SEARCH

SEARCH BY CITATION

Keywords:

  • agriculture;
  • aquatic invertebrate communities;
  • boosted regression trees;
  • intensive land use;
  • structural equation models

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

1. Riparian management has been embraced by water and land managers globally to offset the deleterious effects of intensive agricultural land use on aquatic ecosystems. However, the documented responses of stream communities to riparian management have been variable, particularly in highly degraded systems.

2. We used boosted regression trees and structural equation models to assess the effects of riparian condition and stream size on the invertebrate communities of 64 agricultural waterways on the Canterbury Plains, New Zealand. We hypothesized that small streams would be more degraded than larger waterways but would show a greater increase in the abundance of pollution-sensitive aquatic invertebrates in response to riparian management. We also predicted that land-use legacies of poor in-stream habitat would reduce the effectiveness of current riparian management. The two strongest determinants of community structure were primarily in-stream habitat, where sedimentation and low water velocity had negative impacts on stream communities, and stream size, with smaller waterways generally more impacted than large waterways. Not surprisingly, with >150 years of agriculture and patchy riparian management on the plains, current management has not greatly improved in-stream habitat and thus had little effect on the abundance of sensitive aquatic insect (EPT) taxa.

3. Managed streams did, however, have more pollution-sensitive communities in general. This was largely mediated by decreased stream temperature, narrower/deeper channels and greater organic matter resources in streams with riparian planting and restricted stock access. Thus, if water velocity and sedimentation issues can be mitigated, then riparian management should become more effective.

4.Synthesis and applications. Within the context of a degraded agricultural landscape, we identified factors limiting the effectiveness of riparian management for stream invertebrate communities. Riparian management should primarily target and protect small streams and those without degraded in-stream habitat. Intensive management, such as in-stream habitat or channel morphology modification, may be needed to address historical factors (e.g. low velocity and sedimentation), which otherwise may continue to limit community recovery.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Conversion of land to agriculture and intensification of land use are global phenomena linked to population expansion (Matson et al. 1997; Tilman 1999). Land-use intensification is one of the predominant changes of the last century and is likely to have serious consequences for biodiversity and ecosystem functioning at local, regional and global scales (Matson et al. 1997). Aquatic ecosystems are particularly vulnerable to these changes, and land-use intensification consistently has deleterious effects on water quality and aquatic communities (Harding & Winterbourn 1995; Harding, Winterbourn & McDiffett 1997; Allan 2004; Gordon, Peterson & Bennett 2007). With land-use intensification set to continue, it is imperative to develop practical ways of managing and protecting aquatic systems in agricultural landscapes.

Management of the riparian zone, including tree planting and the exclusion of stock, is one such tool promoted universally to reduce the negative impacts of land-use activities on stream systems. In general, riparian management, particularly planting, reduces stream temperatures, lessens sediment and nutrient run-off, and enhances in-stream habitat and nitrogen processing (Mayer et al. 2005; Craig et al. 2008). However, the responses of stream invertebrate communities to these efforts have been mixed and highly context-dependent (Parkyn et al. 2003; Rhodes, Closs & Townsend 2007; Wilcock et al. 2009; Palmer, Menninger & Bernhardt 2010). Moreover, the responsiveness of invertebrate communities to riparian management is likely to be determined by the dimensions and condition of the riparian buffer, by the hydrology, geology and land-use history of the area, and by the size and network position of the river reach (Parkyn et al. 2003; Craig et al. 2008; Wilcock et al. 2009).

Despite the proliferation of guidelines for riparian management techniques (Lee, Smyth & Boutin 2004; Mayer et al. 2005; Craig et al. 2008), there have been surprisingly few tests of the effectiveness of riparian management (Richardson & Danehy 2007; but see Quinn & Stroud 2002). Furthermore, expectations of the effectiveness of riparian management should be treated with caution, particularly in degraded systems, as historical legacies of land use may limit the benefits contemporary management can have on stream invertebrate communities (Harding et al. 1998; Zhang, Richardson & Pinto 2009). The absence of suitable reference sites in a degraded landscape can also make comparative assessments of the success of riparian management difficult.

The effectiveness of riparian management in protecting or improving stream invertebrate communities has seldom been investigated explicitly in the context of large-scale degraded ecosystems. In addition, there have been few direct tests of the role of stream size and network location within a landscape with a long history of agricultural land use. This is particularly important because of the high biodiversity value of headwater streams (Meyer et al. 2007) and because of how they influence larger downstream reaches (Dodds & Oakes 2008), indicating their management is particularly crucial for maintaining ecosystem health. This study provides the first investigation into direct and indirect (mediated) effects of riparian management and stream size on invertebrate communities within the context of an ecosystem degraded by agricultural land use over large temporal (∼160 years) and spatial (100s kilometres) scales. In particular, through the use of boosted regression trees and structural equation modelling, we identify the mechanisms that are likely to limit positive responses in communities to current riparian management, and we provide recommendations for improving riparian management effectiveness in degraded landscapes.

