In addition, agricultural field boundaries and land cover/rotations were obtained for each watershed (Fig. 3). Fields were delineated beginning with field boundaries publicly released by the USDA in 2005. All ownership- and county-level attribution was removed, and aerial imagery collected in 2009 under National Agricultural Imagery Program (Farm Service Agency, 2012) was used to edit field boundaries to reduce the number of polygons containing multiple crops, as described by Tomer et al. (2015). Each field was assigned a major land use and a crop rotation based on classified crop cover for 2007 through 2011 (NASS, 2012). Major classes included corn and soybean rotations, rotations including other crops, and perennial cover (Fig. 3). Soil survey information was also extracted from the National Soil Survey database (Soil Survey Staff, 2013) for each watershed.

The field boundary and elevation data were used to estimate contributing areas along each stream. The DEMs were subjected to pit filling and flow routing analysis (Wilson and Gallant, 2000) using the watershed delineation and analysis processes in the ArcGIS Spatial Analyst (ESRI, 2014). These analyses assumed that overland flow from each grid cell is directed to the single neighboring cell that is determined by the steepest downslope gradient (i.e., D8 algorithm; Wilson and Gallant [2000]). Streams in each watershed were assumed present wherever a minimum contributing area of 100 ha was exceeded (Tomer et al., 2013a). However, on field review, this stream initiation threshold was increased to 150 ha for watershed 5, in which subsurface tile drainage mains conveyed tile drainage from large areas. Flow routing was altered for grid cells immediately adjacent to the stream network to force flow directly into the channel from bank cells, preventing the occurrence of parallel flow lines adjacent to any stream. Results of the flow routing analyses provided raster coverages of flow-area accumulation and the stream network for each watershed. A raster of estimated depth to water table was also developed for each watershed by calculating elevation differences between each grid cell and the stream-channel grid cell that would receive overland flow from that cell. This approach assumes the stream channel elevations provide an estimate of the local water table depth that can be extended laterally into the riparian zone. Although this assumption clearly loses validity as distance from the channel increases, the mapped result should depict where exchange between stream water and shallow riparian groundwater is most likely to occur. The flow-area accumulation and estimated water table depth rasters provided the key input data for classifying the riparian zone according to buffer design types described in Fig. 1.

Two additional rasters were developed to provide a tile-drained and a non–tile-drained mask for each watershed. Cropped fields with low slopes (<5%) and/or a consistent area of hydric soils were assumed to be artificially (tile) drained (Tomer et al., 2013b). These two drainage masks were used to calculate flow-area accumulation rasters that included only upstream grid cells that are tile drained and only non–tile-drained grid cells.

Delineation of Riparian Assessment Polygons

Riparian assessment polygons (RAPs) are analytical units created along the stream network that represent discrete riparian zones on each side of the riparian corridor. The RAPs had to be centered along the stream and then divided into left and right elements to evaluate each side of the stream independently. The riparian site-type classification routines developed for this study use the RAPs as the basic unit for collation and analysis of derived terrain data on runoff contributing and SWT areas.

Individual reaches of stream had to be segregated before the RAPs could be delineated in each watershed. The stream network was converted to a polyline coverage to give a unique identifier to each length of stream between confluences. The separation of stream reaches was done to avoid topological complexity at stream confluences. To delineate the riparian zone, a set of rectangular 250- by 180-m polygons was first individually centered and placed along each stream-segment polyline. The “strip map index features” tool contained in the ArcGIS Cartography toolset was applied to create these rectangles. The choice to designate 250-m lengths of riparian zone as sites was arbitrary but was considered to approximate a length of riparian corridor that could be managed separately. This delineation led to riparian zones near confluences and sites of stream initiation (100–150 ha) being omitted where 250-m-length rectangles could not be fitted. To produce the RAPs, the rectangular polygons were converted to polylines, merged with the stream reach polylines, and then converted back into polygons, which divided the rectangles in half to independently represent the riparian zone on each side of the channel. Some overlap of the original polygons was unavoidable, and these feature conversions were conducted for each individual polygon to prevent overlap areas from becoming separate polygons. After these feature conversions, each RAP had a constant length of 250 m, and, because stream meanders defined one side of each polygon, a width that averaged 90 m (Fig. 4).

