Abstract:

We conducted 1000 Atlantis simulations, using permutated diet matrices created by drawing from a beta distribution that represented uncertainty in predator diet. We also employed a statistical emulator which simulates the production of many more Atlantis runs than could be achieved through numerical simulation. This dataset includes 3 products: statistical emulator output, diet matrices, and Atlantis simulations. These are each contained in their own zip file.

Suggested Citation:

Ainsworth, Cameron, ainsworth@usf.edu. 2018. Evaluation of diet matrix uncertainty for the Atlantis model. Distributed by: GRIIDC, Harte Research Institute, Texas A&M University–Corpus Christi. doi:10.7266/N71G0JSG

Publications:

Morzaria-Luna, H. N., Ainsworth, C. H., Tarnecki, J. H., & Grüss, A. (2018). Diet composition uncertainty determines impacts on fisheries following an oil spill. Ecosystem Services. doi:10.1016/j.ecoser.2018.05.002

Purpose:

These are data from an evaluation of the uncertainty surrounding previous estimates made of the ecosystem impacts of the Deepwater Horizon oil spill. We tested the impact of uncertainty on the diet matrix.

Data Parameters and Units:

Product 1: Statistical emulator output. Two types of data are reported annually for the statistical emulators, the fits (error) for each model type (e.g., file "365_varfit.cscv"), and the predictions of the best fitting model (e.g., file "365_predobs.csv"). In both cases, the number at the front of the file name represents the time step (day) being evaluated. The "varfit" files specify functional group (abbreviated), best fitting models (model_fit), root mean squared error (RMSE), variable (catch or biomass), time step (day) reported, and an index number for convenience in data analysis that increments with functional group or age stanza. The functional group abbreviations are as follows: Macroalgae (ALG), Greater amberjack (AMB), Sediment bacteria (BB), Blue crab (BCR), Black drum (BDR), Benthic feeding sharks (BEN), Other billfish (BIL), Bioeroding fish (BIO), Bivalves (BIV), Blue marlin (BMR), Brown shrimp (BSH), Bluefin tuna (BTN), Crustose coralline algae (CCA), Carnivorous macrobenthos (CMB), Stony corals (COR), Diving birds (DBR), Carrion detritus (DC), Deep diving odontocetes (DDO), Toxic dinoflagellates (DIN), Labile detritus (DL), Dolphins and porpoises (DOL), Refractory detritus (DR), Deep serranidae (DSR), Deep water fish (DWF), Herbivorous echinoderms (ECH), Epiphytes (EPI), Filter feeding sharks (FIL), Flatfish (FLT), Gag grouper (GAG), Sea grass (GRS), Infaunal meiobenthos (Inf), Jacks (JCK), Jellyfish (JEL), King mackerel (KMK), Kemps ridley (KMP), Ladyfish (LDY), Large sharks (LGS), Crabs and lobsters (LOB), Loggerhead (LOG), Large pelagic fish (LPL), Large phytoplankton (LPP), Large reef fish (LRF), Little tunny (LTN), Lutjanidae (LUT), Large zooplankton (LZP), Manatee (MAN), Menhaden (MEN), Microphytobenthos (MPB), Medium pelagic fish (MPL), Mullets (MUL), Mysticeti (MYS), Octocorals (OCT), Other demersal fish (ODF), Other shrimp (OSH), Other tuna (OTN), Oysters (OYS), Bacteria (PB), Pinfish (PIN), Pompano (POM), Protists (PRO), Pink shrimp (PSH), Skates and rays (RAY), Red drum (RDR), Red grouper (RGR), Red snapper (RSN), Spanish sardine (SAR), Surface feeding birds (SBR), Sciaenidae (SCI), Scamp (SCM), Stone crab (SCR), Small demersal fish (SDF), Seatrout (SEA), Sessile filter feeders (SES), Sheepshead (SHP), Spanish mackerel (SMK), Small sharks (SMS), Snook (SNK), Sponges (SPG), Small pelagic fish (SPL), Small phytoplankton (SPP), Squid (SQU), Small reef fish (SRF), Shallow serranidae (SSR), Swordfish (SWD), Small zooplankton (SZP), Blacktip shark (TIP), Other turtles (TUR), Vermilion snapper (VSN), White marlin (WMR), White shrimp (WSH), Yellowfin tuna (YTN). The "predobs" files specify the emulator iteration number (x1), the time step being evaluated (day), the variable being evaluated (catch or biomass), functional group code, the functional group name, guild (a general grouping scheme that can be used to interpret results across similar types of taxa), the best fitting model, the observed data from Atlantis, the predicted data from emulator, and the residual error. The best fitting models correspond to the following categories: generalized linear model (glm), bagged multivariate adaptive regression splines (BMARS), bagged classification and regression trees (BCART), random forest (RF), boosted regression trees (BRT), extreme learning machine (ELM). Product 2. Diet matrices. We include every one of the 10000 diet permutations used in the uncertainty analysis. There is one comma separated file per diet matrix. The headers resemble the following code "pPrey1AMB1_DWF". This code is broken down as follows "pPrey" only indicates this is a diet value, "1" indicates that the prey is in the juvenile cohort (adult=2), "AMB" indicates the group code of the prey item (e.g., amberjack, AMB), "1" indicates that the predator is in the juvenile cohort (adult=2), "DWF" indicates the group code of the predator (e.g., deepwater fish, DWF). The data value for pPrey1AMB1_DWF would therefore indicate the amount of juvenile amberjack that is consumed by juvenile deepwater fish. If either the prey or predator cohort number is missing, this indicates that age stanzas are not tracked for that functional group in Atlantis (this is the case for most invertebrate groups). Note: These values are availabilities (not diet percentage). Availability is a variable in the functional response used by Atlantis (see Atlantis documentation at https://research.csiro.au/atlantis/ for more detail). The functional group abbreviations are defined as in Product 1 and are identical for both predator and prey. Product 3: These are the catch and biomass output text files from 1000 Atlantis simulations (corresponding to diet matrices 1-1000 from product 2). Both catch and biomass are in tons and reported annually (e.g., end of first year is 3.65E+02 days) Atlantis diet availabilities (unitless), catch (tons), biomass (tons), root mean standard error (tons)

Individual Files: