The emphasis on restoring ecological functions is an evolving field in stream and river restoration research [1
] and practice. Functions characterize an ecosystem’s processes, roles, services, or state of trajectory [5
To improve the success of restoration, researchers and practitioners have sought to make connections between landscape and instream variables and biological condition [6
]. Schueler [7
] compiled several studies that associate declines in diversity of aquatic insects and fish, and instream habitat quality, when watershed imperviousness exceeds 10%–15% e.g., [8
]. Riparian deforestation leads to loss of large woody debris, leaf litter, and organic dissolved inputs, elevates stream temperatures, and narrows stream channels, due to encroachment of herbaceous vegetation that would have otherwise been shaded out by the forest canopy [13
]. A study of 16 streams in eastern North America concluded that forested channels exhibited lower average channel velocities, higher bed roughness, greater abundance of macroinvertebrates, and a greater amount of organic matter processing compared to deforested channels [15
]. Researchers have predicted macroinvertebrate assemblages with watershed-scale variables, such as catchment-wide geology and land cover [16
] and reach-scale variables, such as riparian corridor land use change [17
Research on methods to implement and evaluate the functional uplift resulting from restoration actions is a relatively new path of study [2
] that involves identifying functions that support desired ecosystem services and how best to recover those lost functions [20
]. Recent studies have evaluated the tools for gauging ecological change resulting from restoration efforts [21
]. Other studies have focused on evaluating the response of macroinvertebrates and fish communities to restoration efforts [24
]. In an assessment of 79 stream reaches across North Carolina (NC), Doll et al. [24
] showed that EPT taxa (Ephemeroptera, Plecoptera, and Trichoptera) were positively correlated with accessible floodplain width (i.e., entrenchment ratio), substrate size, and mean bankfull depth, and negatively correlated with width-to-depth ratio and sinuosity. Results indicated that deliberate site selection, restoration activities, and design can optimize biotic condition and ecological function. Further, results indicated that larger streams with steeper valleys, larger substrate, and undeveloped watersheds were expected to have higher numbers of EPT taxa [24
]. At 28 predominantly agricultural watershed sites in Ohio, D’Ambrosio et al. [25
] also found that macroinvertebrate communities were driven by floodplain connection, as well as stream slope and size. In a study of 31 wadeable Piedmont Georgia streams, Walters et al. [17
] found that macroinvertebrate metrics were best predicted by urban cover, specific conductivity, fines in riffles, and local relief.
D’Ambrosio et al. [26
] found that fish assemblages in 32 Ohio streams were influenced by watershed- and reach-scale factors, including stream order, percent (%) wooded riparian zone, drainage area, instream cover quality, substrate quality, average substrate size, stream slope, stream size (i.e., cross-sectional area), and width of the flood prone area. These results indicated that site selection, geomorphic features, and instream habitat should be carefully considered for biological restoration projects. Similarly, Walters et al. [17
] found fish metrics were best predicted by embeddedness, turbidity, stream slope, and forest cover. In a study of 21 Piedmont streams in Alabama, Helms et al. [27
] found that measures of fish assemblage and crayfish size were strongly predicted by watershed size (i.e., drainage area) and geomorphic channel dimensions related to stream size, including bankfull cross-sectional area, width, mean depth, and discharge. Conversely, Walters et al. [17
] found that both fish and macroinvertebrate metrics correlated poorly with measures of stream size, such as bankfull channel width, depth, and cross-sectional area, and drainage area and discharge.
Despite knowledge advancements in stream function relationships, measuring differences between pre- and post-restoration conditions and targeting restoration actions that improve biology is complicated by the interrelatedness of function variables [24
], such that effects on one function can initiate a cascade of effects on other functions [28
]. Understanding the relationships between these physical, chemical, and biological variables is challenged by the drastic disturbance often associated with restoration, such as changes in channel geometry, perturbation of soils, and inclusion of constructed structures [29
The United States (US) requires compensatory mitigation projects to restore natural or historic aquatic resource functions [30
]. The US defines functions as physical, chemical, and biological processes of ecosystems, which are also considered services when these processes benefit humans [30
]. To help practitioners implement functional restoration projects, in 2006, Fischenich [28
] proposed a functional framework that identified and organized a hierarchy of basic stream and riparian corridor functions described in the literature. Building on Fischenich’s framework [28
], in 2012, the Stream Functions Pyramid Framework (Pyramid Framework) was developed [31
]. The Pyramid Framework asserts that functions are interrelated such that they build on each other in a specific order, whereby watershed hydrology is the fundamental support for hydraulic, geomorphic, physicochemical, and biological functions, and biology relies on all underlying functions (Figure 1
). The Pyramid Framework was designed to help regulators and practitioners visualize how physical watershed and fluvial hydromorphological factors support physicochemical and biological functions and identify which factors must be addressed to improve these higher-level functions. The Pyramid Framework also identifies specific parameters and metrics (indicators of function or direct measures of function) that can be measured for pre- and post-restoration condition monitoring.
The Pyramid Framework is the basis of the Stream Quantification Tool (SQT) [32
], a stream mitigation crediting tool in use in Minnesota, Tennessee, Wyoming, and Georgia [33
] and in development in Colorado, Michigan, and Alaska to date [37
]. The first SQT was developed for NC [32
] but is not formally adopted as a mitigation crediting tool at this time. The SQT aims to support mitigation site selection, determine a site’s restoration potential, and quantify functional change associated with restoration activities to aid in mitigation debit and credit determination [32
]. The SQT consists of a conditions assessment and an excel workbook tool that uses embedded reference (performance) curves to assign function index scores to measured field and desktop values. Function index scores are multiplied by linear feet of pre-restoration, proposed, and/or restored stream to calculate “functional feet,” which are used to determine the quantity and quality of impacted, proposed, and/or restored streams and inform credit and debit awards. However, the relationships proposed in the Pyramid Framework and that underpin the SQT have not yet been statistically evaluated. Further, the ability of the SQT conditions assessment protocol (SQT protocol) to predict biological function is untested.
