Journal list menu

Volume 53, Issue 5 p. 1543-1553
Standard Paper
Free Access

A before-and-after assessment of patch-burn grazing and riparian fencing along headwater streams

Danelle M. Larson

Corresponding Author

Danelle M. Larson

Division of Biology, Kansas State University, Manhattan, KS, 66506 USA

Correspondence author. E-mail: [email protected]Search for more papers by this author
Walter K. Dodds

Walter K. Dodds

Division of Biology, Kansas State University, Manhattan, KS, 66506 USA

Search for more papers by this author
Matt R. Whiles

Matt R. Whiles

Department of Zoology, Southern Illinois University, Carbondale, IL, 62901 USA

Search for more papers by this author
Jessica N. Fulgoni

Jessica N. Fulgoni

Department of Zoology, Southern Illinois University, Carbondale, IL, 62901 USA

Search for more papers by this author
Thomas R. Thompson

Thomas R. Thompson

Missouri Department of Conservation, Grassland Systems Field Station, Clinton, MO, 64735 USA

Search for more papers by this author
First published: 10 May 2016
Citations: 10

Summary

  1. Fire and grazing are common in grasslands world-wide to maintain grass cover and cattle production. The effects of fire, cattle grazing and riparian fencing efficacy on prairie stream ecology are not well characterized at catchment scales.
  2. We examined alterations to stream water quality and biology from patch-burn grazing (PBG) in tallgrass prairie during a five-year, replicated, catchment scale experiment that used a Before-After/Control-Impact (BACI) design and was analysed by mixed-effects models. Treatments included two patch-burned control catchments (fire but no grazers) and PBG in two riparian-fenced and two unfenced catchments. We assessed the effectiveness of riparian fencing for mitigating potential water quality impacts by monitoring water quality and riparian usage by cattle via Global Positioning System.
  3. Riparian fences effectively excluded cattle; however, in unfenced pastures, cattle aggregated along streams 10–20% of the grazing season.
  4. After initiation of PBG, we detected large increases in some nutrients, Escherichia coli, algal biomass, primary productivity and community respiration in all catchments with PBG. Some water quality variables, such as E. coli concentrations, recovered quickly after cattle were removed from pasture, which indicated resiliency.
  5. Riparian fencing moderately reduced the impacts to stream variables, indicating either overland flow and/or subsurface flow allowed nutrients and bacteria to enter the streams.
  6. Synthesis and applications. Patch-burn grazing is a measurable disturbance that can alter the ecological condition of streams. Riparian fencing lessened the degree of impact, yet some water quality variables still exceeded regional reference conditions. Managers will need to assess the costs of riparian fencing compared to the moderate benefits that fencing provides to water quality.

Introduction

Riparian protection is promoted to buffer the effects of land management on stream water quality (Muscutt et al. 1993). Grassland headwater streams and riparian areas may be influenced by fire and grazing, which are dominant ecosystem processes and management tools for conservation and/or livestock production (e.g. Biondini, Patton & Nyren 1998; Knapp et al. 2004). Patch-burn grazing (PBG) is a grassland management technique where a subset of the pasture is burned and cattle are attracted to the new growth and graze heavily while the unburned pasture recovers from past grazing. The PBG approach mimics the heterogeneous historical fire and grazing regime in grasslands and is being progressively adopted on public and private lands world-wide because it can increase plant and animal diversity (Brockett, Biggs & Van Wilgen 2001) and safeguard against woody plant encroachment (Briggs et al. 2005). On private grasslands, PBG is increasingly considered a viable alternative to other strategies of range management (e.g. Winter, Fuhlendorf & Goes 2014).

