The use of multiscale stressors with biological condition assessments: a framework to advance the assessment and management of streams

Abstract

Incorporating information on landscape condition (or integrity) across multiple spatial scales and over large spatial extents in biological assessments may allow for a more integrated measure of stream biological condition and better management of streams. However, these systems are often assessed and managed at an individual scale (e.g., a single watershed) without a larger regional multiscale context. In this paper, our goals were: (1) To develop a conceptual framework that could combine stream biological condition to abiotic landscape integrity (or, conversely, stressor) data at three spatial scales: watershed, catchment and stream-reach scale, to enable more targeted management actions. Measures of landscape integrity and stressors are negatively related, i.e., integrity on a 0–1 scale is equal or equivalent to stressors on a 1–0 scale. (2) To develop the framework in such a way that allows operational flexibility, whereby different indicators can be used to represent biological condition, and landscape integrity (or stressors) at various scales. (3) To provide different examples of the framework’s use to demonstrate the flexibility of its application and relevance to management. Examples include stream biological assessments from different regions and states across the U.S. for fish, macroinvertebrates and diatoms using a variety of assessment tools (e.g., the Biological Condition Gradient (BCG), and an Index of Biotic Integrity (IBI)). Landscape integrity indicators comprise U.S. EPA’s nationally available Index of Watershed Integrity (IWI) and Index of Catchment Integrity (ICI), and state and regional derived watershed and stream-reach scale integrity indicators. Scatterplots and a landscape integrity map were used to relate samples of stream condition classes (e.g., good, fair, poor) to watershed, catchment and stream-reach scale integrity. This framework and approach could provide a powerful tool for prioritizing, targeting, and communicating management actions to protect and restore stream habitats, and for informing the spatial extent at which management is applied.

1. Introduction

Traditionally, methods to assess and manage the condition, or integrity¹, of streams and rivers have focused on the cumulative responses of aquatic biological communities to any and all stressors impacting them at the reach or site-scale (Flotemersch et al., 2006; USEPA, 1996). However, in the last two decades, initiatives to incorporate landscape information to address watershed- or catchment-scale influences on stream condition have increased (Association of Fish and Wildlife Agencies, 2012; Nature Conservancy of West Virginia, 2013; USEPA, 2012; USFS, 2011), particularly with growing access to geospatial data (Tsang et al., 2014). We define catchment-scale as only the area that contributes flow directly to particular stream segments (excluding upstream contributions), and a watershed-scale, which incorporates all upstream contributions. While the terms “catchment” and “watershed” have been used synonymously, their use here is consistent with how they are defined in the National Hydrography Dataset Plus Version 2 (NHDPlusV2; McKay et al., 2012), which we used as a geospatial framework. Incorporating information on landscape condition adjacent to and upstream of aquatic resources in aquatic biological assessments may allow for a more integrated appraisal of condition. Communities of fish, macroinvertebrates, and diatoms, commonly used to measure the biological condition of streams (Griffith et al., 2005; Karr, 1981; Stevenson et al., 2013; USEPA, 2016a), are responsive to a variety of stressors at both the local scale (e.g., point-source pollutants from sewage discharge) and watershed-scale (e.g., agricultural and urban runoff). For example, upstream (watershed-scale) degradation in water quality and habitat in response to increasing agriculture and urban development have been known to produce local scale stress on diatom communities (which are good indicators of water quality changes), and on macroinvertebrate and fish communities (which are better related to habitat degradation) (Kovalenko et al., 2014; Woo et al., 2019). Thus, combining stream biological condition data with information on local and landscape condition at varying scales could provide valuable information for interpreting the response of biological communities to stressors and developing more effective management strategies (Wang et al., 2011).

In a recent effort to help combine the assessment and management of aquatic resources at the larger watershed and catchment-scale, the U.S. Environmental Protection Agency (EPA) developed the Index of Watershed Integrity (IWI) and Index of Catchment Integrity (ICI) to quantify and map integrity for 2.6 million stream segments across the conterminous U.S. (CONUS) (Flotemersch et al., 2016; Thornbrugh et al., 2018). The IWI and ICI were built on the conceptual model by Flotemersch et al. (2016), who defined watershed integrity as “the capacity of a watershed to support and maintain the full range of ecological processes and functions essential to the sustainability of biodiversity and of the watershed resources and services provided to society.” This definition also applies to defining catchment integrity. Landscape integrity was mapped at two spatial scales: the catchment- and watershed-scale. Stream segments were defined using the NHDPlusV2 (McKay et al., 2012). This is a medium resolution geospatial framework at a nominal 1:100K scale which contains geospatial information on streams and their related catchments. Based on the framework of the NHDPlusV2, the IWI and ICI integrate CONUS available data from EPA’s StreamCat dataset (Hill et al., 2016). These indices incorporate 23 anthropogenic stressors taken from StreamCat (e.g., % of watershed composed of urban or agricultural land uses, % impervious land cover) that are known to impact six key watershed functions: hydrologic regulation, regulation of water chemistry, sediment regulation, hydrologic connectivity, temperature regulation, and habitat provision (see Flotemersch et al., 2016). The IWI and ICI were developed using simple negative linear approximations which assume that all stressors are negatively related to watershed function (Thornbrugh et al., 2018); thus, as stressors increase within a watershed or catchment, its integrity declines (Flotemersch et al., 2016). Both the IWI and ICI have a 0 to 1 scaling system, where 1 represents high integrity and no stressors. Note, the IWI and ICI can either be represented as integrity on a 0–1 scale, or as stressors on a 1–0 scale, i.e., 1 minus the integrity value.

Hill et al. (2017) demonstrated how watershed-scale data such as the IWI could be combined with instream estimates of biological condition (using National Rivers and Streams Assessment (NRSA) data; USEPA, 2016a) to prioritize protection and restoration based on specific management criteria. The use of multiscale landscape integrity data at the watershed- and catchment-scale, and at a smaller, “stream-reach scale” (defined as measurements taken from the water column), combined with instream biological condition data, could enhance our interpretation of biological responses to stressors and allow for distinct management actions. For example, identifying and elucidating how different taxonomic assemblages (e.g., fish, macroinvertebrates, and diatoms) respond to different stressors at different scales (watershed-, catchment- or stream-reach scale), could create a more informative, integrated assessment of biological condition and recommendations for the most effective management actions (Johnson and Hering, 2009; Kovalenko et al., 2014). Used in a diagnostic mode, such an analysis could help determine the scale(s) at which biological condition is responding to stress, which could inform the scale(s) of subsequent management actions. However, a common framework for assessing assemblage-specific responses to anthropogenic stress at multiple spatial scales is lacking.

