11/27/2018 Stock Assessment Workshop 19th June -25th June 2008 SPC Headquarters Noumea New Caledonia

Day 2 Session 1 Parameter estimation – Recruitment 11/27/2018 Day 2 Session 1 Parameter estimation – Recruitment

11/27/2018 “The most important and generally the most difficult problem in the biological assessment of fisheries is the relationship between stock and recruitment” Hilborn and Walters (1992) The first of the four processes we will look at is recruitment. As Hilborn and Walters state…..By the end of this session you will hopefully have a better understanding of what it is, why its so important to stock assessments, and understanding different ways of estimating.

11/27/2018 Session overview This session is going to review the theory behind recruitment dynamics and then demonstrate how recruitment is accounted for in stock assessment models, with a particular focus on MULTIFANCL and tuna assessments. You were introduced to the concept of recruitment on Monday. Today we are going to revise and expand on that description What is recruitment? Processes effecting recruitment Estimating recruitment Stock recruitment hypothesis Estimation of recruitment by MULTIFANCL for WCPO assessments Estimation of recruitment outside of MULTIFANCL

Bt+1=Bt+R+G-M-C Our conceptual model Death (Natural mortality) 11/27/2018 Our conceptual model Bt+1=Bt+R+G-M-C Death (Natural mortality) Recruitment (+) Whole population (-) Catch (Fishing mortality) (-) Growth (+) As you will remember now, and hear many times during this workshop, there are 4 key processes for fish populations which will effect the size (or biomass) of a population over time, these being R,G,M and F. This morning we looked at the initial “additive process”, or process that adds biomass to a population, that being recruitment. Once fish have recruited to the fishery, they continue to grow and in doing so add further biomass to the population. As such the second process critical to population dynamics we will consider is growth!

What is Recruitment? Our preferred definition: 11/27/2018 Bt+1=Bt+R+G-M - C What is Recruitment? Our preferred definition: The number of fish alive at a specified stage after hatching. In fisheries science, that stage is typically determined by when we are first able to detect the species, for example, in the fishery catch. Hence we effectively are considering recruitment to not just the assessable population (we cant assess what we cant detect!), but to the fishery also. e.g. WCPO Skipjack, yellowfin and bigeye tunas all recruit to the purse seine fishery at around 0.5-1 years of age. Before that point it would appear that they are too small to be caught by any commercial gear

What is recruitment? Adults Eggs Larvae Spawning and fertilisation 11/27/2018 What is recruitment? Adults Spawning and fertilisation Maturation Eggs 2 year olds Hatching 1 year olds = recruits Larvae Metamorphosis

What is Recruitment? Recruitment versus Natural Mortality 11/27/2018 Bt+1=Bt+R+G-M - C What is Recruitment? Recruitment versus Natural Mortality The number of fish that will recruit to a population is predominantly a function of two main factors: a) fecundity – the amount of eggs the females can produce b) survival/mortality – of eggs, larvae, and juveniles up untill the point of recruitment (e.g. become susceptible to the fishery) While mortality plays a role in the number of recruits to a population, its near impossible to get an estimate of natural mortality in fish of a size that we cant even catch in the fishery. Note that in stock assessment models, Natural Mortality (M) is defined as the death of fish by natural causes at any point in time after the fish have recruited to the fishery

Key processes in recruitment 11/27/2018 Key processes in recruitment Adults Spawning and fertilisation FECUNDITY Maturation Eggs 2 year olds SURVIVAL Hatching 1 year olds = recruits Larvae Metamorphosis

Key processes in recruitment 11/27/2018 Bt+1=Bt+R+G-M Key processes in recruitment Processes that effect egg production Total stock egg production per time period (e.g. year) will be a product of: Number of spawnings per year Number of mature females Age (e.g. older adults might produce more eggs) Eggs per spawning Many of which might be also effected by 1. Adult condition: e.g. Nutritional condition and stress factors 2. Environment (e.g. water temperature) 3. Other

Key processes in recruitment 11/27/2018 Bt+1=Bt+R+G-M Key processes in recruitment Processes that effect egg, larval and juvenile survival Biotic (e.g.) Starvation Predation Disease Abiotic (e.g.) Temperature Salinity Oxygen Small variations in survival = big variations in recruitment Offspring survival (per female adult “tuna”) 107 106 Eggs/Larvae 105 Juveniles Adults Numbers 104 103 102 101 Days Months Years

Key processes in recruitment 11/27/2018 Bt+1=Bt+R+G-M Key processes in recruitment In summary, many different factors can impact both egg production and the early life history survival of marine fish prior to recruitment …. This is an important observation…. …….and we’ll see why in a moment.

