2007 Progress Report: Ecological Thresholds and Responses of Stream Benthic Communities to Heavy MetalsEPA Grant Number: R832441
Title: Ecological Thresholds and Responses of Stream Benthic Communities to Heavy Metals
Investigators: Clements, William , Noon, Barry , Wang, Haonan
Institution: Colorado State University
EPA Project Officer: Packard, Benjamin H
Project Period: July 31, 2005 through December 30, 2007 (Extended to December 30, 2008)
Project Period Covered by this Report: July 1, 2006 through June 30,2007
Project Amount: $295,760
RFA: Exploratory Research: Understanding Ecological Thresholds In Aquatic Systems Through Retrospective Analysis (2004) RFA Text | Recipients Lists
Research Category: Aquatic Ecosystems , Water
The primary goal of our research is to identify and validate ecological thresholds for stream communities impacted by heavy metal contamination from historic mining operations (Fig. 1). We hypothesize that the location of ecological thresholds along a well defined gradient of metal contamination varies among state variables and is influenced by local (reach-scale) and landscape-level environmental drivers. We integrated results of 16 microcosm experiments and two spatially extensive surveys of 79 Colorado streams to identify ecological thresholds along a gradient of heavy metal contamination. For the purposes of this proposal we define an ecological threshold (Xr) as a metal concentration that, when exceeded, results in pronounced changes in the values of one or more state variables. Using parametric and nonparametric regression techniques, we identified thresholds for state variables across several levels of biological organization. To validate and test the ecological significance of these thresholds, we examined long-term changes in response variables in a mining impacted stream where metal concentrations were recently reduced below threshold levels.
Specific Hypotheses to be tested
Our research will test the following hypotheses:
- Ecological thresholds in the abundance and diversity of benthic macroinvertebrates occur along a gradient of heavy metal contamination in Rocky Mountain streams impacted by mining.
- The location and characteristics of ecological thresholds along a gradient of metal contamination vary among response variables.
- Reach-scale and landscape-level environmental drivers influence the location of ecological thresholds along a gradient of metal contamination.
- Reduction in metal concentrations below threshold levels results in recovery of state variables.
- Recovery of state variables to pre-stressor levels may be characterized by significant time lags.
Figure 1. Overview of the data to be used for identification and validation of ecological thresholds in metal-contaminated streams.
1. Identification of Thresholds based on Microcosm Experiments
We completed our analysis of 16 separate microcosm experiments examining effects of heavy metals (Cd, Cu, and/or Zn) on benthic macroinvertebrate communities. Metal concentrations were characterized using the cumulative criterion unit (CCU), which is defined as the ratio of the measured metal concentration to the hardness-adjusted criterion value and summed for each metal (Clements et al. 2000). CCU = ∑ mi/ci, where mi = the measured concentration of the ith metal and ci = the hardness-adjusted criterion value for the ith metal. Results showed that ecological thresholds derived from these relatively short-term microcosm experiments were comparable to EC20 values but generally greater than thresholds derived from a spatially extensive field surveys (Table 1).
Table 1. Comparison of microcosm EC20 values with ecological thresholds derived from microcosm experiments and U.S EPA R-EMAP data.
2. Identification of Threshold based on Spatially Extensive Field Data
Results of threshold analyses from a spatially extensive survey of 73 Colorado streams along a gradient of metal pollution (Clements et al. 2000) were reported in the previous annual report. Our analyses showed statistically significant ecological thresholds for most measures of abundance and species richness. Both piecewise linear regression and a monotone spline technique identified a threshold for total mayfly abundance of approximately ln[CCU] = 1.4, corresponding to a CCU value of 4.0. These results suggest that significant changes in benthic communities will likely be observed when metal concentrations exceed 4X the CCU.
3. Threshold Validation: Long-term Assessment of Recovery in the Arkansas River, CO
A third dataset was used to assess the ecological significance of the threshold derived from the spatially extensive survey. Our primary goal was to determine if this threshold was related to recovery of the Arkansas River, a metal polluted stream in central Colorado (Fig. 2). Long-term data collected from the Arkansas River before and after remediation were used to test this hypothesis. Results of these analyses were reported in our 2006 annual report. Briefly, results showed that ecological thresholds from the spatially extensive survey were consistent with results of the long-term monitoring. Piecewise regression identified a distinct threshold in 1995, indicating that benthic communities recovered within approximately 2 years after metals were reduced below approximately 4.0 CCU (Fig. 3).
Figure 2. Relationship between resistance and resilience thresholds in ecosystems. Resistance thresholds indicate stressor levels (e.g., metal concentrations) that result in rapid changes in ecological responses. Resilience thresholds indicate the amount of time required for a system to recover after removal of a stressor.
Figure 3. Long-term changes in metal concentrations (as CCU) and abundance of mayflies in the Arkansas River, CO. Horizontal line shows the approximate threshold based on a spatially extensive survey of 73 Colorado streams.
Identifying ecological thresholds in multivariate space
Multivariate analysis of macroinvertebrate data collected from the Arkansas River provided an opportunity to investigate ecological thresholds in community composition over the duration of the 17 year monitoring project. Canonical discriminant analysis was used to examine differences among years based on abundance of the 20 dominant taxa. A threshold response in this example represents an abrupt shift in community composition over time. Macroinvertebrate communities in the Arkansas River showed several distinct shifts along canonical variable 1, which explained most of the variation in community composition (Fig. 4). Simple piecewise regression identified a single threshold in 1998; however, it is obvious from the scatterplot that the relationship is much more complex and that 3 potential thresholds exist in these data.
Figure 4. Multivariate analyses of community composition based on abundance of the 20 dominant taxa. The figure shows mean values (+ s.e.) in multivariate space (1989 to 2006) (left) and piecewise regression analysis of the relationship between year and canonical variable 1. Arrows indicate potential thresholds where communities changed abruptly.
We have developed a new derivative-based approach that uses locally weighted polynomial regression to detect and quantify multiple ecological thresholds (Sonderegger et al. 2008). The Significant Zero crossings (SiZer) approach (Chaudhuri and Marron 1999; Hannig and Marron 2006) applies a nonparametric smoother to the stressor-response data and then examines derivatives of the smoothed curve to identify a threshold. The approach makes relatively few model assumptions and is therefore applicable to a broader range of ecological applications. When the weight function (also called a kernel function) is the normal density curve, the level of smoothing is controlled by the standard deviation of the kernel. For a given tuning parameter (h), and a given point (x0), the function fˆ′(x0) is obtained by weighting the data points according to a normal curve centered at x0 with a standard deviation σ = h. Data close to x0 (within ± h) have a large influence over the smoothing function, data between h and 2h are less influential, and data farther than 2h from x0 have only a slight influence. The novel aspect of SiZer is that it considers all reasonable bandwidths and exploits the notion that different values provide different information about the data. More importantly, this approach also allows us to quantify multiple thresholds over time.
We applied SiZer to the first canonical axis, which explained 58.2% of the total variation in the multivariate analysis described above (Fig. 5). Using a tuning parameter (h) of 1.5, we fit a locally weighted polynomial regression model to these data. The associated SiZer map from this analysis is also shown. To read the SiZer map, first notice that the y-axis represents the bandwidth parameter h, displayed in units of log(h) for visual clarity. The dashed horizontal line represents the bandwidth parameter used in the polynomial regression model. Wherever the SiZer map is yellow, the derivative is significantly increasing; wherever it is purple the derivative is possibly zero, and wherever it is red the derivative is significantly decreasing.
The first derivative shows a generally increasing function, but there is a sharp decreasing trend between 1995 and 1997 (color change to red). These results reflect macroinvertebrate community responses to changes in water quality from 1989 to 2006. Heavy metal concentrations declined from 1989-1994, increased abruptly in 1995 and1996, and then declined again as a result of ongoing restoration in the Arkansas River (Clements, 2004). The SiZer map shows that the decrease between 1995 and 1997 was statistically significant and was defined as an ecological threshold. A threshold near 2000, which was a result of macroinvertebrate community recovery after improvements in water quality, was also statistically significant.
Figure 5. Figure 3: Results of locally weighted polynomial regression (left) and the associated SiZer map (right) of the Arkansas River multivariate data. The SiZer map categorizes the 1st derivative as either positive (yellow), negative (red) or possibly zero (purple). The horizontal dashed line is the associated bandwidth from the locally weighted polynomial regression.
Future activities during the next year of this project will include 1) applying SiZer to additional macroinvertebrate datasets to examine responses to other stressors; and 2) developing mathematical approaches to estimate the importance of error associated with stressor gradients. In addition to applying SiZer to a large-scale, spatially extensive dataset of 150 Colorado streams, we are working with researchers from the National Park Service to quantify the effects of geothermal discharges on macroinvertebrate communities. Preliminary results indicate that SiZer identified a distinct and statistically significant ecological threshold associated with natural geothermal releases in Yellowstone streams. These results will assist resource managers in developing monitoring plans designed to separate natural variation associated with geothermal influences from variation associated with anthropogenic stressors. In September, 2006 Clements collaborated with the National Park Service to develop a special session on applications of ecological thresholds at workshop on communicating science to managers. At this session Clements presented a talk entitled “Thresholds in aquatic and terrestrial ecosystems.”
We are also investigating the effect of measurement error of predictor variables on regression inference. While this phenomenon is well known in the statistical literature, the influence of measurement error on our ability to quantify relationships between stressors and responses has received little attention in ecotoxicology. We will demonstrate that by ignoring measurement error, researchers risk introducing a strong bias in slope estimates that consistently underestimate the strength of the relationship between stressor and response variables. This bias is relatively small if the measurement error is small compared to overall variability, but increases as measurement errors increase. We will also propose a solution to this problem that smoothes the predictor variable with respect to another covariate (e.g., time) and then uses the smoothed predictor to estimate the response variable.
Chaudhuri, P., and J. S. Marron. 1999. SiZer for exploration of structures in curves. Journal of the American Statistical Association 94:807–823.
Clements, W.H. 2004. Small-scale experiments support causal relationships between metal contamination and macroinvertebrate community responses. Ecological Applications 14: 954-967.
Clements, W.H., D.M. Carlisle, J.M. Lazorchak, and P.C. Johnson. 2000. Heavy metals structure benthic communities in Colorado mountain streams. Ecological Applications 10:626-638.
Hannig, J., and J. S. Marron. 2006. Advanced distribution theory for SiZer. Journal of the American Statistical Association 101:484–499.
Sonderegger, D.L., H. Wang, W.H. Clements, B.R. Noon (in press). Using SiZer to detect thresholds in ecological data. Frontiers in Ecology and the Environment.