Since the U.S. Clean Water Act (CWA) was enacted, different methods and indicators have been developed to quantitatively define and measure aspects of the CWA integrity goal, which is primarily to restore and protect the biological integrity of the nation’s waters. This paper is one of the first to conceptually and practically link these different indicators for practical application. In this paper, our goal was: (1) To develop a conceptual framework that combines information on stream biological condition with landscape integrity at three different spatial scales, i.e., stream-reach scale integrity, catchment-scale (local) integrity, and watershed-scale (upstream cumulative) integrity, in order to allow managers to focus on the spatial scale(s) that may have the greatest influence on condition and its management. (2) To develop the framework in such a way that it enables operational flexibility, in that different datasets can be substituted. For example, a watershed integrity indicator could be represented by the nationally available IWI, or by a more locally (e.g., state or regional) derived watershed integrity index. Further, indicators of stream biological condition could be represented by national, regional, or statewide indicators, including the Biological Condition Gradient (BCG). (3) To provide different examples of the framework’s use to demonstrate the flexibility of its application and relevance to stream protection and restoration management.

We accomplish these objectives by first describing this multiscale framework and then illustrating how the framework can be used with biological condition information. We provide four case study examples of comparisons of stream biological condition assessments from different regions and states across the U.S. that represent a variety of spatial and temporal scales and employ various biological response indicators. We also demonstrate the utility of the IWI and ICI, which can both be used as generalized landscape integrity gradients since they are nationally available. In the first three case studies we apply landscape integrity indicators at two spatial scales, the catchment- and watershed-scale, and in the fourth case study, we introduce a landscape integrity indicator at a third spatial scale, i.e., the stream-reach scale, along with catchment- and watershed-scale integrity indicators.

2. Conceptual Framework

Biological assessment methods and metrics used to quantify the biological condition of aquatic resources can vary between states, territories, tribes (herein defined as states) and regions, making comparisons of condition assessments difficult and challenging to communicate to the public. To address this issue and achieve goals set by CWA, the BCG was developed to allow for more consistent interpretation of biological condition across states, even when methods to measure condition were different (Davies and Jackson, 2006). The BCG is a conceptual model that has been used to classify the biological condition of sampled rivers and streams based on the response of biological attributes for fish, macroinvertebrates or algae, to increasing levels of anthropogenic stress (from undisturbed to completely degraded) and is now being applied to large rivers, estuaries and coral reefs (USEPA, 2017; Yoder et al., 2015). The attributes represent 10 key features of ecological condition that are associated with different spatial scales, including taxonomic composition and community structure at the site-scale, and physical‐biotic interactions at the larger catchment- and watershed-scales. The BCG model divides the condition gradient into 6 levels of biological condition, from (1) undisturbed to (6) highly disturbed (Fig. 1; Davies and Jackson, 2006; USEPA, 2016b). Typically, states have used the BCG model in combination with metrics, indices or models to more clearly communicate condition along a gradient of natural, pristine biological conditions to severely altered. Since its conception, BCG models have been developed mostly for freshwater perennial streams in over 15 states and regions across the US (USEPA, 2016b). Thus, the methodology for defining a BCG is well-established. However, a standard approach that relates a well-defined stressor gradient to assess changes in condition would enhance its application and advance a more integrated approach to use of biological, stressor and land use data and information to manage and protect our aquatic resources. Here, we demonstrate how two already established landscape integrity (or stressor gradients), the IWI and ICI, can serve this function. Fig. 1 shows the conceptual BCG model which describes incremental biological changes along a gradient of anthropogenic stressors, in which the IWI and ICI represent a generalized landscape integrity gradient. Since the IWI and ICI are generated from CONUS-wide stream data, they can be applied with BCG models and/or other types of site-scale biological condition models (e.g., NRSA’s macroinvertebrate multimetric index (MMI; Hill et al., 2017) for streams in any state or region in the CONUS.

Fig. 1.

Conceptual model of the BCG, showing stages of change in biological conditions from Level 1 to 6 in response to increasing landscape stressors or decreasing integrity. The landscape integrity gradients can be represented by the Index of Watershed Integrity (IWI) and Index of Catchment Integrity (ICI) that incorporate anthropogenic stressors known to degrade six key watershed functions. The figure is modified from USEPA (2016b) where BCG is shown in reverse slope from degraded to improved biological conditions.

Examining stressor-response relationships using a multiscale landscape integrity gradient approach (e.g., catchment and watershed-scale) could improve our interpretation of site-scale biological condition (e.g., as represented by a BCG, one of the NRSA biological response indicators, or an indicator defined by a state), and provide a framework that could help link management goals to strategic management actions. In a 2D scatterplot, we can relate the IWI and ICI with estimates of biological condition, represented by synthetic data, to illustrate how site-scale biological condition data could be combined with multiscale landscape integrity data to inform management decisions on appropriate actions to protect or restore stream habitat (Fig. 2).

Fig. 2.

A 2D scatterplot showing the hypothetical relationship between a generic site-scale biological condition indicator and generic watershed- and catchment-scale integrity indicators. The points represent site condition estimates from poor condition (red) to good condition (pale blue). Various indicators could be used to represent biological condition (e.g., the Biological Condition Gradient (BCG), a NRSA indicator, an Index of Biotic Integrity (IBI or an MMI) and landscape integrity gradients (e.g., the nationally available Index of Watershed Integrity (IWI) and Index of Catchment Integrity (ICI), or state or regional watershed- and catchment-scale indicators). Watershed- and catchment-scale indicators can either be plotted on a scale of integrity (as shown here), or on a scale of stressors which would be plotted as 1 minus the index value (e.g., 1 – IWI value) so that high stressors are at the origin, and low stressors are associated with the upper right quadrant.

The four quadrants in Fig. 2 were obtained by drawing two perpendicular lines positioned at the 0.5 value of each integrity index. Identifying the quadrant that a good or poor condition site is located in could provide valuable information for distinct management actions, dependent on the specific management goals and resources available. For example, if the goal is to identify areas of high conservation value to protect, then sites in good condition that have high integrity or low stressors (pale blue dots in upper right quadrant) would be good candidates for protection, since natural infrastructure (i.e., minimally impacted land) is generally in good condition and requires minimal restoration at those scales, and protection would help maintain biological functions. If the goal is to identify high priority sites for conducting restoration activities, then, in general, targeting poor condition sites that have high watershed and catchment integrity may increase restoration feasibility. Identifying poor condition sites that have high watershed but lower catchment integrity (red dots in upper left quadrant) would be the next best area to target, since restoring and sustaining (or maintaining) catchment integrity may be more cost-effective than restoring integrity over the larger upstream watershed (Hill et al., 2017). A healthy watershed can support restoration efforts responding to stressors at the catchment level, while restoration within a watershed with low integrity (dots in lower two quadrants) could require considerable effort and resources responding to upstream (watershed-scale) stressors (Aho et al., 2020). It is important to keep in mind, however, that in certain scenarios a watershed could be less expensive to restore than a catchment. For example, a stream that has a small watershed vs. a relatively large catchment, or a watershed that has a discrete, point source impact that would be easy to remedy vs. a catchment that has a major impact that would require expensive restoration.

