Lower Columbia Coho Adult Abundance Monitoring Design
Spatial Design
Temporal Design
Response Design
Inference Design
Evaluation Thresholds
References
Spatial Design
The design is similar to that created for Oregon coast coho (Stevens, 2002). A GIS coverage of streams was used as a frame for the population. The coverage was based on USGS 1:25,000 topographic maps, modified by ODFW to correspond to the target population of streams for each population. The goal was to provide annual estimates for each of the assessment units with a 95% confidence interval of +30%. To do this, a minimum of 30 sites or 30% (whichever comes first) of the available spawning habitat for coho salmon was sampled in each of the assessment units.
Inference is made for Independent Populations. Click here to see the current coho spawner assessment units monitored by.ODFW’s Oregon Adult Salmonid Inventory and Sampling (OASIS) project (ODFW monitors only the portions in Oregon). Generalized Random Tessellation Stratified designs (GRTS) achieve a spatially-balanced random point distribution (Stevens, 1997; Stevens & Olsen, 1999; Stevens & Olsen, 2003; Stevens and Olsen, 2004) and this approach was used.
Temporal Design
A rotating panel design was used to accommodate the need for repeat visits while continuing to expand the scope of the sample every year. Sets of panels are visited on different cycles. The design consists of 40 panels, with one panel defining sites visited every year, 3 panels defining sites visited on a 3-year cycle, 9 panels defining sites visited on a 9-year cycle, and 27 panels defining sites visited on a 27-year cycle. An equal number of sites were allotted to each panel. The three-year cycle parallels the coho life cycle.
Response Design
Spawning ground surveys are conducted from October through January. Visits at a site are conducted at least once every 10 days throughout the course of the spawning survey season. Surveyors either walk upstream or boat downstream, depending of the size of the stream being surveyed. On average, sample sites are roughly 1 mile long. The exact length of samples sites is determined by proximity to landmarks (such as bridge crossings and tributary junctions).The surveyors count redds, live and dead fish (by species), and sample carcasses for gender, length (mm MEPS), scales, fin clips and tags. For a full description of the field methods used to conduct the spawning surveys click here.
Analytical procedures to calculate metrics:
1) Area-Under-the-Curve estimates of site abundance.
Coho salmon can be found spawning in Oregon coastal streams over a period of three or more months. Because it is too costly and impractical to conduct counts every day of the number of adult coho spawning at all randomly selected spawning survey sites during this extended time period, it is necessary to employ statistical techniques to expand periodic counts (i.e. those conducted once every 10 day or less throughout the spawning season) to an estimate of the total number of coho spawning at the site.
The observed numbers of live fish are plotted over time. An additional point with a value of zero is added seven days before or after the first and last observations to close the curve. An integral is used to calculate the area under the curve to estimate fish days. This statistic is then divided by an estimate of the average length of time coho spend on the spawning grounds to estimate the number of fish.
An average spawning life (D) of 11.3 days was used for coho salmon spawning in survey streams (Willis 1952, Beidler and Nickelson 1980, and Perrin and Irvine 1990). Survey data were screened to avoid making spawning density estimates for stream segments where few data points were available or significant portions of the run were missed. These qualification criteria pertained to: (1) the duration of the spawnng season over which counts needed to be made, (2) the number of counts that needed to be conducted for each survey and (3) the number of times that the interval between successive counts could exceed ten days. Additionally, water visibility had to be acceptable (bottoms of riffles were visible) over the majority of the survey area. AUC estimates were not made for surveys that did not meet these criteria. If the first or last count in the index area was greater than zero, a count of zero was assumed to occur seven days before or after the actual count.
2) The percentage of spawning fish that are of natural origin.
During visits to the site surveyors examine carcasses to determine whether they are hatchery-origin fish (adipose fin has been clipped) or if they are natural origin fish. Surveyors also attempt to see marks on live fish and record these as well. If there are insufficient numbers of carcass examinations, live fish observations are added in. The number of marked and unmarked fish is totaled by population to calculate the proportion of total spawning fish that is made up of natural-origin fish.
