SEARCH

SEARCH BY CITATION

Keywords:

  • Ecological risk assessment;
  • Exposure analysis;
  • Spring Chinook;
  • Spatial modeling;
  • Temporal modeling;
  • Pesticides

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

In this paper, we present a novel approach for determining the probable co-occurrence of juvenile salmon or steelhead with agricultural pesticides and apply it to spring Chinook (Oncorhynchus tshawytscha) salmon in the Willamette Basin, Oregon. We adapted a published exposure analysis framework by explicitly considering fish migration among habitat units and assuming that habitat use is proportional to habitat quality. Temporal variability in habitat use was accounted for via biweekly time steps over the entire period when a single brood was expected to spawn until the last juvenile migrated to sea. Spatial variability was accounted for at the watershed and reach scale. Exposure to 6 acetylcholinesterase-inhibiting insecticides at any life stage was expressed in terms of the future adults (adult-equivalents; AEQ). Several datasets were available to inform our framework with input values on extent of spring Chinook fish use, habitat quality preferred by juvenile spring Chinook, choice of juvenile life-history pathways, timing of emergence, and timing of migration either in-stream or to sea. We used insecticide concentration profiles constructed from available monitoring data to demonstrate the effect of accounting for variation in space and time on predicted exposure to chemical residues. In contrast to the assumption commonly used in screening-level risk assessments that the entire population in a watershed is exposed, available data applied to our model framework indicate that a small fraction of AEQ juveniles in the Willamette Basin would co-occur with detectable concentrations of the 6 insecticides. Overall, our results indicated that the use of a spatially and temporally explicit framework yields a better understanding of the proportion of organisms potentially impacted by agricultural pesticides. Integr Environ Assess Manag 2012; 8: 271–284. © 2011 SETAC


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

Migratory organisms present a challenge when predicting exposure to environmental stressors given their uneven distribution in time and space. This challenge has been acknowledged for Pacific salmon (Oncorhynchus sp.), of which 26 evolutionarily significant units (ESUs) across the western United States had been listed by 2002 as threatened or endangered under the US Endangered Species Act (ESA) (Ruckelshaus et al. 2002). Salmon are often used as indicator species in ecological risk assessments because they are high on the food chain, they have high esthetic, recreational, and commercial values, and much data are available on their biology and habitat. Salmon are diadromous; generally, they spawn and rear in freshwater, migrate to sea where they grow and mature, and then return when mature to their natal area in freshwater to spawn. The timing of salmon movements in freshwater tends to follow repeatable patterns, as do their spatial habitat preferences (Quinn 2005). Similarly, the intensity of environmental stressors varies but tends to follow repeatable patterns across time and locations. Quantitative frameworks for integrating these numerous and often disparate pieces of information are needed to assess risks to salmonids (Ruckelshaus et al. 2002).

When considering the exposure of salmonids to pesticides, the need for such a framework is critical. Because the presence of both pesticides and fish vary independently over time and space, we would expect the exposure to fish in their natural habitats to vary widely over time and space. If we do not account for the spatial and temporal variability in fish and pesticides, we are often left making the simplifying assumption that 100% of the population is exposed to a single pesticide concentration. This assumption, although unlikely to be true, is often carried forward in risk assessments (e.g., Baldwin et al. 2009). However, a substantial body of information on fish migration and stream habitat quality is available that enables better characterizations of exposure. Therefore, the primary objective of our study was to develop a framework that integrates this information and leads to more accurate estimates of the percentage of the population exposed to environmental stressors, and to agricultural pesticides in particular.

In this paper, we present a spatially and temporally explicit model for determining the exposure of juvenile spring Chinook (Oncorhynchus tshawytscha) salmon to pesticides in the Willamette Basin, Oregon. A simple mathematical framework is presented that supports evaluation at ecologically relevant scales. For demonstration purposes, we analyze the exposure of juvenile spring Chinook salmon through contact with the organophosphorous insecticides chloropyrifos, diazinon, and malathion, and the carbamate insecticides carbaryl, carbofuran, and methomyl, subjects of recent Biological Opinions by NMFS (2008, 2009). In a companion paper in this issue, Poletika et al. extend this framework to evaluate indirect effects on juvenile salmon survival through reduction of the prey base as a result of toxic effects of pesticides on aquatic invertebrates. Overall, this framework is presented as a novel approach to ecological risk assessment for Pacific salmon in freshwater habitats.

METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

We built our demonstration model to evaluate spawning and rearing use of juvenile spring Chinook salmon in the Upper Willamette River Chinook (CKUWR) evolutionary significant unit (ESU). As defined by the National Oceanic and Atmospheric Administration (NOAA), the CKUWR ESU includes natural populations of spring-run Chinook salmon spawned in streams of the Willamette River Basin, Oregon (NOAA 2005a) (Figure 1). Seven geographically distinct populations of spring-run Chinook have been identified within the Willamette Basin by NOAA (2005b); we incorporated an eighth—Coast Fork Willamette River—which has marginal spawning capacity. The Willamette Basin drains a large area (approximately 31 000 km2) and is heavily populated on the valley floor (about 1.9 million people or about 68% of Oregon's population). About 70% of the basin is forested (largely tributary basins), about 22% is farmed, and the remaining 8% is urbanized or in other uses. Agricultural land use in the Willamette Basin is concentrated in the Willamette Valley (Figure 2). Using the National Land Cover Database (Homer et al. 2004), we determined that about 600 km of stream that is designated in the CKUWR ESU as “critical habitat” (NOAA 2005a) occurs within 300 m of lands classified as agricultural (pasture-hay or cultivated crops); this represents about one-fifth of all critical habitat designated within the Willamette Basin.

thumbnail image

Figure 1. The extent of spring Chinook salmon spawning-rearing reaches (highlighted lines) and rearing-migration reaches in the Willamette River Basin, Oregon.

Download figure to PowerPoint

thumbnail image

Figure 2. The extent of agricultural land cover (shaded area) and spring Chinook critical habitat in the Willamette River Basin, Oregon.

Download figure to PowerPoint

Assessment endpoints relevant to recovery of spring Chinook are the production of juvenile fish and reproduction of adults. We assumed that agricultural chemical stressors are most significant for juvenile fish reared in freshwater, and that successful out-migration to the ocean precludes impairment of reproduction. Reported concentrations of the 6 insecticides from USGS monitoring in the Willamette River Basin (USGS 2010) indicate fish acute lethality is unlikely, based on acute toxicity endpoints available from rainbow trout regulatory studies (USEPA 2010), either considered singly or as concentration additions. The recent Biological Opinions by NMFS (2008, 2009) indicated that consideration of concentration additions is appropriate for this purpose. Episodic exposure patterns for this type of pesticide are generally observed in agriculturally dominated reaches (Giesy et al. 1999; Poletika et al. 2002), suggesting chronic effects are also unlikely. Sublethal effects related to behavior such as those assessed by Baldwin et al. (2009) are observed only at elevated, constant exposure levels and are, therefore, unlikely to occur at environmentally relevant concentrations in the Willamette River Basin.

