There is a critical need for tools and methodologies capable of managing aquatic systems in the face of uncertain future conditions. We provide methods for conducting a targeted watershed assessment that enables resource managers to produce landscape-based cumulative effects models for use within a scenario analysis management framework.
There is a critical need for tools and methodologies capable of managing aquatic systems within heavily impacted watersheds. Current efforts often fall short as a result of an inability to quantify and predict complex cumulative effects of current and future land use scenarios at relevant spatial scales. The goal of this manuscript is to provide methods for conducting a targeted watershed assessment that enables resource managers to produce landscape-based cumulative effects models for use within a scenario analysis management framework. Sites are first selected for inclusion within the watershed assessment by identifying sites that fall along independent gradients and combinations of known stressors. Field and laboratory techniques are then used to obtain data on the physical, chemical, and biological effects of multiple land use activities. Multiple linear regression analysis is then used to produce landscape-based cumulative effects models for predicting aquatic conditions. Lastly, methods for incorporating cumulative effects models within a scenario analysis framework for guiding management and regulatory decisions (e.g., permitting and mitigation) within actively developing watersheds are discussed and demonstrated for 2 sub-watersheds within the mountaintop mining region of central Appalachia. The watershed assessment and management approach provided herein enables resource managers to facilitate economic and development activity while protecting aquatic resources and producing opportunity for net ecological benefits through targeted remediation.
Anthropogenic alteration of natural landscapes is among the greatest current threats to aquatic ecosystems throughout the world1. In many regions, continued degradation at current rates will result in irreparable damage to aquatic resources, ultimately limiting their capacity to provide invaluable and irreplaceable ecosystem services. Thus, there is a critical need for tools and methodologies capable of managing aquatic systems within developing watersheds2-3. This is particularly important given that managers are often tasked with conserving aquatic resources in the face of socioeconomic and political pressures to continue development activities.
Management of aquatic systems within actively developing regions requires an ability to predict likely effects of proposed development activities within the context of pre-existing natural and anthropogenic landscape attributes3, 4. A major challenge to aquatic resource management within heavily degraded watersheds is the ability to quantify and manage complex (i.e., additive or interactive) cumulative effects of multiple land use stressors at relevant spatial scales2, 5. Despite current challenges, however, cumulative effects assessments are being incorporated into regulatory guidelines throughout the world5-6.
Targeted watershed assessments designed to sample the full range of conditions with respect to multiple land use stressors can produce data capable of modeling complex cumulative effects7. Moreover, incorporating such models within a scenario analysis framework [predicting ecological changes under a range of realistic or proposed development or watershed management (restoration and mitigation) scenarios] has the potential to greatly improve aquatic resource management within heavily impacted watersheds3, 5, 8-9. Most notably, scenario analysis provides a framework for adding objectivity and transparency to management decisions by incorporating scientific information (ecological relationships and statistical models), regulatory goals, and stakeholder needs into a single decision-making framework3, 9.
We present a methodology for assessing and managing cumulative effects of multiple land use activities within a scenario analysis framework. We first describe how to appropriately target sites for inclusion within the watershed assessment based on known land use stressors. We describe field and laboratory techniques for obtaining data on the ecological effects of multiple land use activities. We briefly describe modeling techniques for producing landscape-based cumulative effects models. Lastly, we discuss how to incorporate cumulative effects models within a scenario analysis framework and demonstrate the utility of this methodology in aiding regulatory decisions (e.g., permitting and restoration) within an intensively mined watershed in southern West Virginia.
1. Target Sites for Inclusion in Watershed Assessment
Figure 1. Hypothetical scatter plot of NHD catchments with respect to influence from 2 land use activities. Magnitude of influence of 2 land use activities across all NHD catchments within the hypothetical watershed (n=4,229) (A). Selected study sites (n=40) that represent the full range of observed conditions within the watershed with respect to independent and combined stressor gradients (B). Please click here to view a larger version of this figure.
2. Field Protocols for Collection of Physicochemical and Biological Data
Note: All data for each site should be collected during the same site visit at normal base flow conditions. Protocols presented herein represent standard operating procedures for the West Virginia Department of Environmental Protection (WVDEP)13. It may be more appropriate to use state or federally recognized procedures for the specific watershed being assessed.