To address gaps in our understanding of the universal applicability of the Pyramid Framework as a schema for explaining biological function restoration, we apply the NC SQT protocol to test its predictive ability. We hypothesize that the Pyramid Framework is generalizable, in that hydrologic variables explain the most variance in biological function variables, while other higher-level variables (e.g., hydraulics, geomorphology, and physicochemistry) explain relatively less variance. The specific objectives of this study are to: (1) document stream functional data in the Piedmont ecoregion of NC (USA); (2) evaluate the SQT protocol’s ability to predict biological function; and (3) determine variables most important to predicting biological function. Results from objectives (2) and (3) will evaluate the hierarchical premise of the Pyramid Framework, which in theory underlies the SQT protocol. Identifying improvements to the SQT protocol for NC and other states is also an anticipated outcome of this effort.
To gain insight into the applicability of the conceptual Pyramid Framework as the premise of the SQT, the SQT protocol was implemented in Piedmont, NC (Southeast, USA). Statistical analyses revealed stepwise and ridge as the best predictive models for the biotic integrity data collected in this study. The best performing stepwise and ridge models identified key variables that influence stream macroinvertebrates communities: floodplain width (ER), channel shape (W/D), channel dimension (dbkf
), bank erosion stress and susceptibility (NBS and BEHI), active streambank erosion, pool depth ratio, buffer width, adequate extent of riffle and run habitat (% riffle), substrate size (D84
), and stream summer temperature. Results suggest that restoration activities such as increasing floodplain and buffer widths, planting riparian vegetation to shade streams, and reducing streambank erosion and erosion potential will improve macroinvertebrate community composition. Results also suggest that creating appropriate local conditions such as an ample riffle habitat with a suitable substrate type will help recruit and support EPT taxa, although the presence of contaminants [60
] and the existence and diversity of the regional species pool [13
] are also critical factors for improving community composition. Nevertheless, practitioners should take care to design streams that will support not only adequate riffles, but also other habitat units such as pools, stable undercut banks, leaf packs, overhanging vegetation, and point bars, to support a diversity of aquatic species. Our findings help elucidate the complex relationships between watershed- and reach-scale variables and biological function and can be used to help practitioners establish feasible restoration goals with appropriate success metrics to improve stream restoration success.
Statistical analyses revealed that the SQT protocol reasonably predicts biological function, and cross-validation and comparisons of model performance indicated that stepwise and ridge models should reasonably predict EPT richness and the NCBI in NC and the greater southeast Piedmont (USA). Results provided moderate support for the hierarchical Pyramid Framework: highly predictive ridge models included metrics from all Pyramid levels, while highly predictive stepwise models included metrics from higher Pyramid levels, excluding watershed hydrology variables. Reach-scale metrics were more important than watershed hydrology metrics to predict macroinvertebrates, suggesting that successful biological functional restoration is possible despite watershed condition. However, the removal of several additional watershed variables from the SQT+ models, due to multicollinearity, underscores that watershed variables directly and indirectly affect higher-level variables, reinforcing relationships depicted in the Pyramid Framework. Further, that so few variables were needed to predict biological function suggests that importance should be placed on measuring these metrics in the NC and southeast Piedmont. Considering limited time and money, practitioners can focus efforts on measuring the most important metrics to understand the integrity of the biotic community and to inform restoration goals and design decisions.
Although prediction of macroinvertebrate biota with the SQT protocol was reasonable, the fact that 36% to 47% of variability amongst EPT richness and the NCBI remains unexplained by the best SQT models and that the SQT+ models improved predictions for most models suggests the SQT protocol is lacking critical measures of function. Future endeavors to repeat this approach with (1) a larger sample size with equal representation of restored, degraded, and geomorphic and biological reference sites; (2) robust monitoring data, especially for highly variable physiochemical parameters; (3) additional physicochemical variables, such as biochemical oxygen demand and pH; and (4) fish and other fauna as indicators of biological function, will augment findings.
The SQT excel workbook tool applies arbitrary, implicit weighting by averaging all metrics within each function category (e.g., hydrology). As the number of metrics within a function category (e.g., geomorphology) increases, the significance that each metric carries decreases. Improved regression models could result in multipliers for individual metrics that more accurately reflect their importance to the biological function, thus improving the accuracy of the overall function quantification. However, the SQT allows users some choice in metric and assessment method selection. This flexibility challenges the creation of multipliers because the relationship of each metric to biological function will change with respect to other metrics measured. Thus, as the number of metrics change, the ability of the SQT protocol to reasonably predict biological function will also change.
Regional testing of the SQT in other states resulted in revisions, additions, and deletions of some metrics and assessment methods [33
]. Thus, this paper lays out an approach to evaluate the Pyramid Framework and test regionalized versions of the SQT protocol. Further, our results can help inform updates to the NC SQT protocol [32
], which is currently under revision.
Our findings indicate that the SQT displays promise and merit. With further refinement, the SQT can be a beneficial tool for practitioners and regulators. The future of stream restoration remains hopeful, as conceptual frameworks and tools for practitioners and regulators, like the SQT, are created and evaluated to ultimately improve the practice of stream restoration.