We specifically hypothesized that small streams would be more degraded than larger waterways, owing to reduced dilution and flushing of pollutants, but would show larger improvements in aquatic invertebrate communities in response to riparian management. In addition, we predicted that the effectiveness of modern riparian management within the context of a degraded agricultural landscape is likely to be limited by historical and ongoing negative effects on in-stream habitat.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Study area and site selection

Agricultural development has been occurring on the Canterbury Plains, South Island, New Zealand, for more than 150 years, with mixed sheep, dairy, beef, deer, cropping and lifestyle farms replacing forest, scrub and wetlands. This land-use intensification has been especially rapid in the last two decades (Taylor & Smith 1997; MacLeod & Moller 2006) and has been associated with degradation of water quality (Smith et al. 1993). The majority of riparian plantings on the Canterbury Plains are narrow strips of exotic vegetation planted as wind breaks or for bank stabilization, e.g. willow Salix spp., poplars Populus spp., pines Pinus spp. macrocarpa Cupressus macrocarpa and gorse Ulex europaeus. In some areas, active programmes of native vegetation planting have been undertaken, but these programmes are currently limited in size, distribution and age.

We sampled 64 waterways on the Canterbury Plains. The network of waterways on the plains is complicated by the presence of extensive stock water and irrigation race networks, and thus, we only sampled waterways that included natural meandering stream reaches and were connected to a downstream river system flowing into a lake or the ocean. We stratified our sampling so that sites varied in width (<1–10 m), degree of riparian management (bare banks to fenced and planted) and land-use intensity. All sites were below 200 m a.s.l (Figs 1 and S1, Supporting Information) and were not connected to one another by surface flow (see Appendix S1 and Table S1, Supporting Information).

image

Figure 1.  Study site locations on the Canterbury Plains, South Island, New Zealand. Only major rivers are shown.

Download figure to PowerPoint

Field methods

At each waterway, a 50-m sampling reach was delineated using GPS and sampled between 2 December 2008 and 12 January 2009 (austral summer). Within each reach, three Surber samples (0·09 m2, 500 μm mesh) were collected where flow was the fastest and substrate largest. A composite kick-net sample (0·25-m-wide triangular opening, 500-μm mesh) was taken from five additional microhabitats. Three leaf packs (6 g abscised willow leaves dried for 48 h at 60 °C, 1-cm mesh; Benfield 2007) were also installed in pools or backwaters at each site for 2 weeks in early February 2009.

All macroinvertebrates collected in Surber samples were counted and identified to species or genus where possible (using Winterbourn, Gregson & Dolphin 2000). Taxa from the kick-net samples and leaf packs were identified to species or genus, where possible, and added to the taxa list collected in Surber samples.

Reach-scale stream size was estimated from average channel wetted widths (five transects), discharge and depths (one transect; for details, see Appendix S2, Supporting Information). Riparian condition was measured using habitat assessment Protocol Two in Harding et al. (2009), which involved assessment of 13 variables (Appendix S3, Supporting Information). Spot measurements were made of water chemistry including dissolved oxygen and specific conductivity, while grab samples were taken to measure ammonia, soluble reactive phosphorus (SRP) and nitrite–nitrate nitrogen (Appendix S2, Supporting Information). Water temperature, shading and light were measured with loggers (Appendix S2, Supporting Information). To quantify stream habitat quality and heterogeneity, habitat type (riffles, runs, pools and backwaters), substrate size, macrophyte cover and channel shape variables were measured along the reach. Algal food resources along with inorganic sediment accumulation were assessed on colonization frits and measured using a fluorometer and calculation of ash-free dry mass (AFDM), respectively. In addition, woody debris cover was estimated visually, and coarse particulate organic matter (CPOM) was collected in the Surber samples (Appendix S2, Supporting Information).

Analyses

Five response variables were derived from the invertebrate community data: taxonomic richness, percentage abundance of Ephemeroptera, Plecoptera and Trichoptera (EPT), Quantitative Macroinvertebrate Community Index (QMCI) and the first two axes of a multivariate community composition ordination (Detrended Correspondence Analysis, DCA; Canoco for windows version 4·55; Ter Braak & Smilauer 2006; Fig. S2, Supporting Information). Taxonomic richness was calculated from the combined Surber, kick-net and leaf pack samples, while the other metrics used the average of the three Surber samples per site. Percentage EPT was arcsine-square-root-transformed before analysis to meet assumptions of normality. The QMCI (Stark 1998) is calculated on scores assigned to taxa according to their sensitivity to organic pollution and is widely used to assess the health of streams in New Zealand, particularly in response to varying land use (e.g. Townsend et al. 1997). A larger QMCI score indicates the presence of more pollution-sensitive taxa.

We used boosted regression trees (BRT) to test the main and interactive effects of variables related to riparian management, stream size and the environmental factors measured on the five community structure response variables (Appendix S4, Supporting Information). BRTs are an advanced regression modelling technique using machine learning (Friedman 2002) that are robust to irrelevant predictors and can assess higher-order interactions, making them well suited to ecological data (Elith, Leathwick & Hastie 2008). Pairwise co-linearity (r > 0·9) was assessed among the predictor variables prior to BRT analysis, and channel cross-sectional area and catchment area were removed, leaving 46 variables in the initial analysis (Table S1).