Previous studies emphasized impacts of fire or grazing on streams but rarely both practices. In mesic areas, fire is almost always used in concert with seasonal grazing to favour grass production (e.g. Bryant 1982; Belsky, Matzke & Uselman 1999). Fire alters streams draining forests and shrublands (Minshall, Robinson & Lawrence 1997; de Koff et al. 2006), but information is sparse for tallgrass prairie streams (except see Dodds et al. 1996; Larson et al. 2013a,b). Livestock grazing alters water quality in forests (Kauffman & Krueger 1984; Agouridis et al. 2005), deserts (Belsky, Matzke & Uselman 1999) and pastures (Hooda et al. 2000). Few studies address fire–grazer interactions or riparian fencing efficacy using explicit experimental designs with appropriate reference and control sites. Further, few studies encompass full catchments (the natural experimental unit for water quality) because replication is difficult (Rinne 1988; Larsen et al. 1998; Sarr 2002).

Successful environmental management requires sound grounding in ecological principles, such as disturbance and resiliency. Currently, most ecological state changes from disturbance are inferred from observational data (Schröder, Persson & De Roos 2005), but we tested for changes and resiliency experimentally. We used the term ‘resiliency’ as the tendency of ecosystem state variables to quickly return to pre-treatment conditions following disturbance (Ludwig, Walker & Holling 1997). ‘Pulse’ and ‘press’ terminology are used in the ecological literature to refer either to the disturbance or to the response type (Lake 2000). We viewed the repeated pasturing of cattle over three years as both a multiyear press disturbance and also within a year as a pulsed disturbance.

We assessed the influences of PBG and riparian fencing on tallgrass prairie water chemistry and whole-stream metabolism using a Before-After/Control-Impact (BACI) design. We hypothesized PBG would increase mean concentrations of nutrients, sediments and coliform bacteria, which would subsequently increase microbial production and respiration; thus, we predicted PBG would alter the structure and functions of streams. We further predicted that the strongest effects would be observed when cattle were present and these effects would diminish when cattle were removed from pasture (i.e. the system would be resilient and exhibit a pulse response; Lake 2000). We expected the riparian fencing would fully buffer water quality changes because cattle would not have stream access for direct inputs and the ungrazed riparian vegetation would absorb or slow overland flow.

Materials and methods

Description of study sites

We studied six, first-order catchments at Osage Prairie in south-western Missouri, USA, which were completely encompassed by native tallgrass prairie (Fig. 1). Osage Prairie was owned and managed by the Missouri Department of Conservation and The Nature Conservancy. Past land use included cattle grazing from the early 1900 to 1987, midsummer haying every 3 years, and prescribed burning every 3–5 years. See Table S1 in Supporting Information for more catchment attributes.

Details are in the caption following the image
A catchment map during a patch-burn grazing (PBG) experiment in 2009–2013. Prescribed fire occurred in one-third of each catchment each year as indicated by burn breaks. We repeatedly sampled water on 38 occasions at the sampling site at the most downstream location of each catchment.

Experimental design

We used a Before-After/Control-Impact design with samples paired in time (BACIP) to test for stream disturbances from PBG. The BACIP design evaluated changes among the control and two types of impact sites after experimental treatments were applied (Downes 2002). Six catchments were allocated to three treatments, which were randomized to a catchment (Fig. 1). The treatments included as follows: fire but no grazers (‘Control’; = 2 control catchments), grazing where cattle had free access to the riparian areas and streams (‘PBG–Unfenced riparian’; = 2 impact catchments) and grazing with riparian fencing (‘PBG–Fenced riparian’; = 2 impact catchments). The ‘impact’ results represent the combined effects of fire and grazing, with or without, riparian fencing. The comparison of PBG–Unfenced to the Controls represents impacts from PBG, and the comparisons of PBG–Fenced to PBG–Unfenced and Controls represent the impacts from riparian fencing. The experiment was conducted from September 2009 to July 2013 (45 months).

The ‘before’ (or pre-treatment) phase occurred September 2009–March 2011, in which all catchments had no fire or grazing within the last 3 years. Pre-treatment results are described in detail in Larson et al. (2013a), but all six streams exhibited similar characteristics. The ‘after’ (or treatment) phase followed from April 2011–July 2013 when we implemented PBG (fire and grazing). There is a diagram of the design structure in Appendix S1.