Therefore, we chose to demonstrate our framework relative to several concentration thresholds that are useful as screening level indicators with the understanding that these thresholds do not account for the spatial and temporal patterns of exposure that are the focus of our framework. Thus, they may overstate the ecological risk. Poletika et al., this issue, apply the framework in a risk assessment incorporating spatially and temporally explicit indicators of exposure and effects. For our demonstration, we chose to use these thresholds simply to provide an indication of the potential magnitude of effect by considering spatial and temporal variability and to demonstrate model sensitivity. We chose as our lowest threshold the value 0.1 µg/L. The recent Biological Opinions by NMFS (2008, 2009) indicated that population growth (“lambda” parameter) of Chinook salmon remains relatively unaffected by these contaminants below this level. This value is slightly higher than the sum of the limits of detection for the 6 acetylcholinesterase (AChE)-inhibiting contaminants reported by USGS (2010) (0.08 µg/L). We chose as our highest threshold the value of 1.0 µg/L, a concentration beyond which the percent change in lambda was usually more than 1 standard deviation from control populations (NMFS 2008, 2009). Finally, we chose as an intermediate threshold the value of 0.5 µg/L to demonstrate the sensitivity of model outcomes across a range of concentration thresholds.

Quantitative framework

Our framework follows USEPA guidance (e.g., USEPA 1998; Sample et al. 1997) for estimating the proportion of organisms potentially impacted by environmental stressors. Originally developed for terrestrial wildlife, this guidance has since been adapted for aquatic organisms (e.g., Linkov et al. 2002; von Stackelberg et al. 2002, 2005). We defined exposure as the probability of stressor and receptor co-occurrence at any point in time and space, calculated:

  • equation image(1)

where P(Exposure)i,j is the joint probability of stressor-receptor co-occurrence in the ith time step and jth stream reach; P(Receptor)i,j is the independent probability that a receptor will occur in the ith time step and jth stream reach; and, P(Stressor)i,j is the independent probability that a stressor will occur in the ith time step and jth stream reach. Major steps and assumptions associated with quantifying each term in this equation are described below.

Equation (1) is a generalized adaptation of the exposure analysis framework presented by Hope and Peterson (2000), which explicitly considers habitat use proportional to habitat quality, via habitat patches, and spatial distribution of contaminants among habitat patches. Their models were constructed to address terrestrial ecosystems. In our framework, we use stream reaches as the aquatic ecosystem analog to habitat patches. A stream reach generally corresponds to a continuous length of stream with relatively homogenous features of gradient, flow, and morphology. Temporal variability is also accounted for by Hope and Peterson (2000) via a temporal utilization factor that quantifies the relative amount of time a receptor is in a habitat patch. Our framework explicitly accounts for differences in habitat use by time steps in order to account for repeatable patterns of juvenile salmon migration as they move from reaches in which they were spawned to reaches in which they rear. Finally, Hope and Peterson (2000) account for multiple stressors that vary spatially, but they do not vary stressors temporally. Our framework explicitly accounts for the spatial and temporal variability of pesticides.

We expressed P(Exposure)i,j in terms of adult-equivalent (AEQ) juveniles, defined as the number of spawning adults that would naturally result from a given number of fish at an earlier life-stage. We did this in order to standardize the expression of exposure, at any time step, in terms of the future adults. Because fish are potentially exposed to environmental stressors at different times and locations during their freshwater life history and there are differences in expected survival from the point of co-occurrence to the culmination of the life cycle as spawning adults (excluding any effect of agricultural pesticides), we converted the number of fish in any exposure event to their number in terms of AEQ juveniles. Six life-history pathways have been described for spring Chinook in the Willamette Basin (Schroeder et al. 2005, 2007) (see Figure 3). These pathways differ by 2 factors: 1) the timing of emigration from natal reaches to nonnatal reaches; and, 2) the timing of outmigration to the ocean. Because of these differences, each pathway encounters a unique combination of environmental stressors. Survival assumptions used to calculate AEQ juveniles are available online as Supplemental Data.

thumbnail image

Figure 3. Juvenile spring Chinook life-history pathways in the Willamette River Basin, Oregon.

Download figure to PowerPoint

Probability of fish use

Our accounting of the probability that a juvenile spring Chinook salmon will occur in the ith time step and jth stream reach, P(Receptor)i,j, differed between spawning-rearing (sr) and rearing-migration (rm) reaches designated in the Critical Habitat database compiled by NOAA (2005a) for the Willamette Basin. Probabilities within spawning-rearing reaches reflected the relative production of each of the 8 geographically distinct populations we modeled within the Willamette Basin, the different emigration timing within each, and reach-scale differences in habitat quality. Probabilities within rearing-migration reaches reflected the immigration of juveniles from upstream spawning reaches, the different outmigration timing of each distinct population, and reach-scale differences in habitat quality. In total, we accounted for 859 j distinct stream reaches in the NOAA dataset and 42 biweekly i time steps to cover the entire period during which a single spring Chinook brood is expected to emerge from the gravels as fry in late winter to early spring until the last juvenile migrates to sea the following spring.

For spawning-rearing (sr) reaches, or what we also refer to as natal reaches, the independent probability was calculated in terms of adult equivalents (AEQ) as:

  • equation image(2)

where P(AEQsr)i,j is the probability that adult equivalents of juvenile spring Chinook salmon will be present in the ith time step and jth spawning-rearing (natal) stream reach; P(AEQsr)k is the probability that AEQ juveniles will originate in natal reaches in the kth subbasin; P(AEQsr)i|k is the probability that AEQ juveniles will be present in the ith time step given they originated in the kth subbasin; and, P(AEQsr)j|k is the probability that AEQ juveniles will be present in the jth natal stream reach given they were spawned in the kth subbasin.

For rearing-migration (rm) stream reaches, also referred to as nonnatal reaches because no spawning occurs there, the overall probability was calculated as:

  • equation image(3)

where P(AEQrm)i,j is the probability that adult equivalents of juvenile spring Chinook salmon will be present in the ith time step and jth rearing-migration (nonnatal) stream reach; P(AEQrm)i|k is the probability that AEQ juveniles will be present in the ith time step in nonnatal reaches in the kth subbasin; and, P(AEQrm)j|k is the probability that AEQ juveniles will occur in the jth nonnatal stream reach given that they moved into the kth subbasin.

Because natal and nonnatal streams are mutually exclusive, the probability that a juvenile spring Chinook salmon will occur in the ith time step and jth stream reach was calculated simply as the union of Equation (2) and Equation (3):

  • equation image(4)

The combination of Equations (2) and (3) is represented schematically in Figure 4 as are the essential steps to combine the spatial and temporal components within each equation. Further description of these components follows below.

thumbnail image

Figure 4. System of equations to determine the probability of fish use by reach and time-step.

Download figure to PowerPoint

Spatial variability of fish use

We followed models introduced by Sample et al. (1997) and adapted by others (e.g., Hope and Peterson 2000) in assuming that fish will use habitat in proportion to habitat qualities preferred by the fish. In our framework, habitat quality and quantity were calculated in terms of reach-level juvenile carrying capacity. Reach-level carrying capacity was quantified using the Unit Characteristic Method (UCM) (Cramer and Ackerman 2009). The UCM uses established patterns of fish preference for specific habitat features to predict the maximum density that a specific unit of stream will support, given the habitat features that exist in that unit. The supportable density of fish in a habitat unit multiplied by the unit area estimates the carrying capacity for that unit. The UCM uses the classification of channel unit types (e.g., pool, riffle, or glide per Hawkins et al. 1993) as the basal spatial stratum for quantifying rearing capacity. Because the UCM is based on patterns of habitat preference demonstrated by juvenile spring Chinook, we assume that juvenile Chinook will distribute within a stream reach into the available habitat in accordance with their preferences.