3. Laboratory Protocols for Physicochemical and Biological Data
Note: Describing laboratory protocols for quantifying water chemistry attributes is outside the scope of this manuscript. However, the current study used standard chemical methods for water and wastes14.
4. Statistical and Scenario Analyses
Forty 1:24,000 NHD catchments were selected as study sites within the Coal River, West Virginia (Figure 2). Study sites were selected to span a range influence from surface mining (% land area24), residential development [structure density (no./km2)], and underground mining [national pollution discharge elimination system (NPDES) permit density (no./km2)] such that each major land use activity occurred both in isolation and in combination to the extent possible (Figure 3). At each site, data on physicochemical conditions and macroinvertebrate community structure were collected.
In a previous study, these data were used to construct cumulative effects models for predicting West Virginia Stream Condition Index (WVSCI), a family-level multi-metric index of biotic integrity developed for West Virginia25, and specific conductivity with a high degree of precision and accuracy7. Herein, these models are used to predict current and future conditions for two sub-watersheds of the Coal River [Drawdy Creek (Figure 4A) and Laurel Fork (Figure 4B)] under various land use development scenarios. Drawdy Creek and Laurel Fork have nearly identical levels of surface mining and % development (Table 1). However, Drawdy Creek is influenced by residential structures and underground mining, whereas Laurel Fork is not. Consequently, these two watersheds offer a unique opportunity to assess and compare the extent to which cumulative effects of multiple land use activities control current aquatic conditions and the outcome of future land use development scenarios.
Laurel Fork was not predicted to exceed chemical (specific conductivity > 500 µS/cm26) or biological criteria (WVSCI < 6825), suggesting it can assimilate additional land use activity without risking impairment (Table 1). A series of scenarios were then assessed to quantify the maximum amount of additional surface mining, underground mining, and residential development Laurel Fork can likely assimilate before its outflow crosses each criterion. To do this, specific conductivity and WVSCI were predicted under the full range of each land use activity while holding the other landscape metrics constant. Scenario analysis suggests Laurel Fork can assimilate 14% (25% total) and 21% (32% total) increases in surface mine lands prior to crossing the specific conductivity and WVSCI criteria, respectively (Figure 5A, 5B). Laurel Fork can also assimilate 8 underground mine NPDES permits and 22 residential structures before crossing the specific conductivity and WVSCI criteria, respectively (Figure 5A, 5B).
In contrast, the outflow of Drawdy Creek is predicted to exceed both the chemical and biological criteria, suggesting an inability to assimilate any additional land use development without first mitigating effects of current stressors (Table 1). Consequently, mitigation scenarios that reduce the overall effect size of pre-existing land use activities (e.g., a 10% reduction in the effect of 100 structures would be equivalent to 90 structures) were simulated. Fully mitigating the effect of residential development and underground mining did not result in a respective increase in WVSCI above 68 or decrease in specific conductance below the 500 µS/cm criterion (Figure 6A, 6B). However, the outflow of Drawdy Creek was predicted to exceed a WVSCI score of 68 and decrease below 500 µS/cm with simultaneous reductions in both residential development and underground mining of a 94 and 75%, respectively.
Figure 2. Map of the Coal River watershed. The Coal River watershed is shown with respect to its location within West Virginia. Locations of study sites (n=40) and Laurel Fork and Drawdy Creek sub-watersheds are also presented. Please click here to view a larger version of this figure.
Figure 3. Coal River study sites. Magnitude of surface mining and residential development for selected study sites (n=40) within independent stressor gradients and their combination. Symbol size is relative to the number of underground mining national pollution discharge elimination system (NPDES) permits. Please click here to view a larger version of this figure.
Figure 4. Maps depicting land use activities within Drawdy Creek (A) and Laurel Fork (B). These watersheds represent patterns of land use geography typical throughout the MTR-VF region. Residential development [land cover (as defined by the NLCD) and structures] and mining (underground mining NPDES permits and surface mine extent) land use activities are shown. Additional un-mined permits used in scenario analysis are shown. Refer to Figure 2 for watershed location within West Virginia. Please click here to view a larger version of this figure.
Figure 5. Example scenario analysis results predicting in-stream response to simulated land use development within Laurel Fork. Predicted WVSCI scores following simulated increases in surface mining and residential development (A) and predicted specific conductance following simulated increases in surface mining and underground mining (B) within the Laurel Fork watershed. Horizontal lines represent WVSCI (68) and specific conductance (500 µS/cm) criteria. Vertical lines represent additional levels of mining resulting in crossing of each criterion. Units for the x-axis vary depending on the landscape attributes changed under each scenario and correspond to units specified in the legend. Please click here to view a larger version of this figure.
