Marine Heatwaves are modelled off their terrestrial namesakes and are prolonged periods of anomalously high sea surface temperature (SST). In effect, they are categorized SST anomalies (c.f., the NOAA Coral Reef Watch (CRW) daily global 5km SST Anomaly product). They also are a more general version of the CRW daily global 5km Coral Bleaching HotSpot product, but rather than being aimed specifically at corals, they provide a more generalized description of marine heat stress that is likely applicable to a broader range of marine life. The next important step, therefore, is to develop a generalized accumulation of heat stress, much like the daily global 5km Degree Heating Week (DHW) product accumulates HotSpots.

This daily global 5km-resolution Marine Heatwave Watch (MHW) product is derived by applying the Marine Heatwave algorithm of Hobday et al. (2018)¹ to the daily global 5km 'CoralTemp' SST data product. 'CoralTemp' is one of the best and most internally consistent daily global 5km SST products available, which allowed for the creation of an accurate climatology. This enabled the derivation of an accurate, consistent MHW product (comprised of near real-time and historic components), stretching back over three decades.


Methodology

Climatology

CoralTemp SST values from January 1, 1985 to December 31, 2012 were used to derive the climatology. For each day of the year, both an average and a 90th percentile SST value were created. These two values were calculated from the SST in an 11-day window centered on the day of the year extended over the 28 years of the climatology. This means that each day's average and 90th percentile values were calculated from 308 values (i.e., 11 x 28).

Note that the SST value for February 29 in each leap year was left out of the calculation of the climatology so as to simplify the calculation of the daily climatology. When needed, the climatology for February 29 is derived by averaging the climatology values for February 28 and March 1.

Since January 1, 1985 is the first day of the dataset, special consideration was given to calculating the climatology average and 90th percentile values for January 1, 2, 3, 4, and 5 of that year. As one example, the average and 90th percentile values for January 1, 1985 should have included the SST values from December 27 to January 6 (the 11-day window around January 1) for each year of the climatology. However, since there are no SST values available for December 27-31, 1984, the average and 90th percentile values for January 1, 1985 could only be calculated from the 303 values that are available (i.e., 6 values in 1985 and 11 values from each of the other 27 years of the climatology, giving n=303). It follows then that n=304 for January 2, 1985; n=305 for January 3, 1985; and so on, until n=308 for January 6, 1985 and for every other day of the climatological year.

Calculation of Heatwave Category

For each day from January 1, 1985 to the present, the Marine Heatwave category was calculated for each 5km satellite pixel grid in the v3.1 daily global 'CoralTemp' SST dataset.

A Marine Heatwave was identified if the SST for a particular day was greater than the 90th percentile value for that location. Once a Marine Heatwave was identified, it was categorized based on its intensity, after Hobday et al. (2018)¹. Intensity categories were defined based on the difference between the average and 90th percentile values for each 5km pixel (diff). If the SST for a particular day was ≥ (average + diff) and < (average + 2diff), it was categorized as being Marine Heatwave Category 1. If the SST for a particular day was ≥ (average + 2diff) and < (average + 3diff), it was categorized as being Marine Heatwave Category 2; and so on (Figure 1).

Figure 1: Pictorial description of Marine Heatwave categories (from Hobday et al., 20181).

Since the climatology does not have a value for February 29 in each leap year, when this value was needed, both the average and 90th percentile SST values were calculated using an average of the SST values from February 28 and March 1.

Dealing with Sea Ice

Due to the methodology used in the derivation of the MHW product, its interpretation becomes problematic if all pixels are not approximately normally distributed. At the very least, if the methodology is applied to a pixel with non-normally distributed temperatures (e.g., a pixel with any sea ice) and then applied to a pixel with approximately normally distributed temperatures (e.g., a pixel with only water temperatures), then the interpretation of each of these pixels will be different, creating internal inconsistencies within the MHW product. For this reason, if sea ice was present in a 5km satellite grid location within the 11-day window surrounding any date in the 28-year climatology period, then the pixel was classified as ice and given a flag value of 1 in the mask array of the climatology. The sea ice mask in the MHW product is therefore static and is a conservative estimate of the ice-free satellite pixels within the climatology. It is not an indication of sea ice for that pixel for a given day.


NetCDF files

Each day of the global 5km MHW product is output as a separate NetCDF file. Each daily file contains the Marine Heatwave Category values for all 5km satellite pixels for that day (Table 1).

There are three arrays of data in each file: latitudes, longitudes and Marine Heatwave Category.

Relevant references:

1. Hobday AJ, Oliver ECJ, Sen Gupta A, Benthuysen JA, Burrows MT, Donat MG, Holbrook NJ, Moore PJ, Thomsen MS, Wernberg T, Smale DA. 2018. Categorizing and naming marine heatwaves. Oceanography 31, 2. doi: 10.5670/oceanog.2018.205.

2. Hobday AJ, Alexander LV, Perkins SE, Smale DA, Staub SC, Oliver ECJ, Benthuysen JA, Burrows MT, Donat MG, Feng M, Holbrook NJ, Moore PJ, Scannell HA, Sen Gupta A, Wernberg T. 2016. A hierarchical approach to defining marine heatwaves. Progress in Oceanography 141, 227-238.

3. Oliver ECJ, Donat MG, Burrows MT, Moore PJ, Smale DA, Alexander LV, Benthuysen JA, Feng M, Gupta AS, Hobday AJ, Holbrook NJ, Perkins-Kirkpatrick E, Scannell A, Straub SC, Wernberg T. 2018. Nature Communications 9, 1324. doi: 10.1038/s41467-018-03732-9.