The document summarizes several models that were used to simulate and analyze the salmon aquaculture industry in Chile, including:
1. A hydrodynamic model that simulated water flow, temperature, salinity and other factors. It was validated against observed data from 2014-2016.
2. A catchment model that simulated freshwater flows from major river systems. It was compared against gauged river data.
3. A biogeochemical model that simulated nutrients and primary production. It was validated spatially and temporally against field observations from 2015-2016.
4. Additional models simulated sediment transport, connectivity between farms, farm-scale epidemiology, and the bioeconomic impacts on the industry. Most were
The document summarizes several models that were used to simulate and analyze the salmon aquaculture industry in Chile, including:
1. A hydrodynamic model that simulated water flow, temperature, salinity and other factors. It was validated against observed data from 2014-2016.
2. A catchment model that simulated freshwater flows from major river systems. It was compared against gauged river data.
3. A biogeochemical model that simulated nutrients and primary production. It was validated spatially and temporally against field observations from 2015-2016.
4. Additional models simulated sediment transport, connectivity between farms, farm-scale epidemiology, and the bioeconomic impacts on the industry. Most were
The document summarizes several models that were used to simulate and analyze the salmon aquaculture industry in Chile, including:
1. A hydrodynamic model that simulated water flow, temperature, salinity and other factors. It was validated against observed data from 2014-2016.
2. A catchment model that simulated freshwater flows from major river systems. It was compared against gauged river data.
3. A biogeochemical model that simulated nutrients and primary production. It was validated spatially and temporally against field observations from 2015-2016.
4. Additional models simulated sediment transport, connectivity between farms, farm-scale epidemiology, and the bioeconomic impacts on the industry. Most were
Supplementary material. Table 1: Summary of SIMA models, content, data sources and state of validation. References are in manuscript.
Model Type Content Data sources Status
Hydrodynamic Process based Equations for momentum, continuity and River inflow from gauged rivers. Hindcast for Dec 2014-Jan 2016. Herzfeld (2018) conservation of heat and salt, employing the Model-based boundary and initial Near real time mode post Jan hydrostatic and Boussinesq assumptions conditions taken from models 2016. Skill assessed versus applied on a curvilinear grid. Outputs from maintained by the Australian observed sea levels, sea surface the model include three-dimensional Bureau of Meteorology. temperature and salinity. distributions of velocity, temperature, salinity, density, passive tracers, mixing coefficients and sea-level. Inputs required by the model include forcing due to wind, atmospheric pressure gradients, surface heat and water fluxes and open-boundary conditions such as tides and low frequency ocean currents. Catchment Process based Flow model of the instantaneous flows of all Topography, precipitation, ET, Compared against gauged major freshwater systems discharging into temperature and snow/ice cover rivers. the area of interest. Sediment Process based Multiple layers with a thin layer of easily Parameter values have where Predicted sediment Wild-Allen et al (2018) resuspendable material overlying thicker possible been sourced from characteristics were found to Margvelashvili (2009) layers of consolidated sediment. It deals literature reporting local most closely match the with viscosity, remineralization, and observations in Chile or otherwise characteristics of locally resuspension. found from model parameter data observed particles. libraries (Robson et al., 2018) or modelling studies in Australian temperate waters (Wild-Allen et al., 2013; Wild-Allen & Andrewartha 2016). Biogeochemical Process based Tracks the flux of N and P through three size As for the sediment model. Validated (spatial and temporal) Wild-Allen et al (2018) structured phytoplankton and two against observations collected zooplankton groups, as well as in the region during 2015-16 by microphytobenthos, macroalgae, shellfish IFOP and from 4 mooring beds, dissolved nutrient pools and 2 detritus locations maintained by the classes. It includes diffusion, gas exchange, Chilean MUSEL project. Optical adsorption, nitrification/denitrification and model products are compared the main ecological processes associated against MODIS images. with primary production, consumption and mortality. A full optical model is also included. Connectivity Statistical and Statistical particle tracking using flows from Parameters for particle behavior No independent drifter data Gorton and Condie Process based the hydrodynamic model. Current fields taken from parameter repository was available for validation, but (2018) from