As a further example of the utility of this framework, we provide examples of how multiscale landscape integrity data, when visualized in a map, could provide managers with a powerful tool for identifying sites that meet certain protection and restoration criteria (Fig. 3). Such a map could allow managers to easily identify priority areas for protection, i.e., those with high IWI and high ICI values (represented by blue catchments), and areas for restoration, i.e., high IWI and low ICI values (green catchments, Fig. 3). Including site condition on such a map would further identify areas that meet specific management criteria.

Fig. 3.

A landscape integrity map for the Western Balkans showing high and low watershed- and catchment-scale integrity based on values derived from the Index of Watershed Integrity (IWI) and Index of Catchment Integrity (ICI) (from Aho et al., 2020).

The multiscale landscape integrity framework could also include a third, more localized scale of integrity, i.e., a stream-reach scale integrity (Fig. 4), which could be represented by a recognized major regional or statewide stressor, e.g., relative bed stability or % sand/fines impacting macroinvertebrates in streams across Washington State (Larson et al., 2019). The 3D scatterplot in Fig. 4 relates the biological condition of streams to three landscape integrity gradients: watershed, catchment and stream-reach scale. Including three different scales of integrity may allow for a more precise interpretation of biological responses to stressors and possibly allow for distinct management actions based on the scales at which those stresses occur. For example, the response could show that stream biological condition is most responsive to catchment integrity, indicating that the catchment could be the best scale for management attention.

Fig. 4.

A 3D scatterplot showing the hypothetical relationship between a generic site-scale biological response indicator and three stressor gradients: watershed-scale, catchment-scale, and stream-reach scale. The color of the plotted data represents different classes of stream biological condition. Biological condition and the three scales of stressors can be represented by multiple national, regional or state indicators, as shown in the scatterplot.

The multiscale stressor framework enables operational flexibility (see Figs. 2 and 4). Watershed and catchment integrity could be represented by the nationally available IWI and ICI, or a more locally available dataset, while stream biological condition could be represented by the Biological Condition Gradient (BCG), a NRSA index, or an Index of Biotic Integrity (IBI) or MMI (as will be demonstrated in this paper). Similar to the map in Fig. 3, a landscape integrity map displaying catchments according to the level (high or low) of watershed, catchment, and stream-reach scale integrity could be highly valuable for prioritizing areas for protection and restoration.

3. Materials and methods

3.1. Study areas

The four case study areas (Fig. 5) include the following: (1) The Puget Sound Region (PSR), which is a coastal area of the Pacific Northwest in the state of Washington (WA). The study area encompasses the following EPA Level III ecoregions (Omernik and Griffith, 2014): the Puget Lowland, the Cascades, North Cascades, and the Coast Range. King County (KC), located within the PSR, is the most populated county in WA, and contains 39 towns and cities including Seattle. We used data from PSR and KC to demonstrate the utility of combining stream biological condition data with two different scales of landscape integrity (catchment- and watershed-scale integrity) in order to support strategic stream management actions. (2) The high gradient region of New Jersey (NJ) is a case study area that we used to illustrate how relationships between biological indicators and multiscale landscape integrity indicators can be compared among different biological groups (diatoms and macroinvertebrates); and demonstrate the flexibility of this approach by applying different types of biological indicators. (3) Data from the state of Connecticut (CT) were applied to further show the flexibility of the approach – in this case, demonstrating the transferability of the multiscale landscape integrity indicators used to examine stream biological condition data. (4) The Central Appalachia (CAPP) EPA Level III ecoregion 69 (Woods et al., 1996), which includes parts of West Virginia (WV), and Virginia (VA). This case study illustrates the utility of relating stream biological condition to an indicator of landscape integrity at a third spatial scale, i.e., the stream-reach scale, along with catchment- and watershed-scale integrity indicators.

Fig 5.

Clockwise from top left - map of the conterminous United States showing case study dataset locations; map of the case study areas - Connecticut, New Jersey - High Gradient Region, Central Appalachia, and Puget Sound Region and King County (inside black border), showing land use and land cover features using 2011 National Land Cover Data.

3.2. Case studies

In the sections below, we use case studies in four different regions of the U.S. as examples of specific ways in which our framework can be applied (Table 1).

Table 1.

Summary of case study datasets.

Case study dataset/State	n	Sampling dates	Biotic endpoints	Biological Condition Indicator	Mean Catchment/Watershed Length (km)	Mean Catchment/Watershed Area (km2)	Landscape Integrity Indicator
							Watershed-scale	Catchment-scale	Stream-reach scale
Puget Sound Region (WA)	199	Smr/Fall 2009–2017	Macro-invertebrates	Benthic Index of Biological Integrity (B-IBI)	3.6/59.1	5.9/92.2	Index of Watershed Integrity (IWI)	Index of Catchment Integrity (ICI)	-
King County (WA)	583	Smr/Fall 2012–2017	Macro-invertebrates	B-IBI	3.4/46.7	7.7/48.1	IWI	ICI	-
New Jersey (NJ)	33	Smr/Fall 2007–2009	Macro-invertebrates	High Gradient Macro-invertebrate Index (HGMI)	3/32.3	6.7/49.6	IWI	ICI	-
New Jersey (NJ)	33	Fall 2009	Diatoms	Biological Condition Gradient (BCG)	3/32.3	6.7/49.6	IWI	ICI	-
Connecticut (CT)	634	Smr/Fall 2013–2016	Fish	BCG	2.8/52	4.3/87.7	Watershed Integrity*	ICI	-
Central Appalachia (V, WV)	1011	Sprg/Smr 2000–2017	Macro-invertebrates	BCG	3.1/52.1	5.1/54.2	IWI	ICI	Stream-reach Scale Integrity**

^*Watershed Integrity indicator is based on statewide Hydrological Stressor Index classes assigned to CT BCG sample sites.

^**Stream-reach Scale Integrity indicator is based on in-situ measures of conductivity (μS/cm) at CAPP BCG sample sites. Length of stream reaches range between lateral, 0.4–40 m; and longitudinal, 20–100 m; with at least 2 riffles within the reach.

3.2.1. Relating a biological condition indicator with IWI and ICI: Puget Sound Region and King County biological datasets

This analysis combines the nationally available indicators of watershed and catchment integrity with a regional Benthic Index of Biotic Integrity (B-IBI). The PSR is a coastal area of the Pacific Northwest in the state of Washington (WA). The Puget Lowlands B-IBI, a macroinvertebrate MMI (Karr, 1998; Morley and Karr, 2002), is routinely used to assess the biological condition of streams across the PSR, including KC. We combined two sets of B-IBI scores into one dataset (Table 1): the first set (n = 199) was collected from streams throughout the PSR by the Environmental Assessment Program at the Washington State Department of Ecology (https://ecology.wa.gov/Research-Data/Monitoring-assessment/River-stream-monitoring/Habitat-monitoring/Stream-biological-monitoring), and the second set (n = 583) was collected from KC streams by the KC Water and Land Resources Division (WLRD) as part of their Freshwater Benthic Macroinvertebrate Monitoring Program. Macroinvertebrate assemblage data used to produce the B-IBI scores were collected once per year between July and October from 2009 to 2017 following standard protocols (Larson et al., 2019). Data and details of both assessments are available at: https://www.pugetsoundstreambenthos.org. The combined dataset (n = 782) was based on 500-count samples from 280 unique wadeable and non-wadeable stream sites. We categorized samples into good, fair, and poor biological condition based on condition class thresholds for B-IBI scores in the Western WA region where the sample sites were located (one of three regions in Washington defined by precipitation gradients; Larson et al., 2019). B-IBI scores ≤ 49.98 were classified poor, ≥ 73.73 scores classified as good, and between 49.98 and 73.73 were fair. In a 2D scatterplot (Fig. 6a), we showed the relationship between samples of B-IBI condition and corresponding IWI and ICI values for each sample site. Finally, we classified IWI and ICI values into High (≥ 0.5) or Low (< 0.5) and, on a map, displayed sample sites of B-IBI condition, and catchments for the PSR and KC associated with High and Low IWI and ICI.

