conceptualmodels.Rmd

---
title: "Conceptual models"
output:
  html_document: default
  word_document: default
  pdf_document: default
bibliography: conceptmods.bib
---

```{r setup,  echo = F, message=F}

#Load packages
library(knitr)
library(rmarkdown)

data.dir  <- './data'
image.dir <- './images'

```

<!--The purpose of this form is to provide the methods used to collect, process, and analyze derived data prior to data submission for all EDAB reports. This information will be compiled into the State of the Ecosystem Tech Memo, AND be used to meet metadata requirements put forth by the Public Access to Research Results (PARR) directive. 

For the tech memo, we will be documenting the methods behind each indicator to the level of a peer-reviewed document.-->
# Conceptual Models

## Contributor name(s)
Sarah Gaichas, Patricia Clay, Geret DePiper, Gavin Fay, Paula Fratantoni, Robert Gamble, Sean Lucey, Charles Perretti, Patricia Pinto da Silva, Vincent Saba, Laurel Smith, Jamie Tam, Robert Wildermuth 

## Data name
Conceptual models for the New England (Georges Bank and Gulf of Maine) and Mid-Atlantic regions of the Northeast US Large Marine Ecosystem

### Indicator category 
<!-- 1. Database pull -->
<!-- 2. Database pull with analysis -->
3. Synthesis of published information
4. Extensive analysis; not yet published
<!-- 5. Published methods -->


## Methods
Conceptual models were constructed to facilitate multidisciplinary analysis and discussion of the linked social-ecological system for integrated ecosystem assessment. The overall process was to first identify the components of the model (focal groups, human activities, environmental drivers, and objectives), and then to document criteria for including groups and linkages and what the specific links were between the components.

The prototype conceptual model used to design Northeast US conceptual models for each ecosystem production unit (EPU) was designed by the California Current IEA program. The California Current IEA developed an overview conceptual model for the Northern California Current Large Marine Ecosystem (NCC)  (https://www.integratedecosystemassessment.noaa.gov/Assets/iea/california/conceptual-models/Integrated-SocioEcological-System-Overview6.png), with models for each focal ecosystem component that detailed the ecological, environmental, and human system linkages (https://www.integratedecosystemassessment.noaa.gov/regions/california-current-region/components/focal-components/coastal-pelagic-overview.html#, https://www.integratedecosystemassessment.noaa.gov/regions/california-current-region/components/focal-components/coastal-pelagic-environmental.html#, https://www.integratedecosystemassessment.noaa.gov/regions/california-current-region/components/focal-components/coastal-pelagic-ecological.html#, https://www.integratedecosystemassessment.noaa.gov/regions/california-current-region/components/focal-components/coastal-pelagic-human.html#). Another set of conceptual models outlined habitat linkages (https://www.integratedecosystemassessment.noaa.gov/regions/california-current-region/components/mediating-components/habitat.html). 

An inital conceptual model for Georges Bank and the Gulf of Maine was outlined at the 2015 ICES WGNARS meeting. It specified four categories: Large scale drivers, focal ecosystem components, human activities, and human well being. Strategic management objectives were included in the conceptual model, which had not been done in the NCC. Focal ecosystem components were defined as aggregate species groups that had associated US management objectives (outlined within WGNARS for IEAs, see @depiper_operationalizing_2017): groundfish, forage fish, fished invertebrates, living habitat, and protected species. These categories roughly align with Fishery Managment Plans (FMPs) for the New England Fishery Management Council. The Mid-Atlantic conceptual model was developed along similar lines, but the Focal groups included demersals, forage fish, squids, medium pelagics, clams/quahogs, and protected species to better align with the Mid Atlantic Council's FMPs.

```{r draftmod, echo = F, eval = T}

knitr::include_graphics(file.path(image.dir, 'GBGOMconceptual1.png'))
```