All six catchments were burned in the treatment period; specifically, in mid-April 2011, 2012 and 2013, a prescribed patch-burn was carried out in one-third of each catchment. Our control sites were also burned because periodic fire is a reference condition (Anderson 2006) and minimally influences water chemistry (Larson et al. 2013a,b).

The riparian fencing in the two PBG–Fenced riparian catchments had two-strand poly-electric tape installed for 10 m on each side of the geomorphically active stream channel. We acknowledge riparian width is a complex aspect of riparian protection, but we chose a 10-m riparian width: (i) according to regional standards and (ii) other studies suggested a 10-m riparian zone captures most nonpoint–source run-off in small grassland catchments (e.g. Daniels & Gilliam 1996; Lee, Smyth & Boutin 2004).

The four catchments with PBG had yearling stocker calves stocked at a density of ~0·825 calves ha−1 (where one calf is 159–228 kg). This is a low-to-moderate stocking rate for the region. Cattle were present from 1 May to 31 July in the three treatment years. The animals were always provided with drinking water tanks located in the upper catchment away from the stream.

Water quality field sampling and analyses

We sampled each stream monthly at the base of each catchment (when flowing). We collected and analysed water samples for total suspended solids, nutrients (total phosphorus, total nitrogen, ammonium, nitrate and soluble reactive phosphorus) and Escherichia coli bacteria in acid-washed bottles (APHA 1995; Environmental Protection Agency 2005). We collected rocks in situ for chlorophyll a determination and analysed them fluorometrically (Sartory & Grobbelaar 1984, Welschmeyer 1994). We measured dissolved oxygen concentration dynamics and modelled whole-stream metabolism estimates using the single-station method (Holtgrieve et al. 2010; Dodds et al. 2013). We have varying sample sizes across variables, sites and years due to slight differences in flow, stream freezing and collection of E. coli data at a later starting date. See Appendix S1 for detailed procedures.

Cattle locations

We monitored the position of one stocker calf per catchment throughout the grazing season each year using LOTEK GPS_3300 wireless collars that recorded at 1-h intervals (Lotek, Newmarket, Canada). Using ArcGIS v.10.0 (Environmental Systems Research Institute, Redlands, CA, ON, USA), we digitized streams using the 2012 orthoimagery of Vernon County, Missouri (USDA 2012). We mapped and analysed cattle location data to identify the frequency of cattle within the 10-m riparian areas and burned patches. We measured riparian vegetation structures and reported methods and supporting results in Table S1.

Detecting responses to PBG

We searched for changes in stream variables after the implementation of PBG. We conducted two typical analysis procedures for detecting ecological changes (Andersen et al. 2009): (i) a principal component analysis as a visualization method to observe simultaneous changes in multiple variables and (ii) linear mixed-effects models with repeated measures (Laird & Ware 1982). We included data from ~36 sampling dates of baseflow conditions. Log-transformation successfully normalized the data. We performed all statistical analyses using the software R 3.0.2 (R Core Team, 2013) and the packages nlme (Pinheiro et al. 2011) and phia (Rosario-Martinez 2013).

We ran principal components analysis to graphically illustrate the relationships among treatments and water chemistry (Andersen et al. 2009). The variables were standardized using a Z-score (Yi-Ȳ/s) in order to compare measurements that are not in the same units and to down-weight high values (Gotelli & Ellison 2004). The biplot contained a component score associated with a sampling date in the treatment period in order to differentiate periods when cattle were present and absent in relation to water quality.

We also conducted a balanced design, linear mixed-effects model for each response variable (Laird & Ware 1982; Pinheiro, Bates & Lindstrom 1995; Faraway 2005). The mixed-effects, repeated-measures model was a substitute for traditional Before-After/Control-Impact (BACI) analysis for our data set because we had treatment replication (= 2), two ‘impact’ treatments (PBG–Fenced and PBG–Unfenced) and the use of within-subject factor, Time, allowed for correlated data from repeated measures (Laird & Ware 1982). We ran a model that included year, which was not a significant factor (Appendix S1), so we reported results of a reduced model to increase the degrees of freedom:
urn:x-wiley:00218901:media:jpe12692:jpe12692-math-0001