A detailed description of the UCM model used in our framework is available online as Supplemental Data. Table 1 summarizes the system of equations used to derive the spatial components within Equations (2) and (3). The building block of this system of equations is the determination of unit-level parr carrying capacity in Equation (5). Carrying capacity was quantified based on habitat conditions measured in mid to late summer when rearing capacity for parr is typically the bottleneck that is limiting freshwater production (Cramer and Ackerman 2009). Our UCM model for spring Chinook calculated carrying capacity as a function of stream area, percent of stream area within each of 3 habitat unit types, standard parr densities for each habitat unit type, and stream width. Standard parr densities were derived from 3 years of observations of spring Chinook parr in the Coldwater River, British Columbia. Spring Chinook parr densities differed by type of habitat unit (pools and log jams; riffles and glides; backwater) and by depth (Table 2). The preference of pool habitat has been observed by several other researchers (Bjornn and Reiser 1991; Mendel et al. 1993; Bumgarner et al. 1994). The Coldwater River data also indicated that average densities in backwater habitat can be relatively high.

Table 1. Summary of the system of equations used to calculate spatial components of the probability of fish use by reach and time-step
Spatial Variability Components
EquationTermsDescription
Equation (5)StreamAreajTotal stream area in the jth reach.
equation imagePercentHabitatUnitlPercent of stream area in the lth habitat unit (pools and log jams; riffles and glides; or backwater), where L = 3, within the jth reach.

Reach-level juvenile spring Chinook carrying capacity calculated using the Unit Characteristic Method (UCM) (Cramer and Ackerman 2009)

.
  
 DensitylStandard spring Chinook parr densities in the lth habitat unit (pools and log jams; riffles and glides; and, backwater), where L = 3, within the jth reach.
 WidthScalarjPercent decrement in juvenile spring Chinook carrying capacity in non-backwater habitat units in the jth reach (see Equation 6).
Equation (6)StreamWidthjTotal stream width in the jth reach.
WidthScalarj = 10.111 * StreamWidthj−0.755  
A scalar to discount the midsection area of large channels that are not used by juvenile spring Chinook salmon based on observations in Allen (2000).  
Equation (7)CCsrj|k|srReach-level carrying capacity of the jth reach (see Equation 5) in the kth subbasin given it is a spawning-rearing (natal) reach.
equation image  
Relative carrying capacity in natal reaches in the kth subbasin compared to total capacity in natal reaches in all M subbasins in the Willamette Basin.  
Equation (8)CCsrj|k|rmReach-level carrying capacity of the jth reach (see Equation 5) in the kth subbasin given it is a rearing-migration (non-natal) reach.
equation image  
Relative carrying capacity in the jth rearing-migration reach in the kth subbasin compared to the total nonnatal carrying capacity in the subbasin.  
Table 2. Parr densities used to calculate juvenile spring Chinook salmon carrying capacity in the Willamette Basin, Oregon
Habitat unit groupDepth class (cm)Parr density (#/100 m2)
  • a

    Deep, slow-water main channel habitat formed by channel scour or behind log debris.

  • b

    Fast-water main channel habitat that is highly turbulent (riffle) or low turbulent (glide).

  • c

    Relatively shallow, slow-water off-channel or side-channel habitat connected to the main channel.

Pools and log jamsa<300
30–600–49.73
> 6023.13
Riffles and glidesb0–100
10–400–16.5
>4025.1
Backwaterc31.5

Existing aquatic habitat inventories were used to quantify aquatic habitat parameters in the UCM model for most reaches. Where data were absent, we used average values from spawning-rearing or rearing-migration reaches. Where the coverage of datasets overlapped, we used the set likely to have measured or estimated the value of interest with greatest accuracy. As first preference, we used direct measurements of width, depth, and percent composition by channel unit type from reaches surveyed by Oregon Department of Fish and Wildlife (ODFW 2004). The ODFW surveyed mostly spawning-rearing reaches, and covered most producing subbasins. As second preference, we used data on stream width and percent composition of habitat unit types from the Ecosystem Diagnosis and Treatment (EDT) process for the Willamette River main stem, the Clackamas River subbasin, and the McKenzie River subbasin (StreamNet 2003). EDT data were developed by local Subbasin Planning teams to spatially define habitat units and reach attributes in support of planning analyses as mandated by the Northwest Power and Conservation Council. In reaches not covered by the EDT and/or ODFW data sets, we used predicted width and depth from habitat suitability modeling by the Pacific Northwest Ecosystem Research Consortium (PNWERC 2002). PNWERC models covered most wadeable streams within the Willamette Basin.

Temporal variability of fish use

In our framework, migration patterns were determined for each life-history strategy employed by each geographically distinct population we modeled within the Willamette Basin. For each population, we accounted for the proportion moving from and staying in natal and nonnatal reaches at each biweekly time step. These proportions were approximated based on juvenile and adult monitoring conducted in the Willamette Basin. A detailed description of the major assumptions underlying our migration model is available online as Supplemental Data. Table 3 summarizes the system of equations used to derive the temporal components within Equations (2) and (3). Terms in these equations account for: 1) emergence of fry from gravel beds within spawning-rearing (natal) reaches; 2) emigration from natal reaches to downstream rearing-migration (nonnatal) reaches; and, 3) outmigration to sea from natal and nonnatal reaches. In these equations, we assumed that juvenile spring Chinook moved into nonnatal streams within their natal subbasin and to downstream nonnatal reaches in the Willamette River. We also assumed that emigration and outmigration was “instantaneous”, i.e., movement would occur within a time step; we did not account for any lag.

Table 3. Summary of the system of equations used to calculate temporal components of the probability of fish use by reach and time-step
Temporal Variability Components
EquationTermsDescription
Equation (9)P(AEQsr)n|kThe probability of AEQ juveniles produced by the nth life history pathway in the kth subbasin.
equation imageP(AEQsr.emerge)i|n|kThe cumulative proportion of adult-equivalent fry that emerged through the ith time step and followed the nth life-history pathway in the kth subbasin.
The proportion of AEQ juveniles that will be present in the ith time step in a spawning-rearing (natal) reach of the kth subbasin.P(AEQsr.disperse)i|n|kThe cumulative proportion of adult-equivalent fry that dispersed from the kth subbasin through the ith time step via the nth life-history pathway.
Equation (10)P(AEQsr.disperse)i|n|kThe cumulative proportion of adult-equivalent fry that dispersed from the kth subbasin through the ith time step via the nth life-history pathway.
equation image  
The proportion of AEQ juveniles that will be present in the ith time step in a rearing-migration (non-natal) reach of the kth subbasin.P(AEQrm.outmig)i|n|kThe cumulative proportion of AEQ juveniles outmigrating from the kth subbasin through the ith time step from the nth life history pathway.

Our model framework was able to account for different life-history strategies (see Figure 3). Of the 8 populations we modeled, 4 had sufficient data to characterize the proportion exhibiting “stream” type (outmigrating after overwintering) versus “ocean” type behavior (outmigration to sea in their first year of life) (Table 4). Where we did not have sufficient data, we assigned proportions of either stream or ocean type based on proportions in 1 of the 4 streams sampled that was most similar in flow regime and subbasin physiographic setting. Other researchers (e.g., Beechie et al. 2005) have observed that flow regimes and physiographic setting are correlated with the predominance of either stream type or ocean type life histories in a given stream. We found that there were moderate to high proportions AEQ juveniles with stream type life histories in the Willamette Basin. The relatively low levels of fry emigrants rearing in nonnatal reaches during the summer in the Willamette Basin have been observed by several other researchers, too (e.g., Hughes and Gammon 1987; Waite and Carpenter 2000; Landers et al. 2002; Friesen et al. 2007).

Table 4. Summary of juvenile spring Chinook salmon emigration and outmigration observed in selected subbasins in the Willamette Basin, Oregon
SubbasinEmigration from natal streamsOutmigration to the ocean (% AEQ juveniles/peak timing)
Emigrant TypePeak TimingSpring subyearlingFall subyearlingSpring yearling
Clackamas RiverFryJan 1–150.1%0.0%0.0%
Jun 1–15
ParrNov 16–3017.6%9.3%
Dec 1–15Mar 16–31
SmoltMay 1–1573.0%
May 16–31
No. Santiam RiverFry33.9%0.0%7.7%
Parr1.7%35.1%
Smolt21.6%
So. Santiam RiverFry59.4%0.0%13.4%
Parr0.8%16.3%
Smolt10.0%
McKenzie RiverFryFeb 1–1510.2%0.6%8.9%
Jun 1–15Dec 1–15Mar 1–15
ParrNov 1–152.4%48.3%
Nov 16–30Mar 16–31
SmoltMar 1–1529.7%
Apr 16–31

Our framework also accounted for differences in migration timing. In the Willamette Basin, fry generally emerge from February to March, although sometimes as late as June. Juveniles emigrate continuously out of natal reaches and into lower tributaries and into the main-stem Willamette River as fry (late winter to early spring), parr (fall to early winter), and yearlings (late winter to spring). Of the 8 populations we modeled, only 2 had sufficient data to characterize migration timing with biweekly precision (see Table 4). Recognizing that water operations in the Clackamas River system could account for unique timing sequences (personal communication Nick Ackerman, Portland General Electric, Estacada, Oregon), and absent any other information from other subbasins, we assumed that peak migration timing in the McKenzie River system was representative of other watersheds. We assumed a normal distribution about the peaks that encompassed 1 month either side of the peak. We could expect slight variability in timing given slight differences in flow and temperature regimes. However, the resulting high proportion of AEQ juveniles calculated to outmigrate during late winter and spring was consistent with observations in the Portland harbor area, which indicates that peak spring Chinook densities occurred from February through May (Friesen et al. 2007). Similar timing has been described in other systems (e.g., Beechie et al. 2005). Overall, the migration timing we employed represented an average; actual timing would be expected to vary from year to year.

Probability of pesticide occurrence

We accounted for the probability that agricultural pesticides will occur in the ith time step and jth stream reach, P(Stressor)i,j, in order to account for the spatial and temporal distribution of pesticide use. Our investigation was focused on use of the AChE-inhibiting insecticides carbaryl, carbofuran, chloropyrifos, diazinon, malathion, and methomyl, subjects of 2 recent Biological Opinions by NMFS (2008, 2009). For demonstration purposes, a contaminant profile for the 6 insecticides was developed using USGS NAWQA monitoring data from the monitoring station with the longest period of record and greatest number of samples collected (Zollner Creek, described below). This provided a basis for assigning the temporal pattern of concentrations within the basin. Because agricultural land was unevenly distributed in the basin, the actual concentration in a given reach was scaled by the percent of agricultural land within a 300-m buffer relative to that for Zollner Creek, as determined from the National Land Cover Database (NLCD) (Homer et al. 2004). The Biological Opinions by NMFS (2008, 2009) indicated that aerial application within this buffer width could result in higher pesticide concentrations caused by spray drift into adjacent water bodies. If the relative percent of agricultural lands along any reach was greater than or less than that occurring along Zollner Creek, we adjusted the reference concentrations proportionally.

The USGS NAWQA monitoring station with the longest period of record in the Willlamette River Basin is on Zollner Creek within the Mollala-Pudding subbasin (shown in Figure 1). Most agricultural activity in the Basin is concentrated in the low-gradient valley floor, and Zollner Creek (Station 14201300, Zollner Creek, Near Mt. Angel, OR; at 45.1° N, -122.8° W) is the NAWQA station with the highest concentration of agriculture lands in its adjacent and upstream vicinity. Samples were collected from 1993 to 2008. Total number of observations was 174, with a range of 4 to 13 years of observations in any given month. Sampling frequency ranged from around semimonthly to 1.5 times per month over the entire period of record. All reported surface water values were used, and where results from 2 analytical methods were reported for the same sample, the values were averaged. This was the case for many of the CAR and CFN values. Figure 5 summarizes the monthly average concentrations for the 6 compounds. Three of the chemicals were found in 50% of the samples, 4 in 10%, and 5 in 1%. All 6 were never found in combination in any 1 sample.

thumbnail image

Figure 5. Average monthly concentrations (bold line), with the upper 95% tolerance limits for the 50th (solid line), 75th (dashed line), and 95th (dotted line) percentiles, of 6 AChE-inhibiting pesticides in samples collected at Zollner Creek, Oregon, between 1993–2008.

Download figure to PowerPoint

The recent Biological Opinions by NMFS (2008, 2009) stated that “salmonids rely upon a variety of non-main channel habitats that would be expected to yield higher pesticide concentrations…” Zollner Creek is a small agriculturally dominated stream and most closely resembles the off-channel habitat type of side channel, where stream geometry and flow rates are considerably smaller than in the associated main channel in a reach, but constant flow occurs. Measurements of daily discharge taken for the period of record of the chemical monitoring give a median flow of 0.13 m3 s−1 and a 90th percentile flow of 1.7 m3 s−1 (USGS 2010). In a recent article describing trends in pesticide concentrations in western US streams, Zollner Creek is shown to have one of the lowest mean daily stream discharges of all sampled water bodies (Johnson et al., 2010). Although not representative of every off-channel habitat type, Zollner Creek is probably the best available surrogate for our demonstration. The contaminant profile from this monitoring station was applied to all reaches occurring along agricultural lands.

The probability that agricultural pesticides will occur in the ith time step and jth stream reach, P(Stressor)i,j, was evaluated as a binomial (0,1) probability relative to several concentration thresholds that we selected for demonstration purposes (described above). We used several statistics derived from the monitoring data at Zollner Creek to further demonstrate the sensitivity of the model to different characterizations of contaminant profiles. Monthly averages were calculated using the log-transformed values of the observed concentrations and scaled by the percent of agricultural land within a 300-m buffer (as described above). In addition to the monthly averages, we also estimated co-occurrence at several tolerance limits (see Figure 5): the upper 95% confidence limit (UCL) of the 50th percentile (i.e., median); the 95% UCL of the 75th percentile; and the 95% UCL of the 95th percentile. These tolerance limits represent progressively infrequent occurrences of pesticide concentrations within each month. If the monthly average or tolerance limit was above the threshold value, then P(Stressor)i,j was set to 1; otherwise, the probability that agricultural pesticides would occur was set to 0.

