Skip to contents

Coral Reef Watch Heat Stress Products in Google Earth Engine

Implementation of Skirving et al. (2020) using NOAA OISST

Overview

This document describes five Google Earth Engine (GEE) JavaScript scripts that implement the Coral Reef Watch (CRW) heat stress product suite following the methodology of @skirving2020. The scripts compute, for each grid cell within a defined region of interest:

  1. Monthly Mean (MM) Climatology — 12 regression-adjusted SST values
  2. Maximum Monthly Mean (MMM) Climatology — the annual thermal maximum
  3. Daily Climatology (DC) — 366-day interpolated baseline
  4. SST Anomaly — departure from the daily climatological mean
  5. Coral Bleaching HotSpot (HS) — instantaneous stress above MMM
  6. Degree Heating Weeks (DHW) — accumulated thermal stress over 12 weeks

All scripts use the NOAA CDR OISST v02r01 dataset (NOAA/CDR/OISST/V2_1) at 0.25° spatial resolution, available in the GEE public data catalogue.

Region of Interest

The region of interest covers the Great Barrier Reef:

xmin: 141.0958  ymin: -24.70584
xmax: 153.2032  ymax:  -8.926405

Defined in each script as:

var ROI = ee.Geometry.Rectangle([141.0958, -24.70584, 153.2032, -8.926405]);

All output images are clipped to this ROI via .clip(ROI) to restrict computation and display.

Important Note on OISST Resolution

The operational CRW products use CoralTemp at 0.05° (~5 km) resolution, blending geostationary and polar-orbiting satellite SST. This implementation uses OISST at 0.25° (~25 km), which is the highest-resolution daily SST product available in GEE. Results will differ from official CRW products in absolute values but follow the identical methodology.

Table 1: Comparison of operational CRW products with this GEE implementation.
Feature CRW v3.1 This Implementation
SST source CoralTemp (5 km) NOAA OISST (25 km)
Resolution 0.05° 0.25°
Gap filling Multi-sensor blend OI interpolation
Platform Custom pipeline Google Earth Engine

Script 01: Monthly Mean and Maximum Monthly Mean Climatology

File: 01_mm_mmm_climatology.js

Method

The Monthly Mean (MM) climatology is a set of 12 SST values representing the regression-adjusted average SST for each calendar month, computed over the period 1985–2012.

For each month \(m \in \{1, 2, \ldots, 12\}\):

  1. Compute the monthly mean SST for each year \(y \in \{1985, \ldots, 2012\}\), yielding 28 values.
  2. Apply ordinary least squares (OLS) linear regression: \[\overline{\text{SST}}_{m,y} = \beta_0 + \beta_1 \cdot y + \varepsilon\]
  3. Evaluate the regression at the target year \(y^* = 1988.2857\): \[\text{MM}_m = \hat{\beta}_0 + \hat{\beta}_1 \cdot 1988.2857\]

The target year 1988.2857 is the mean of the years used in the original CRW climatology (1985–1990 and 1993), chosen to maintain consistency across product versions.

The Maximum Monthly Mean is then:

\[\text{MMM} = \max_{m=1}^{12} \text{MM}_m\]

The MMM represents the warmest monthly climatological temperature at each pixel and serves as the threshold above which thermal stress is defined.

Memory Optimisation

OISST images are global single-tile rasters. A naive nested mapping over 28 years × 12 months creates a 336-image collection that exceeds GEE’s user memory limit. The script avoids this by processing each month independently in a client-side loop:

function computeMMForMonth(month) {
  // Loads only 28 images (one per year for this month)
  // Runs regression and returns a single image
}

var mmImages = [];
for (var m = 1; m <= 12; m++) {
  mmImages.push(computeMMForMonth(m).rename('mm_' + ...));
}

This ensures GEE never holds more than 28 images in memory simultaneously.

Outputs

  • mmMultiBand: 12-band image (mm_01 to mm_12), one per calendar month
  • mmm: single-band image, the per-pixel maximum of the 12 MM values

Both are exported as EE assets and GeoTIFFs for reuse in subsequent scripts.

Script 02: Daily Climatology

File: 02_daily_climatology.js

Method

The 12 MM values are interpolated to produce a daily climatology (DC) of 366 values (one per day of year, including leap day). Each MM value is assigned to the 15th of its corresponding month:

Table 2: Day-of-year anchor points for the 15th of each month (non-leap year).
Month Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Anchor DOY 15 46 74 105 135 166 196 227 258 288 319 349

For any day-of-year \(d\) falling between anchor points \(d_{\text{lo}}\) (month \(m_{\text{lo}}\)) and \(d_{\text{hi}}\) (month \(m_{\text{hi}}\)):

\[\text{DC}_d = \text{MM}_{m_{\text{lo}}} + \left(\text{MM}_{m_{\text{hi}}} - \text{MM}_{m_{\text{lo}}}\right) \cdot \frac{d - d_{\text{lo}}}{d_{\text{hi}} - d_{\text{lo}}}\]

The interpolation wraps around from December (DOY 349) to January (DOY 15) by introducing virtual anchor points at DOY −16 (December of the “previous year”) and DOY 380 (January of the “next year”).