Fig. 4.
Fig. 4.

Illustration of scheme to discretize riparian zone into 250-m riparian assessment polygons to enable data collation for each side of a stream.


To populate the RAPs with classification data, the stream channel raster was expanded by one cell on each side to identify stream-bank grid cells (Fig. 4). Each of the stream bank grid cells was within and assigned to a single RAP. Up-gradient contributing areas (i.e., total area and areas of tile-drained and non–tile-drained cropland) were then extracted to each bank grid cell. For each RAP, the following information was tabulated: (i) average width of shallow water table (for both the <1.5-m and the 1.5- to 3-m depths) based on the number (area) of grid cells with <1.5 and 1.5 to 3.0 m estimated water table depths within each RAP divided by polygon length (250 m); (ii) areas of potential runoff contribution, including total contributing areas, and contributing areas of tile-drained and non–tile-drained cropland, based on summations of data extracted from the bank cells; (iii) total watershed size above each RAP; and (iv) stream length within each RAP, which allows sinuosity to be calculated by dividing the stream length by the polygon length (250 m).

Classifying Riparian Assessment Polygons

The RAP data were classified to assess the varying potentials for stabilizing stream banks, intercepting shallow groundwater flows, and trapping runoff and sediment within and among the six watersheds. First, RAPs averaging less than a 10-m width with <3 m depth to water table (i.e., <1.5 m plus 1.5–3 m) were highlighted as areas where banks could be >3 m high and possibly be subject to bank undercutting. These RAPs would show where toe stabilization might be needed to reduce bank erosion.

The RAP data were classified into three classes for both surface and subsurface flows as defined in the previous section and assigned a buffer design type as illustrated in Fig. 1. This process involved assignment of two possible buffer widths to each RAP, one to address potential runoff/sediment trapping opportunities (“runoff width”) and one to address potential opportunities to influence shallow groundwater (“SWT width”). The larger of these two values was designated as the “optimal buffer width,” which was used in subsequent calculations to evaluate and compare precision conservation management schemes within and among the six watersheds.

To assign a buffer width for runoff interception to each RAP, a “runoff width” was calculated as 0.02 times the contributing area and then divided by the RAP length (250 m). For RAPs in the “High” runoff class (conveying runoff from large contributing areas comprising half the watershed), the calculated runoff width sometimes exceeded 90 m; a maximum width of 90 m was assigned in these cases. The calculated runoff widths were assigned to the remaining RAPs, and demarked the “Medium” from “Low” runoff classes, depending on whether the runoff width was greater or less than 10 m.

The “SWT width” was then assigned to each RAP based on the width of the zone where the water table is expected to be <1.5 m depth. For the “Wide” class with at least a 50-m-wide SWT zone, the SWT width was 50 m. For the “Moderate” SWT–width RAPs, the calculated width of the SWT zone was assigned (i.e., between 25 and 50 m). For “Narrow” or “Absent” SWT zones, a minimum width of 6 m was assigned where the SWT was <25 m wide, given that the opportunity to enhance riparian denitrification was uncertain. This meant that a minimum “optimal width” would always be at least 6 m to ensure bank stability in areas where the SWT zone was the narrowest.

The larger of the two buffer widths, considered an optimal width for water quality benefits, was tabulated for all RAPs and summed by riparian design type (Fig. 1). These summed areas suggest an “optimized” distribution of buffer treatments that can be prioritized according to function and can define for planners and landowners the total fraction of the watershed that would be involved with total (or partial) implementation. Results were assessed through visual map assessments and through field reviews, during which photographs were taken of each type of riparian setting. Comparisons of the distributions of riparian types were made among watersheds, particularly among the three landform regions represented, using t tests where appropriate, with p values <0.10 reported to indicate significance of each comparison.