Probability of exposure

Our demonstration focused on the co-occurrence of juvenile Chinook AEQs and pesticide concentrations exceeding demonstration thresholds in backwater, off-channel habitats occurring along agricultural lands. The recent Biological Opinions by NMFS (2008, 2009) stated that “salmonids rely upon a variety of non-main channel habitats that would be expected to yield higher pesticide concentrations…” These included alcoves, channel edge sloughs, overflow channels, backwaters, terrace tributaries, off-channel ponds, and braids. Our best representation of these habitats was provided by the EDT and ODFW aquatic habitat inventories, which record these as backwater, alcoves, backwater pools, and isolated pools. These data sets report that the average stream area covered by backwater, off-channel habitat was about 1.4% in spawning-rearing reaches and 4.5% in rearing-migration reaches. This is consistent with the extent of backwater habitats mapped by PNWERC (2002). Because of the preference of parr for backwater habitats (Table 2) and our discounting of the midsection area of large channels that are not used by juveniles, the relative carrying capacity is higher. Therefore, our first step in determining exposure was to scale the joint probability of stressor-receptor co-occurrence in the ith time step and jth stream reach, P(Exposure)i,j, by the proportion of total carrying capacity in each jth stream reach in backwater habitat unit (i.e., l = backwater) that was within 300 m of agricultural lands. Agricultural land was determined through spatial analysis of the NLCD (Homer et al. 2004) where agricultural land was Classes 81 and 82.

To interpret the probability of exposure across time steps, we then assumed that once an individual fish dispersed into an individual habitat unit within an individual reach, it remained there until predicted to move in a future time step. That is, as fish disperse and occupy habitat proportional to habitat quality, they stay in a specific reach and channel unit until the next redistribution event. Several lines of evidence support this assumption. First, the consistent patterns of juvenile migration timing, as determined from trapping of migrants, shows there is little movement of fish during summer, but substantial movements during spring and fall (Healey 1991, Quinn 2005). Second, juvenile salmonids captured in backwater habitats often show differences in growth rate and size from fish rearing in the main stream channel, which indicates these fish must have remained in these environments long enough to display the differences in growth (Sommer et al. 2001).

Given this assumption, calculation of the total co-occurrence over all time steps is a 2-step process. Within each of 3 redistribution events—fry emigration, fall emigration, and spring outmigration—we first determined the maximum co-occurrence of AEQ juveniles with pesticide concentrations exceeding 1 of several evaluation thresholds (described above). Because we assumed that juvenile spring Chinook salmon stay in a specific reach and channel unit until the next redistribution event, this maximum reflects the total AEQ juveniles moving into backwater habitat along reaches where the pesticide concentrations could exceed threshold amounts, either chronically or episodically. With each redistribution event, we assume that a new set of AEQ juveniles move into backwater habitats. In the fall redistribution event, this is likely a conservative assumption because a high proportion of AEQ juveniles that emigrated as fry to nonnatal streams will overwinter there (and not be replaced) rather than outmigrate to the sea (Table 4). Then, we sum the maxima from each of the 3 redistribution events. We are essentially assuming 3 independent exposure opportunities. This is likely not the case and leads to some double-counting. Given the low level of co-occurrence we found (below), this was not an immediate concern. However, further refinement of this calculation would be warranted if exposure probabilities were much higher.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

The fraction of AEQ juveniles from a single brood year that is present in the entire Willamette Basin at any time step is summarized in Figure 6. Generally, the number of AEQ juveniles begins to increase in mid-November as fry emerge from the gravel, and by mid-February, 100% of future adults have emerged. About 70% “stay” in natal reaches to rear while the other 30% “move.” Most fry emigrants subsequently outmigrate to the ocean as subyearlings and only about 10% of all AEQ juveniles rear in nonnatal reaches during the summer (mid-June to end September). Summer rearing in nonnatal reaches occurs either within the natal subbasin or downstream in the Willamette River. In the fall, over half of the AEQ juveniles remaining in natal reaches emigrate; most of these find overwintering habitat in nonnatal reaches. Only about 10% of all AEQ juveniles migrate to the ocean in the fall. Most AEQ juveniles outmigrate as yearlings in the following spring. The timing of departure is extended over 3 months; the central 50% occurs between April 15 and May 15. Spring outmigration from natal reaches occurs slightly earlier than from nonnatal reaches. The entire brood has left by mid-July.

thumbnail image

Figure 6. The proportion of AEQ juveniles, from a single brood, predicted to occur in total (solid line), natal (dashed), and nonnatal (dotted) reaches in the Willamette River Basin, Oregon.

Download figure to PowerPoint

The probability of exposure to pesticides is summarized by time step in Figures 7 and 8. Figure 7 depicts the reduction in exposure through limiting the spatial extent of potential co-occurrence. The coarsest resolution, if we had not accounted for spatial and temporal distinctions, would have employed the simplifying assumption that 100% of the population is exposed; this is the fraction of AEQ juveniles from a single brood year that are present in the entire Willamette Basin throughout the brood's 18-month span of freshwater rearing. Accounting for the spatial distribution of juvenile salmon and agricultural lands indicated that less than 20% of all AEQ juveniles produced in the basin rear in reaches within 300 m of agricultural land cover during any time step. When considering distribution of fish among backwater habitat units, less than 5% were estimated to rear in backwater habitat along agricultural lands at any one time. When we summed the maximum probable occurrence of fish use associated with each redistribution event—fry emigration, fall emigration, and spring outmigration—we estimated that up to about 13% of all AEQ juveniles produced in the Willamette River Basin reared in agricultural backwater habitats.

thumbnail image

Figure 7. The proportion of total AEQ juveniles (solid line), from a single brood, predicted to co-occur within 300 m of agricultural lands (dashed) and in backwater habitat units within 300 m of agricultural lands (dotted) in the Willamette River Basin, Oregon.

Download figure to PowerPoint

thumbnail image

Figure 8. The proportion of total AEQ juveniles, from a single brood, in backwater habitat units within 300 m of agricultural lands predicted to co-occur with average concentrations of 6 AChE-inhibiting pesticides exceeding 0.1 µg/L (solid line) and 0.5 µg/L (dashed line) in the Willamette River Basin, Oregon.

Download figure to PowerPoint

The probability of exposure to pesticides is reduced when contaminant thresholds are considered. Figure 8 depicts differences in exposure using thresholds of 0.1 µg/L and 0.5 µg/L. Nearly all AEQ juveniles rearing in backwater habitat along agricultural lands would co-occur with average pesticide concentrations exceeding the 0.1 µg/L threshold. However, only about 2% would co-occur with concentrations exceeding 0.5 µg/L. At this threshold, co-occurrence was greatest in the late fall and winter during periods of greater precipitation and runoff, which coincide with peak movement of juvenile spring Chinook to overwintering habitat in nonnatal stream reaches. No fish co-occurred with average pesticide concentrations exceeding 1.0 µg/L; this is because monthly average concentrations of the 6 AChE-inhibiting chemicals never exceeded this threshold. When we considered statistics other than the mean, we found that exposure levels at this upper threshold increased. Using the upper 95% confidence limit (UCL) of the 50th percentile (i.e., median), we found that estimated exposure increased slightly (<<1%); exposure at the 95% UCL of the 75th percentile was estimated to be about 4%; exposure at the 95% UCL of the 95th percentile was about 12%. Overall, this demonstrates that exposure occurs above this threshold level (1.0 µg/L); however, it is infrequent (<5%). Given the episodic nature of such concentrations, the probability would likely be lower.