Output

  • dcImage: 366-band image (dc_001 to dc_366), each band being the interpolated SST climatology for that day of year.

Script 03: SST Anomaly

File: 03_sst_anomaly.js

Method

The daily SST Anomaly is the departure of the observed SST from the daily climatological mean:

\[\text{SST Anomaly}_i = \text{SST}_i - \text{DC}_d\]

where \(i\) is the observation date and \(d\) is the corresponding day of year.

The script uses server-side array indexing to select the correct DC band for each image’s day of year:

var dcArray = dcAllBands.toArray();  // 366-element array image

var sstAnomalyCollection = oisstAnalysis.map(function(img) {
  var doy = ee.Date(img.get('system:time_start'))
    .getRelative('day', 'year').add(1).min(366);
  var dcForDay = dcArray
    .arraySlice(0, doy.subtract(1), doy)
    .arrayProject([0])
    .arrayFlatten([['dc_sst']]);
  return img.subtract(dcForDay).rename('sst_anomaly');
});

Output

  • sstAnomalyCollection: daily SST anomaly images for the analysis period. Positive values indicate warmer-than-average conditions; negative values indicate cooler-than-average conditions.

Script 04: Coral Bleaching HotSpot

File: 04_hotspot.js

Method

The Coral Bleaching HotSpot (HS) measures the instantaneous thermal stress above the climatological maximum:

\[\text{HS}_i = \max(\text{SST}_i - \text{MMM},\; 0)\]

The HotSpot is always zero or positive. When the observed SST is below the MMM, the HotSpot is set to zero — there is no negative thermal stress in this framework.

The MMM is a static field (one value per pixel), whereas the HotSpot varies daily with the observed SST.

Interpretation

Table 3: HotSpot interpretation thresholds.
HotSpot (°C) Interpretation
0 No thermal stress
0 – 1 Below bleaching watch threshold
≥ 1 Significant thermal stress; contributes to DHW

Output

  • hotspotCollection: daily HotSpot images (°C above MMM).

Script 05: Degree Heating Weeks

File: 05_dhw.js

Method

The Degree Heating Week (DHW) product accumulates thermal stress over a rolling 12-week (84-day) window. Only HotSpot values ≥ 1°C contribute to the summation. Daily values are divided by 7 to convert from degree-days to degree-weeks:

\[\text{DHW}_i = \sum_{n=i-83}^{i} \frac{\text{HS}_n}{7}, \quad \text{where } \text{HS}_n \geq 1\,°\text{C}\]

Worked Example

If the 84-day window ending on 1 June 2024 contained only four daily HotSpot values above zero — 1.0, 2.0, 0.8, and 1.2°C — then:

  • 0.8°C is excluded (below the 1°C threshold)
  • Contributing values: 1.0, 2.0, 1.2°C
  • \(\text{DHW} = (1.0 + 2.0 + 1.2) / 7 = 0.6\,°\text{C-weeks}\)

Bleaching Thresholds

Table 4: DHW bleaching thresholds following @liu2014.
DHW (°C-weeks) Alert Level Expected Impact
0 No stress —
0 – 4 Watch Possible bleaching
4 – 8 Warning Significant bleaching likely
≥ 8 Alert Level 2 Severe bleaching and mortality likely

Computational Considerations

Computing DHW for every day of a year-long analysis period creates 365 derived images, each requiring the summation of an 84-image window. For long periods or large regions, consider:

  • Processing in monthly batches
  • Exporting the daily HotSpot collection first, then computing DHW offline
  • Using the automated pipeline (see companion document)

Output

  • dhwCollection: daily DHW images (°C-weeks) for the analysis period.

Data Flow Summary 

flowchart LR
    OISST[NOAA OISST<br/>Daily SST] --> MM[Monthly Mean<br/>12 values]
    MM --> MMM[Maximum<br/>Monthly Mean]
    MM --> DC[Daily Climatology<br/>366 values]
    OISST --> ANOM[SST Anomaly<br/>SST − DC]
    DC --> ANOM
    OISST --> HS[HotSpot<br/>max SST−MMM, 0]
    MMM --> HS
    HS --> DHW[Degree Heating Weeks<br/>84-day Σ HS/7, HS≥1]
Figure 1: Data flow from raw OISST to DHW products.

Running the Scripts

  1. Open the GEE Code Editor
  2. Paste 01_mm_mmm_climatology.js → Run → Execute export tasks in the Tasks tab
  3. Update asset paths in scripts 02–05 to point to your exported MM/MMM assets
  4. Run scripts 02–05 in sequence

For the asset path, replace the placeholder:

// var mmm = ee.Image('projects/YOUR-GEE-PROJECT/assets/coral_dhw/mmm_climatology');

with your actual asset path after exporting from Script 01.