Modelling the impacts of climate change on thermal habitat suitability for shallow-water marine fish at a global scale

Edward Lavender^1,2*, Clive J. Fox¹ and Michael T. Burrows¹

¹ The Scottish Association for Marine Science, Scottish Marine Institute, Dunstaffnage, Oban, Scotland, PA37 1QA
² Current Address: Centre for Research into Ecological and Environmental Modelling, The Observatory, University of St Andrews, Fife, Scotland, KY16 9LZ

^* This repository is maintained by Edward Lavender (el72@st-andrews.ac.uk).

Introduction

This repository contains methods, written in R and organised as an R Project, for Lavender, Fox and Burrows (2021). Modelling the impacts of climate change on thermal habitat suitability for shallow-water marine fish at a global scale. PLOS ONE 16(10): e0258184. https://doi.org/10.1371/journal.pone.0258184

In this work, we developed an approach based on species’ thermal affinities to predict changes in thermal habitat suitability for more than 2,000 shallow-water marine fish species under future temperature change scenarios (Figure 1).

Figure 1. An overview of the methods used to predict change in thermal habitat suitability for shallow-water fish under future temperature change. We started with a list of over 2000 shallow-water fish species. For each species, we overlaid a model of occurrence with baseline sea surface temperature (SST) or sea bottom temperature (SBT) data to estimate quantiles of observed variation in occupied temperatures, from which we inferred thermal affinities. For example, we took the median temperature across the species’ range (the species’ thermal index, STI) as the optimum temperature for that species and the 10th (T10) and 90th (T90) percentiles to define the species’ thermal range (T90 - T10). We used these to parameterise a climate response curve: a Gaussian model relating temperature to an index of thermal habitat suitability, termed the climate response curve thermal habitat suitabilty index (CRCS), with a mean equal to the optimum temperature and a variance based on the thermal range. The central premise of this model is that species have a thermal optima around which habitat suitability declines, an idea that is closely related to the abundance-centre hypothesis. To predict change in CRCS, we fed this model with baseline and future temperatures derived from CMIP5 (Coupled Model Intercomparison Phase Project Phase 5) models. For populations living below the optimum temperature, in the blue half of the Gaussian distribution shown, an increase in temperature is assumed to lead to an increase in thermal habitat suitability. And for populations living above the optimum temperature, in the red half, an increase in temperature is assumed to lead to a decline in thermal habitat suitability. In this way, for every species, we predicted the thermal habitat suitability in each grid cell in which it currently occurs under baseline and future temperatures and calculated the change in thermal habitat suitability. We then synthesised patterns across all species and Exclusive Economic Zones.

Structure

1. data-raw contains raw data for the project.
   • spatial contains spatial information.
     • ne_110m_coastline contains world coastline data from Natural Earth.
   • sdm_aquamaps contains raw Aquamaps species distribution model (SDM) predictions (maps) for modelled species, obtained via get_sdm_aquamaps.R (see below).
   • occurrence contains raw occurrence data for modelled species, obtained via process_sdm_aquamaps.R (see below).
   • temperature contains sea surface temperature (SST) and sea bottom temperature (SBT) global-scale CMIP5 ensemble average projections for (a) historical (1956-2005), (b) mid-century (2006-2055) and (c) late-century (2050-2099) time scales under Representative Concentration Pathways (RCPs) 4.5 and 8.5. All projections were obtained from NOAA’s Climate Change Web Portal. For SST, for the historical scenario, the HADISST dataset is also included for comparison to modelled scenarios.
     
     
2. data contains processed data.
   • spptraits.rds defines modelled species.
   • spatial contains processed spatial data.
     • coastline contains world coastline data generated by process_coastline.R.
     • eez contains Exclusive Economic Zone (EEZ) boundary data from the Flanders Marine Institute and datasets defining the number of modelled cells and species in each EEZ, generated by process_eez.R.
     • species contains species richness data; namely, map_species_richness.asc, a raster of the number of modelled species, generated by analyse_spptraits.R.
     • sdm_aquamaps contains processed Aquamaps SDMs for modelled species, generated by process_aquamaps.R.
     • occurrence contains processed occurrence data for modelled species, generated by process_sdm_aquamaps.R.
   • temperature contains processed SST and SBT temperature projections for the time periods and scenarios described above, generated by process_temp.R.
   • sensitivity contains processed ‘sensitivity’ rasters for modelled species, generated by process_sensitivity.R.
   • abundance contains thermal habitat suitability predictions* for SST and SBT for all temperature scenarios, generated by project_abund_1.R.
     