Sensitivity analysis

Overall, we recognized there is a series of assumptions underlying the model calculations and that it is necessary to understand the importance of these assumptions on model outcomes. We conducted local sensitivity analysis to evaluate the influence of several key variables on exposure estimates. In constructing this framework, it was readily apparent that 3 factors had the potential to greatly influence exposure estimates. Foremost was the relatively low co-occurrence of spawning-rearing habitat with agricultural lands. Graphically, this is evident when comparing Figures 1 and 2, and it accounts for a great separation of fish use and pesticide use. Also seemingly important was the characterization of relatively high proportions of stream type spring Chinook salmon. Stream type Chinook are produced predominantly in natal streams, and these streams were seldom within 300 m of agricultural lands (Figure 2). Finally, we recognized that the proportion of carrying capacity within backwater habitat in nonnatal streams could influence the intersection of fish use with elevated pesticide concentrations when spring Chinook were present in nonnatal reaches. Only about 5% of stream area in nonnatal streams was assumed to be backwater habitat.

Of the 3 parameters we evaluated, the extent of natal reaches (spawning habitat), and the proportion of stream area in backwater habitat had the greatest influence on exposure estimates. A 35% increase in the downstream extent of spawning habitat (to about 1538 k) resulted in the percentage of fish co-occurring with monthly average concentrations > 0.5 µg/L increasing to about 3% (∼50% increase). A 20% increase in backwater habitat (to about 6% of stream area) led to an estimated co-occurrence of 2.4% (∼20% increase). Both of these results are plausible. The downstream extension of spawning habitat has the effect of increasing the co-occurrence of juvenile stream type salmon with agricultural lands. The increase in backwater habitat has the effect of increasing the co-occurrence of juvenile AEQs with low-velocity habitats where higher pesticide concentrations might be expected. There was no change in exposure when we varied the percentage of fry emigrating by 10%. This is plausible given that monitoring data indicated that most fry emigrants also migrate to sea in the spring as subyearlings, thus largely avoiding exposure in backwater habitats.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

Our results demonstrate that by accounting for spatial and temporal variation in fish use and pesticide concentrations, the co-occurrence of fish and pesticides is far less than might be inferred simply by accounting for their presence in a river basin. Our model framework made it possible to integrate readily available information to improve this accounting. Probability of fish use could be predicted by integrating information on the distinction between natal and nonnatal reaches, timing of fish use, and habitat quality. Similarly, the probability of pesticide occurrence could be distinguished by integrating information on land use, habitat unit types, and water quality. Land use interpreted from the NLCD database (Homer et al. 2004) indicated that only about 20% of spring Chinook critical habitat was within 300 m of agricultural land cover. Most of this was along nonnatal streams and of this a small fraction of the stream area occurred in backwater habitat units. Finally, given the relatively high proportion of stream type spring Chinook and their migratory behavior, a large fraction of AEQ juveniles avoid these reaches altogether. Co-occurrence with pesticides is further reduced when considering the concentrations and frequency at which they were detected. Overall, accounting for known spatial and temporal variation in fish use and pesticide occurrence appears to lead to substantially lower estimates of exposure than the assumption of universal co-occurrence.

In our demonstration, emphasis was placed on quantifying the timing and patterns of fish use. Several aspects of the framework and data available to inform it facilitated our efforts to achieve this objective. The classification of reaches into either spawning-rearing or rearing-migration uses offered a ready and reliable means to distinguish where spawning occurs within the basin. Such classifications are compiled by NMFS in the critical habitat databases for each ESU of salmonids listed as threatened or endangered under the ESA. When considered in combination with available data on migration timing at different life stages, we were able to account for proportions that “stay” in natal (spawning) reaches and those that “move” to nonnatal (rearing) reaches. Whereas migration data were not complete and several assumptions were needed to extrapolate patterns across the entire Willamette Basin, reliable observations were available for subbasins producing at least two-thirds of spring Chinook salmon. Such data are collected in most ESUs. Overall, we were able to calculate the probability that juveniles representing future adult salmon will be present in a given reach in a given time step, P(Receptor)i,j. In turn, we were able to demonstrate, with reasonable confidence, that this distribution is not uniform over time and space. This ability to account for the unequal distribution of juvenile production across time and space was meaningful in estimating co-occurrence.

Plausible factors for the unequal distribution of AEQ juveniles in the Willamette Basin are the extent of spawning habitat and the timing of use of nonnatal habitat. Gradient and elevation are low in the mainstem Willamette River and lower tributaries, substantially less than in spawning reaches. This results in differences in stream morphology, and a lack of suitable spawning habitat (Hughes and Gammon 1987). The relatively low levels of fry emigrants rearing in nonnatal reaches during the summer in the Willamette Basin has been observed by several researchers (e.g., Hughes and Gammon 1987, Waite and Carpenter 2000, Landers et al. 2002, Friesen et al. 2007). A probable cause for low use of nonnatal reaches during the summer is elevated stream temperatures. Nearly all reaches not designated for spawning of spring Chinook in the Willamette Basin (i.e., rearing-migration reaches) are listed as thermally impaired on the 303(d) list (ODEQ 2006). Summer temperatures in these low-elevation streams exceed 18°C; thermal stress to Chinook salmon occurs above this level (USEPA 2004). Temperature is also correlated to channel width, which results in a decreasing percentage of the stream providing the edge habitat that juvenile salmon prefer. Our findings are congruent with these field observations.

From a modeling perspective, our use of AEQ juveniles was a novel solution that enabled us to simplify our quantitative framework. From a practical standpoint, expression of juvenile abundance at each stage in terms of AEQ juveniles made it possible to easily roll up effects at different life stages into a common scale. This scale also had the added benefit of allowing us to evaluate co-occurrence and consequences in terms of future spawner abundance, a key metric in conservation planning. Overall, the use of AEQ juveniles made it possible to integrate outcomes across subbasins to predict exposure to future spawners at the ESU scale. It also made it possible to integrate outcomes across time steps to predict exposure at the brood scale. Whereas this metric provides convenience for estimating exposure on a common, meaningful scale, it does not preclude interpretation of co-occurrence by life stages. This life stage context is valuable, because effects on fry can be different than on parr or yearling smolts. In the case of quantifying population co-occurrence with pesticide, we were able to use the framework to differentiate exposure that occurred to fry emigrants and to fall emigrants. This differentiation could enable further accounting for differences in pesticide effects on different life stages of fish.