Results and Discussion

Results showed that each of the six watersheds had unique distributions of riparian settings and buffer-design types that matched those settings (Tables 2 and 3). The input data in terms of runoff pathways and SWT widths are illustrated in Fig. 5 for part of each watershed, and the resulting classification of riparian areas is shown in Fig. 6. Riparian corridors in watersheds 1, 2, and 6 were dominated by SBS-type riparian settings (43–47% of RAPs) where narrow buffers (i.e., <10 m) would suffice to maintain (or increase) stream bank stability and intercept runoff from limited upslope areas. This suggests that relatively more channel incision has occurred in these three watersheds. The other three watersheds (3, 4, and 5) still had substantial extents of SBS-type riparian settings (16–22% of RAPs), where narrow buffers could filter limited amounts of runoff and maintain/improve bank stability. The SBS-type riparian settings were found more frequently (p < 0.01) in watersheds in the older, well-incised Illinoian landform region (i.e., watersheds 1 and 2) than those in the moderately incised Illinoian landform regions (i.e., watersheds 3 and 4). However, the two younger Wisconsinan watersheds (watersheds 5 and 6) had dissimilar frequencies of SBS-type settings (i.e., 22 and 42%, respectively), which probably resulted from drainage ditches being maintained at varying depths within and among watersheds on the younger Wisconsinan landform regions. Among the six watersheds, most of the stream banks suggested for evaluation of the need for toe stabilization occurred in watersheds 2 and 6 (5.5 of total 6.6 km of stream bank; see Table 3). This suggests that the greatest channel incision follows not only the landform region but also the extent of ditching and depth at which drainage ditches are maintained.

View Full Table | Close Full ViewTable 2.

Contingency tables for numbers of riparian assessment polygons classified according to opportunities to influence surface runoff and subsurface flow pathways for each of six watersheds , following scheme summarized in Fig. 1.

Width of shallow water table
Rank H† M L Total H M L Total H M L Total
Watershed 1 (Walnut Creek South)
Watershed 3 (Beaver Creek)
Watershed 5 (Headwaters Beaver Creek)
H 5CZ 14MSB 17SSG 36 12 10 4 26 18 10 5 33
M 7MSB 16MSB 46SSG 69 15 11 10 36 46 35 16 97
L 26DRV 47DRV 145SBS 218 45 69 34 148 104 76 86 266
Total 38 77 208 323 72 90 48 210 168 121 107 396
Watershed 2 (Squaw Creek)
Watershed 4 (Headwaters Blackhawk Creek)
Watershed 6 (Lime Creek)
H 11 10 8 29 18 15 8 41 11 10 17 38
M 7 17 22 46 30 19 9 58 11 12 32 55
L 32 43 134 209 76 96 61 233 63 80 176 319
Total 50 70 164 284 124 130 78 332 85 102 225 412
H (high), >50 m; M (medium), 50–25 m; L (low), <25 m.
CZ, critical zone; DRV, deep-rooted vegetation; MSB, multispecies buffer; SBS, streambank stability; SSG, stiff-stemmed grasses; TS, toe stabilization.

View Full Table | Close Full ViewTable 3.

Total lengths of stream bank and areas of buffer suggested, based on buffer design classification scheme for riparian assessment polygons (RAPs) for each of six watersheds. Numbers of polygons contributing to each class are tallied in Table 2.

Watershed Classification for type of riparian buffer design†
Total length of streambank
1. Walnut Creek South 1.4 11.5 20.1 23.0 44.7 0.3 100.7
2. Squaw Creek 3.2 9.7 8.7 21.2 37.4 2.8 77.5
3. Beaver Creek 3.8 11.9 4.8 36.5 11.2 0.0 68.3
4. Headwaters Blackhawk Creek 5.4 20.3 5.5 51.3 19.0 0.3 101.6
5. Headwaters Beaver Creek 5.2 27.8 6.1 54.2 25.1 0.5 118.5
6. Lime Creek 3.1 9.5 14.1 40.5 50.3 2.7 117.5
Total area of buffers (contributing area to each buffer type)
1. Walnut Creek South 8.9 (428) 43.2 (1520) 49.2 (2464) 74.5 (353) 23.1 (680) 198.9 (5445)
2. Squaw Creek 16.0 (658) 35.9 (1030) 18.4 (918) 77.5 (320) 21.3 (475) 169.1 (3401)
3. Beaver Creek 16.3 (677) 40.4 (1009) 8.6 (431) 122.0 (531) 5.6 (156) 192.9 (2804)
4. Headwaters Blackhawk Creek 24.3 (959) 74.1 (1909) 13.8 (691) 187.5 (817) 9.6 (295) 309.4 (4671)
5. Headwaters Beaver Creek 31.5 (1747) 111.3 (3126) 19.3 (1023) 200.8 (662) 13.2 (250) 376.1 (6808)
6. Lime Creek 15.8 (720) 37.9 (1113) 32.1 (1618) 152.3 (654) 27.1 (489) 265.2 (4594)
CZ, critical zone; DRV, deep-rooted vegetation; MSB, multispecies buffer; SBS, streambank stability; SSG, stiff-stemmed grasses; TS, toe stabilization.
Because RAPs do not include stream threshold points, total RAP contributing areas are less than watershed areas.
Fig. 5.
Fig. 5.