After the initial draft model was outlined, working groups were formed to develop three submodels following the CCE example: ecological, environmental, and human dimensions. The general approach was to specify what was being included in each group, what relationship was represented by a link between groups, what threshold of the relationship was used to determine whether a relationship was significant enough to be included (we did not want to model everything), the direction and uncertainty of the link, and documentation supporting the link between groups. This information was recorded in a spreadsheet (*make available on ERDDAP*). Submodels were then merged together by common components using the "merge" function in the (currently unavailable) desktop version of Mental Modeler (http://www.mentalmodeler.org/#home; @gray_mental_2013). The process was applied to Georges Bank (GB), the Gulf of Maine (GOM), and the Mid-Atlantic Bight (MAB) ecosystem production units (see EPU section). 

### Data source(s) (All types)
<!--Please provide a text description of data sources.-->
####Ecological submodels
Published food web (EMAX) models for each subregion [@link_documentation_2006; @link_northeast_2008], food habits data collected by NEFSC trawl surveys [@smith_trophic_2010], and other literature sources [@smith_consumption_2015] were consulted. Expert judgement was also used to adjust historical information to current conditions, and to include broad habitat linkages to Focal groups. 

####Environmental submodels
Published literature on the primary environmental drivers (seasonal and interannual) in each EPU was consulted. 
Sources for Georges Bank included @backus_georges_1987 and @townsend_oceanography_2006. 
Sources for the Gulf of Maine included @smith_mean_1983, @smith_interannual_2001, @mupparapu_role_2002, @townsend_oceanography_2006, @smith_regime_2012, and @mountain_labrador_2012.  
Sources for the Mid Atlantic Bight included @houghton_middle_1982, @beardsley_nantucket_1985, @lentz_climatology_2003, @mountain_variability_2003,   @glenn_biogeochemical_2004, @sullivan_evidence_2005, @castelao_seasonal_2008, @shearman_long-term_2009, @castelao_temperature_2010, @gong_seasonal_2010, @gawarkiewicz_direct_2012, @forsyth_recent_2015,@fratantoni_description_2015, @zhang_dynamics_2015, @miller_state-space_2016, and @lentz_seasonal_2017.

####Human dimensions submodels
Fishery catch and bycatch information was drawn from multiple regional datasets, incuding the Greater Atlantic Regional Office Vessel Trip Reports & Commercial Fisheries Dealer databases, Northeast Fishery Observer Program & Northeast At-Sea Monitoring databases, Northeast Fishery Science Center Social Sciences Branch cost survey, and the Marine Recreational Informational Program database. Further synthesis of human welfare derived from fisheries was drawn from @fare_adjusting_2006, @walden_productivity_2012, @lee_inverse_2013, @lee_hedonic_2014, and @lee_applying_2017. Bycatch of protected species was taken from @waring_us_2015, with additional insights from @bisack_measuring_2014. The top 3 linkages were drawn for each node. For example, the top 3 recreational species for the Mid-Atlantic were used to draw linkages between the recreational fishery and species focal groups. A similar approach was used for relevant commercial fisheries in each region.

Habitat-fishery linkages were drawn from unpublished reports, including:
1. Mid-Atlantic Fishery Management Council. 2016. Amendment 16 to the Atlantic Mackerel, Squid, and Butterfish Fishery Management Plan: Measures to protect deep sea corals from Impacts of Fishing Gear. Environmental Assessment, Regulatory Impact Review, and Initial Regulatory Flexibility Analysis. Dover, DE. August, 2016. 
2. NOAA. 2016. Deep sea coral research and technology program 2016 Report to Congress. http://www.habitat.noaa.gov/protection/corals/deepseacorals.html retrieved February 8, 2017.  
3. New England Fishery Management Council. 2016. Habitat Omnibus Deep-Sea Coral Amendment: Draft. http://www.nefmc.org/library/omnibus-deep-sea-coral-amendment Retrieved Feb 8, 2017.
4. Bachman et al. 2011. The Swept Area Seabed Impact (SASI) Model: A Tool for Analyzing the Effects of Fishing on Essential Fish Habitat. New England Fisheries Management Council Report. Newburyport, MA.

