65 Thermal Habitat Projections
Description: Species Thermal Habitat Projections
Found in: State of the Ecosystem - Gulf of Maine & Georges Bank (2018), State of the Ecosystem - Mid-Atlantic (2018)
Indicator category: Published methods
Contributor(s): Vincent Saba
Data steward: Vincent Saba, vincent.saba@noaa.gov
Point of contact: Vincent Saba, vincent.saba@noaa.gov
Public availability statement: Source data are available to the public. Model outputs for thermal habitat projections are available here.
65.1 Methods
This indicator is based on work reported in Kleisner et al. (2017).
65.1.1 Data sources
65.1.1.1 Global Climate Model Projection
We used National Oceanographic and Atmosheric Administration’s Geophysical Fluid Dynamics Laboratory (NOAA GFDL) CM2.6 simulation consisting of (1) a 1860 pre-industrial control, which brings the climate system into near-equilibrium with 1860 greenhouse gas concentrations, and (2) a transient climate response (2xCO2) simulation where atmospheric CO2 is increased by 1% per year, which results in a doubling of CO2 after 70 years. The climate change response from CM2.6 was based on the difference between these two experimental runs. Refer to Saba et al. (2016) for further details.
65.1.1.2 Modeling Changes in Suitable Thermal Habitat
The NOAA Northeast Fisheries Science Center, U.S. Northeast Shelf (NES) bottom trawl survey, which has been conducted for almost 50-years in the spring and fall, provides a rich source of data on historical and current marine species distribution, abundance, and habitat, as well as oceanographic conditions (Azarovitz 1981). The survey was implemented to meet several objectives: (1) monitor trends in abundance, biomass, and recruitment, (2) monitor the geographic distribution of species, (3) monitor ecosystem changes, (4) monitor changes in life history traits (e.g., trends in growth, longevity, mortality, and maturation, and food habits), and (5) collect baseline oceanographic and environmental data. These data can be leveraged for exploring future changes in the patterns of abundance and distribution of species in the region.
65.1.2 Data analysis
65.1.2.1 Global Climate Model Projection
The CM2.6 80-year projections can be roughly assigned to a time period by using the International Panel on Climate Change (IPCC) Representative Concentration Pathways (RCPs), which describe four different 21st century pathways of anthropogenic greenhouse gas emissions, air pollutant emissions, and land use (IPCC 2014). There are four RCPs, ranging from a stringent mitigation scenario (RCP2.6), two intermediate scenarios (RCP4.5 and RCP6.0), and one scenario with very high greenhouse gas emissions (RCP8.5). For RCP8.5, the global average temperature at the surface warms by 2C by approximately 2060-2070 relative to the 1986-2005 climatology (see Figure SPM.7a in IPCC, 2013). For CM2.6, the global average temperature warms by 2C by approximately years 60-80 (see Fig. 1 in Winton et al. (2014)). Therefore, the last 20 years of the transient climate response simulation roughly corresponds to 2060-2080 of the RCP8.5 scenario.
Here, the monthly differences in surface and bottom temperatures (‘deltas’) for spring (February-April) and fall (September- November) are added to an average annual temperature climatology for spring and fall, respectively, derived from observed surface and bottom temperatures to produce an 80-year time series of future bottom and surface temperatures in both seasons. The observed temperatures come from the NEFSC spring and fall bottom trawl surveys conducted from 1968 to 2013 and represent approximately 30,000 observations over the time series.
65.1.2.2 Modeling Changes in Suitable Thermal Habitat
We modeled individual species thermal habitat across the whole U.S. NES and not by sub-region because we did not want to assume that species would necessarily maintain these assemblages in the future. Indeed, the goal here is to determine future patterns of thermal habitat availability for species on the U.S. NES in more broad terms. We fit one generalizaed additive model (GAM) based on both spring and fall data (i.e., an annual model as opposed to separate spring and fall models) and use it to project potential changes in distribution and magnitude of biomass separately for each season for each species. By creating a single annual model based on temperature data from both spring and fall, we ensure that the full thermal envelope of each species is represented. For example, if a species with a wide thermal tolerance has historically been found in cooler waters in the spring, and in warmer waters in the fall, an annual model will ensure that if there are warmer waters in the spring in the future, that species will have the potential to inhabit those areas. Additionally, because the trawl survey data are subject to many zero observations, we use delta-lognormal GAMs (S. Wood 2011), which model presence-absence separately from logged positive observations. The response variables in each of the GAMs are presence/absence and logged positive biomass of each assemblage or individual species, respectively. A binomial link function is used in the presence/absence models and a Gaussian link function is used in the models with logged positive biomass.
The predictor variables are surface and bottom temperature and depth (all measured by the survey at each station), fit with penalized regression splines, and survey stratum, which accounts for differences in regional habitat quality across the survey region. Stratum may be considered to account for additional information not explicitly measured by the survey (e.g., bottom rugosity). Predictions of species abundance are calculated as the product of the predictions from the presence-absence model, the exponentiated predictions from the logged positive biomass model, and a correction factor to account for the retransformation bias associated with the log transformation (Duan 1983; and see Pinsky et al. 2013).
We calculated the suitable thermal habitat both in terms of changes in ‘suitable thermal abundance’, defined as the species density possible given appropriate temperature, depth and bathymetric conditions, and changes in ‘suitable thermal area’, defined as the size of the physical area potentially occupied by a species given appropriate temperature, depth and bathymetric conditions. Suitable thermal abundance is determined from the predictions from the GAMs (i.e., a prediction of biomass). However, this quantity should not be interpreted directly as a change in future abundance or biomass, but instead as the potential abundance of a species in the future given changes in temperature and holding all else (e.g., fishing effort, species interactions, productivity, etc.) constant. Suitable thermal area is determined as a change in the suitable area that a species distribution occupies in the future and is derived from the area of the kernel density of the distribution. To ensure that the estimates are conservative, we select all points with values greater than one standard deviation above the mean. We then compute the area of these kernels using the gArea function from the rgeos package in R (Bivand et al. 2011).
catalog link No associated catalog page