Specifically, Treatment (Control, PBG–Fenced or PBG–Unfenced) was a fixed effect and Time (before or after the initiation of PBG) was a fixed effect with repeated measures every sampling month. The Error variance is unexplained variance and is calculated as the ratio of between-subject variance, Catchment, and the within-treatment variance, Time. Catchment was a random effect, which allowed inference to tallgrass prairie sites beyond our study. A significant interaction term (Treatment*Time) indicated a probable change in the after period at the impact sites relative to the control sites. If the interaction term was statistically significant (α = 0·10), we proceeded with pairwise comparisons using post hoc interaction analysis, corrected using Holm–Bonferroni for multiple tests (Rosario-Martinez 2013). We determined ‘significance’ at a level of α = 0·10 or a moderate to large effect size (Cohen's D > 0·4, combined with graphic trends; Murtaugh 2002; Cohen 2013).

Results

Cattle locations

Riparian areas comprised 5–15% of the total catchment areas. Cattle spent an average of about 50% of their time in the most recently burned patch (33% of the pasture), with variation-dependent on catchment and year (Fig. 2; Table S2). The riparian fencing in PBG–Fenced catchments effectively excluded the cattle. In the PBG–Unfenced catchments, cattle spent 2–22% of their time in the riparian areas. Cattle use was greatest in the lower one-third of riparian areas (i.e. most downstream locations), irrespective of burn location and study year, likely due to the presence of water and shade. Vegetation structure was well developed in PBG–Fenced riparian areas and lacking in the PBG–Unfenced areas (Table S1), which highlighted the efficacy of riparian fencing at exclusion, as well as grazing activity in the PBG–Unfenced riparian areas.

Details are in the caption following the image
Cattle positions every 2 h in May–July at Osage Prairie, MO, USA, during a three-year patch-burn grazing experiment (PBG). Fire occurred in the lowest one-third of the catchment in 2011, the middle in 2012 and the upper catchment in 2013 as indicated by the fire break lines. Treatments included two catchments under PBG with riparian fencing (‘Fenced’) and two catchments that allow cattle full access to the riparian areas (‘Unfenced’).

Water quality after patch-burn grazing

Stream concentrations of nutrients, suspended sediments and E. coli were consistently low in the pre-treatment phase. Following PBG, the greatest concentrations of compounds used to indicate water quality were in PBG catchments, with and without riparian fencing, when cattle were present (Tables 1 and S3; Figs 3 and 4). The first two PCA axes explained 81% of the data set's variation, and the eigenvectors showed a gradient of water quality concentrations in relation to when cattle were present and absent. When cattle were removed from pasture, the concentrations declined but typically did not return to pre-treatment concentrations. For most variables, the treatment effect size comparing PBG–Unfenced streams to Controls was moderate to large, indicating a strong effect of PBG without riparian fencing on these streams. Additionally, the treatment effect size comparing PBG–Unfenced to PBG–Fenced streams was small, indicating the riparian fencing only moderately mitigated impacts from PBG.