Fig. 6.

(a) A 2D scatterplot using macroinvertebrate B-IBI assessment data for Puget Sound Region and King County streams and rivers displaying the relationship between samples of B-IBI condition (n = 782) and corresponding IWI and ICI values for each sample site from (0) low integrity to (1) high integrity. The different color of points denotes the class of site condition from poor (red) to good condition (blue). The pie charts display the distribution of each B-IBI class in each quadrant. Dashed line represents the 1:1 relationship between IWI and ICI. (b) A landscape integrity map displaying the catchments of the study area according to the level of watershed and catchment integrity (high or low IWI and ICI), along with locations and most recent conditions of B-IBI sample sites.

3.2.2. Comparing two types of biological condition indicators with IWI and ICI: New Jersey biological datasets

Here we compare results from two separate biological indicators when combined with nationally available indicators of watershed and catchment integrity. We compared two sets of biological data (n = 33 per set) collected from the same thirty-four wadeable stream and river sites in the high gradient region of NJ: (1) diatom BCG levels and (2) macroinvertebrate High Gradient Macroinvertebrate Index (HGMI) condition classes (Table 1). The diatom BCG levels were derived from diatom assemblages sampled from rocks in September and October 2009 (Hausmann et al., 2016). NJ Department of Environmental Protection (NJDEP) Bureau of Freshwater Biological Monitoring, collected diatom samples using the top-rock scrape protocol (ANSP, 2005), and 600 valves were counted per sample in accordance with USGS’s National Water-Quality Assessment (NAWQA) guidelines (Charles et al., 2002). The HGMI is an MMI developed specifically for high gradient streams of New Jersey (Jessup, 2007) as part of NJDEP’s Ambient Macroinvertebrate Network (AMNET) stream assessment program (https://www.nj.gov/dep/wms/bfbm/amnet.html). We obtained publicly available HGMI condition class data (excellent, good, fair, poor) for macroinvertebrate samples collected as close as possible to the time of diatom sampling, which was sampling data collected from July-November, 2007–2009 (https://www.nj.gov/dep/wms/bfbm/downloads.html#top). Macroinvertebrate sampling and analysis was performed in accordance with NJDEP Field Procedures Manual (NJDEP, 2005), USEPA Rapid Bioassessment Protocol (RBP) guidelines (Barbour et al., 1999) and Standard Operating Procedures of the NJDEP Aquatic Biomonitoring Laboratory (NJDEP, 2007). We produced two 2D scatterplots for comparison: the first plot (Fig. 7a) showing the relationship between samples of diatom BCG condition and IWI and ICI values associated with each sample site, and the second plot (Fig. 7b) relating samples of HGMI condition to the same IWI and ICI values. For this NJ project, BCG level 4 was understood as representing biological conditions that still supported aspects of a naturally expected biological community and its ecological function while BCG levels 5 and 6 represented severely altered biological communities and loss of function. We used this as a guideline to make general comparisons between diatom BCG and HGMI site classifications. Thus, BCG Level 2 and 3 was compared to HGMI excellent and good, Level 4 to HGMI fair, and Level 5 to HGMI poor.

Fig. 7.

2D scatterplots for high gradient New Jersey streams and rivers relating a) samples of diatom BCG condition (n = 33) and b) samples of macroinvertebrate HGMI condition (n = 33), to corresponding IWI and ICI values for each sample site, where 1 represents high integrity. Each site has an associated number. BCG and HGMI samples were from the same sites and thus, the IWI and ICI values were the same for both plots. The points correspond to site condition estimates from disturbed BCG Level 5 (orange) and HGMI poor condition (red) to minimally disturbed BCG Level 2 and HGMI excellent condition (both light blue). The pie charts display the distribution of each biological condition class in each quadrant. There were no BCG Level 1 and Level 6 sites. Dashed line represents the 1:1 relationship between IWI and ICI.

3.2.3. Relating a biological condition indicator with state-derived watershed integrity indicator and ICI: Connecticut biological and watershed datasets

In this analysis, we substitute a state watershed integrity indicator for our national IWI indicator and use this and the ICI to examine a fish BCG score. The CT dataset (n = 634) consists of BCG condition levels derived from fish assemblages in high gradient wadeable streams of CT (Gerritson and Jessup, 2007; Stamp and Gerritson, 2013). Samples were collected from 548 unique sites between June-September, 2013–2016 by Connecticut Department of Energy and Environmental Protection’s (CT DEEP) Bureau of Water Protection and Land Reuse (WPLR) following CT DEEP standard operating procedures (CT DEEP, 2013). To comply with CT Stream Flow Standards and Regulations, CT DEEP developed a watershed indicator, the Hydrologic Stressor Index (HSI), to quantify and map stream flow integrity for 36,000 stream segments across the state (www.ct.gov/deep/streamflow). The HSI includes watershed flow integrity metrics where HSI values are divided into three stream condition classes: Class 1 (natural conditions), Class 2 (near natural conditions), and Class 3 (altered conditions) (https://www.ct.gov/deep/lib/deep/water/watershed_management/flowstandards/streamflowclass_process_050216update.pdf). We obtained HSI condition classes for each BCG sample site (see Connecticut Final Streamflow Classifications: https://www.ct.gov/deep/cwp/view.asp?a=2698&q=322898&deepNav_GID=1707). We equally divided the three HSI condition classes along a Watershed Integrity scale of 0 (altered conditions) to 1 (natural conditions). In a 2D scatterplot, we related samples of fish BCG condition and corresponding ICI and Watershed Integrity values. Overplotting can be an issue with 2D and 3D scatterplots, particularly when using large datasets and/or non-continuous variables such as the Watershed Integrity indicator. Many points may be plotted onto the same location which makes it difficult to determine the true distribution of data. To address this, we produced a bar chart displaying the frequency of occurrence of samples for each BCG condition level at each Watershed Integrity value to provide additional information about the sample distribution.