We ran separate BRT analyses for each of the five response variables using the ‘gbm’ library of Ridgeway (2004) in r (version 2·9·0; R Development Core Team 2009). The relative influence of each environmental variable was determined using code provided by Elith, Leathwick & Hastie (2008), and fits of the BRT models to the data were assessed using 10-fold cross-validation.

We then used structural equation modelling (SEM; amos v 16·0; Arbuckle 2007) to investigate the proximate mechanisms mediating the effects of riparian management and stream size that were identified in the BRT analyses. We ran separate SEMs on the invertebrate response variables. Full models were constructed in which all direct and indirect effects of stream size, riparian management, a size–management interaction, in-stream habitat and food resources were included (Fig. S3, Supporting Information). Each full model was then simplified to a most parsimonious reduced model (Appendix S5, Supporting Information).

The number of variables included in the SEMs was reduced using principal component analyses (PCA) to collapse related variables into axes for stream size: (axis 1: reach-scale size, axis 2: network position), riparian management: (axis 1), food resources: (axis 1) and in-stream habitat: (axis 1: water velocity and substrate size; axis 2: water temperature and channel shape). Only one axis of an ordination was included if the second explained little additional variation (Table 1: PCA variables; Appendix S6: PCA results, Supporting Information).

Table 1.   Contribution (%) of stream size, riparian management and other environmental factors to boosted regression tree (BRT) models explaining two axes of community composition (from ordination; see Fig. S2, Supporting Information), and taxa richness Thumbnail image of

To test for potential interactions between stream size and riparian management in the SEM and to avoid collinearity, we calculated a size × riparian management interaction variable by multiplying deviation scores (Kline & Dunn 2000). A more positive number was associated with large well-managed and small unmanaged streams, while a negative value was associated with smaller managed and larger unmanaged reaches.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The streams sampled had a large size range (width: 0·6–9·5 m, discharge: 0·0005–1·0 m3 s−1, Table S1). Many had high levels of fine bed sediment (<2 mm), with more than half having >50% bed cover in fine sediment. At only 17 of the 64 sites was >30% of the streambed comprised of cobbles. Four sites dried completely between visits.

Riparian management was generally poor with a median buffer width score of 1·5 (out of a possible 5), indicating that vegetation buffers generally ranged from <1 m to 5 m wide. The median vegetation composition score was 2, indicating the buffer vegetation was mainly exotic weedy shrubs or high grasses. Stock access was better with a median value of 4, indicating low ability of stock to reach the water, with streams either fenced or with naturally limited access.

A total of 84 invertebrate taxa were collected, with abundances from 460 ± 314 m−2 (mean ± SE) to 19 250 ± 7465 m−2, and taxa richness ranged from 14 to 31 (Table S2; taxa list). The QMCI indicated that >89% of the sites were in ‘fair’ or ‘poor’ condition (QMCI 4·99–4·0 and <4·0, respectively), with only one ‘excellent’ site (Fig. 2b). The most common taxa were pollution-tolerant oligochaetes (all 64 sites), orthoclad midges (63 sites), the hydrobiid snail Potamopyrgus (58 sites) and ostracods (53 sites). However, 40 sites contained the relatively pollution-sensitive leptophlebiid mayfly Deleatidium, and it was the most abundant invertebrate at six sites.

image

Figure 2.  Axis 1 sample scores from a community composition ordination, detrended correspondence analysis (DCA; Fig. S2), against the percentage EPT (arcsine-square-root-transformed) at each site (a) and axis 2 scores from the community composition DCA against the QMCI scores (b). A larger QMCI score indicates the presence of more pollution-sensitive taxa. Dashed lines indicate water quality guidelines based on the QMCI score (Stark 1998).

Download figure to PowerPoint

Community composition

The first two axes of the invertebrate community composition ordination (DCA) cumulatively explained 43% (Axis 1 = 27%, Axis 2 = 16%) of the variation in community composition (Fig. S1, Supporting Information). Axis 1 was highly positively related to %EPT (R2 = 0·74, F1,62 = 168·4, = <0·001, Fig. 2a); thus, a more positive DCA1 score indicated communities dominated by sensitive EPT taxa. DCA axis 2 was significantly positively related to the QMCI score (F1,62 = 17·2, = 0·001, R2 = 0·22, Fig. 2b), indicating that sites with a positive DCA2 axis score were dominated by pollution-sensitive communities (Fig. 2b). Subsequent analyses were conducted on DCA axes 1 and 2 (hereafter community composition 1 and 2) as these axes are orthogonal and include additional community information than the %EPT and QMCI scores alone.

The final BRT models for community composition axes 1 and 2 and taxa richness each identified 40 covariates (six were removed) and had average (±SE) cross-validation correlation coefficients of 0·79 ± 0·01, 0·58 ± 0·04 and 0·49 ± 0·01, respectively, where 1 indicated perfect model fit (Table 1). Variation in community composition axis 1 (∼%EPT) was related to in-stream habitat, catchment land-use and stream size, with riparian management accounting for 7% of variation (Table 1). For community composition axis 2 (∼QMCI), in-stream habitat, particularly average water temperature and variability, stream size, riparian management and food resources explained most of the variation (Table 1). Variation in taxa richness was most strongly related to in-stream habitat, stream size, food resources and adjacent land-use, with just over 6% of the variation accounted for by riparian management. No interactions between variables were identified in any BRT model. Thus, the BRT models explained substantial portions of community structure, but most of that variation was accounted for by in-stream habitat and stream size, rather than riparian management.