Table 1. A summary of the statistical changes in water quality following patch-burn grazing (PBG) at Osage Prairie, USA, from 2009 to 2013. Treatments included two fenced (F) and two unfenced (U) riparian catchments with PBG and two control catchments (C). We focused on the interactive term Treatment*Time, which indicated a significant change at the impact sites compared to the controls after the treatment was applied. We conducted a post hoc interaction analysis corrected for multiple tests if the interaction was significant (Rosario-Martinez 2013); BA stands for before/after (Time); and CF, CU and FU are treatment contrasts (C = control, F = fenced riparian, U = unfenced riparian)
Water quality characteristic Model variable Degrees of freedom (numerator, denominator) F statistic P-value Post hoc interaction analysis Effect size (Cohen's d) Effect size category (Cohen 2013)
Total suspended solids (mg L−1) Time 1,27 0·04 0·84 U-C: 0·3 Small
Treatment 2,3 0·31 0·76 F-C: 0·4 Moderate
Time*Treat 2,27 0·44 0·65 U-F: −0·1 Small
Total nitrogen (μg L−1) Time 1,26 69·65 <0·01 U-C: 0·9 Large
Treatment 2,3 0·27 0·77 F-C: 0·7 Moderate
Time*Treat 2,26 0·95 0·49 U-F: 0·2 Small
Ammonium (μg N L−1) Time 1,27 11·22 <0·01 U-C: 0·1 Small
Treatment 2,3 1·05 0·45 F-C: 0·1 Small
Time*Treat 2,27 0·12 0·89 U-F: 0·1 Small
Nitrate (μg N L−1) Time 1,27 157·36 <0·01 BA*UC; F1  = 9·77, < 0·01 U-C: 0·8 Large
Treatment 2,3 0·91 0·52 BA*FC; F1  = 2·71, P = 0·14 F-C: 0·6 Moderate
Time*Treat 2,27 5·06 0·02 BA*UF; F1  = 3·29, P = 0·14 U-F: 0·3 Small
Total phosphorus (μg L−1) Time 1,21 3·96 0·06 U-C: 0·4 Moderate
Treatment 2,3 0·22 0·82 F-C: 0·2 Small
Time*Treat 2,21 0·81 0·46 U-F: 0·2 Small
Soluble reactive phosphorus (μg L−1) Time 1,25 0·43 0·52 U-C: 2·9 Large
Treatment 2,3 0·46 0·67 F-C: 2·6 Large
Time*Treat 2,25 1·37 0·27 U-F: −0·1 Small
Escherichia coli (colony forming units) Time 1,14 25·43 <0·01 BA*UC; F1  = 5·59, P = 0·05 U-C: 0·9 Large
Treatment 2,3 2·28 0·25 BA*FC; F1  = 0·88, P = 0·35 F-C: 0·4 Moderate
Time*Treat 2,14 2·88 0·09 BA*UF; F1  = 2·22, P = 0·27 U-F: 0·4 Moderate
Benthic chlorophyll a (μg cm−2) Time 1,19 0·16 0·70 BA*UC; F1  = 6·13, P = 0·04 U-C: 0·6 Moderate
Treatment 2,3 6·75 0·08 BA*FC; F1  = 1·62, P = 0·40 F-C: −0·3 Small
Time*Treat 2,19 3·08 0·07 BA*UF; F1  = 1·65, P = 0·39 U-F: 0·5 Moderate
Gross primary production (g O2 m−2 day−1) Time 1,15 1·36 0·26 U-C: 0·7 Moderate
Treatment 2,3 0·94 0·48 F-C: 0·7 Moderate
Time*Treat 2,15 2·02 0·16 U-F: 0·1 Small
Ecosystem respiration (g O2 m−2 day−1) Time 1,15 1·87 0·19 U-C: 0·8 Large
Treatment 2,3 1·44 0·36 F-C: 0·9 Large
Time*Treat 2,15 0·82 0·46 U-F: 0·1 Small
Net ecosystem production (g O2 m−2 day−1) Time 1,15 2·13 0·16 U-C: 0·6 Moderate
Treatment 2,3 1·92 0·29 F-C: 0·9 Large
Time*Treat 2,15 0·64 0·54 U-F: 0·6 Moderate
Details are in the caption following the image
A principal components analysis showing the relationship of treatments to gradients of several water quality variables. Data are from Osage Prairie, MO, USA, in 2011–2013 and include three treatments: patch-burn grazing (PBG) with riparian fencing (F), PBG with grazer access to streams (G) and control sites without grazing (C). The light grey symbols are sample dates when cattle were absent and dark grey symbols indicate when cattle were present.
Details are in the caption following the image
Time-series plots of water quality from Osage Prairie, MO, USA, before and after a patch-burn grazing (PBG) experiment in years 2009–2013. The data are averages of two catchments per treatment for ease of visualizing trends. The dashed vertical line shows the separation of the before-and-after periods of PBG. The grey panels indicate sampling dates when cattle were present from 1 May to 31 July.