3.2.4. Relating biological condition indicator with IWI, ICI and stream-reach scale integrity indicator: Central Appalachia ecoregion biological and chemical datasets

This analysis includes a local scale integrity stressor with the national IWI and ICI scales to examine macroinvertebrate BCG values. The CAPP dataset (n = 1011) is comprised BCG condition levels based on 200 count macroinvertebrate assemblage samples and measures of conductivity from wadeable streams and rivers across the CAPP EPA Level III ecoregion 69 (Woods et al., 1996). Macroinvertebrate assemblages were sampled from 749 unique sites from March–June, 2000–2017 by Virginia Department of Environmental Quality (VDEQ; n = 32) following EPA RBP guidelines (Barbour et al., 1999), and West Virginia Department of Environmental Protection (WVDEP; n = 979) in accordance with WVDEP standard protocols (WVDEP, 2018).

Elevated conductivity in streams (up to ~ 2000μS/cm) – as a result of coal mine drainage (mostly alkaline) throughout this region – is a major stressor that has critically impacted stream biota and ecological attributes (USEPA, 2010). Given that negative relationships between macroinvertebrates and conductivity have been established (e.g., Cormier and Suter II, 2013; Zalack et al., 2010) and a complete set of conductivity measurements collected while sampling BCG sites was readily available, we used conductivity to demonstrate the use of a third integrity scale (stream-reach) that macroinvertebrate BCG classes might be related to: stream-reach scale integrity represented by measures of conductivity, and catchment- and watershed-scale integrity represented by the ICI and IWI. We used existing conductivity thresholds reported for streams within the CAPP region (Pond et al., 2011; USEPA, 2010) in order to develop the Stream-reach Integrity indicator. Based on the literature, conductivity <180μS/cm represents least impaired (reference) sites, and >500μS/cm could indicate that the ionic concentration is exceeding background concentrations for the area, and, as such, is a flag for further investigation (Pond et al., 2011). Conductivity values were then transferred to a Stream-reach Integrity scale from 0–1, in which BCG samples with associated conductivity values <180μS/cm = 0 and >500μS/cm = 1. Stream-reach Integrity values were calculated for conductivity between 180–500μS/cm using a linear interpolation (because of the linear response between BCG samples and conductivity), and the remaining BCG samples and their related conductivity measures were assigned an Integrity value between 0 and 1. Finally, we plotted the relationship between samples of macroinvertebrate BCG condition and corresponding Stream-reach Integrity, ICI, and IWI values as a 3D scatterplot.

3.3. Obtaining IWI and ICI values

For each case study dataset, we checked that each site was located on the NHDPlusV2 1:100K stream network (http://www.horizon-systems.com/NHDPlus/NHDPlusV2_home.php). To do this, we calculated the distance between the site and the nearest NHDPlusV2 streamline using R statistical software (R Development Core Team, 2017). Sites that had distances >50m to the streamline were examined in ArcGIS (v10.6.1. ESRI: Redlands, CA, USA) to confirm that the site was located on the 1:100K network. Sites that were not on this network were dropped from the analysis. Each sample was assigned a COMID value, a unique NHDPlusV2 identifier associated with the streamline where sites were sampled. We checked that each sample was assigned the correct COMID since a sample could be given an incorrect COMID if the sample site was associated with the wrong streamline; e.g., if the site was near a confluence. To verify this, we created a 50m buffer for each site in R and, using ArcGIS, checked those sites that had more than one COMID within the buffer to ensure they were assigned the correct COMID. A few sites had incorrect COMIDs and were corrected accordingly. Finally, we matched each COMID value to the corresponding FEATUREID value (unique identifier for a defined catchment) in the StreamCat database to obtain Versions 2.1 of the IWI and ICI values for each sample (https://www.epa.gov/national-aquatic-resource-surveys/streamcat (Johnson et al., 2019). An important caveat to note when obtaining index values is that sites that are on NHDPlusV2 1:100K stream lines with no defined flow direction (e.g., sites on an isolated flowline, or a canal or ditch that has no flow direction) have no catchment defined and therefore no FEATUREID assigned. This was the case for a few sites in our study. Our approach, until further work is done to resolve this issue, was to assign those sites with an ICI value that was the same as the IWI value for the watershed that the site was in.

It is also important to mention that the location of the sample site within the catchment could introduce some noise to the analysis. For example, in the case where the site is near the upper portion of the catchment and the entire catchment is used to calculate the catchment integrity. To address this issue, we are planning an analysis where we use different portions of the catchment to represent the sub-catchment associated with the sampling point, and we expect that this will give us better guidance on how to minimize noise while maximizing computational efficiency.

A series of box plots representing the distribution of landscape integrity values across biological condition classes for each case study are provided in the supplementary materials, Figs. S1a-e.

4. Results

4.1. B-IBI with IWI and ICI: Puget Sound Region and King County

The 2D scatterplot in Fig. 6a related samples of PSR and KC stream condition based on B-IBI scores to watershed- and catchment-scale integrity (IWI and ICI). The scatterplot showed a third of all samples were from areas where watershed conditions were the same as the catchment (samples plotted directly on the 1:1 line), meaning that both the watershed- and catchment-scales had similar effects on stream condition. A larger proportion of samples were from areas where watershed integrity was greater than catchment integrity (44% of samples above the 1:1 line), while a smaller number of samples were from areas with better catchment integrity than watershed integrity (23% of samples below the line, Fig. 6a). In the former case, poor catchment integrity may have a more negative effect on stream condition, while poor watershed integrity may have a more negative effect in the latter case. Further, the scatterplot (Fig. 6a) and map (Fig. 6b) revealed that, overall, samples were more abundant in areas with high watershed and high catchment integrity (82% of samples, upper right quadrant Fig. 6a; green areas Fig. 6b, Fig. S1a). Of those samples, almost equal proportions were in good and fair condition (40%, and 38%, respectively). Less than 10% of samples were from areas with high watershed but low catchment integrity (upper left quadrant Fig. 6a, yellow areas in Fig. 6b), of which the majority were in fair condition (57% of samples), followed by poor (28%) and good (15%). Similarly, samples from areas with high catchment but low watershed integrity (5%, lower right quadrant) were mostly in poor (66%) and Fair (29%) condition. Likewise, samples from low watershed and low catchment integrity (7%, lower left quadrant) were generally classified as being in fair (61%) and poor condition (31%). Overall, the large proportion of sites in the upper right quadrant generally suggests that management efforts focus on protection of existing, high integrity watershed and catchment infrastructure, with local restoration of fair and poor condition sites. Also, some limited restoration opportunities (<10%) exist in high watershed/low catchment integrity areas. Information on streams where the watershed and/or the catchment condition scores are very low (e.g., lower two quadrants) provide managers highly useful information on where investments in restoration may not be feasible nor sustainable, and provide low return on investment of public resources. The integration of biological data with landscape integrity data at different scales provides the framework to inform and support these management decisions.

