Is it worth counting redds to assess Atlantic salmon spawning escapement?
This chapter has been submitted as a scientific article in Fisheries Research and is currently under review. The WinBUGS code corresponding to this model can be found in Annex 2.
Redd counts are widely used as an indicator of spawner population size for salmonids (Hay, 1987; Emlen, 1995; Rieman and Myers, 1997; Rieman and Allendorf, 2001; Isaak et al., 2003; Al-Chokhaky et al., 2005; Gallagher and Gallagher, 2005). One of the advantages of redd counting is its relatively low cost. With limited resources it is possible to monitor different rivers of a large catchment over long periods of time. Therefore long time-series redd count data are frequently available for salmonid populations (ICES, 1995; Thurow, 2000; Crozier et al., 2003). Redd counts are often criticised because they lack assessment of their associated uncertainty while several authors have documented that sampling error is common in redd counts (Rieman and McIntyre, 1996; Bonneau and LaBar, 1997; Dunham et al., 2001). A number of factors may affect a redd census such as redd age, redd size, redd density, physical characteristics of the redd location (substrate color, amount of flow, etc.), vegetation cover, year to year climatic variation (storm/flood events), and observer experience. Even though, redd counts are often significantly correlated to the spawner abundances (Beland, 1996; Dunham et al., 2001; Gallagher and Gallagher, 2005).
More accurate measures of spawner abundances can be obtained by means of fish counters. Resistivity or infra-red devices provide abundance data for upstream migrating adults returning from the sea (Dunkley and Shearer, 1982; Shardlow and Hyatt, 2004). Counts are usually automated and allow counting migrating fish over the whole year. There are some uncertainties associated to these counts especially during flood events for resistivity counters (Dunkley and Shearer, 1982; Hendry et al., 2007) or at high adult densities for infrared counters (Shardlow and Hyatt, 2004). Since this is a relatively new technology, available fish counter datasets are usually limited in time. Until they become more common and extensively used, it is important to find other sources of information to quantify adult abundance.
When redd counts and fish counters data are simultaneously available, it may be possible to build a calibration relationship between the two datasets. This calibration can then be used to derive spawner estimates and assess their associated uncertainty when only redd counts data are available. With this aim, a generic Bayesian hierarchical model is developed here. The proposed methodological approach is illustrated using an Atlantic salmon (Salmo salar L.) data set from three tributaries of the Foyle river system, Ireland. Redd counts and fish counter data were gathered simultaneously over a short period of time (2001 to 2006) and redd count data only were collected over an earlier and much longer period (1959 to 2000).
Material and Methods
The Foyle River basin drains a 4500 km2 catchment on the North coast of Ireland (55°00’ N; 07°20’ W). The Foyle supports a large population of Atlantic salmon with commercial catches reaching more than 80 000 fish in the mid 1960’s. The catchment also supports Arctic charr (Salvelinus alpinus L.), brown trout (Salmo trutta L.), European eel (Anguilla Anguilla L.), Lampreys (Petromyzon marinus L., Lampetra fluviatilis L. and Lampetra planeri Bloch) and three spined stickleback (Gasterosteus aculeatus L.) (Loughs agency, 2009a,b,c).
Field data collection
Atlantic salmon redds have been consistently counted in the Foyle catchment from 1959 to 2006, except for the Faughan and the Roe sub-catchments in 1997. In addition, the wetted area of river surveyed for redd counts is also available since 2001. Fish counters provide adult abundances for three tributaries of the Foyle River basin (Faughan, Finn and Roe rivers) from 2001 to 2006 (Table 3.1).
During reproduction, which occurs usually during winter (November to February), female salmon dig a hole in the gravel where she lays her eggs. After egg deposition and fertilisation, she covers the eggs with gravel to complete the nest (redd). The gravel recently moved and cleaned are visible from the bank of the river and thus redds can be counted. During spawning time, Loughs Agency Fishery Officers aim to survey annually 460km of river channel. Depending on environmental conditions, some areas are visited one or several times during the spawning season. Training in salmon redd identification is provided by experienced Fishery Officers.