Direct and mediated effects of riparian management and stream size

The final boot-strapped structural equation models (SEM) for community composition axes 1 and 2 had acceptable goodness-of-fit statistics (Fig. 3) and had no significant differences between the predicted and observed covariance structures of the model and the data (community composition 1: = 0·42; community composition 2: = 0·34). Although the SEM for taxa richness also fit the data well (P = 0·52), the relatively low taxonomic richness in all sites (14–31) probably contributed to the fact that only 18% of the variation in taxa richness was explained by the SEM, and thus, the results for this analysis are not reported.

image

Figure 3.  The most parsimonious structural equation models testing the direct and indirect effects of riparian management, stream size (size 1: reach-scale size and size 2: network position or upstream length) and associated environmental variables on: (a) the percentage community composition axis 1 (∼%EPT) and (b) community composition axis 2 (∼QMCI). Arrows represent causal pathways from predictor to response variables. The number associated with each arrow is the unstandardized partial regression coefficient for the direct effect with the sign indicating whether the relationship is positive or negative. The statistical significance of individual regression coefficients is indicated by the degree of shading of the line (black  0·05, dark grey  0·10, light grey and dashed > 0·10). The thickness of the line is proportional to the magnitude of the standardized path coefficients, which relates to the effect sizes presented in Table 2. For the four endogenous variables in the final model, squared multiple correlations (R2) indicate the variance explained by all the associated pathways linking that variable. Goodness-of-fit statistics for the models were as follows: community composition 1; RMSEA = 0·02 (90% confidence limits = 0·00–0·12), AIC default model = 56·4, saturated model = 144·9, CMIN/d.f. = 1·02, and community composition 2; RMSEA = 0·04 (90% confidence limits = 0·00–0·12), AIC default model = 57·7, saturated model = 144·9, CMIN/d.f. = 1·11. CSA = cross-sectional area.

Download figure to PowerPoint

The SEM on community composition axis 1 (∼EPT) indicated that both components of stream size (reach-scale, i.e. width, depth and discharge, and network position, i.e. upstream length) had a multitude of effects on this component of the community. The higher proportion of EPT in the community in larger and further downstream reaches was mediated through faster water velocities and bigger substrate sizes in these sites, while a direct negative effect also indicated increased EPT community dominance in smaller reaches (Fig. 3a, Table 2). When these relationships were combined, they led to no overall significant effect of reach-scale size and only a weakly positive overall effect of network position on community composition 1 (∼EPT; Table 2). Network position was also negatively related to in-stream habitat, with warmer temperatures and wider, shallower stream channels further downstream. Riparian management had no significant effect on community composition 1 (Fig. 3a, Table 2) because, although riparian management had a strong, positive relationship with food resource availability, this had no detectable effect on the community (Fig. 3a). The riparian management × stream size interaction had negative effects on community composition 1, both directly and through decreases in water velocity and substrate size.

Table 2.   Standardized path coefficients from structural equation models (Fig. 3a,b) showing the direct, indirect and total effects (sum of direct and indirect effects) of variables influencing (a) community composition axis 1 (∼%EPT) and (b) community composition axis 2 (∼QMCI)
Causal variableDirectIndirectTotal
  1. sed, sediment; temp, temperature; channel, channel shape.

  2. Dashed lines indicate paths that were removed from the final, most parsimonious structural equation model. Standardized path coefficients show the number of SD of change in the dependent variable for every one SD of change in the independent variable. Significance levels based on bootstrapped confidence intervals, estimated across 500 random samples; *< 0·1, **< 0·05, ***< 0·001.

(a) DCA1
 Size 1 (reach-scale)−0·18**0·27**0·09
 Size 2 (network position)−0·21*0·43**0·23*
 Riparian management−0·12−0·12
 Size 1*Management−0·21**−0·14−0·34**
 Food resources
 Habitat 1 (velocity, sed)0·89**0·89**
 Habitat 2 (temp, channel)
(b) DCA2
 Size 1 (reach-scale)0·35**0·07*0·41**
 Size 2 (upstream length)−0·20**0·12−0·08
 Riparian management0·19**0·19**
 Size 1*Management0·03*0·03*
 Food resources0·36**0·36**
 Habitat 1 (velocity, sed)0·22*0·22*
 Habitat 2 (temp, channel)0·14

Reach-scale stream size directly affected community composition 2 (∼QMCI), with more pollution-sensitive communities in larger reaches. Only a small amount of this relationship was explained through positive size-effects on in-stream habitat (water velocity and substrate) (Fig. 3b, Table 2). Network position had no total effect on community composition owing to a negative direct effect (more pollution-sensitive communities in the headwaters) and slightly weaker positive effect mediated through better in-stream habitat in downstream sites cancelling each other out (Fig. 3b). The strong, positive effects of riparian management leading to more pollution-sensitive communities were entirely mediated through increases in the abundance of food resources, decreases in water temperature and narrower and deeper river channels in more managed stream reaches (Fig. 3b, Table 2).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The success of management initiatives to offset the effects of land-use intensification on stream ecosystems will depend on understanding how riparian management mitigates the negative effects of factors limiting stream communities. By identifying the relative impacts of limiting factors and the role that riparian management and stream size have in mediating their effects, our results allow more robust recommendations for improving riparian management techniques, particularly in landscapes experiencing long-term degradation.

