Climate-informed multispecies assessment model methods for determining biological references points and Acceptable Biological Catch

Kirstin Holsman (  kirstin.holsman@noaa.gov ) National Oceanic and Atmospheric Administration, Alaska Fisheries Science Center, Seattle, WA. USA https://orcid.org/0000-0001-6361-2256 Alan Haynie National Oceanic and Atmospheric Administration, Alaska Fisheries Science Center, Seattle, WA. USA Anne Hollowed National Oceanic and Atmospheric Administration, Alaska Fisheries Science Center, Seattle, WA. USA Jonathan C. P. Reum National Oceanic and Atmospheric Administration, Alaska Fisheries Science Center, Seattle, WA. USA Kerim Ayind National Oceanic and Atmospheric Administration, Alaska Fisheries Science Center, Seattle, WA. USA Albert Hermann Joint Institute for the Study of the Atmosphere and Ocean, University of Washington, Seattle, WA Wei Cheng Joint Institute for the Study of the Atmosphere and Ocean, University of Washington, Seattle, WA Amanda Faig School of Aquatic and Fishery Sciences, University of Washington, Seattle, WA 98195, U.S.A James Ianelli National Oceanic and Atmospheric Administration, Alaska Fisheries Science Center, Seattle, WA. USA Kelly Kearney Joint Institute for the Study of the Atmosphere and Ocean, University of Washington, Seattle, WA Andre Punt School of Aquatic and Fishery Sciences, University of Washington, Seattle, WA


Introduction
Here we present a protocol for deriving climate-informed multi-species reference points for sheries management. This approach builds of the sloping harvest control rule and maximum sustainable yield biomass proxy approaches currently implemented in the Bering sea (Alaska, USA) and modi ed to derive harvest recommendations as part of the CEATTLE multi-species stock assessment model 1 (https://archive.afsc.noaa.gov/refm/docs/2019/EBSmultispp.pdf) for walleye pollock (Gadus chalcogrammus), Paci c cod (G. microcephalus), and arrowtooth ounder (Atheresthes stomias). The CEATTLE model has been updated annually and included as an appendix to the BSAI walleye pollock stock assessment since 2016 as part of the Bering Sea shery stock assessment process.
The CEATTLE model is t to historical survey and shery data as well as hindcasts (1979-2017) of oceanographic conditions from a high resolution ROMSNPZ model for the Bering Sea (see 1 for more detail). We used hindcasts (1979-2017) (coupled regional oceanographic -nutrient-phytoplanktonzooplankton model; see2).The same ROMSNPZ model is then projected to derive harvest recommendations under future climate scenarios . Details of projection realizations that composed the ensemble members are available in [2][3][4] .  7 .These three models were selected because they projected a broad range of global patterns for precipitation and SST, and provided contrasting views of future ocean conditions in the EBS. Output from these models under two representative concentration pathways or RCPs (4.5 and 8.5; 8) were used to drive the Bering10K regional ROMSNPZ model. RCP 8.5 and 4.5 represent a high-baseline carbon emission scenario and a moderate mitigation scenario, respectively. This resulted in a suite of 6 projections of Eastern Bering Sea conditions including bottom and surface temperature, summer "cold pool", and large zooplankton (key prey resource) abundance during spring and fall (critical periods for juvenile pollock and cod survival). We additionally included a persistence scenario as our "null" climate-constant scenario (i.e., average of 2006-2016 conditions).
We use these scenarios to derive 1) projections without harvest (un shed spawning biomass) and 2) projections where in each year harvest was set to the sustainable limit using current management approaches (sloping harvest control rule and harvest rate that results in 40% of un shed spawning biomass). This approach is detailed below and is also used in 1 .

Reagents
No reagents were used.

Equipment
The CEATTLE and recruitment models are programed in AD Model Builder release version 11.6 (http://www.admb-project.org); the ROMSNPZ model is programmed in Regional Ocean Modeling System version 3.2; AIC analyses were conducted with R version 3.5.3 (2019-03-11) https://www.r-project.org.
Custom code was created for the multispecies stock assessment model (CEATTLE), recruitment model and projections, ROMSNPZ CMIP5 projections, and threshold analyses and plotting. Details can be found in 1,2,9 . All code is publicly available at the following github site and will be archived via Zenodo upon publication: https://github.com/kholsman/EBM_Holsman_NatComm.

Procedure
To determine ABC given the sloping harvest control rule for pollock, Paci c cod, and arrowtooth ounder in each simulation year project the population forward using estimated parameters from the multispecies mode of the CEATTLE model t to data from 1979-2017 and retrospective modeled data; recruitment in each projection year is dynamic and is based on biomass in simulation year and future environmental covariates from the Bering10K model downscaled projections.
For each annual timestep in the projection period: 1. For each environmental covariate create the "persistence" projection using the average of the last 10 years of the hindcast ROMSNPZ model (i.e., constant future climate conditions). 3. Iteratively solve for F 40%, i.e., the harvest rate that results in an average spawning biomass (B40 % ) during 2095-2099 that is 40% of B 0 for pollock and Paci c cod simultaneously with arrowtooth ounder set to the historical average (as arrowtooth ounder is a major predator of pollock and historical F for arrowtooth ounder is much lower than F 40% ). 4. Once F 40% for pollock and Paci c cod are found, iteratively solve for F 40% for arrowtooth ounder. where B y is the spawning biomass at the start of the simulation year for each climate scenario based on climate effects on recruitment, predation mortality, and growth, --i.e., the climate-informed B y.

Time Taken
Anticipated Results