the near real time hydrodynamic and the literature on the behavior the model utilises the validated model were interpolated onto a regular of Chilean pathogens (as hydrodynamic model (see latitude-longitude grid (resolution ~ 0.005 summarized in Table 2.1 of Gorton above). Experts on the system degrees), along with wind fields. These and Condie 2018). agree with the forecast drivers were then used to drive the dispersal dispersal pathways, which also model – seeded with a100 particles per grid concur with those from other cell per day (in the user selected source/sink operational oceanography location) over the user specified release products in the region ( period. Final particle distributions are https://www.ifop.cl/chonos/) presented in two ways: either as a percentage (per grid cell) of all the particles released ending up in that grid cell; or as a cumulative exposure of cells to particles over the dispersal time, again expressed as a percentage of all the particles released. These dispersal statistics were used to calculate network connectivity metrics between the aquaculture farms in the region. Farm scale Minimum realistic This farm-scale epidemiological model The model was parameterised Model skill was high for Los Richards (2018) simulates fish growth, fish mortality, and using data reported by industry to Lagos and Aysén regions, but disease dynamics. The model considers SERNAPESCA as part of standard poorer for Magallanes due to three fish species Atlantic salmon (Salmo production, mortality and disease the sparse data for the region. salar), Coho salmon (Oncorhyncus kisutch), treatment reporting between Nevertheless, the fits of the and rainbow trout (Oncorhyncus mykiss). August 2010 and September 2017 model predictions are quite The growth and mortality models are tight around the observations deterministic, while the disease model is (Richards 2018). stochastic. Bioeconomic Econometric Bioeconomic model of the production of The biological components of the The results of the standard van Putten et al (2018) Atlantic salmon (Salmo salar), Coho salmon model are parameterized from parameterization of the model (Oncorhyncus kisutch), and rainbow trout industry data held by SERNAPESCA have been verified by industry (Oncorhyncus mykiss). The model marries a (spanning August 2010 to experts in Chile. simplified form of the farm-scale growth and September 2017). Economic data is mortality models with an economic model drawn from FAO (2018b), of production costs and revenue generation. production and treatment data held by SERNAPESCA , SalmonChile 2017, and Chile Aduanas Customs 2018. Benthic impacts Statistical An empirical biphasic impact-response curve The curves were developed using While the relationships are Bravo (2018) for benthos state in response to organic the database of the Informes noisy (see data plots in Bravo loading. This was possible as both Sanitarios y Ambientales (2018)) the final curves are a abundance and diversity correlate with Acuicultura (INFA) monitoring good representation of the organic matter, pH and redox in sediments. program of SERNAPESCA (from median and capture well the The formulation of the curve was modified 2012 to 2017). nonlinear relationships between from the standard Hill equation according benthic biomass, organic Veroli et al., (2015). enrichment and redox – the relationships also capture the thresholds in the response. System The system model represents the entire The model draws on very many The overall model produces Fulton et al (2018) Input socioecological system of each of the data sources as detailed below plausible dynamics that regions Los Lagos, Aysén and Magallanes. It (with full parameterization given in resonate with local residents (as includes many individual components as Fulton et al (2018)). discussed in workshops held in detailed below. the region) and regional statistics for production Environmental drivers The first components are the environmental The physical drivers are taken from recorded by government drivers (temperature, rainfall, irradiance and the physical, biogeochemical and agencies such as IFOP, current flows). connectivity models as well as SERNAPESCA, SUBPESCA and observational time series available Instituto Nacional de at Estasísticas HABs Data-based A footprint of harmful algal blooms with all The footprints are represented by As the footprints are based on components in this footprint suffering stochastically drawing from the observations, they match mortality or modified rate parameters. historical locations of previous historical patterns of blooms (using data supplied by distribution. local oceanographers) and the intensity of the bloom based on historical observations or scenario definition. Impact rates are taken from literature and scaled based on the intensity of the bloom (see Fulton et al (2018) for more details). Benthos Statistical Seabed beta biodiversity characterization Biodiversity survey data was The resulting biodiversity maps map calculated from empirical observations