Are small streams the most degraded?

Small and headwater streams often contribute a significant proportion of biodiversity within a river network (Meyer et al. 2007); however, when intensive agriculture homogenizes headwater habitats over a wide area, regional diversity can be constrained (Harding, Winterbourn & McDiffett 1997). During the 150 + years of agriculture on the Canterbury Plains, many headwater streams have been deforested, diverted, channelized or buried in pipes to create more land area for agriculture. Those that do remain are often managed as ‘farm drains’: not fenced, ‘cleaned’ annually with diggers, substrate removed and riparian vegetation sprayed, mowed or grazed.

Streams on the Canterbury Plains that were either closer to the headwaters (network position) or smaller (narrower, lower discharge) had poorer in-stream habitat with slower water velocities and more stream-bed sedimentation than larger or further downstream sites, possibly due to a reduced ability to dilute and flush out sediments. Associated with the poor in-stream habitat, these reaches had communities with fewer pollution-sensitive taxa, particularly less dominance of EPT than reaches that were wider or further downstream. However, the direct negative relationship between network position and both aspects of community composition indicated that headwater streams supported greater EPT abundances and more sensitive taxa than downstream reaches. This seemingly contradictory effect of more pollution-sensitive communities in both larger and downstream reaches (i.e. streams with low sediment and high flow levels) and also in smaller, headwater reaches can be explained by the presence of cool headwater springs, which are common in the coastal region of the Plains and support EPT taxa.

The dual effects of more pollution-sensitive taxa, particularly EPT, in some headwater reaches combined with larger and downstream reaches containing better habitat indicates that management of streams is likely to be complicated by the nature of the water source and the network position of the targeted reach. Generally, however, both small streams and those closer to the headwaters were the most impacted by agricultural land use on the Canterbury Plains. Headwater springs provide an exception to this, but because they contain particularly pollution-sensitive communities they also deserve special attention.

Thus, management of small streams is likely to be particularly critical as they are more vulnerable to adjacent land-use effects because of their size and often have a low status in protective policy owing to their relative obscurity in the landscape and on maps (Meyer & Wallace 2001; Baker, Weller & Jordan 2007). In addition, because of their prevalence, the negative effects of upstream land use can accumulate in downstream, larger reaches (Dodds & Oakes 2008), which can overwhelm any local riparian management (Parkyn et al. 2003).

Is riparian management effective?

Currently, there is little active riparian planting occurring on the Canterbury Plains, with most management being patchy or primarily designed as ‘wind breaks’ rather than for stream management. Many waterways had poor in-stream habitat (>50% sites had >50% fine sediment covering the streambed) and were dominated by pollution-tolerant communities, with relatively depauperate diversity and low abundances of sensitive Ephemeroptera, Plecoptera and Trichoptera (EPT), similar to invertebrate communities seen in pastoral land use elsewhere (Harding & Winterbourn 1995; Harding, Winterbourn & McDiffett 1997; Zweig & Rabeni 2001; Niyogi et al. 2007; Matthaei, Piggot & Townsend 2010).

Surprisingly, given the poor ecological state of many of the waterways and the limited extent of riparian plantings, we did see positive associations of riparian management with more sensitive aquatic invertebrate communities, largely mediated through better managed streams having more food resources (organic matter) and slightly better in-stream habitat (cooler water temperature and narrower channels).

However, although managed streams had more pollution-sensitive communities in general, riparian management was not related to an increase in the prevalence of the most sensitive taxa (EPT; Ephemeroptera, Plecoptera, Trichoptera; community composition axis 1). The use of SEM allowed further dissection of these effects: EPT taxa were most abundant in fast water velocities and large substrate sites; however, perhaps unsurprisingly, riparian management did not correlate with a reduction in stream-bed sedimentation or faster water velocities. Thus, the effectiveness of the current riparian management was likely to be limited by sedimentation and reduced stream flow in many sites. Nevertheless, it is promising that we identified some improvements in stream communities associated with riparian management, given the limited extent of active management and the relatively degraded aquatic habitat of waterways on the Canterbury Plains.

Is the effectiveness of riparian management dependent on stream size?

Riparian management should be more effective in small waterways than larger ones (Parkyn et al. 2003; Craig et al. 2008), as streams with smaller channels are easier to shade (Richardson & Danehy 2007; Quinn & Wright-Stow 2008), better retain organic matter inputs (Chen et al. 2006; Quinn, Phillips & Parkyn 2007), are often naturally more dependent on allochthonous inputs (Wallace et al. 1997) and have higher rates of nitrogen cycling than larger rivers (Peterson et al. 2001). However, although riparian management techniques are often scaled to stream size (e.g. Lee, Smyth & Boutin 2004; Correl 2005), there have been few direct tests of how stream size, either local channel size or network position, alters the effectiveness of riparian management (Correl 2005).