Compared to other spatially and temporally explicit models, Wickwire et al. (2011) identifies several favorable qualities which our framework seems to embrace. Most notably, our framework accounts for variability in habitat, integrates actual fish behaviors, avoids use of site-wide average-based estimates, and considers exposure at an ecologically relevant scale (i.e., the ESU scale). Despite this, Wickwire et al. (2011) also identify several impediments to the use of such models, which are likely to be experienced with our model. Foremost, there is little to no precedent for the use of a framework such as ours in modeling the exposure of salmon to pesticides. There are also likely to be misperceptions as to the purpose of this model. Some, as Wickwire et al. (2011) points out, may incorrectly perceive such models as a means to “dilute” exposure estimates. This is not the case with our modeling framework; rather, we formulated this model to provide a better understanding of the proportion of organisms potentially impacted by agricultural pesticides. Finally, Wickwire et al. (2011) note that technical concerns can be an impediment to use. To the extent possible, this paper discloses the details of our model formulation and demonstrates the sensitivity of the model to key assumptions underlying it. This is an important step in gaining acceptance and use, but further technical transfer would be of value.

We recognize that another potential impediment to use of this model is the series of assumptions, and underlying uncertainties, used in its construction. We identified several key uncertainties underlying our characterization of the probability of fish use and reported on a sensitivity analysis to gauge the relative importance of our assumptions. In the Willamette Basin, the downstream extent of spawning habitat, the amount of backwater habitat along agricultural lands, and the proportion of fish choosing to rear in nonnatal streams are critical. The values we chose for simulation are central tendencies based on the best available information from the Willamette Basin. We could expect these parameters to vary over time; however, there are limitations to the amount by which they would vary. Extent of spawning habitat, for instance, is limited by the availability of suitable substrate and cool temperatures, which generally decrease from the headwaters to the mouth (Vanotte et al. 1980). Generally, our model predictions of the spatial and temporal distribution of juvenile salmon production throughout the stream network were congruent with results of independent sampling of juvenile fishes in the basin. A detailed discussion of this is available online as Supplemental Data. Overall, we improve upon the simplifying assumption that 100% of the population could co-occur with pesticides.

Regarding pesticide concentrations, we recognize key uncertainties underlying our assumptions, too. Pesticide use will vary among crops and monitoring programs may not detect peaks in pesticide concentrations. To address this, we chose the best available data to represent high concentrations in the Willamette Basin. The median velocity recorded at Zollner Creek is 0.40 ft sec−1; the 90th percentile is 1.12 ft sec−1. While free-flowing, this is comparable to velocities in backwater, off-channel habitat evaluated in Biological Opinions by NMFS (2008, 2009). Zollner Creek is surrounded by diverse crop systems; it ranks at the 90th percentile in terms of percentage of agricultural land use within 300 m of a stream. This station has been monitored frequently for over a decade. To account for uncertainty, we conducted simulations using the 95% UCL of the 95th percentile. Pesticide concentrations were characterized on a monthly time-step and individual fish residence in a backwater unit could span multiple months; this increased the chances of an individual fish co-occurring with high pesticide concentrations. Even though this is the best available data and we approach this simulation conservatively, we could improve predictions by integrating intensive monitoring, throughout the basin, especially of backwater, off-channel habitats. Our framework is set up to accommodate such a data set.

CONCLUSIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

Accounting for the detailed aspects of exposure (presence of fish and of the pesticide stressor) in space and time within a Pacific salmonid ESU resulted in a refined exposure analysis that provided a realistic description of the co-occurrence of juvenile spring Chinook salmon with varying concentrations of 6 AChE-inhibiting insecticides. We found it possible to inform model parameters using readily available information from the Willamette River Basin that would be generally available in other ESUs. We found the model responsive to patterns and relationships in the literature and conforming to expected spatial and temporal trends in the Willamette River Basin. Using available data, the model indicated that the average mixture of 6 AChE-inhibiting insecticides likely co-occur with juvenile spring Chinook salmon less than commonly assumed. The major factor accounting for this finding is that a relatively small percentage of juvenile salmon were found to rear in off-channel, backwater habitats that occur along agricultural lands where pesticide concentrations are of greatest concern. Pesticide concentrations within agricultural backwater habitats were important, but only affected juvenile salmon when they were found to occur in those habitats. Applied in other ESUs, our findings suggest that it is important to account for the extent of spawning and rearing habitat, the extent of backwater habitat along agricultural lands, life-history strategies employed, and spatiotemporal variability in pesticide use. Overall, such spatially and temporally explicit understanding of co-occurrence improves risk management. Applied in an adaptive management framework, testable hypotheses can be formulated from this model to validate outcomes, address key uncertainties, and ultimately improve predictions through field study.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