Figure 6. Example scenario analysis results predicting in-stream response to simulated mitigation activities within Drawdy Creek. Predicted WVSCI scores (A) and specific conductivity (B) following simulated decreases in the effect size of existing residential development and underground mining, respectively. Predicted conditions following simultaneous reductions in the effect size of both residential development and underground mining are also shown for each response. Horizontal lines represent WVSCI (68) and specific conductance (500 µS/cm) criteria. Vertical lines indicate mitigation activities resulting in improvements beyond each criterion. Please click here to view a larger version of this figure.
Current landscape | |||
Drawdy Creek | Laurel Fork | ||
Land use characteristics | |||
Surface mining (%) | 10.7 | 10.9 | |
Underground mining (# NPDES permits) | 9 | 0 | |
Development (%) | 4.1 | 4.8 | |
Structure density (#) | 470 | 0 | |
Observed conditions | |||
Specific conductance (µS/cm) | 686 | 156 | |
WVSCI | 65 | 68.8 | |
Predicted conditions | |||
Specific conductance (µS/cm) | 831 | 279 | |
WVSCI | 60.9 | 73.1 |
Table 1. Landscape characteristics and observed and predicted aquatic conditions for Drawdy Creek and Laurel Fork. Land use characteristics (surface mining, underground mining, and residential development) and predicted chemical and biological conditions for Drawdy Creek and Laurel Fork under current landscape conditions and the additional mining scenario.
We provide a framework for assessing and managing cumulative effects of multiple land use activities in heavily impacted watersheds. The approach described herein addresses previously identified limitations associated with managing aquatic systems in heavily impacted watersheds5-6. Most notably, the targeted watershed assessment design (i.e., sampling along individual and combined stressor axes) produces data that are well suited for quantifying complex cumulative effects at relevant spatial scales (i.e., watershed scale) via easily interpretable and implementable modeling techniques3, 7. Moreover, these models are readily incorporated into a scenario analysis framework that enables accurate prediction of future management (e.g., restoration and mitigation) and development outcomes. Consequently, the presented approach will likely be of value to aquatic resource managers who increasingly rely on forecasting conditions under various land use scenarios to aid in regulatory decisions27.
The contrast between Drawdy Creek and Laurel Fork highlights the utility of the presented framework when managing aquatic systems within actively developing and socioeconomically important regions. Scenario analysis suggested that Laurel Fork, which is impacted solely by surface mining (10.9%), can assimilate additional land use development without exceeding chemical and biological criteria. Drawdy Creek, which is affected by equivalent levels of surface mining (10.7%), is predicted to not meet either criterion as a result of cumulative effects associated with underground mining and residential structures. However, simulated mitigation of non-surface mining stressors (e.g., underground mine effluent and residential wastewater) improved ecological conditions, suggesting strategic management activities could enable further development to occur. Consequently, the presented approach makes it possible to facilitate economic and development activity while also producing opportunity for net benefits through remediation of other stressors28.
Successful identification and sampling of dominant land use stressors is a critical step in successfully implementing the methodologies presented herein. It is also critical that sampling and subsequent data analyses are based upon the best available and most up-to-date land cover and use information. Temporal consistency between land cover and in-stream data help ensure accurate statistical relationships and subsequent ecological predictions3, 9. If conducted appropriately, the presented watershed assessment technique produces data that are largely unbiased (i.e., minimizes specification error and omitted variable biases) and unaffected by multicollinearity. Consequently, these data are well suited for predictive modeling via traditional regression techniques. One potential limitation of the current approach, however, is that strong ability to empirically predict spatial pattern does not guarantee an ability to predict change over time. Notably, studies have observed interactions between climate and land use change on physicochemical and biological conditions29-31. Thus, adaptive management approaches that test temporal predictions and update spatial predictive models will be an important component of management efforts. This should involve incorporating climate change into statistical models and subsequent scenario analyses.
