Table 1. Metrics measuring the thermal properties of warm-season MHWs.
All of the annual warm-season MHW properties are defined for a given heat stress year (HSY). The HSY starts from a month in the climatological cold season in the given grid cell (Li & Donner, 2022). This spatially varying definition of the HSY is necessary because the warm season in parts of the ocean overlaps across two calendar years. In the remainder of the manuscript, we use the word “year” instead of HSY for clarity, and label them according to the first calendar year of a HSY. For example, the last year of the historical heat stress analysis refers to HSY 2013, including data from the calendar years 2013 and 2014.
To test the effects of theoretical acclimation or adaptation to warming by marine ecosystems for the end of the century, we repeat the analysis using the end point of rolling climatology (Logan et al., 2014), in which the MMM is calculated over the 2041-2100 period.
2.2 Coral reef and kelp forests distribution
Projected changes of warm-season MHW properties are also specifically assessed over the global region of coral reef and kelp systems according to the high resolution global maps of warmwater corals (UNEP, 2010) and the Laminarian kelp biome (Jayathilake and Costello 2020). The maps are converted to global 1° x 1° latitude-longitude grids using ArcGIS for consistency with the grid of model outputs.
3 Sea surface temperature datasets
We use 1° x 1° latitude-longitude global daily SST outputs from the simulations of historical (1985-2014) and future climate (2015-2100) in the Coupled Model Intercomparison Project Phase 6 (CMIP6, Eyring et al. 2016). All the simulation outputs are from the first ensemble member (r1i1p1) of three models: GFDL-ESM4 (Dunne et al., 2020), MRI-ESM2 (Yukimoto et al., 2019) and CESM2-WACCM (Danabasoglu et al., 2020). These three models are employed because i) they have low, medium and high climate sensitivities relative to the range of values in the CMIP6 ensemble (2.7K, 3.4K and 4.8K, respectively); ii) they have relatively strong performance in simulating key natural modes of climate variability (e.g., ENSO, Dunne et al. 2020; Beobide-Arsuaga et al. 2021; Danabasoglu et al. 2020), which is critical to simulating warm-season MHW frequency and severity (Sen Gupta et al., 2020; Holbrook et al., 2019; Oliver et al., 2018). The future projections are examined for three future scenarios, SSP 1-2.6, SSP 2-4.5 and SSP 3-7.0, used in the CMIP6 that represent a low, medium and high level of radiative forcing in the range of the future emission pathways (O’Neill et al., 2016). Note that most of the CMIP6 model simulation outputs are available for download with the original native tripolar grid in unit of kilometer. For the consistency, we regridded the CESM2-WACCM outputs using the same 1st order conservative algorithm that data centers used for the regridded GFDL-ESM4 and MRI-ESM2 outputs (Jones 1999).
We use 0.05° x 0.05° latitude-longitude global daily satellite-derived SST dataset CoralTemp v3.1 from the National Oceanic and Atmospheric Association (NOAA) Coral Reef Watch (CRW) program (W. Skirving et al., 2020), to contrast simulations and observations for the 1985-2014 historical period. For the consistency of the spatial resolution between the observed and simulated SST data, we regridded the observed SST datasets to 1° by 1° using the 1st order conservative algorithm (i.e., a common method used for upscaling dataset resolution; Jones 1999).
4. Results
4.1 Evaluation of model biases in the thermal properties of warm-season MHW over the historical (1985-2014) period