Funding for this project was provided by Dow AgroSciences LLC, DuPont Crop Protection, and Bayer CropScience. We would like to thank Mark Schocken, Schocken Consulting, and Matt Kern, Bayer CropScience, for their detailed review. We would also like to thank Larry Dominguez and Phil Gaskill, Cramer Fish Sciences, and Randy Ericksen, the Wild Salmon Center, for their advice in preparing this manuscript. Finally, we would like to thank 2 anonymous reviewers for their helpful feedback and suggestions on our manuscript.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information
  • Baldwin D, Spromberg J, Collier T, Scholz N. 2009. A fish of many scales: Extrapolating sublethal pesticide exposures to the productivity of wild salmon populations. Ecol Appl 19: 20042015.
  • Beechie TJ, Liermann M, Beamer EM, Henderson R. 2005. A classification of habitat types in a large river and their use by juvenile salmonids. Transact Am Fish Soc 134: 717729.
  • Bjornn TC, Reiser DW. 1991. Habitat requirements of Salmonids in streams. Am Fish Soc (Special Publication) 19: 83138.
  • Bumgarner J, Mendel G, Ross L, Milks D, Dedloff J. 1994. Tucannon River spring Chinook salmon hatchery evaluation program. 1993 Annual Report. Washington DC: Washington Department of Fish and Wildlife AFF 1/LSR-94-09, submitted to US. Fish and Wildlife Service, Boise, Idaho.
  • Cramer SP, Ackerman NK. 2009. Linking stream carrying capacity for salmonids to habitat features. In: Knudsen EE, Michael JH Jr, editors. Pacific salmon environmental and life history models: Advancing science for sustainable salmon in the future. Bethesda (MD): American Fisheries Society. Symposium 71. p 223254.
  • Friesen TA, Vile JS, Pribyl AL. 2007. Outmigration of juvenile Chinook salmon in the Lower Willamette River, Oregon. Northwest Sci 81: 173190.
  • Giesy JP, Solomon KR, Coates JR, Dixon KR, Giddings JM, Kenega EE. 1999. Chlorpyrifos: ecological risk assessment in North American aquatic environments. Rev Environ Contam Toxicol 160: 1129.
  • Hawkins CP, Hawkins CP, Kershner JL, Bisson PA, Bryant MD, Decker LM, Gregory SV, McCullough DA, Overton CK, Reeves GH, Steedman RJ. 1993. A hierarchical approach to classifying stream habitat features. Fisheries 18: 312.
  • Healey MC. 1991. Life history of Chinook salmon (Onchorhynchus tshawytsca). In: Groot C, Margolois L, editors. Pacific salmon life histories. Vancouver (BC): UBC Press. p. 313393.
  • Homer C, Huang C, Yang L, Wylie B, Coan M. 2004. Development of a 2001 National Landcover Database for the United States. Photogrammetr Eng Remote Sens 70: 829840.
  • Hope BK, Peterson JA. 2000. A procedure for performing population-level ecological risk assessments. Environ Manag 25: 281289.
  • Hughes RM, Gammon JR. 1987. Longitudinal changes in fish assemblages and water quality in the Willamette River, Oregon. Transact Am Fish Soc 116: 196209.
  • Johnson HM, Domalgalski JL, Saleh DK. 2010. Trends in pesticide concentrations in streams of the western United States, 1993–2005. JAWRA 122.
  • Landers D, Fernald A, Andrus C. 2002. Off channel habitats. In: Hulse D, Gregory S, Baker J, editors. Willamette River Basin Atlas: Trajectories of environmental and ecological change. Corvallis (OR): Oregon State Univ Press. 192 p. [cited 2011 November 29]. Available from: http://www.fsl.orst.edu/pnwerc/wrb/Atlas_web_compressed/PDFtoc.html
  • Linkov I, Burmistrov D, Cura J, Bridges TS. 2002. Risk based management of contaminated sediments: consideration of spatial and temporal patterns in exposure modeling. Environ Sci Technol 36: 238246.
  • Mendel G, Bumgarner J, Peterson K, Bugert R, Ross L, Milks D, Dedloff J, Shaklee JB, Knutson C. 1993. Tucannon River spring Chinook salmon hatchery evaluation program. 1992 Annual Report. Washington DC: Washington Department of Fish and Wildlife AFF 1/LSR-93-06, submitted to US Fish and Wildlife Service, Boise, Idaho.
  • [NMFS] National Marine Fisheries Service. 2008. National Marine Fisheries Service Endangered Species Act Section 7 Consultation, Bological Opinion, Environmental Protection Agency Registration of Pesticides Containing Cchlorpyrifos, Diazinon, and Malathion. [cited 2009 21 December]. Available from: www.nmfs.noaa.gov/pr/pdfs/pesticide_biop.pdf
  • [NMFS] National Marine Fisheries Service. 2009. National Marine Fisheries Service Endangered Species Act Section 7 Consultation, Biological Opinion, Environmental Protection Agency Registration of Pesticides Containing Carbaryl, Carbofuran, and Methomyl. [cited 2009 December 21]. Available from: www.nmfs.noaa.gov/pr/pdfs/carbamate.pdf
  • [NOAA] National Oceanic and Atmospheric Administration. 2005a. Habitat areas for the Upper Willamette River Chinook salmon ESU. GIS database downloaded June 2009. [cited 2009 June 1]. Available from: http://www.nwr.noaa.gov/Salmon-Habitat/Critical-Habitat/CH-GIS-Data.cfm
  • [NOAA] National Oceanic and Atmospheric Administration. 2005b. Endangered and threatened species; designation of critical habitat for 12 evolutionarily significant units of West Coast Salmon and Steelhead in Washington, Oregon, and Idaho; Final Rule. Federal Register 70(170): 5263052678.
  • [ODEQ] Oregon Department of Environmental Quality. 2006. Oregon's 2004/2006 integrated water quality assessment report. GIS database and report. [cited 2009 June 1]. Available from: http://www.deq.state.or.us/wq/assessment/rpt0406.htm
  • [ODFW] Oregon Department of Fish and Wildlife. 2004. Aquatic habitat inventories project. Habitat and reach data. GIS database. [cited 2009 June 1]. Available from: http://oregonstate.edu/dept/ODFW/freshwater/inventory/index.htm
  • [PNWERC] Pacific Northwest Ecosystem Research Consortium. 2002. Stream data sets developed for the Willamette Basin alternative futures project. GIS database. [cited 2009 June 1]. Available from: http://www.fsl.orst.edu/pnwerc/wrb/access.html
  • Poletika NN, Woodburn KB, Henry KS. 2002. An ecological risk assessment for Chlorpyrifos in an agriculturally dominated tributary of the San Joaquin River. Risk Anal 22: 291308.
  • Quinn TP. 2005. The behavior and ecology of Pacific salmon and trout. Seattle (WA): Univ Washington Press.
  • Ruckelshaus MH, Levin P, Johnson JB, Kareiva PM. 2002. The Pacific salmon wars: What science brings to the challenge of recovering species. Annu Rev Ecol Systemat 33: 665706.
  • Sample BE, Asplin MS, Efroymson RA, Sutter GW II, Welsh CJE. 1997. Methods and tools for estimation of the exposure of terrestrial wildlife to contaminants. Oak Ridge (TN): Oak Ridge National Laboratory. ORNL/TM-13391. October.
  • Schroeder RK, Kenaston KR, Krentz LK. 2005. Spring Chinook salmon in the Willamette and Sandy Rivers, October 2004 through September 2005. Corvalis (OR): Oregon Department of Fish and Wildlife. F-163-R-10.
  • Schroeder RK, Kenaston KR, McLaughlin LK. 2007. Spring Chinook salmon in the Willamette and Sandy Rivers, October 2005 through September 2007. Corvalis (OR): Oregon Department of Fish and Wildlife. F-163-R-11/12.
  • Sommer TR, Nobriga ML, Harrell WC, Batham W, Kimmerer WJ. 2001. Floodplain rearing of juvenile Chinook salmon: Evidence of enhanced growth and survival. Can J Fish Aquat Sci 58: 325333.
  • StreamNet. 2003. Columbia Subbasin Planning EDT reach files and reach attributes. GIS data set. [cited 2009 June 1]. Available from: http://www.streamnet.org/subbasin_planning.html
  • [USEPA] US Environmental Protection Agency. 1998. Guidelines for ecological risk assessment. Office of Research and Development. Washington DC: USEPA. 20460. EPA/630/R-95/002F. April.
  • [USEPA] US Environmental Protection Agency. 2004. Support document for EPA's action reviewing new or revised water quality standards for the State of Oregon. March 2004.
  • [USEPA] US Environmental Protection Agency. 2010. Pesticide Registration Status, Risk Assessment Documents. [cited 2010 November 24]. Available from: www.epa.gov/pesticides/reregistration/status.htm
  • [USGS] US Geological Survey. 2010. USGS National Water Quality Assessment Data Warehouse. [cited 2010 November 28]. Available from: infotrek.er.usgs.gov/apex/f?p=NAWQA:HOME:0
  • Vannote RL, Minshall GW, Cummins KW, Sedell JR, Cushing CE. 1980. The river continuum concept. Can J Fish Aquat Sci 37: 130137.
  • Von Stackelberg K, Burmistrov D, Linkov I, Cura J, Bridges TS. 2002. The use of spatial modeling in an aquatic food web to estimate exposure and risk. Sci Total Environ 288: 97110.
  • Von Stackelberg K, Wickwire WT, Burmistrov D. 2005. Spatially-explicity wildlife exposure modeling tools for use in human health and ecological risk assessment: SEEM and FISHRAND-migration. In: Transactions of the Wessex Institute, Vol 85, Environmental Exposure and Health. 10 p.
  • Waite IR, Carpenter KD. 2000. Associations among fish assemblage structure and environmental variables in Willamette Basin streams, Oregon. Transact Am Fish Soc 129: 754770.
  • Wickwire T, Johnson M, Hope B, Greenberg M. 2011. Spatially explicit ecological exposure models: A rationale for and path toward their increased acceptance and use. Integr Environ Assess Manag 7: 158168.

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

Additional supporting information may be found in the online version of this article.

FilenameFormatSizeDescription
ieam_1266_sm_SupplData.doc62KSupplementary Data

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