Estimating Recruitment The stock-recruitment hypothesis 11/27/2018 Bt+1=Bt+R+G-M Estimating Recruitment The stock-recruitment hypothesis In a stock assessment model we need to be able to estimate recruitment.. How? One theory states that there is a relationship between stock size and recruitment, and that recruitment can, to a greater or lesser degree, be estimated using this relationship. A number of different forms of the relationship have been proposed, e.g. Density independent Compensatory Depensatory

Estimating Recruitment The stock-recruitment hypothesis 11/27/2018 Bt+1=Bt+R+G-M Estimating Recruitment The stock-recruitment hypothesis Density Independance Recruitment Stock The number of recruits per spawner does not change regardless of the population size, hence recruitment increases with stock size …….doesn’t take into account limited habitat/resources

Estimating Recruitment The stock-recruitment hypothesis 11/27/2018 Bt+1=Bt+R+G-M Estimating Recruitment The stock-recruitment hypothesis Stock Recruitment Compensation The number of recruits per spawner declines as stock size increases, hence recruitment plateaus after initial phase of increase …….takes account of resource limitation (increasing competition = lower fecundity and lower survival, increased cannabalism by adults, disease transmission, predation etc)

11/27/2018 Bt+1=Bt+R+G-M Recruitment (R) Stock-Recruitment Curves – Commonly used compensatory models Beverton and Holt Biological Assumption: That juvenile competition (for resources, habitat etc) results in a mortality rate that is linearly related to the number of fish alive in the cohort (at any point in time)….Compensatory model. The most common form to the equation is: R = (aS)/(b+S) R = Recruitment a = Maximum recruitment (over all stock sizes) S = Stock size b = Stock size when recruitment is half the maximum recruitment (i.e.= a/2)

Estimating Recruitment The stock-recruitment hypothesis 11/27/2018 Bt+1=Bt+R+G-M Estimating Recruitment The stock-recruitment hypothesis Depensation Recruitment Stock Under some conditions recruits per spawner may increase as spawning stock size increases (to a point)……e.g. at extremely low stock size, adults may have trouble finding one another. As stock size increases above this, recruits/spawner will increase.

Estimating Recruitment The stock-recruitment hypothesis 11/27/2018 Bt+1=Bt+R+G-M Estimating Recruitment The stock-recruitment hypothesis Stock Recruitment Overcompensation At very high stock sizes (where resources become limited), adults may be forced into cannabalism on the juveniles to such a degree that recruitment is actually reduced below that associated with lower stock sizes (density based disease transmission could also result in this relationship)

Recruitment (R) Ricker model (for overcompensation) 11/27/2018 Bt+1=Bt+R+G-M Recruitment (R) Ricker model (for overcompensation) Biological Assumption: Mortality of the cohort is stock dependant (i.e. dependant on the initial cohort size which is assumed proportional to the spawning stock size) rather than density dependant. Potential mechanisms include cannabalism by adults at high stock (and cohort) sizes and increased disease transmission/spread. Models key feature is declining recruitment at higher stock sizes 2 common forms to the equation are R = Sea(1-S/b) R = aSe-bS R = Recruitment ea = initial slope of the curve S = Stock size b = value of S where R=S a = productivity parameter proportional to fercundity b = parameter describing the rate at which R/S declines as stock size increases a = recruits per spawner at low stock size

Estimating Recruitment 11/27/2018 Bt+1=Bt+R+G-M Estimating Recruitment Beverton and Holt model of compensation Compensation R = (aS)/(b+S) R = Recruitment a = Maximum recruitment (over all stock sizes) S = Stock size b = Stock size when recruitment is half the maximum recruitment (i.e.= a/2) Recruitment Stock Key factor in a stock recruitment relationship……steepness of the curve! This will be related to b, the stock size when recruitment is half the maximum recruitment This will effectively determine the capacity of a stock at low size to recover quickly or not.