Portions of the stream network in each of six hydrologic unit code 12 watersheds, identifying contributing-area runoff pathways and estimated areas with shallow water table, which are used as input data to classify the discretized riparian assessment polygons. See Fig. 6 for classification results for these areas.

Fig. 6.
Fig. 6.

Areas of each watershed with riparian assessment polygons classified to indicate functional opportunities for water quality improvement using riparian buffer vegetation. Input data used to arrive at these classifications are illustrated in Fig. 5.


Riparian sites appropriate for multispecies buffers (MSB- and CZ-type) occurred in only 11 to 15% of the RAPs among the three more incised watersheds (1, 2, and 6), compared with 23 to 28% of RAPs among watersheds 3, 4, and 5. The watersheds in the older Illinoian landform region (1 and 2) had a lower frequency (p = 0.02) of MSB- and CZ-type RAPs than watersheds 3 and 4, which were from the less-incised Illinoian landform region. Opportunities to intercept runoff using >10-m-wide buffers, including stiff-stemmed grasses, did not vary greatly but were most prevalent in watershed 1 (i.e., 20% of RAPs were classified as SSG vs. 5–12% of the RAPs among the other five watersheds). Riparian areas where interception of runoff would require buffers >10-m wide to meet a buffer area ratio of 0.02 (i.e., CZ, MSB, and SSG design types) showed little variation among watersheds and summed to comprise 23 to 33% of RAPs among all six watersheds. These limited extents and variation were not necessarily expected but point out that the need to focus on runoff interception as a primary factor determining riparian buffer design is most critical only along parts of the stream network and that this is true across a range of landscape regions.

Opportunities to primarily manage riparian vegetation to influence shallow groundwater (i.e., DRV-type RAPs) occurred in 23 to 54% of RAPs among the six watersheds. There was evidence that the frequencies of occurrence of DRV-type RAPs varied among landform regions. The frequency of DRV-type RAPs was 52 to 54% in the younger Illinoian landform (watersheds 3 and 4), which was significantly greater (p = 0.07) than in the youngest Wisconsinan watersheds, as DRV-type RAPS occurred in 45% of RAPs in Watershed 5 and 35% in Watershed 6. The older Illinoian landform watersheds (1 and 2) showed the lowest frequency of DRV-type RAPs (23–26%), which was significantly less that the frequencies found in the Wisconsinan watersheds (5 and 6; p = 0.06) and in the younger Illinoian watersheds (3 and 4; p < 0.01).

Areas of “Optimal” Buffer Plantings

The distributions of riparian buffer types carry specific recommendations for buffer widths, which affects the proportion of each watershed suggested for riparian buffer vegetation using our classification and mapping system. However, on balance, this effect was only important when considering the recommended extent of DRV-type buffers. That is, among the six watersheds evaluated, the total area of buffers suggested for CZ, MSB, SSG, and SBS design types was consistent, comprising 2.3 to 2.7% of the area of each watershed while potentially buffering the stream from overland flows contributed from 81 to 94% of the watershed areas contributing to the RAPs. The narrow range in buffer area was largely driven by the 0.02 area-buffer ratio used to determine the “runoff-width” for each RAP. The remaining 6 to 19% of watershed areas contributing to the RAPS were located above riparian zones classed as DRV, where a wide zone of shallow groundwater could be influenced by deep-rooted vegetation. These DRV-type buffers, if planted to the 25- to 50-m widths suggested, would comprise an additional 1.4 to 4.4% of the areas contributing to the RAPs, yet would intercept overland flows from relatively small areas of adjacent uplands. These DRV-type riparian zones could serve as nutrient sinks by implementing novel practices that could provide denitrifying treatment of tile-drainage water. For example, in watershed 6, we observed that riparian sites identified as possible locations for nutrient removal wetlands (Tomer et al., 2013a) were dominantly DRV-type settings.