accessed from the Fundacion match regional water column using the R package gradientForest (method Huinay (Los Lagos), with data on and bathymetry patterns as well as of Ellis et al 2012; Pitcher et al 2012). sediment and water quality as major geographic breaks. The conditions from samples collected patterns concur with expert by Instituto de Fomento Pesquero knowledge of distributions in (IFOP). the region. Sediment & water Statistical Scalars on input values are used to reflect Scalars are derived from Soto and As the relationships are based quality indices of the dynamic state of water Norambuena (2004) and Leipe et on observations, they match column and sediment conditions – al (2011). historical patterns. specifically nutrients, organic material and oxygen levels. Kelp beds Cellular automata A simple carrying capacity-based The model is parameterized based The patterns concur with expert formulation is used to represent the on Buschmann et al. (1994) and knowledge of distributions in biomass of kelp beds per cell. Abreu et al (2009). the region and their evolution through time. Wild mussel banks Cellular automata Biomass model per area adapted from a Wild mussel biomass was The patterns concur with expert model for Mytilus galloprovincialis. initialised based on Flores et al knowledge of distributions in (2015), with growth model the region and their evolution parameterized based on through time. Bustamante and Branch (1996). Sealions and whales Agent based Agent based model of key marine mammals Rates were taken from The numbers and distribution in the region. These animals move toward observations shared by concur with observations taken the highest prey biomasses in their researchers working in the region, in the region and census counts immediate vicinity (which can include SeaLifeBase over the past 15 years. salmon pens for the sealions). Food ( https://www.sealifebase.org/) demand is based on what is needed given an and Rosas et al (1993), SUBPESCA age dependent growth rate (which in turn is (1999), Venegas et al (2002), Vilata set as the increment needed to meet et al (2010) and Reyes et al (2013). observed size at age). Mortality is applied stochastically, with the probability of death conditioned on: (i) a background level of 0.004 per month; (ii) the level of shipping traffic; and (iii) condition. Wild fish and Age structured Biomass based metapopulation model of Parameterised based on stock The biomass (and catch levels invertebrates cellular automata key target species in the region - hake, assessments supplied by IFOP and supported) concur with the sardine, urchin, crab. Punt et al (2016) IFOP assessments. The spatial distributions and depletion patterns through time concur with local expert knowledge. Terrestrial vegetation Cellular automata Metapopulation model of land cover types, Initialised using data from The evolution of land use including agriculture, plantations and www.ide.cl and through time matches the grassland-forest succession and disturbance. en.datachile.io/explore/geo. general patterns observed for Parameterisation based on Gray the regions (as stored in et al (2006) and Heckbert (2013). historical maps and GIS layers held by www.ide.cl). Salmon farms Agent based The growth, production and mortality (due In addition to parameters taken The validated state is as of the minimum realistic to pathogens and background mortality) from the bioeconomic and farm- farm-scale and bioeconomic were based on the simplified for the of the scale models. Caligus (parasite) models, with projected farm-scale model used in the bioeconomic levels were estimated using numbers per week of parasites model. Schools of wild/escaped salmonids SERNAPESCA held data from a close fit to 2017 observations. are also tracked (including breeding and August 2010 and December 2017. naturalization if they survive to appropriate ages). Mussel farms Agent based The mussel farm model tracks growth and Mussel farms were initialized Production levels agree with production per farm (via tracking size and based on lease data from those recorded by IFOP for numbers per string on the farm) using the SUBPESCA. Mussel production was 2012-2017. formulations of Marambio et al (2012). Spat parameterized using Toro et al is collected from wild banks – via a (2004) and Marambio et al (2012). ‘harvesting’ rate. Spat collection were tuned to reproduce the volume of production observed (reported by IFOP) in 2012-2017. Algal farms Agent based Biomass production per algal farm – using Algal farms were initialized based Production levels agree with the same equations as for wild kelp beds but on lease data from SUBPESCA. those recorded by IFOP for with seeding and harvesting. Algal production on farms was 2012-2017. parameterized based on literature values (spanning numerous Chilean and international studies) as detailed in Fulton et al (2018). Fishing Agent based Commercial, artisanal and indigenous fishing Parameterised based on IFOP Production (catch and effort) are represented by a simple