4.2. Comparison between diatom BCG and statewide HGMI with IWI and ICI: New Jersey

The 2D scatterplots in Fig. 7 focused on comparisons between two types of biological condition assessment data, the NJ diatom BCG and the HGMI, in relation to the IWI and ICI. Unlike the PSR scatterplot, the NJ scatterplots showed only a small proportion of sample sites were in areas where watershed and catchment integrity were similar (18% of sites on the 1:1 line, Figs. 7a, b). The proportion of sites within watersheds with higher integrity than the catchment (42% of sites above the l:1 line) was approximately equal to sites in watersheds with lower integrity than the catchment (40% of sites below the line). The scatterplots also revealed more than a third of sites were from areas with high watershed and high catchment integrity (36% of all samples, upper right quadrant, Figs. 7a, b), of which most were considered by the BCG in good condition (Level 2, 17%; Level 3, 41%, Fig. S1b). Similarly, the HGMI rated the majority of sites in good or excellent status (excellent, 25%, good, 25%, Fig. S1c), while the remaining sites were mostly in moderate condition (fair, 42%). Of those sites, the BCG classified more sites in poor condition than the HGMI (BCG Level 5, 25%; HGMI poor, 8%). Only a few sites were from areas with high watershed and low catchment integrity (15%, upper left quadrant Figs. 7a, b), which comprised fair to poor quality waters (BCG Level 4, 25%; and Level 5, 75%, and HGMI fair, 40%; and poor, 40%; only one site was rated good). Sites within watersheds that had low integrity but high catchment integrity (18%, lower right quadrant) ranged from good to poor condition (BCG Levels 2–5; HGMI excellent-poor). The remaining sites had low integrity at both the watershed- and catchment-scale (31%, lower left quadrant Figs. 7a, b), of which most were rated by the BCG as being in either poor or good condition (BCG Levels 5 and 3, 40% each) while the HGMI generally assessed sites in good condition (excellent, 30% and good, 40%, respectively). There were differences in scoring between the two biological indicators, and in such cases, the diatom BCG generally rated sites in poorer condition than the macroinvertebrate HGMI. In three such instances, sites were assigned widely different condition classes. Site 6 and Site 32 had low watershed and low catchment integrity (lower left quadrant) and were rated a BCG Level 5 (disturbed) while the HGMI classified them as good and excellent, respectively (Figs. 7 a, b). Site 33 had higher watershed and catchment integrity (upper right quadrant) and was rated good by the HGMI while the BCG assigned it a Level 5 (Figs. 7 a, b). Despite these differences, site conditions were, for the most part, classified similarly between the diatom BCG and macroinvertebrate HGMI (Figs. 7a, b). Overall, this area has less landscape integrity than the PSR and KC (i.e., fewer percent of the points in the upper right quadrant), and more scale dependence (i.e., less points lie on the 1:1 line), so more cases of the watershed-scale integrity higher than the catchment-scale integrity, and vice versa (i.e., IWI > ICI or ICI > IWI).

4.3. Fish BCG with ICI and statewide watershed condition indicator: Connecticut

The 2D scatterplot in Fig. 8 related fish BCG condition for CT streams to the ICI and the Watershed Integrity indicator (based on state derived stream flow condition data). The scatterplot revealed a large proportion of samples were from sites in areas with high catchment integrity (83% of samples, upper and lower right quadrant, Figs. 8a, S1d). A third of samples associated with low catchment and high watershed integrity (10%, upper left quadrant) were identified as being in good condition (Level 3, 30%; and Level 2, 3%), while almost two-thirds (in the same quadrant) were considered moderately disturbed to disturbed (Level 4, 38%; Level 5, 26%; and Level 6, 3%). The remaining samples were associated with low catchment and low watershed integrity (7%, lower left quadrant), and were from waters in good condition to disturbed (Level 2, 5%; Level 3, 33%; Level 4, 40%; and Level 5, 22%).

Fig. 8.

(a) A 2D scatterplot using fish BCG assessment data for Connecticut streams illustrating the relationship between samples of BCG condition (n = 634) and corresponding ICI and Watershed Integrity (HSI). The points correspond to site condition estimates from highly disturbed Level 6 (red) to undisturbed Level 1 (yellow). The pie charts display the distribution of each BCG class in each quadrant. (b) A bar chart displaying the frequency of occurrence of samples for each BCG Level at each Watershed Integrity value. High integrity values represent better condition.

Samples were unevenly distributed along the Watershed Integrity scale and across BCG classes, as illustrated in the bar chart (Fig. 8b). Ninety percent of all samples were almost equally divided between two Watershed Integrity values, that being, samples from streams with natural conditions (Watershed Integrity of 1), of which almost two-thirds represented waters in good condition (Level 1, 11%; Level 2, 17%; and Level 3, 33%); and samples from sites with near natural conditions (Watershed Integrity of 0.5), of which nearly two-thirds reflected more disturbed conditions (Levels 4, 35%; and Level 5, 25%; Fig. 8b). The remaining 10% of samples were from sites with altered stream flow (Watershed Integrity of 0) and were from waters of the poorest condition of all the Watershed Integrity values (Level 4, 44%; and Level 5, 31%; Fig. 8b). Samples from highly disturbed sites (Level 6) were not common throughout the dataset (3% of all samples). In contrast to the other case studies, there was not an obvious 1:1 relationship here between the distribution of the sites and the ICI and Watershed Integrity (HSI). This was because of the categorical nature of the HSI, which led to a large variance within each class. However, looking at the distribution of the yellow and blue values among the three classes, it is apparent that there is a trend that slopes to the right.

4.4. Macroinvertebrate BCG with IWI, ICI and Stream-reach Integrity: Central Appalachia

The 2D and 3D scatterplots in Fig. 9 display the results of macroinvertebrate BCG condition for CAPP streams and rivers against three scales of landscape integrity: watershed- and catchment-scale (IWI and ICI, Fig. 9a) as well as a stream-reach scale, based on a measure of conductivity (Stream-reach Integrity, Fig. 9b). To help distinguish where the points in the point cloud are on the three axes, we also produced scatterplots that relate BCG condition to Stream-reach Integrity and the ICI, and on a separate plot, Stream-reach Integrity and the IWI (Figs. S2 a, b). Approximately one-third of all samples were from areas where watershed integrity was: the same as catchment integrity (30% of samples on the 1:1 line); in better condition than the catchment (37% of samples above the 1:1 line); or in poorer condition (32% of samples below the 1:1 line, Fig. 11a). Moreover, both 2D and 3D scatterplots revealed that 95% of all samples were from areas with high watershed and high catchment integrity (far right quadrant, Figs. 9a, b, S1e). Of those samples, more than a third were classified as being in good condition (BCG Levels 2 and 3 comprised 39% of samples) while a larger proportion of samples were considered impaired (45% of samples were equal to BCG Levels 5 or 6). However, these impaired samples associated with high watershed and catchment integrity were related to low integrity at the stream-reach scale, as shown in the 3D scatterplot (lower far right quadrant, Figs. 9b, S1e). This is a good example of how the different landscape integrity axes can be used to suggest spatial scales which may be associated with impairment: viewing the 2D plot only (Fig. 9a), the impaired samples in the upper right quadrant seem anomalous; by adding the stream-reach scale data (Fig. 9b), we observe that half of the impaired samples are associated with low stream-reach scale integrity. In this way, the inclusion of a third integrity indicator at the stream-reach scale can provide a separation of half the impacted sites that the IWI and ICI were unable to provide. Only 1% of samples were from areas with high watershed but low catchment integrity (far left quadrant Figs. 9a, b), which had a range of conditions from Levels 3 to 6; a similar proportion of samples were in areas of high catchment but low watershed integrity, of which most were rated as impaired (Level 5 and 6 contributed up to 71%). The few outstanding samples were from areas with poor watershed and catchment integrity (near left quadrant) and were mostly assessed as impaired (BCG Levels 4 (27%), 5 (32%), 6 (20%) while some were less impaired (21% BCG Level 3).