Variation among Landform Regions

The three landform regions were differentiated by the extents of the DRV-type riparian buffers among the watersheds under a scenario of full buffer implementation (Table 3). The DRV-type riparian plantings would be most extensive in watersheds 3 and 4, occupying 4.0 to 4.4% of RAP contributing areas in these two watersheds and intercepting flows from 17 to 19% of these areas. Watersheds 3 and 4 are on the less-incised but older (Illinoian) landscapes. The DRV-type riparian plantings would be of intermediate extent in watersheds 5 and 6, occupying 2.9 to 3.3% of RAP contributing areas and intercepting flows from 10 to 14% of these areas. Watersheds 5 and 6 are on the younger (Wisconsinan) landforms. The DRV-type riparian buffers would be of least extent in watersheds 1 and 2 and occupy 1.4 to 2.3% of RAP contributing areas, while intercepting flows from 6 to 9% of these areas, which are located on the more incised region of older (Illinoian) landforms. Despite these regional differences, the DRV-type buffers, if fully implemented, would occupy a larger area than any other buffer design type in all six watersheds (Table 3). These DRV-type buffers would ostensibly intercept hydrologic flows from relatively small portions (no more than 19%) of these watersheds. However, this information is based on an assumption that the areas contributing surface runoff and groundwater flows to any given riparian zone are identical. This assumption probably underestimates the potential influence of DRV-type buffers on groundwater in a scenario of full implementation, considering riparian-zone processes such as bank return flows and the substantial extent of the DRV-type buffer-type setting in all six watersheds. The DRV-type setting occupied 50 to 53% of the streambank lengths in watersheds 3 and 4, 34 to 46% of the streambanks in watersheds 5 and 6, and 23 to 26% of the streambanks in watersheds 1 and 2 (Table 3). Therefore, even in the most incised of these watersheds, nearly a quarter of the streambank lengths could be planted with wide (>25 m) buffers to help treat shallow groundwater.

Distributions of Buffer Types within Watersheds

In these headwater catchments, the DRV-type riparian settings were concentrated along first-order streams; watershed 1 appeared to be an exception (see Supplemental Fig. S1–S6). However, the SBS-type riparian settings tend to be concentrated lower in the watershed (along second- and third-order streams), particularly in watersheds 2, 4, and 6. We hypothesize that distributions of buffer types within watersheds may vary according to postsettlement patterns of erosion and sediment accretion as much as landform region. For example, observations during our field review indicated that the stream in watershed 4 had clearly cut through accreted postsettlement sediment to form the bank heights found in the lower parts of the watershed. Further information on historical sedimentation and its effects on fluvial processes in the upper Mississippi River basin can be found in Knox (2006), Yan et al. (2010), and papers cited therein.