Three Logie Aquantic 2100c resistivity counters (Fewings, 1994) were used to quantify the number of adult salmon returning from the sea from 2001 to 2006. They are located at the downstream end of the Faughan, Finn, and Roe rivers. During the period of time considered, the fish counters provided uninterrupted counts. Important flood events occurred in 2007 and 2008 leading to inconsistent counts. These years are therefore, not included in this study. The total river width is covered in order to count all upstream and downstream movements of any fish with a body length greater than 45 cm passing the fish counter. This detection threshold is lower than the minimum size of a returning adult Atlantic salmon in the Foyle system and well above the average size of returning sea trout which is the only species that could also move across the fish counter. Validation of the fish counters is undertaken by two ways; firstly pseudo-graphical signals are collected and examined for the characteristic fish signal, secondly, video footage of fish passing the counting station is related to the signals generated by the counter and counts either accepted or rejected on that basis. Total annual fish counts from January to December of a given year are obtained by subtracting total downstream movements from total upstream movements (Table 1). Some downstream movements may correspond to kelts (adults migrating downstream after spawning) while the upstream movements may include the small number of sea trout bigger than 45cm (Elson and Tuomi, 1975). These errors are considered minor relative to the number of adult salmon returns. Given the methodological focus of the present paper, they are ignored in the following. These potential errors are incidental with regard to the modelling approach presented thereafter. It would apply equally well to more refined measures of adults returns.
Henceforth unobservable quantities will be denoted by Greek lower case letters and observable quantities by capital Roman letters. The notation a|b ~ f(b) means the random variable a (whether unobservable or observable) conditionally on b is distributed according to the probability distribution function (pdf) f. Unobservable quantities are necessarily unknown but observable quantities may be unknown as well if not observed (missing data). In our case, adult counts are observable quantities but they have been observed only for the last 6 years of the time series (2001-2006).
Table of contents :
CHAPTER 1: FISHERIES AND ATLANTIC SALMON
1.1. Fisheries management
1.1.1 Precautionary approach and management
1.1.2 Sources of uncertainty
1.2. Dealing with uncertainty using a Bayesian approach
1.3. A case study: Atlantic salmon in the Foyle catchment (Ireland)
1.3.2 Conservation status of wild Atlantic salmon
1.3.3 Foyle catchment description and data available
1.4. Bayesian modelling framework
1.4.1 Frequentist vs. Bayesian statistics
1.4.2 Bayes’ theorem
1.4.3 Prior distributions
1.5. Hierarchical modelling and state-space models
1.5.1 Hierarchical modelling
1.5.2 State space models
1.5.3 Bayesian inference for complex model in practice
CHAPTER 2: ESTIMATING 0+ JUVENILE PRODUCTION FROM 5 MINUTES ELECTRIC FISHING ABUNDANCE INDICES
2.1. General introduction
2.2. Building a relationship between 0+ juvenile densities and 5 minutes electric fishing abundance indices
2.2.2 Material and methods
2.3. Estimating average 0+ juvenile densities for several geographical unit
2.3.1 Data available
CHAPTER 3: IS IT WORTH COUNTING REDDS TO ASSESS ATLANTIC SALMON SPAWNING ESCAPEMENT?
3.2. Material and Methods
3.2.1 Study site
3.2.2 Field data collection
3.2.4 Bayesian inference and posterior computations
3.2.5 Posterior model checking
3.2.6 Estimating adult returns from redd counts alone
CHAPTER 4: A LIFE STAGE STATE-SPACE MODEL FOR ATLANTIC SALMON POPULATION DYNAMICS IN THE FOYLE CATCHMENT.
4.2. Material and methods
4.2.3 Parameters of management interest
4.2.4 Inter-generation replacement ratio (IGRR)
4.2.5 Bayesian inference and posterior computation
4.2.6 Posterior checking
4.3.1 Observation models
4.3.2 Exploitation and populations dynamic model
4.4.1 Modeling or dealing with limitations and constraints
4.4.2 Outputs of the model for management advice
4.4.3 Beyond the case study
CHAPTER 5: GENERAL DISCUSSION
5.2. Population dynamics in the Foyle catchment
5.2.1 Main results
5.2.2 Perspectives and improvements