Fig. 9.

2D and 3D scatterplots using macroinvertebrate BCG assessment data for Central Appalachia streams and rivers displaying the relationship of BCG condition samples (n = 1011) with (a) corresponding IWI and ICI values, and (b) IWI, ICI and Stream-reach Integrity values (1 being high integrity). The points represent site condition estimates from highly disturbed BCG Level 6 (red) to minimally disturbed BCG Level 2 (light blue). The pie charts display the distribution of each BCG class in each quadrant. Note, the pie chart in the upper right quadrant conceals some points. There were no Level 1 samples in the dataset. Dashed line represents the 1:1 relationship between IWI and ICI.

5. Discussion

The value of landscape integrity information at multiple spatial scales over large areas is being increasingly stressed in stream research and management (Brown et al., 2011, Tsang et al., 2014). Stream studies have, over many years, tended to focus on individual spatial scales e.g., a single watershed or stream site (Boyero and Bosch, 2004). While there have been efforts to develop approaches that could assist stream management at multiple regional and national scales (e.g., Thornbrugh et al., 2018; Wang et al., 2011), until now, there has not been a simple analytical framework that links watershed-, catchment- and reach-scale integrity information with stream biological condition to help prioritize streams for conservation or restoration.

In many instances, site selection for restoration or protection projects is based on land availability rather than areas that may increase the likelihood of project success, which can result in the selection of suboptimal sites (Palmer, 2009). Incorporating a multiscale perspective into stream management could provide insight into the likelihood of restoration or protection success at a stream site (Palmer et al., 2010). For example, restoration of stream condition in an impaired catchment but relatively intact watershed is more likely to succeed (based on specific criteria to evaluate success) than similar efforts where both the watershed and catchment are impaired. In this regard, the spatial context of a project is fundamental to restoration success. We think that this multiscale landscape integrity framework, that can be relatively easy to implement using our national available datasets or statewide datasets, could provide important spatial information by way of scatterplots and landscape integrity maps, that could help managers determine the relative feasibility and benefits of restoration or protection, and allow the comparison of information for multiple stream sites across large geographic regions.

5.1. Identify priority sites for restoration or protection

The main technique we used to visualize results from our framework was through the use of scatterplots and a landscape integrity map. This yields several advantages that could benefit stream restoration and protection programs. Resource managers can use the scatterplot and map as a site-screening tool to help identify and prioritize candidate sites for protection or restoration within high integrity watersheds, and effectively convey those decisions to stakeholders. For example, in a 2D scatterplot and map that relates macroinvertebrate condition data (B-IBI classes) for PSR and KC streams and rivers to the IWI and ICI, we could identify samples from sites associated with high IWI and high ICI values (upper right quadrant of Fig. 6a, and green areas in Fig. 6b), as having the most restoration or protection potential. The tools could address inquiries specific to a management goal. For example, if the goal is to identify a pool of candidate stream sites within intact watersheds for protection or restoration, selecting sites with IWI values > 0.5 will help identify stream sites that meet that criterion. This way, managers could increase the feasibility of achieving successful restoration of biologically impaired sites or conservation of those in good biological condition, whilst reducing risk of directing limited funds to suboptimal sites (i.e., low IWI values in the lower two quadrants, Fig. 6a, and orange and red areas in Fig. 6b). The IWI and ICI have been included in other optimization analysis that reduces the coverage of watersheds spatially. For example, Hill et al. (2017) used the IWI, ICI and interpolated macroinvertebrate MMI values to map streams across the US that met a criteria of having predicted probabilities of good biological condition < 0.5, ICI < 0.6, but with IWI > 0.75, in order to identify candidate streams for restoration.

PSR and KC samples with low IWI values were from watersheds mostly comprised agricultural lands and high urban development (Fig. 5). For example, the Seattle-Bellevue metropolitan area is located within KC. In this highly urbanized area, restoration efforts to improve riparian vegetation and natural flow regime to reduce fine sediment inputs to the stream channel (Larson et al., 2019), are far more likely to succeed if restoration activities target existing, high integrity infrastructure that is more capable of supporting those efforts(e.g., areas that have more intact, older forested land and less development). Similarly, efforts to maintain the biological function of streams in the PSR may have a higher probability of success if protection efforts target high integrity areas such as the less-developed, forested lands where several sample sites were located (Figs. 5, 6b). In this regard, our scatterplots and maps provide important information that could be used to support decisions of where to implement protection and restoration efforts to achieve the greatest gains. In a similar context, the plots and maps could also be used as site-screening tools to help identify candidate reference sites to guide restoration and monitoring programs. One caution regarding this use is that the National Land Cover Dataset that we used to identify agricultural and urban lands has sub-categories that could have been used that would have further distinguished between higher and lower integrity (e.g., pasture and hay vs. row crops as agricultural sub-categories and low intensity residential vs. commercial/industrial/transportation for urban). While the use of these general categories was appropriate for a national analysis, it could overgeneralize integrity categories at the scale of our case studies. Finer categorical detail could therefore be appropriate if data and resources are available.

The biological condition of a sample or site in the scatterplots and map were not always consistent with the landscape integrity information for each case study. For example, a proportion of poor condition sites were related to high IWI and ICI values in the upper right quadrant. These anomalies are an indication that further investigation is required in order to identify the factor(s) causing these inconsistencies (e.g., anthropogenic stressors not accounted for by the landscape integrity indicator or taxonomic misidentification that could affect the biological condition score) and, as such, provides an opportunity for further management and developing more informed mitigation practices at these sites. The inclusion of an additional landscape integrity indicator at a more localized spatial scale may elucidate the lack of correlation between biological condition and integrity at the watershed- and catchment-scale.