Field Review of Results

Comparison of the extent of riparian buffer design types among watersheds depends on consistency in the collection and processing of LiDAR data and the use of that data to consistently identify runoff pathways and SWT areas. Therefore, we visited each watershed to view and photograph differently classed riparian settings to check the consistency of our classification scheme. During these visits we found that distinctions among CZ-, MSB-, and SSG-type riparian settings varied according to the length, area, and shape of slopes above the riparian zone. That is, above CZ-type riparian settings, we consistently found ephemeral waterways that drained significant subwatershed areas that comprised more than one cropped field. Roadside ditches also provided surface flow pathways for large contributing areas leading to the CZ designation (see Supplemental Fig. S1–S6). Above MSB-type riparian settings, ephemeral (grassed) waterways draining smaller subwatersheds or convergent hillslopes without grassed waterways were consistently found. Riparian areas designated as CZ and MSB consistently had relatively wide flat areas adjacent to the channel where one would expect an SWT to occur, at least on a seasonal basis. Above SSG-type riparian settings, slopes were either relatively long and without convergent features or were shorter with gently convergent features. Where level areas occurred adjacent to the channel, these were not wider than about 20 m. Because many of these distinctions are based on features of the contributing area, these observations were difficult to represent clearly in photographs, especially at publication scale. However, the contrast between SBS-type versus DRV-type riparian settings was easier to depict with photographs, and this contrast is shown in Fig. 7. The paired photographs depict a consistent difference in riparian settings, with SBS-type riparian zones exhibiting relatively high banks and the DRV-type riparian zones exhibiting lower banks and wide, low-lying areas adjacent to the channel. The SBS-type classification does not indicate where stream banks are being actively eroded. Although all unstable banks we found during our field reviews occurred where the riparian zone had an SBS-type designation, not all SBS-type riparian areas had eroding banks. The SBS designation simply indicates where the extent of any SWT is narrow and where a narrow (<10-m-wide) buffer would adequately intercept and infiltrate runoff from a limited upslope contributing area.

Fig. 7.
Fig. 7.

Example photographs of riparian sites typed as deep-rooted vegetation and stream bank stabilization in each watershed.


Several caveats of this approach occurred due to the nature of the data and the landscapes being evaluated, which were noted during the field reviews. The data were detailed but were interpolated to a 3-m grid, meaning that features smaller than 3 m were not necessarily well represented. For example, where the channel was narrow, the extent of high water table zones may have been overestimated because the channel elevation becomes somewhat elevated due to interpolation. Field review suggested this issue only became problematic for small drainage ditches, and SWT zones designated in upper headwaters of ditched watersheds may have been overestimated. Nevertheless, it was clear in our review that these headwater riparian areas would exhibit a wide SWT zone at least on a seasonal basis. The upper half of watershed 5 and the lower half of watershed 6 presented very little topographic relief and had shallow, narrow ditches, such that classifications that were readily interpreted in watersheds 1 through 4 were less obvious in watersheds 5 and 6. Although interpretation was not problematic in the field, landscape features above the riparian zone that influenced the classifications were most subtle in the youngest and least-incised terrain of watersheds 5 and 6.

Summary and Conclusion

We devised an approach to match the design of riparian buffers to riparian zone opportunities to stabilize streambanks, intercept surface runoff, and influence shallow groundwater throughout a stream network. Two watersheds from each of three glaciated landform regions common to the agricultural Midwest were used to demonstrate the approach. Riparian buffers that could be placed to effectively intercept runoff and stabilize streambanks would occupy 2.3 to 2.7% of the six watersheds and intercept overland flows from 81 to 94% of uplands above the RAPs. However, the three landform regions were differentiated (p < 0.10) by the extent of riparian settings with wide zones of shallow water tables (DRV-type buffers). The DRV-type settings were most common (52–54% of RAPs) in the least-incised watersheds and least common (23–26% of RAPs) in the most-incised landform regions. Many of these areas could serve as natural bioreactors in tile-drained landscapes (Tomer et al., 2015).

We conclude that this scheme can provide a reasonable basis to identify riparian management alternatives along headwater streams in the glaciated Midwest. This scheme should be applied flexibly, and mapping results should be examined to ensure that LiDAR data provide consistent representation of land surfaces and flow paths. ArcGIS tools to conduct this riparian mapping are part of the Agricultural Conservation Planning Framework (Tomer et al., 2015; USDA–ARS, 2015) and will be available online by 1 Oct. 2015.


This study was conducted by the USDA–ARS and was partly supported through a Conservation Innovation Grant awarded by the USDA–NRCS to the Environmental Defense Fund. The authors thank the North Central Region Water Network for agreeing to host the ACPF ArcGIS toolbox on its website (http://northcentralwater.org/acpf/).





Be the first to comment.

Please log in to post a comment.
*Society members, certified professionals, and authors are permitted to comment.