The adult peak count per mile (Hi) and total number of adult coho salmon per mile (Ni) in a given stream segment are adjusted to eliminate the contribution of hatchery fish using the following equations:
Hi = (Hi)( PSk)
and
Ni = (Ni)( PSk)
where
PSk = estimated proportion of total adult coho salmon spawners in population k that originated from natural production.
Values of PSk were estimated from fin-mark recoveries. Adipose fin-marking occurred for all adult coho production at coastal facilities, thus the ratio of naturally produced coho could be calculated by dividing the number of unmarked coho carcasses by the total number of coho carcasses encountered. Fin-mark ratios were calculated for each population, and data were pooled within each Stratum. Only recoveries on random surveys were used. Values were calculated as follows:
where
CuK = number of unmarked (naturally produced) adult coho carcasses in population K, and
CmK = number of adipose fin-marked ( produced) adult coho carcasses in population K.
The average total fish per mile (T) spawning in a given set of stream segments was calculated as follows:
where
n = number of stream segments surveyed, and
Ni = estimated total number of spawning fish per mile in stream segment i (see below).
The estimated spawning density (total fish per mile) for a given stream segment (Ni) was calculated as follows:
Ni = (Oi)/(mi)
Where
Oi = The total number of adult coho salmon spawning in a given stream segment (Oi) throughout the course of the spawning season and
mj = miles of accessible coho salmon spawning habitat in reach segment j.
where
a = number of periods,
Chi = mean count in period h,
thi = number of days in period h, and
D = average spawning life (days) of coho salmon in survey segments.
Peak count per mile in a given stream segment (Hi) was calculated as follows:
where
Pi = peak count of live and dead fish in stream segment i, and
mj = miles surveyed in stream segment i.
Inference Design
The recommended estimator is the Horvitz-Thompson or π-estimator. Briefly, the estimator weights the observation collected at si by the reciprocal of the inclusion density function π(si). Annual status estimates are obtained from all sites visited in that year. The design selects points on streams. Suppose point si falls on a stream segment with length li. The observation collected represents an aggregate over the entire length of the segment. Thus, if the observation is spawner count, then the entire segment is examined for spawners, and all spawners in the segment are counted. Let yi be the aggregated observation. Then an estimator of the total number of spawners over the entire stream network is
where n is the number of samples. Moreover, within an assessment unit, π(s) is constant, so, letting
, the estimator becomes .
Refer to epa.gov/nheerl/arm for R scripts (spsurvey library; function CDF.compute) to calculate population abundance estimates and confidence intervals.
The proportion of natural-origin fish found on the spawning grounds calculated by population is multiplied by the total estimate to determine the estimate of natural-origin (wild) fish.
There are occasions where estimates from non-statistical designs for isolated regions are added to this probability estimate. Where this occurs it is not possible to calculate confidence intervals for the non-statistical design. Likewise, when the non-statistical estimate is added to statistical estimates to obtain an estimate for the entire Stratum or ESU, the confidence for the total estimate is unknown since the confidence is unknown for portions of the estimate.
Evaluation Thresholds
Abundance and Productivity Evaluation Thresholds (delisting)
Pass – The observed spawner abundance is > the abundance modeled for delisting at least six times in any 12-year period and the average observed spawner abundance is > the average modeled abundance for delisting over that same time period. (12 years was selected because it represents roughly three to four brood cycles for salmon and steelhead and should thus provide a reasonable snapshot in time of the trend in status of a population.)
Fail – The observed spawner abundance is > the abundance modeled for delisting less than six times in any 12-year period or the average observed spawner abundance is < the average modeled abundance for delisting over that same time period.