5.2. Establish the best spatial scale(s) for management action

Using our multiscale approach, managers could also incorporate a third, more localized scale of landscape integrity, i.e., the stream-reach scale, in a 3D scatterplot. In this way, the three spatial scales of landscape integrity information (watershed-, catchment- and reach-scale) could more effectively determine the spatial scale(s) that has the greatest influence on biological condition. As an illustration, we showed that most impaired CAPP samples associated with unexpectedly high IWI and ICI in the 2D scatterplot (Fig. 9a), were associated with low reach-scale integrity in the 3D scatterplot (Fig. 9b). The inclusion of a Stream-reach Integrity indicator provided a separation of half of the impaired samples (i.e., BCG Level 5 and 6 samples) associated with high integrity in the upper right quadrant that the IWI and ICI were unable to provide. In this case, the biological impairment in these streams was likely due to conditions at the reach-scale. Identifying the spatial scale(s) that stream biological condition is most responsive to could enable managers to focus on the scale(s) in which restoration activities are more likely to succeed. It is worth noting that stream biological condition could also be responding to influences at the watershed- and catchment-scale that were not accounted for in the IWI and ICI. For example, underground mining in the CAPP region may not have been well represented by the nationwide mining geospatial layers included in the IWI/ICI, and as a result, lead to an overestimate of IWI/ICI values. Further analysis of the data is required to determine whether there was an overestimate of integrity.

Conductivity was used as a stressor indicator to represent reach-scale integrity, as we assumed it was a good indicator of human disturbances (mostly from mining) in CAPP streams because undisturbed streams in this region have low background conductivities and freshwater organisms are adversely affected with increasing conductivity (Griffith, 2014). Yet in areas with naturally high background conductivity attributed to underlying geology, we advise the selection of a different stressor indicator. By including an individual or multiple stressor indicator that incorporates, for example, sediment, nutrients and conductivity, to represent the reach-scale integrity, managers or scientists could apply the same framework to examine relationships between biological condition and a specific stressor of interest (e.g., Vander Laan et al., 2013). Note, the IWI and ICI are aggregate indicators that summarize information based on multiple landscape stressors as opposed to a specific contaminant.

When developing the reach-scale integrity indicator, we observed a negative linear response between conductivity and CAPP macroinvertebrate BCG classes. However, other types of stressors (e.g., nutrients) could have a nonlinear response to biological condition data, e.g., an asymptotic or bell-shaped response with an optimal range. To determine stressor-response relationships in order to develop the reach-scale integrity indicator, the publicly available data from NRSA and the U.S. EPA’s Environmental Monitoring and Assessment Program (EMAP) could be a valuable source of information, particularly in data deprived areas, to help define stressor value ranges for reference and non-reference sites that can be used to model stressor-responses (e.g., Bryce et al., 2010; Jessup et al., 2014).

5.3. Identify the scale(s) of the biological response to stressors

An additional advantage of the multiscale framework using a scatterplot and map tool to visualize results from the approach, is that it allows direct comparisons of plotted relationships between different types of biological condition data with multiscale landscape integrity information. This could be an effective way for managers to more accurately identify the scale(s) of the biological response to stressors and ensure that it is consistent with the scale of management actions (Niemi and McDonald, 2004). For example, we compared scatterplots of the diatom BCG vs. macroinvertebrate HGMI condition classes in relation to the IWI and ICI for the same NJ stream sites (Figs. 7a, b) and, in doing so, identified sites that had been assigned vastly different condition classes among the biological groups. Site 33 had high IWI and ICI values and was classified as good by the HGMI while the diatom BCG rated it as impaired (Level 5). In this case, the diatoms appear to be more responsive to conditions at the reach-scale than the macroinvertebrates however further analysis is required to support this. Other factors such as differences in the response time of macroinvertebrate and diatoms to disturbances (in which diatoms have a faster response time), should also be considered. Further, Site 6 and Site 32, located in low integrity areas (low IWI and ICI), were both considered impaired by the diatom BCG (Level 5), while the HGMI rated them in good condition. Differences between diatom and macroinvertebrate responses have been found in other river studies (e.g., Chessman et al., 2006; Hasselquist et al., 2018). Yet again, the diatoms seem to be more responsive than the macroinvertebrates, in this instance, to impairment at much larger scales; i.e., the upstream watershed and catchment. Given that different organisms respond to stressors in different ways, these results could provide more robust condition assessments and insight on the predominant stressors and impact pathway. In the past, states have used this biological information independently, where a degradation signal from one biological assemblage but not another is sufficient to signal degradation.

Reasons for discrepancies in biological condition classes here are unclear but may be elucidated by adding a reach-scale integrity indicator (that could be represented by a known major regional stressor) with the IWI and ICI. Here, we simply illustrate how this comparison approach could assist those involved in stream research in formulating hypotheses about responses to types and spatial scales of stressors among different biological groups. While some studies have found fish and macroinvertebrate community composition to be more responsive to a specific stressor at a particular scale (e.g., Townsend et al., 2003; Walsh et al., 2005), these studies did not account for watershed-scale influences on biological communities, as we have done here. Thus, the plot comparisons using different types of biological indicators could provide a more precise assessment of the scale(s) that most influences biotic conditions and the spatial extent at which management is applied.

5.4. Analytical flexibility

The multiscale approach, using a scatterplot and map to illustrate the results from our approach, enables investigative flexibility since these tools are not limited to any specific dataset and, as such, can be applied to biological and landscape condition datasets of any temporal or spatial scale, in any geographical area. The four case studies clearly illustrate the transferability of the datasets. In one example, we plotted the national IWI and ICI data and NJ macroinvertebrate condition data (HGMI) collected once (Fig. 7b). In another example, we replaced the national IWI data with state watershed integrity data and plotted it, and ICI scores, with CT fish BCG data collected over three years (Fig. 8a). We also demonstrate how this standardized approach could enable managers, scientists and stakeholders to compare results of this type of analysis across multiple scales of governance.

6. Conclusion

This paper presents a conceptual framework and approach that could further advance the assessment, communication, and management of stream condition by integrating biological condition data with multiscale landscape integrity information. We applied scatterplots and maps to relate stream condition to three different scales of landscape integrity (stream-reach, catchment- and watershed-scale), and demonstrated the flexibility in our approach by applying different types of indicators to represent biological condition and multiscale landscape integrity data at local-, state- and regional-scales. Moreover, our approach could support both ecologically and economically optimal management decisions; the graphical nature of the scatterplot and landscape integrity map may enable identifying and prioritizing sites for protection and restoration as well as areas to avoid where remediation efforts would be costly and less effective. Further, this framework offers an approach to integrate measures of system resilience; landscape integrity data at the stream-reach, watershed- and catchment-scale could provide valuable insight into the scale of the stressor(s), and, as such, help identify sites where changes could cause the area to flip to a more degraded or improved condition level. While the application of the national IWI and ICI is still relatively new and requires further evaluation and refinement (Aho et al., 2020; Thornbrugh et al., 2018), we show how the existing IWI and ICI could provide an important dataset, especially in data poor areas, to inform management and decision-making at local or large geographic scales. To further assess the application of the national IWI and ICI in stream condition assessments, development of a regional high resolution IWI and ICI using local data and NHDPlus high resolution data (1:24K scale; https://www.usgs.gov/core-science-systems/ngp/national-hydrography/nhdplus-high-resolution) could help by allowing us to compare regional vs. current (medium resolution) IWI/ICI scores. This may provide insight into whether these national indices require further improvement and to what extent refinement is needed.