Total nitrogen (TN) was low in the pre-treatment phase (<300 μg N L−1), but after PBG increased onefold to fourfold in both PBG–Fenced and PBG–Unfenced streams relative to the controls (Tables 1 and S3; Figs 4 and 5). The values of TN were highest when cattle were present and repeatedly exceeded 1800 μg N L−1. The TN concentrations decreased when cattle were removed, but neither to pre-treatment nor to control values. Nitrate values increased 1–2 orders of magnitude when cattle were present in PBG–Unfenced riparian catchments. Mean total phosphorus and soluble reactive phosphorus concentrations increased more than an order of magnitude in both the PBG–Unfenced and PBG–Fenced riparian catchments when cattle were present. Escherichia coli bacterial counts significantly increased by threefold to tenfold in the PBG–Unfenced riparian catchments when cattle were present, but declined to negligible values within 2 months of cattle absence. In contrast, no significant changes occurred in total suspended solids, inorganic suspended solids or organic suspended solids during PBG (Fig. S1); however, the largest total suspended solids values occurred when cattle were present on PBG–Fenced and PBG–Unfenced catchments.

Details are in the caption following the image
The principles and evidence for press and pulse responses from patch-burn grazing (PBG). Panels (a) and (b) show a disturbance at the dashed line and a response type (Lake 2000). Panels (c) and (d) show evidence of a press and pulse response following a PBG experiment. The black dashed line shows the before-and-after period of an applied experiment threshold. ‘Sampling periods’ are 1-month intervals. The coloured dashed lines in panel (c) are the means of each of the response variables after PBG, and panel (d) has a trend line to highlight the pulse dynamics when cattle were present.

Ecosystem functions after patch-burn grazing

Benthic chlorophyll a significantly, but marginally, increased in PBG–Unfenced riparian catchments (Figs 3 and 4; Tables 1 and S3). We detected large treatment effects for gross primary production (GPP), ecosystem respiration (ER) and net ecosystem production (NEP) at both PBG–Fenced and PBG–Unfenced riparian catchments compared to the controls. The treatment effect size for PBG–Fenced compared to PBG–Unfenced was small, indicating the riparian fencing did not mitigate changes to whole-stream metabolism. The streams were slightly net heterotrophic but shifted towards autotrophy following PBG. The GPP and ER rates did not correlate with nutrient concentrations (Fig. S2).

Discussion

The tallgrass prairie streams were strongly influenced by the fire–cattle grazing interaction. The increased nutrient concentrations, bacterial loads, algal production and whole-stream metabolism following PBG are generally considered undesirable, negative impacts to water quality. However, fire and grazing are natural processes in grasslands and given the conservation and economic benefits, it will likely continue. Preserving grasslands world-wide can include livestock grazing to mimic historical processes and conditions created by native grazers, but not without the potential to alter water quality and ecosystem functions.

Contrary to expectations, we showed riparian fencing only moderately mitigated water quality changes associated with PBG. We acknowledged the low replication at the catchment scale (= 2); however, based on the graphics, statistics and effect size, we suspect our study is a good representation of PBG effects. Based on significant changes for several ecosystem state variables (water quality and whole-system metabolic activity (this study)), riparian vegetation (Table S1, Larson 2014) and macroinvertebrate assemblage (Jackson et al. 2015), we concluded that PBG substantially altered the stream ecosystem's structure and functions.

Ecological alterations following PBG

The PCA and mixed-effects modelling indicated substantial alterations in ecological conditions for multiple response variables. During PBG, the increases in mean, median and variance of water quality concentrations and ecological processes indicated a systematic shift, but we do not know whether the system is resilient enough to return to pre-treatment conditions when the driver is removed (Carpenter & Brock 2006).

Disturbance is central to stream ecology, and it is often desirable for managers to differentiate stressors and responses that are chronic (press) or acute (pulse that indicate resiliency). By definition, press responses arise quickly and then maintain a near-constant level, whereas pulse responses are sharply delimited change but short term (a form of resilience; Lake 2000). In this experiment, PBG included both press and pulse response types (Fig. 5), which provided further inferences about ecological change and resiliency. As examples from this data set, the mean, median and variance for total nitrogen remained higher than the pre-treatment and control conditions after removal of cattle, which indicated a press response. Conversely, E. coli and nitrate were strong, pulsed responses to PBG because values spiked in the presence of cattle but returned to baseline conditions shortly after cattle were removed (a cycle that repeated across the three treatment years), indicating resiliency to this disturbance. Some state variables did not change significantly in response to grazing (total suspended solids), highlighting resistance of change from grazing.