Potential for very high variability 11/27/2018 Bt+1=Bt+R+G-M Recruitment (R) Stock-Recruitment Curves – Basic Properties Potential for very high variability Recruitment Stock “When dealing with stock-recruitment curves we are averaging many biological phenomena so it is best to think of the curve as a general statistical description”. (HW1991)

Estimating Recruitment 11/27/2018 Bt+1=Bt+R+G-M Estimating Recruitment “….more commonly the number of recruits is effectively independent of adult stock size over most of the observed range of stock sizes”. (Gulland, 1983) Highly differing recruitments at same stock size can be due to difference in environment and impacts of that on egg production by adults and/or survival of larvae Excel example

Estimating Recruitment 11/27/2018 Bt+1=Bt+R+G-M Estimating Recruitment Recruitment can be measured directly through the analyses of size data in the catch, in other words, looking at the size frequencies and associated CPUEs of the recruiting age class Stock can be measured also through the following indicators No. of eggs = Sum { females per age group X fecundity at each age} Total biomass of reproductive individuals Index of abundance of population in year of egg deposition These are in order of decreasing reliability. Errors in the measurement of spawning stock will make stock and recruitment appear independent of each other (i.e. can mask a real relationship)

Estimation of Recruitment by MULTIFANCL 11/27/2018 Bt+1=Bt+R+G-M Estimation of Recruitment by MULTIFANCL MULTIFANCL is a size based, age structured model. There has been a substantial collection of size data from observer and port sampling programs in the WCPO over many decades. MULTIFANCL estimates recruitment based on both the size of recruiting modes in the size data, and the level of CPUE from catch and effort data, to determine the size of the recruitment coming into the fishery at any point in time Weak recruitment Strong recruitment (if CPUE high)

Estimation of Recruitment by MULTIFANCL 11/27/2018 Bt+1=Bt+R+G-M Estimation of Recruitment by MULTIFANCL MULTIFANCL models the stock recruitment relationship typically using either the Beverton and Holt or Ricker curves, which set the average stock-recruitment relationship. Deviations from this relationship are then determined by the strength of the size modes and the CPUE associated with those size modes. The steepness of the recruitment curve can be estimated by the model (typically restrained by fairly broad priors set by the modellers so that the model doesn’t predict unreasonable steepness values) “The overall potential yield of a simple age structured model depends primarily on the natural mortality rate and the steepness of the stock recruitment curve” (Hilborn and Walters, 1992)

Stock-Recruitment Estimation in MULTIFAN_CL: Bigeye 2006 11/27/2018 Stock-Recruitment Estimation in MULTIFAN_CL: Bigeye 2006

Stock-Recruitment Estimation in MULTIFAN_CL: Yellowfin 2006 11/27/2018 Stock-Recruitment Estimation in MULTIFAN_CL: Yellowfin 2006

Stock-Recruitment Estimation in MULTIFAN_CL: Skipjack 2005 11/27/2018 Stock-Recruitment Estimation in MULTIFAN_CL: Skipjack 2005

Recruitment (R) Recruitment Biomass 11/27/2018 Bt+1=Bt+R+G-M Recruitment (R) Recruitment trends drive biomass trends to a large degree (e.g. for YFT in WCPO) Recruitment Biomass

Estimation of Recruitment outside MULTIFANCL 11/27/2018 Bt+1=Bt+R+G-M Estimation of Recruitment outside MULTIFANCL It is recognised that recruitment, or the survival of larval and juvenile fish through to first appearance in a fishery, can be closely linked to oceanographic processes that effect productivity of oceanic waters. These oceanographic processes (shifting of warm pool, promotion of upwelling, productivity etc) are driven by large scale climatic processes (e.g. ENSO, PDO) The information required to help guide estimations of recruitment in MULTIFANCL is not always easily obtained, leading to greater uncertainty regarding recruitment. There is now an increasing emphasis into research on the development of environmental predictors of recruitment. These work on the theory that certain oceanographic/climatic conditions promote survival of eggs, larvae and juvenile fish, and some conditions do not. Models are being investigated which use oceanographic and climatic data to predict shifts in recruitment for some species

Oceanographic/environmental impacts on fish populations – Recruitment 11/27/2018 Oceanographic/environmental impacts on fish populations – Recruitment Yellowfin tuna recruitment Recruitment estimates for YFT in the WCPO, derived from the SA model, are highly variable seasonally, inter-annually and over decadal periods. A generalised linear model (GLM) was developed that predicts the variation in YFT recruitment in response to a range of oceanographic variables. The final model accounted for 68% of observed variation in quarterly recruitment for the period 1980–2003, with the inclusion of 10 different oceanographic variables derived from two zones within the equatorial region of the WCPO. The robustness of the recruitment model was investigated by cross-validation. The model was then applied to hindcast recruitment for the period 1952–1979. Recruitment predictions from the GLM closely followed trends in recruitment estimates from the assessment model through most of this period. The long-term trend in predicted recruitment was largely driven by sea surface temperature in the northwestern area of the equatorial region. Recruitment estimates for yellowfin tuna in the western and central Pacific Ocean (WCPO), derived from a stock assessment model, are highly variable seasonally, inter-annually and over decadal periods. A generalised linear model (GLM) was developed that predicts the variation in yellowfin recruitment in response to a range of oceanographic variables, computed from different areas and both spatial and temporal scales. The final model accounted for 68% of observed variation in quarterly recruitment for the period 1980–2003, with the inclusion of 10 different oceanographic variables derived from two zones within the equatorial region of the WCPO. The robustness of the recruitment model was investigated by cross-validation. The model was then applied to hindcast recruitment for the period 1952–1979. Recruitment predictions from the GLM closely followed trends in recruitment estimates from the assessment model through most of this period. The long-term trend in predicted recruitment was largely driven by sea surface temperature in the northwestern area of the equatorial region. Source: Adam Langley, Karine Briand, David Seán Kirby, Raghu Murtugudde (In press) Influence of oceanographic variability on recruitment of yellowfin tuna Thunnus albacares in the western and central Pacific Ocean