Tourism and habitat linkages were drawn from unpublished reports, including: 
1. http://neers.org/RESOURCES/Bibliographies.html                                 
2. Great Bay (GoM) resources  http://greatbay.org/about/publications.htm          
3. Meaney, C.R. and C. Demarest. 2006. Coastal Polution and New England Fisheries. Report for the New England Fisheries Management Council. Newburyport, MA.
4. List of valuation studies, by subregion and/or state, can be found at http://www.oceaneconomics.org/nonmarket/valestim.asp.

Published literature on human activities in each EPU was consulted. 

Sources for protected species and tourism links included @hoagland_demand_2000 and @lee_economic_2010. 

Sources for links between environmental drivers and human activities included @adams_uncertainty_1973, @matzarakis_proceedings_2001, @scott_climate_2004, @hess_climate_2008, @colburn_social_2012, @jepson_development_2013, and @colburn_indicators_2016. 

Sources for cultural practices and attachments links included @pauly_putting_1997, @mcgoodwin_understanding_2001, @st_martin_making_2001, @norris-raynbird_for_2004, @pollnac_toward_2006, @clay_defining_2007, @clay_definingfishing_2008, @everett_role_2008, @donkersloot_politics_2010, @lord_understanding_2011, @halpern_index_2012, @wynveen_natural_2012, @cortes-vazquez_identity_2013, @koehn_progress_2013, @potschin_landscapes_2013, @reed_beyond_2013, @urquhart_constructing_2013, @blasiak_paradigms_2014, @klain_what_2014, @poe_cultural_2014, @brown_we_2015, @donatuto_evaluating_2015, @khakzad_role_2016, @oberg_surviving_2016, and @seara_perceived_2016.  

### Data extraction 
<!--Text overview description of extraction methods. What information was extracted and how was it aggregated? Can point to other indicator extraction methods if the same.-->
####Ecological submodels
"Data" included model estimated quantities to determine whether inclusion thresholds were met for each potential link in the conceptual model. A matrix with diet composition for each modeled group is an input to the food web model. A matrix of mortalities caused by each predator and fishery on each modeled group is a direct ouput of a food web model (e.g. Ecopath). Food web model biomasss flows between species, fisheries, and detritus were summarized using algorithms implemented in visual basic by Kerim Aydin, NOAA NMFS Alaska Fisheries Science Center. Because EMAX model groups were aggregated across species, selected diet compositions for individual species were taken from the NEFSC food habits database using the FEAST program for selected species (example query below). These diet queries were consulted as supplemental information. 