We saw significant effects of a reach-scale stream size by riparian management interaction on community composition 1 (related to %EPT). This interaction highlighted that large or better managed waterways of small size had communities with relatively more pollution-sensitive EPT taxa than those that were medium-sized or less well managed. However, the size by management interaction should be interpreted with caution as the network of irrigation and stock-water races on the Canterbury Plains complicates the comparison by introducing irregular water flows and alluvial fines from the larger braided rivers and may artificially increase the upstream catchment of some of the smaller stream channels. In addition, the small improvements in stream health in both downstream and larger reaches may be linked to dilution effects, rather than better management. Thus, the intrinsic ability of larger streams to resist land use impacts will be context-dependent, contingent on stream gradient and land-use intensity (Deschênes, Rodríguez & Bérubé 2007).

Scale of riparian management

Riparian management in the context of the Canterbury Plains was limited to relatively small and disconnected areas, ranging from waterways fenced off from stock to a short vegetation buffer that also provided shade to the waterway. Owing to an inability to distinguish electric fences and narrow riparian strips from aerial photographs, our assessment of riparian management was limited to our local-scale reaches rather than upstream catchment management. However, riparian management was so patchy on all rivers that it is unlikely there were large or consistent differences in the extent of management between catchments that might have led to a detectable larger-scale effect. If riparian management progresses and becomes more contiguous, not only will the impacts of upstream land use lessen in the downstream, ‘healthier’ sites (Dodds & Oakes 2008), but connected riparian plantings may provide corridors of habitat suitable for winged adult stages of both aquatic and terrestrial insects, enabling opportunities for dispersal from distant source populations (Petersen et al. 2004).

Interestingly, the dominant land use in either the riparian zone or catchment had only a slight influence on the stream communities, with less than 10% of the variation in community composition or taxa richness explained by the land use at either scale. Their relatively low explanatory power combined with the complexity of the SEM models relative to replicate stream numbers meant that land-use variables could not be included in the SEM analyses. However, it is likely that the previous >150 years of agricultural land use on the Canterbury Plains has left a legacy of lasting impacts on the stream communities (Harding et al. 1998; Zhang, Richardson & Pinto 2009). The rapid recent increase in conversion to dairy farming may result in our point in-time measure correlating poorly with cumulative effects, thus explaining the relatively low predictive power of land-use type in the boosted regression tree models.

How to improve the effectiveness of Riparian management within a degraded landscape?

Although manipulative or experimental research is need to confirm the mechanisms identified by our modelling approach, our models highlight the vulnerability of small streams to adjacent land use and the constraints that historical and ongoing in-stream habitat degradation places on the effectiveness of riparian management. Within a degraded landscape, it is imperative that small or headwater waterways with relatively un-impacted in-stream habitat are prioritized for riparian management as these sites can contain rare examples of pollution-sensitive headwater communities. In addition, as we found, long-term, widespread and intensive agricultural land use creates a template of poor water quality and truncated regional diversity that can limit the effectiveness of riparian management (Parkyn et al. 2003; Craig et al. 2008; Palmer, Menninger & Bernhardt 2010). Adding plantings or fences to sites that are already affected by poor in-stream habitat is unlikely to cause large local benefits but may reduce downstream effects. By adding riparian management to relatively un-impacted reaches, we can often reduce the decline of in-stream habitat (e.g. sediment input e.g. Parkyn et al. 2003), while more intensive measures, such as channel and in-stream habitat modification, may be needed for improvements in systems with already degraded stream habitat.