Our methodology can also be adapted to utilize existing datasets that may not fulfill assumptions of traditional regression techniques (e.g., multicollinearity and sample independence). The use of pre-existing data is beneficial in situations where managers have limited time or resources. Boosted Regression Tree (BRT) models may be particularly useful when analyzing large, pre-existing datasets because they are largely unaffected by multicollinearity, missing data, statistical outliers, and non-normal data32. Moreover, BRT offers high predictive performance and has demonstrated utility in a scenario analysis framework28.
It is important to note the context within which our methodology was developed. First, our approach was developed for watersheds characterized by clearly defined land use gradients. However, clearly defined land use gradients do not always occur at the watershed-scale (e.g., areas of the Midwestern United States with little variation in agricultural extent). Consequently, other approaches to conservation planning, such as risk-based methods that rank conservation targets based on risks of multiple land use activities, may be more appropriate33-34. Moreover, our approach was designed at the 8-digit HUC watershed scale. In a previous study, we found that models constructed across multiple 8-digit HUC watersheds fail to predict watershed-specific nuances between land use and in-stream conditions7. Constructing models across smaller spatial scales (e.g., 12-digit HUC watersheds) may constrain sample size and limit the ability of models to quantify complex cumulative effects. However, our approach can be used to manage across spatial scales via a house-neighborhood framework2. Under this framework, restoration and protection priorities are set for individual streams within the context of surrounding conditions. For example, restoration potential increases with increasing neighborhood condition because of benefits associated with having good streams nearby (e.g., high re-colonization potential).
We provide and demonstrate protocols for assessing and managing cumulative effects within heavily impacted watersheds. Although the current manuscript focused on construction and implementation of cumulative effects models within a scenario analysis framework, the demonstrated watershed assessment techniques produce data capable of quantifying detailed patterns of physicochemical and biological degradation related to the accumulation of dominant land use activities across larger spatial scales35. Consequently, data produced by the study design and sampling protocols described herein have potential management benefits that extend well beyond those discussed. Perhaps most importantly, this framework is transferable to other watersheds facing ongoing transitions in any number of land use activities.
The authors have nothing to disclose.
We thank the numerous field and laboratory helpers that were involved in various aspects of this work, especially Donna Hartman, Aaron Maxwell, Eric Miller, and Alison Anderson. Funding for this study was provided by the US Geological Survey through support from US Environmental Protection Agency (EPA) Region III. This study was partially developed under the Science To Achieve Results Fellowship Assistance Agreement number FP-91766601-0 awarded by the US EPA. Although the research described in this article has been funded by the US EPA, it has not been subjected to the agency's required peer and policy review and, therefore, does not necessarily reflect the views of the agency, and no official endorsement should be inferred.
Slack Invert Sampling Kit | Wildco | 3-425-N56 | |
HDPE Square Jars | US Plastic Corp | 66188 | 32oz./for storing fixed, composite invertebrate samples |
Ethyl Alcohol 190 Proof | PHARMCO-AAPER | 111000190 | For fixing and storing invertebrate samples |
5in. by 20in. Macroinvertebrate sub-samplilng grid | N/A | N/A | This item cannot be purchased and must be made in house |
Stereomicroscope Stemi 2000 with stand C LED | ZEISS | 000000-1106-133 | For macroinvertebrate sorting and identification |
Thermo Scientific Nalgene Reusable Filter Holders with Receiver | Fisher Scientific | 09-740-23A | |
Immobilon-NC Transfer Membrane | Millipore | HATF04700 | Triton-free, mixed cellulose exters, 0.45um, 47mm, disc |
Actron Vacuum Pump Brake Bleeder Kit | Advanced Auto Parts | CP7835 | |
Nitric Acid Solution | HACH | 254049 | 1:1, 500mL |
Oblong NDPE Wide Mouth Bottles | Thomas Scientific | 1229Z38 | 250 mL/for collection of water samples |
650 Multi-parameter display, standard memory | Fondriest Environmental | 650-01 | |
600XL Sonde with temperature/conductivity sensor | Fondriest Environmental | 065862 | |
pH calibration buffer pack | Fondriest Environmental | 603824 | 2 pints each of pH 4, 7, & 10 |
conductivity standard | Fondriest Environmental | 065270 | 1 quart, 1000 uS |
Flo-Mate 2000 | TTT Environmental | 2000-11 | |
Keson English/Metric Open Reel Fiberglass Tape | Forestry Suppliers | 40025 | 300'/100m |
ArcGIS 10.3.1 | ESRI |