     
3. R contains scripts for data processing, projections and analysis.
   • get_* scripts get raw data, where necessary.
   • process_* scripts implement data processing.
   • project_* scripts implement thermal habitat suitability predictions.
   • analyse_* scripts analyse the data and predictions, including figure creation.
   • helper_* scripts contain helper functions used in multiple scripts.
     
     
4. fig contains figures.

Note that the data-raw, data and fig directories are not provided in the online version of this repository.

Workflow

Data acquisition via get_*

1. get_sdm_aquamaps.R gets Aquamaps SDMs for modelled species, via the aquamapsdata R package.

2. Other raw data are acquired manually (see the links above).

Data processing via process_*

1. process_coastline.R processes the raw coastline data:
   • Forces an extent of {-180, 180, -90, 90} to match species’ distributions and temperature projections;
     
     
2. process_spptraits.R defines a list of modelled species:
   • Focuses on the subset of species with depth ranges found on FishBase;
   • Checks for synonyms;
   • Saves a temporary (reduced) list of species for which Aquamaps SDMs are acquired and then processed (see get_sdm_aquamaps.R and process_sdm_aquamaps.R);
   • Using processed species’ distributions:
     • Checks SDMs;
     • Gets thermal niche parameters from processed temperature projections;
     • Gets the full taxonomic breakdown;
       
       
3. process_sdm_aquamaps.R processes Aquamaps species’ distributions:
   • Forces an extent of {-180, 180, -90, 90} to match temperature projections;
   • Replaces 0 for predicted occurrence with NA;
   • Masks land using processed coastline;
     
     
4. process_temp.R processes temperature projections:
   • Extracts SST and SBT temperature projections from raw files;
   • Forces an extent of {-180, 180, -90, 90} to match species’ distributions;
   • Re-samples temperatures to the same spatial resolution as species’ distributions;
   • Masks land using processed coastline to match species’ distributions;
     
     
5. process_eez.R processes EEZ data:
   • Gets EEZ areas (in units of grid cells);
   • Gets EEZ latitudinal mid-points;
     
     
6. process_sensitivity.R derives species’ sensitivity indices given thermal niche parameters derived from (a) SST or (b) SBT. This includes:
   • Mean thermal niche width (species thermal range: STR) over space;
   • Mean thermal bias (species thermal index (STI) - baseline temperatures) over space;
   • Variability in thermal bias over space;

Projections via project_*()

1. project_abund_1.R predicts changes in species’ thermal habitat suitability* under future climate change scenarios.
2. project_abund_2.R synthesises predictions across species.

Analyses via analyse_*()

1. analyse_spptraits.R analyses modelled species, including:

   • taxonomic breakdown;
   • depth ranges;
   • distribution;
   • commercial importance;

2. analyse_example_thermal_niche.R analyses an example thermal niche and analyse_sdm_aquamaps_example.R analyses an example SDM.

3. analyse_temp.R analyses SST and SBT temperature projections, for all scenarios.

4. analyse_sensitivity.R analyses species’ sensitivity indices.

5. analyse_abund_across_globe.R analyses thermal habitat suitability predictions* across the globe.

6. analyse_abund_across_eezs.R analyses thermal habitat suitability predictions* across EEZs.

7. analyse_scenarios.R analyses the relative severity of predictions under RCP 4.5 and RCP 8.5.

GitHub packages

This repository uses a number of non-default packages, available from The Comprehensive R Archive Network. These can be installed with install.packages(). Three packages that are only available on GitHub are also used:

• aquamapsdata. This package is used to acquire SDMs for modelled species.
• prettyGraphics. This package is used for plotting, e.g., via pretty_map().
• utils.add. The utils.add::basic_stats() function is also sometimes used as a convenient routine for summarising numeric vectors. This could be replaced by base R functions, such as summary().

Citation

Lavender, Fox and Burrows (2021). Modelling the impacts of climate change on thermal habitat suitability for shallow-water marine fish at a global scale. PLOS ONE 16(10): e0258184. https://doi.org/10.1371/journal.pone.0258184

---

*Predictions of the change in thermal habitat suitability are expected to correspond in broad terms to changes in abundance, with species increasing in abundance in areas in which thermal habitat suitability increases and declining in areas in which thermal habitat suitability declines, hence the naming structure of these files.