Oceanographic/environmental impacts on fish populations – Recruitment 11/27/2018 Oceanographic/environmental impacts on fish populations – Recruitment YFT BET SKJ In the last five decades for which tuna fishing data are available, the interannual ENSO signal (SOI) and the related Pacific Decadal Oscillation (PDO) suggest two different regimes characterized by higher intensity and frequency of either El Nin˜o or La Nin˜a events. Recent estimates from a statistical population dynamics model (MULTIFAN-CL) suggest that recruitment of three tuna species in the Pacific are correlated with these climate indices. While tropical tuna species like skipjack (Katsuwonus pelamis) and yellowfin (Thunnus albacares) had higher recruitments during El Nin˜o events, the subtropical albacore species (Thunnus alalunga) showed the opposite pattern with low recruitment during El Nin˜o and high recruitment during La Nin˜ a. The hypothesis that the spatial dynamics of temperature, currents (advection), food availability and predation constrain tuna recruitment is evaluated with an application of SEPODYM to skipjack. Simulation results showed that this hypothesis can reproduce fluctuations in the population that are similar to those estimated from the statistical model.

Oceanographic/environmental impacts on fish populations – Recruitment 11/27/2018 Oceanographic/environmental impacts on fish populations – Recruitment Primary Productivity In the last five decades for which tuna fishing data are available, the interannual ENSO signal (SOI) and the related Pacific Decadal Oscillation (PDO) suggest two different regimes characterized by higher intensity and frequency of either El Nin˜o or La Nin˜a events. Recent estimates from a statistical population dynamics model (MULTIFAN-CL) suggest that recruitment of three tuna species in the Pacific are correlated with these climate indices. While tropical tuna species like skipjack (Katsuwonus pelamis) and yellowfin (Thunnus albacares) had higher recruitments during El Nin˜o events, the subtropical albacore species (Thunnus alalunga) showed the opposite pattern with low recruitment during El Nin˜o and high recruitment during La Nin˜ a. The hypothesis that the spatial dynamics of temperature, currents (advection), food availability and predation constrain tuna recruitment is evaluated with an application of SEPODYM to skipjack. Simulation results showed that this hypothesis can reproduce fluctuations in the population that are similar to those estimated from the statistical model.

Recruitment (R) Recruitment Overfishing 11/27/2018 Bt+1=Bt+R+G-M Recruitment (R) Recruitment Overfishing Definition: Whereby a stock is fished so hard that sufficient recruits can not be produced to replace those lost from population Detection of recruitment overfishing requires determining the relationship between spawning stock size and recruitment

11/27/2018 Bt+1=Bt+R+G-M Session Review Understanding relationship between stock and recruitment is critical to detecting recruitment overfishing, and for predicting impacts of fishing on population size over time Many different factors impact on recruitment, at many points in the early life history of fishes. These factors often make it difficult to detect a stock-recruitment relationship, as they introduce variation into the relationship Measurement error can also mask a true S-R relationship However, S-R relationships are evident in many significantly overfished stocks, we simply havnt reached that point yet in the WCPO (and so the relationships are less evident) A number of stock-recruitment relationships have been proposed, each based on specific biological assumptions, including: Beverton and Holt curve Ricker curve

11/27/2018 Bt+1=Bt+R+G-M Session Review Estimates of recruitment within MULTIFAN-CL rely on strength of size modes in size data (modal detection/tracking) and the associated CPUE associated with those size modes Recruitment can also be estimated outside the model, using predictors based on oceanographic and climatic data. Cross-validation has shown some potential for the latter approach assisting in estimation of recruitment for stock assessment models.