By identifying factors limiting the stream invertebrate communities, we can design effective and targeted management practices. For example, if flow and sedimentation issues in degraded streams on the Canterbury Plains can be reduced, then a positive response to riparian management may be seen, at least by taxa moderately tolerant of organic pollution. Thus, to provide riparian management that is effective at protecting or improving stream communities, it is critical to identify the most vulnerable locations, often small systems, and also the mechanisms that limit increases in sensitive taxa.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We thank Environment Canterbury, and the Waimakariri, Selwyn and Ashburton District Councils for site information and water-race data and the landowners who provided access to their land. Katie McHugh, Amanda Klemmer, Megan Fork, Hamish Greig, Francis Burdon, Hamish Carrad, Rachel Saunders, Danladi Umar, Linda Morris and Tanya Blakely provided invaluable assistance and/or advice. Jonathan O’Brien, Mike Winterbourn, the editor and two anonymous reviewers provided feedback that greatly improved the manuscript. The work was funded by the Mackenzie Charitable Foundation, with the aim of improving the effectiveness of riparian management.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Allan, J.D. (2004) Landscapes and riverscapes: the influence of land use on stream ecosystems. Annual Review of Ecology and Systematics, 35, 257284.
  • Arbuckle, J.L. (2007) Amos 16.0 User’s Guide. Amos Development Corporation, Chicago, IL.
  • Baker, M.E., Weller, D.E. & Jordan, T.E. (2007) Effects of stream map resolution on measures of riparian buffer distribution and nutrient retention potential. Landscape Ecology, 22, 973992.
  • Benfield, E.F. (2007) Decomposition of leaf material. Methods in Stream Ecology (eds F.R. Hauer & G.A. Lamberti), pp. 711720. Academic Press, Burlington, NJ, USA.
  • Chen, X., Wei, X., Scherer, R., Luider, C. & Darlington, W. (2006) A watershed scale assessment of in-stream large woody debris patterns in the southern interior of British Columbia. Forest Ecology and Management, 229, 5062.
  • Correl, D.L. (2005) Principals of planning and establishment of buffer zones. Ecological Engineering, 24, 433439.
  • Craig, L.S., Palmer, M.A., Richardson, D.C., Filoso, S., Bernhardt, E.S., Bledsoe, B.P., Doyle, M.W., Groffman, P.M., Hassett, B.A., Kaushal, S.S., Mayer, P.M., Smith, S.M. & Wilcock, P.R. (2008) Stream restoration strategies for reducing river nitrogen loads. Frontiers in Ecology and the Environment, 6, 529538.
  • Deschênes, J., Rodríguez, M.A. & Bérubé, P. (2007) Context-dependent responses of juvenile Atlantic salmon (Salmo salar) to forestry activities at multiple spatial scales within a river basin. Canadian Journal of Fisheries and Aquatic Sciences, 64, 10691079.
  • Dodds, W.K. & Oakes, R.M. (2008) Headwater influences on downstream water quality. Environmental Management, 41, 367377.
  • Elith, J., Leathwick, J.R. & Hastie, T. (2008) A working guide to boosted regression trees. Journal of Animal Ecology, 77, 802813.
  • Friedman, J.H. (2002) Stochastic gradient boosting. Computational Statistics and Data Analysis, 38, 367378.
  • Gordon, L.J., Peterson, G.D. & Bennett, E.M. (2007) Agricultural modifications of hydrological flows create ecological surprises. Trends in Ecology and Evolution, 23, 211219.
  • Harding, J.S. & Winterbourn, M.J. (1995) Effects of contrasting land use on physico-chemical conditions and benthic assemblages of streams in a Canterbury (South Island, New Zealand) river system. New Zealand Journal of Marine and Freshwater Research, 29, 479492.
  • Harding, J.S., Winterbourn, M.J. & McDiffett, W.F. (1997) Stream faunas and eco-regions in the South Island, New Zealand: do they correspond? Archiv für Hydrobiologie, 140, 289307.
  • Harding, J.S., Benfield, E.F., Bolstad, P.V., Helfman, G.S. & Jones III, E.B.D. (1998) Stream biodiversity: the ghost of land use past. Proceedings of the National Academy of Sciences (USA), 95, 1484314847.
  • Harding, J.S., Clapcott, J.E., Quinn, J.M., Hayes, J.W., Joy, M.K., Storey, R.G., Greig, H.S., Hay, J., James, T., Beech, M.A., Ozane, R., Meredith, A.S. & Boothroyd, I.K.G. (2009) Stream Habitat Assessment Protocols for Wadeable Rivers and Streams of New Zealand. School of Biological Sciences, University of Canterbury, Christchurch.
  • Kline, T.J.B. & Dunn, B. (2000) Analysis of interaction terms in structural equation models: a non-technical demonstration using the deviation score approach. Canadian Journal of Behavioral Science, 32, 127132.
  • Lee, P., Smyth, C. & Boutin, S. (2004) Quantitative review of riparian buffer width guidelines from Canada and the United States. Journal of Environmental Management, 70, 165180.
  • MacLeod, C.J. & Moller, H. (2006) Intensification and diversification of New Zealand agriculture since 1960: an evaluation of current indicators of land use change. Agriculture, Ecosystems and the Environment, 115, 201218.
  • Matson, P.A., Parton, W.J., Power, A.J. & Swift, M.J. (1997) Agricultural intensification and ecosystem properties. Science, 277, 504509.
  • Matthaei, C.D., Piggot, J.J. & Townsend, C.R. (2010) Multiple stressors in agricultural streams: interactions among sediment addition, nutrient enrichment and water abstraction. Journal of Applied Ecology, 47, 639649.
  • Mayer, P.M., Reynolds, S.K., Canfield, T.J. & McCutchen, M.D. (2005) Riparian Buffer Width, Vegetative Cover, and Nitrogen Removal Effectiveness: A Review of Current Science and Regulations. Report EPA/600/R-05/118, US Environmental Protection Agency, Washington DC, USA.