Management implications

PBG compared to other land-use effects on streams

Our data showed that PBG caused streams to exceed nutrient, sediment and E. coli reference conditions for this ecoregion (Table 2). Prior to PBG, the streams had similar water quality compared to reference tallgrass prairie streams, prairie streams with fire (but no grazers) and with bison grazing. With PBG, the mean TN and TSS concentrations doubled and the TP concentrations quadrupled compared to recommended reference conditions. Escherichia coli is safe for recreational contact at <125 CFU (Environmental Protection Agency 1986). When cattle were absent, values were always below this criterion; however, when cattle were present in both PBG–Fenced and PBG–Unfenced catchments, most samples exceeded the reference for 2 months. The whole-stream metabolism rates were similar to published reference values for tallgrass prairie streams (Mulholland et al. 2001), but increased primary production could alter these stream food webs (Jackson et al. 2015).

Table 2. A comparison of Osage Prairie water quality to suggested reference values and common land management. Treatments at Osage Prairie were streams with no grazing (Controls), fenced riparian catchments with patch-burn grazing (PBG–Fenced riparian) and unfenced riparian catchments with patch-burn grazing (PBG–Unfenced riparian)
Avg. total nitrogen (μg N L−1) Avg. total phosphorus (μg P L−1) Avg. total suspended solids (mg L−1) Citations
Reference 430 47 Data not available Environmental Protection Agency (2001), Smith, Alexander & Schwarz (2003), Dodds & Whiles (2004)
Controls and fire 300 20 10 This paper, Larson et al. (2013b)
Bison grazing 350 10 2 Larson et al. (2013b)
PBG–Fenced riparian 301 155 18 This paper
PBG–Unfenced riparian 705 207 16 This paper
Row-crop agriculture 2150 270 200 Dodds & Whiles (2004)

The water quality following PBG indicated substantially less impact than streams under row-crop agriculture (Table 2), a common land-use conversion of formerly tallgrass prairie. For example, regional streams with row crops had an order of magnitude greater TSS and double the TN than the studied PBG streams. However, the average and pulsed TP concentrations with PBG approached those of row-crop streams, which altered stream productivity and community respiration (Dodds, Smith & Lohman 2002). Overall, the seasonal agricultural production of cattle using PBG at moderate grazing rates had lower impact on aquatic systems than row-crop agriculture.

Despite degradation of water quality, PBG may be a management compromise for water quality compared to conventional grazing. The current management regime in many tallgrass prairies consists of annually burning large patches and grazing with high stock densities (0·8–1·8 steer ha−1) for a full season (Owensby et al. 2008). In contrast, PBG reduces fire frequency and concentrates cattle impacts to smaller patches (West et al. 2016). Studies in other ecosystems suggested rest-rotation management could reduce riparian impacts (e.g. Fuhlendorf & Engle 2004; Haan et al. 2010). Future study is needed to adequately compare water quality in catchments with traditional management and PBG.

Alternative management strategies with PBG

Livestock management techniques that are promoted as best management practices may, or may not, reduce impacts to riparian areas. These practices include the following: providing shade tents, water troughs and protein supplements distant from streams (Agouridis et al. 2005; Franklin et al. 2009; Kaucner et al. 2013), as well as the removal of riparian wood to reduce shade that cattle seek (e.g. Briggs et al. 2005; Allred et al. 2011a). In this study, we did not provide any shade or water troughs in the lower catchment and thus suspect cattle preferred the lower catchment because greater riparian canopy cover provided shade, higher stream discharge (water availability) and/or diet selection differences. Despite management efforts, cattle may be attracted to riparian zones due to shade, predator protection and water in a single location (Bryant 1982).