Example FEAST sql script for Cod weighted diet on Georges Bank. Queries for different species are standardized by the FEAST application and would differ only in the svspp code. 
```{sql FEAST, eval = F, echo = T}
Select svspp,year,cruise6,stratum,station,catsex,pdid,pdgutw,pdlen,pdwgt,perpyw,pyamtw,COLLCAT,numlen,pyamtv  from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' order by svspp,year,cruise6,stratum,station,pdid,COLLCAT
Select distinct svspp,year,cruise6,stratum,station from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' order by svspp,year,cruise6,stratum,station
Select distinct svspp,year,cruise6,stratum,station,catsex,catnum from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' order by svspp,year,cruise6,stratum,station
Select distinct COLLCAT from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' order by COLLCAT
Select distinct svspp,year,cruise6,stratum,station,catsex,pdid,pdlen,pdgutw,pdwgt  from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' order by svspp,year,cruise6,stratum,station,catsex,pdid
Select svspp,year,cruise6,stratum,station,catsex,pdid,pdlen,COLLCAT,sum(perpyw),sum(pyamtw),sum(pyamtv)  from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' group by svspp,year,cruise6,stratum,station,catsex,pdid,pdlen,COLLCAT order by svspp,year,cruise6,stratum,station,catsex,pdid,pdlen,COLLCAT
Select svspp,year,cruise6,stratum,station,COLLCAT,sum(pyamtv) sumpvol from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' group by svspp,year,cruise6,stratum,station,COLLCAT order by svspp,year,cruise6,stratum,station,COLLCAT
Select svspp,year,cruise6,stratum,station, count(distinct pdid) nstom  from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' group by svspp,year,cruise6,stratum,station,catsex order by svspp,year,cruise6,stratum,station
Select svspp,year,cruise6,stratum,station,pdlen,numlen,count(distinct pdid) nstom  from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and numlen is not null and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' group by svspp,year,cruise6,stratum,station,pdlen,numlen,catsex order by svspp,year,cruise6,stratum,station,pdlen
Select svspp,year,cruise6,stratum,station,pdlen,COLLCAT,sum(pyamtv) sumpvol  from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' group by svspp,year,cruise6,stratum,station,pdlen,COLLCAT order by svspp,year,cruise6,stratum,station,pdlen,COLLCAT
Select distinct svspp,year,cruise6,stratum,station,pdid,pdlen from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and numlen is null and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB'
Select distinct year,cruise6,stratum,station,beglat,beglon  from fhdbs.allfh_feast where pynam <> 'BLOWN' and pynam <> 'PRESERVED' and pynam <> ' ' and svspp='073' and YEAR BETWEEN '1973' AND '2016' and GEOAREA='GB' order by year,cruise6,stratum,station

```


####Environmental submodels
Information was synthesized entirely from published sources; no additional data extraction was completed for the environmental submodels.

####Human dimensions submodels
Recreational fisheries data were extracted from the 2010-2014 MRIP datasets. Original data can be found at \\net\home4\gdepiper\ICES\IEA\top10_prim1_common_mode.xlsx for each region (New England or Mid-Atlantic as defined by states). 
<!--
Write SQL query here
```{sql, eval = F, echo = T}
SELECT * FROM...
```

```{r r_extract, echo = T, eval = F}
# Extraction code
```
-->

### Data analysis
<!--Text description of analysis methods, similar in structure and detail to a peer-reviewed paper methods section.-->
####Ecological submodels
Aggregated diet and mortality information was examined to determine the type of link, direction of link, and which links between which groups should be inclded in the conceptual models. Two types of ecological links were defined using food web models: prey links and predation/fishing mortality links. Prey links resulted in positve links between the prey group and the focal group, while predation/fishing mortality links resulted in negative links to the focal group to represent energy flows. The intent was to include only the most important linkages between focal groups and with other groups supporting or causing mortality on focal species groups. Therefore, threshold levels of diet and mortality were established (based on those that would select the top 1-3 prey and predators of each focal group): 10% to include a link (or add a linked group) in the model and 20% to include as a strong link. A Primary Production group was included in each model and linked to pelagic habitat to allow environmental effects on habitat to be connected to the ecologial submodel. Uncertainty for the inclusion of each link and for the magnitude of each link was qualitatively assessed and noted in the spreadsheet (*link to ERDDAP*). 

Four habitat categories (Pelagic, Seafloor and Demersal, Nearshore, and Freshwater and Estuarine) were included in ecological submodels as placeholders to be developed further along with habitat-specific research. Expert opinion was used to include the strongest links between each habitat type and each Focal group (noting that across species and life stages, members of these aggregate groups likely occupy many if not all of the habitat types). Link direction and strength were not specified. Environmental drivers were designed to link to habitats, rather than directly to Focal groups, to represent each habitat's important mediation function.

