Table 1 Seventeen remote sensing and 6 field-based variables considered as potential predictive drivers of coral bleaching divided by (a) selected metrics and (b) metrics not selected in the final gradient-boosted model

Depth

The mean depth of the reef in meters.

1.00–26.45 (m)

The sixth most relevant variable in the model. BSI was higher at greater depths.

Diversity

The diversity of corals calculated from Hill’s number one is equal to Shannon’s diversity exponent, this represents true diversity without considering the less abundant or “rare” species⁸⁴.

2.173–21.212 (effective species)

The seventh most relevant variable in the model. BSI was higher at greater diversity.

HS_28days

The sum of HS in 28 days before the sampling date.

0.0–48.55 (°C)

The second most relevant variable in the model. BSI was higher in reefs with higher heat stress in the last 28 days.

SI_reef

Reef sensitivity is based on the relative abundance of species weighted by species response to bleaching^3,23.

0.1438–0.3665 (index value)

The third most relevant variable in the model. A sharp increase in the BSI as the sensitivity index increases.

ROTC_clim

The climatological seasonal-warming rate (ROTC_clim) is the average of the annual ROTC values for the period 1985–2012 (sensu Chollett et al. 2014)⁹. The seasonal warming rate (ROTC) reflects the trend in temperature change over 84 weeks during summer⁹.

0.1931–0.2464 (°C by week)

The most relevant variable in the model. Negative relationship with BSI. Reefs with a higher rate of climatological seasonal warming rate exhibited less bleaching.

Trend of DHW

DHW quantifies heat stress by summing up positive daily anomalies above 1 °C above the MMM over 84 days (12 weeks), divided by 7 to express values per week³⁸. ‘Trend of DHW’ is the trend of annual maximums DHWs from 1985 to the sampling year, a trend obtained from a generalized least square model⁴².

0.00000–0.30064 (°C-weeks by year)

The fifth most relevant variable was selected in the model. Higher BSI in reefs with a higher rate of increase in DHWs.

∆ DHW

Difference between the maximum observed value of DHW in the current event up to the sampling date and the maximum observed value of DHW in the last year (building on Hughes 2019¹⁹)

−3.7787 to 7.7007 (°C-weeks)

The fourth most relevant variable was selected in the model. Higher BSI in reefs with higher recent heat stress (∆DHW ≥ 6).

DCA1

The first axis in a multidimensional ordination analysis based on species composition and applying a Detrended Correspondence Analysis⁸⁵. Considered a potential ecological gradient related to bleaching³.

−1.6264 to 2.2710 (axis position)

Not selected in the final model but high correlation with diversity.

DHW_28days

DHW calculation considering a 28-day window¹⁶.

0.0000−6.9358 (°C-weeks)

Not selected in the final model but high collinearity with different heat stress metrics.

DHW

Conventional DHW calculation considering an 84-day window³⁸.

0.5141–13.7115 (°C-weeks)

Not selected in the final model but high collinearity with different heat stress metrics.

DHW_555days

DHW calculation considering a 555-day window represents the accumulated stress since the beginning of the summer of the previous year.

1.217–32.874 (°C-weeks)

Not selected in the final model but high collinearity with different heat stress metrics.

HS_3days

Hotspots (HS) represent positive daily anomalies above the MMM³⁸. HS_3days is the average HS in the last three days before the sampling date.

0.00000–1.49589 (°C)

Not selected in the final model but high collinearity with different heat stress metrics.

HS1_28days

Number of days with HS values greater than 1 °C in the last 28 days¹⁶.

0–28 (days)

Not selected in the final model but high collinearity with different heat stress metrics.

HS1_84days

Number of days with HS values greater than 1 °C in the last 84 days¹⁶.

3–60 (days)

Not selected in the final model but high collinearity with different heat stress metrics.

HS2_28days

Number of days with HS values greater than 2 °C in the last 28 days¹⁶.

0–4 (days)

Not selected in the final model but high collinearity with different heat stress metrics.

HS2_84days

Number of days with HS values greater than 2 °C in the last 84 days¹⁶

0–4 (days)

Not selected in the final model but high collinearity with different heat stress metrics.

HS_84days

The sum of HS in 84 days before the sampling date.

19.94–108.30 (°C)

Not selected in the final model but high collinearity with different heat stress metrics.

HS_555days

The sum of HS in 555 days before the sampling date.

90.13–332.12 (°C)

Not selected in the final model but high collinearity with different heat stress metrics.

MMM

Monthly average of the hottest month registered during 1985-2012³⁸

28.62–29.06 (°C)

Not selected in the final model but high correlation with some thermal metrics.

RFI

Reef Functional Index based on the relative abundance of species multiplied by a functional coefficient⁷¹

0.1800–0.7679 (index value)

Not selected in the final model.

Richness

Number of coral species.

5–34 (species)

Not selected in the final model but high collinearity with diversity.

ROTC_84days

Maximum ROTC observed 84 days before sampling date⁹.

0.1667–0.3138 (°C by week)

Not selected in the final model but high collinearity with ROTC_clim.

SD_84days

Standard deviations of the SST in the previous 84 days.

0.2091–1.1063 (°C)

Not selected in the final model.