catch-per-unit- fishing data for the regions and levels agree with those effort and economics-based effort allocation numbers of vessel transits as recorded by IFOP for 2012- model. Effort is applied based on distance recorded on 2017. from port and CPUE. Effort levels are re- https://www.marinetraffic.com scaled if quota levels are modified by the management model and zoning can stop access to specific locations. Processing plants Cellular automata A modified form of the manufacturing Initialised using data from Production levels agree with model (see below) is fed by fishing and www.ide.cl and those recorded by IFOP and aquaculture products landed in the local en.datachile.io/explore/geo. Rates industry bodies for 2012-2017. port. of employments, costs and processing rates are from information from plants in the region. Economics Econometric Simple economic input-output flow of costs Initial conditions and relevant rates Gross economic imports, and revenue. With the volume of production for demand taken from Chilean revenues, production and dictated by the individual industry models census and the annual and employment match the (see below) and time series of interest rates, quarterly reports of the Instituto magnitude and seasonal cost of living, and economic demand per Nacional de Estasísticas patterns across industries and industry sector. regions recorded in the quarterly reports of Instituto Nacional de Estasísticas Tourism Agent based This model tracks the usage of the area by Initialised using data from Number of visits and changing different kinds of tourists – ecotourists, www.ide.cl and tourism mix through time comfortable budget travelers and mass en.datachile.io/explore/geo follows the general patterns and tourism. It is based on the formulations of volume recorded by Casagrandi and Rinaldi (2002) as updated by government departmental Fulton et al (2017). statistics (e.g. as available from Instituto Nacional de Estasísticas) Shipping Agent-based Traffic flow model following the movements The transport routes and Volume of vessels using the of 4 classes of vessel in the region (along frequency of operation of well navigation channels and ports navigational routes): boats was taken from data per size class concurs with the I. Large (> 150000 DWT) accessed from records available from the port II. Medium (10,000 – 150,000 DWT) https://www.marinetraffic.com. authorities and III. Small (< 10,000 DWT; including well The spatial location was digitised https://www.marinetraffic.com. boats) into a GIS shapefile manually using IV. Local (outboards, smaller motor cruisers QGIS 2.14. etc) Numbers are scaled based on local demographics and number of instances of the other industries modelled in the region. Terrestrial industries Cellular automata Farming, plantations, logging, mining, Initialised using data from Gross level of production per energy generation, manufacturing and www.ide.cl and industry per region concurs with urban retail and services were modelled en.datachile.io/explore/geo the volumes recorded in the using the equations of Fulton et al (2017) quarterly reports of Instituto modified to reflect local decision-making Nacional de Estasísticas. and operational practices in each of the three regions. Human demography Agent based Simple age structured population model of Initialised using data from Chilean Numbers per major settlement indigenous and other communities in the census, www.ide.cl and and total numbers per region region. Equations modified from Meadows en.datachile.io/explore/geo and are within 10-20% of those et al (1972). Tracks permanent residents, parameterized using values from recorded by the Chilean census temporary residents and Fly-in Fly-out (FIFO) Engelen et al (1997) and Fulton et time series. worker population. al (2017). Social influence and Network Networks of information flow and sharing These networks elicited in direct Applied as is without further information flow between actors within the individual consultation with local validation – the results of the community (demographic) and sector communities. models have been found to be models plausible by local workshop participants with provides confidence in the fidelity of the approach. Social perception Minimum realistic Empirical relationship between activity Empirical relationships taken from Outcomes per modelled social levels and environmental state and the qualitative (signed diagraph) group under scenario ‘happiness and satisfaction’ of the different models of connections elicited perturbations agree with community components. from the local population during expectations from local focus workshops in the regions. groups. Government rules Input Government regulations and decision rules. Details of the individual rules were Input rather than model output. taken from the relevant Chilean However, the resulting model government departments. Spatial dynamics have been found to zoning was sourced from be plausible by workshop www.ide.cl and participants in the regions. en.datachile.io/explore/geo