EMAX model groups were aggregated to focal groups for the Georges Bank (GB), Gulf of Maine (GOM) and Mid-Atlantic Bight (MAB) conceptual models according to the table below. "Linked groups" directly support or impact the Focal groups as described above.

```{r groups,eval = T, echo = F}
#read in EMAXconceptualmodgroups.csv and kable it
emaxgroups <- read.csv(file.path(data.dir, "EMAXconceptualmodgroups.csv"))
names(emaxgroups) <- c("Group Type", "Region", "Conceptual model group", "EMAX group(s)", "Notes")
kable(emaxgroups)
```

An example ecological submodel (Gulf of Maine), visualized in Mental Modeler, is below.
```{r draftGOMeco, echo = F, eval = T}

knitr::include_graphics(file.path(image.dir, 'MM_GoM_Ecological.png'))
```

####Environmental submodels

An example environmental submodel (Gulf of Maine), visualized in Mental Modeler, is below.
```{r draftGOMenv, echo = F, eval = T}

knitr::include_graphics(file.path(image.dir, 'MM_GoM_Climate.png'))
```

####Human dimensions submodels


Identify 
focal groups or species 
human activities and objectives
environmental drivers
Document 
criteria and linkages

An example human dimensions submodel (Gulf of Maine), visualized in Mental Modeler, is below.
```{r draftGOMhuman, echo = F, eval = T}

knitr::include_graphics(file.path(image.dir, 'MM_GoM_Human_Connections.png'))
```

#### Merged models
All links and groups from each submodel were preserved in the full merged model for each system. Mental modeler was used to merge the submodels. Full models were then re-drawn in Dia (http://dia-installer.de/) with color codes for each model component type for improved readability. Examples for each system are below. 

```{r diaGB, fig.cap="Georges Bank conceptual model", echo = F, eval = T}

knitr::include_graphics(file.path(image.dir, 'GBoverview5.png'))
```


```{r diaGOM, fig.cap="Gulf of Maine conceptual model", echo = F, eval = T}

knitr::include_graphics(file.path(image.dir, 'GoMoverview4.png'))
```


```{r diaMAB, fig.cap="Mid-Atlantic Bight conceptual model", echo = F, eval = T}

knitr::include_graphics(file.path(image.dir, 'MAB_3.png'))
```


<!--
What packages or libraries did you use in your work flow?
```{r packages, echo = T}
sessionInfo(package = NULL)


#Use this to output a detailed list of the package information
current.session <- sessionInfo(package = NULL)
current.session$otherPkgs
```


Include accompanying R code, pseudocode, flow of scripts, and/or link to location of code used in analyses.
```{r analysis, echo = T, eval = F}
# analysis code
```
-->
### References
<!--List references here.-->
<!--the following line allows reference placement somewhere other than the default end of the doc, if we move the ref section to the end we won't need this-->
<div id="refs"></div>



### ERDDAP data set
<!--If data are available, adjust code to pull those data and set "eval = T" in chunk header options.-->
Put documentation spreadsheet on ERDDAP as a table?

```{r erddap pull, echo = T, eval = F}
library(rerddap)

comet <- 'http://comet.nefsc.noaa.gov/erddap/'

tab_list <- ed_datasets(url = comet)
tab_list <- data.frame(Title = tab_list$Title,
                        Identifier = tab_list$Dataset.ID,
                        Index = 1:nrow(tab_list))

#Get data set ID (will need to change)
#id <- as.character(tab_list$Identifier[9]) 

#Pull data
data <- tabledap(id,url = comet)

#Print metadata
metadata <- info(id, url = "http://comet.nefsc.noaa.gov/erddap/")$alldata$NC_GLOBAL
cat(metadata[metadata$attribute_name == "lineage processing steps",]$value)

```

### Further metadata
<!--Fill out the following metadata required by PARR-->


#### Public availability statement
<!--Will source data be available to the public? If not, is a Data Access Waiver necessary?-->


#### Point of contact - Whom should be contacted?
Sarah Gaichas

#### Data steward (can be POC; in charge of keeping data updated)
Sarah Gaichas