  • Meyer, J.L. & Wallace, J.B. (2001) Lost linkages and lotic ecology: rediscovering small streams. Ecology: Achievement and Challenge (eds M. C. Press, N.J. Huntley & S. Levin), pp. 295317. Blackwell Science, Malden, MA.
  • Meyer, J.L., Strayer, D.L., Wallace, J.B., Eggert, S.L., Helfman, G.S. & Leonard, N.E. (2007) The contribution of headwater streams to biodiversity in river networks. Journal of the American Water Resources Association, 43, 86103.
  • Niyogi, D.K., Koren, M., Arbuckle, C.J. & Townsend, C.R. (2007) Stream communities along a catchment land-use gradient: subsidy stress responses to pastoral development. Environmental Management, 39, 213225.
  • Palmer, M.A., Menninger, H.L. & Bernhardt, E. (2010) River restoration, habitat heterogeneity and biodiversity: a failure of theory or practice? Freshwater Biology, 55(s1), 205222.
  • Parkyn, S.M., Davies-Colley, R.J., Halliday, N.J., Costley, K.J. & Croker, G.F. (2003) Planted riparian buffer zones in New Zealand: do they live up to expectations? Restoration Ecology, 11, 436447.
  • Petersen, I., Z.Masters, Z.Hildrew, A.G. & Ormerod, S.J.  (2004) Dispersal of adult aquatic insects in catchments of differing land use. Journal of Applied Ecology, 41, 934950.
  • Peterson, B.J., Wollheim, W.M., Mulholland, P.J., Webster, J.R., Meyer, J.L., Tank, J.L., Martí, E., Bowden, W.B., Valett, H.M., Hershey, A.E., McDowell, W.H., Dodds, W.K., Hamilton, S.K., Gregory, S. & Morrall, D.D. (2001) Control of nitrogen export from watersheds by headwater streams. Science, 292, 8690.
  • Quinn, J.M., Phillips, N.R. & Parkyn, S.M. (2007) Factors influencing retention of coarse particulate organic matter in streams. Earth Surface Processes and Landforms, 32, 11861203.
  • Quinn, J.M. & Stroud, M.J. (2002) Water quality and sediment and nutrient export from New Zealand hill-land catchments of contrasting land use. New Zealand Journal of Marine and Freshwater Research, 36, 409429.
  • Quinn, J.M. & Wright-Stow, A.E. (2008) Stream size influences stream temperature impacts and recovery rates after clearfell logging. Forest Ecology and Management, 256, 21012109.
  • R Development Core Team. 2009. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org.
  • Rhodes, H.M., Closs, G.P. & Townsend, C.R. (2007) Stream ecosystem health outcomes of providing information to farmers and adoption of best management practices. Journal of Applied Ecology, 44, 11061115.
  • Richardson, J.S. & Danehy, R.J. (2007) A synthesis of the ecology of headwater streams and their riparian zones in temperate forests. Forest Science, 53, 131147.
  • Ridgeway, G. (2004) GBM: generalized boosted regression models, R package, version 1.3-5.
  • Smith, C.M., Wilcock, R.J., Vant, W.M., Smith, D.G. & Cooper, A.B. (1993) Towards sustainable agriculture: freshwater quality in New Zealand and the influence of agriculture. MAF Policy Technical Paper 93/10, Wellington.
  • Stark, J.D. (1998) SQMCI: a biotic index for freshwater macroinvertebrate coded-abundance data. New Zealand Journal of Marine and Freshwater Research, 32, 5566.
  • Taylor, R. & Smith, I. (1997) The State of New Zealand’s Environment. Ministry for the Environment, Wellington, New Zealand.
  • Ter Braak, C.J.F. & Smilauer, P. (2006) Canoco for Windows Version 4.55. Biometrics- Plant Research International, Wageningen, The Netherlands.
  • Tilman, D. (1999) Global environmental impacts of agricultural expansion: the need for sustainable and efficient practices. Proceedings of the National Academy of Sciences (USA), 96, 59956000.
  • Townsend, C.R., Arbuckle, C.J., Crowl, T.A. & Scarsbrook, M.R. (1997) The relationship between land use and physicochemistry, food resources and macroinvertebrate communities in tributaries of the Taieri River, New Zealand: a hierarchically scaled approach. Freshwater Biology, 37, 177191.
  • Wallace, J.B., Eggert, S.L., Meyer, J.L. & Webster, J.R. (1997) Multiple trophic levels of a forest stream linked to terrestrial litter inputs. Science, 227, 102104.
  • Wilcock, R.J., Betteridge, K., Shearman, D., Fowles, C.R., Scarsbrook, M.R., Thorrold, B.S. & Costall, D. (2009) Riparian protection and on-farm best management practices for restoration of a lowland stream in an intensive dairy farming catchment: a case study. New Zealand Journal of Marine and Freshwater Research, 43, 803818.
  • Winterbourn, M.J., Gregson, K.L.D. & Dolphin, C.H. (2000) Guide to the aquatic insects of New Zealand. Bulletin of the Entomological Society of New Zealand, 13, 102.
  • Zhang, Y., Richardson, J. S. & Pinto, X. (2009) Catchment-scale effects of forestry practices on benthic invertebrate communities in Pacific coastal streams. Journal of Applied Ecology, 46, 12921303.
  • Zweig, L.D. & Rabeni, C.F. (2001) Biomonitoring for deposited sediment using benthic invertebrates: a test on 4 Missouri streams. Journal of the North American Benthological Society, 20, 643657.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Fig. S1. Map showing locations of sampling sites coded by stream size and riparian management.

Fig. S2. Community composition DCA

Fig. S3. Full structural equation models for DCA 1, DCA 2 (community composition), and taxa richness

Table S1. Variables included in Boosted Regression Tree (BRT) analyses

Table S2. Total aquatic invertebrate taxa list from all sites

Appendix S1. Study area description

Appendix S2. Measurement of predictor variables

Appendix S3. Riparian condition index

Appendix S4. Boosted Regression Tree (BRT) methods

Appendix S5. Structural equation modeling (SEM) methods

Appendix S6. Habitat variables PCA results

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JPE_2092_sm_AppendixS1-S6-TableS1-S2-FigS1-S3.doc2797KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.