Abundance and Productivity Evaluation Thresholds (broad sense recovery)
Pass – The observed spawner abundance is > the abundance modeled for broad sense recovery at least six times in any 12-year period and the average observed spawner abundance is > the average modeled abundance for broad sense recovery over that same time period.
Fail – The observed spawner abundance is > the abundance modeled for broad sense recovery less than six times in any 12-year period or the average observed spawner abundance is < the average modeled abundance for broad sense recovery over that same time period.
Stock-recruitment curves were developed for each Oregon lower Columbia River population of salmon and steelhead as a way of determining the abundance and productivity needed to achieve delisting and broad sense recovery. No recruitment curves were developed for LCR chum due to lack of data. Because the abundance and productivity derived from these recruitment curves represent the long term (i.e. 100 year) average, annual benchmarks of abundance and productivity will be developed that will allow managers to provide more timely assessments of the progress being made to achieving recovery goals. To do this, in addition to the stock recruitment curves generated for each population, annual estimates of spawner abundance, harvest of wild fish, age at return, and an index of climate impact are needed. Because natural fluctuations in climate conditions play such a significant role in the annual abundance of salmon and steelhead spawners, it is necessary to scale the average abundance targets to an annual index of climate.
The following illustrates how this information will be used to derive annual benchmarks for abundance and productivity against which progress towards recovery can be assessed. The example in this case is for Clackamas winter steelhead.
Step 1- Obtain recruitment parameters for population (Table 1 below).
Step 2 – Determine the age composition of the returning fish.
Step 3 – Obtain total number of spawners (hatchery + wild) for each brood year.
Step 4- Obtain climatic index for each brood year (can be downloaded from this site).
Step 5 – Calculate recruits for each brood year using the following equation based on recruitment curve for the population.
Step 6 – Calculate recruits after fishery (i.e. spawners) using the fishery impact rates used to determine the “modeled current” abundance.
Step 7 – Multiply each brood year recruits by recovery scalar to obtain the number of spawners needed to meet recovery goals given climate conditions for each brood year. The recovery scalar is the amount that the current survival rate needs to be improved to get the probability of CRT to the threshold for the risk category targeted for the population. Numbers are calculated as the exp (Ln(Rt)) estimated from Eq. 1. (Table 2 below)
Step 8 – Use the age composition for this population to determine how many of each age fish returned using the brood year recruits estimated in Step 7.
Step 9 – Take the forecast total and expand it by 20 percent to provide a buffer for the impacts of climate change.
Step 10 – The number derived in Step 9 is the forecast return if recovery goals have been met. These annual forecasts can be plotted along with the actual number of spawners observed to track progress towards recovery.
Table 1. Recruitment model parameters inferred or estimated from data for lower Columbia River populations of salmon and steelhead with the best matching environmental index and associated time lag relative to parental brood year.
Table 2. Recovery scalar: the amount that the current survival rate needs to be improved
to attain the threshold for the risk category targeted for the population.
References
Beidler, W.M., and T.E. Nickelson. (1980). An evaluation of the Oregon Department of Fish and Wildlife standard spawning survey system for coho salmon. Oregon Department of Fish and Wildlife, Information Reports (Fish) 80-9, Portland.
Perrin, C.J., and J.R. Irvine. (1990). A review of survey life estimates as they apply to the area-under-the-curve method for estimating the spawning escapement of pacific salmon. Canadian Technical Report of Fisheries and Aquatic Sciences 1733.
Stevens, D.L., Jr. (1997). .Variable density grid-based sampling designs for continuous spatial populations’. Environmetrics 8: 167-195.
Stevens, D.L. (2002). Sampling design and statistical analysis methods for integrated biological and physical monitoring of Oregon streams. OPSW-ODFW-2002-07, Oregon Department of Fish and Wildlife, Portland, Oregon.
Stevens, Jr., D. L. and A. R. Olsen. (1999). .Spatially Restricted Surveys Over Time for Aquatic Resources. Journal of Agricultural, Biological, and Environmental Statistics 4:415-428.