An alternative approach using PBG could be to replace cattle with native bison, which show less preference for riparian zones and have minimal effects on water quality (Allred et al. 2011a; Larson et al. 2013b). Bison prefer recently burned watersheds (Coppedge & Shaw 1998), so PBG approaches could work. However, bison are difficult to manage, which may make bison grazing unfeasible in small pastures (Allred, Fuhlendorf & Hamilton 2011b).

Several elements have been proposed for a biodiversity-centred, rangeland-management framework, including: grazing unit size, fire and grazing patterns, woody vegetation removal management, and stream hydrology (e.g. Allred et al. 2011a; Augustine & Derner 2014; Freese, Fuhlendorf & Kunkel 2014; Scasta et al. 2015). Targeted grazing to minimize riparian contact also requires detailed, site-specific knowledge and producer investments to execute (Knight, Toombs & Derner 2011; Macon 2014). Therefore, management needs to consider multiple facets and techniques for reducing riparian and stream impacts.

Riparian fencing modestly benefited streams

Riparian protection is a key to water quality management, but surprisingly few studies demonstrated the effectiveness of riparian fencing for cattle exclusion and protecting water quality impacts based on whole-catchment experiments. For example, prior studies used riparian-fenced catchments as the control treatment (e.g. Miller et al. 2010); however, our study suggested that riparian fencing moderately mitigated water quality impacts but was not a control to compare to unfenced riparian grazing. The assumption that riparian-fenced studies are adequate controls could underestimate the effects of cattle grazing.

Because the riparian fencing was robust at exclusion (Fig. 2) and soil depths were shallow, we suspected overland flow or subsurface flow paths might be a mechanism for how the E. coli and nutrients entered the streams. Riparian vegetation was strongly affected by grazing (Table S1) so it is not clear whether the slightly improved water quality in PBG–Fenced catchments was due to intact riparian vegetation and/or simply because cattle were not able to directly deposit faecal matter and suspend sediments in streams.

Future research could focus on required riparian buffer widths and understanding overland and subsurface flow paths to identify the mechanisms responsible and how these riparian fencing results apply to other geographical locations (West et al. 2016). We suspected that regions with deep soils and minimal overland flow would have the best riparian fencing results. Fencing may greatly reduce fluvial geomorphic changes, such as width-to-depth ratios in streams (B. Grudzinski unpublished data, Belsky, Matzke & Uselman 1999). Riparian fencing has several disadvantages such as economic constraints to livestock producers (Platts & Wagstaff 1984) and the loss of ecological benefits from riparian grazing, such as protection from woody overgrowth and creation of new wildlife habitats (e.g. Briggs et al. 2005; Larson 2014). However, fencing may be feasible and beneficial at sites where water quality protection is a management priority.

Acknowledgements

We thank the Missouri Department of Conservation and Kansas State University for funding. We especially thank K. Sullivan, L. Gilmore, K. Winders, S. Whitaker, J. Persinger, B. Jamison and B. Hrabik for help designing and executing this study. We thank T. Laskowski, K. Jackson, C. Larson, J. Vandermyde, B. Grudzinski, K. Erndt, J. Rogosch, K. Heinrich and L. Bansbach for field and laboratory assistance. Thanks to K. Gido, M. Daniels, J. Fry, J. Whitney, K. Sullivan and anonymous reviewers for thoughtful critiques. Contribution no. 16-310-J from the Kansas Agricultural Experiment Station.

    Data accessibility

    Catchment descriptions: uploaded as online supporting information

    Water chemistry summary statistics: uploaded as online supporting information

    Summary data of watershed, riparian and burn areas: Dryad Digital Repository, http://dx.doi.org/10.5061/dryad.qc264 (Larson et al. 2016)

    Raw water chemistry data: Dryad Digital Repository, http://dx.doi.org/10.5061/dryad.qc264 (Larson et al. 2016)

    Raw data of cattle GPS locations within watersheds and riparian zones: Dryad Digital Repository, http://dx.doi.org/10.5061/dryad.qc264 (Larson et al. 2016)