An interactive dashboard for exploring Sea Surface Temperature (SST) anomalies and Marine Heatwave (MHW) events from global ocean data spanning 1881–2026.
Sea Surface Temperature (SST) is the temperature of the uppermost layer of the ocean (top 1 mm to 10 m). Tracking SST anomalies — deviations from a long-term climatological baseline — reveals warming trends, regional variability, and the growing frequency of extreme ocean heat events.
Marine Heatwaves (MHW) are prolonged periods of anomalously warm ocean temperatures. They are formally defined (Hobday et al., 2016) as ≥5 consecutive days where SST exceeds the local 90th-percentile climatological threshold. MHWs have significant ecological and economic impacts, affecting coral reefs, fisheries, and coastal ecosystems.
The dashboard enables intuitive spatial-to-temporal climate exploration: users can scan global SST anomaly maps through time, detect long-term warming trends and extreme events at any location, and visualise MHW frequency and intensity year by year.
| Layer | Libraries |
|---|---|
| Data & computation | xarray, numpy, pandas, dask, scipy |
| Storage | zarr, netCDF4 |
| Visualisation | hvPlot, HoloViews, Bokeh |
| Dashboard | Panel |
| ML | scikit-learn |
The project uses the NOAA Optimum Interpolation (OI) SST V2 High Resolution dataset. The dataset covers Sea Surface Temperature from 1881 to 2026 with weekly and biweekly resolution.
You will need to download the following datasets and place them in the parent directory inside a data folder:
- sst.week.mean.nc
- all daily data named as followed: sst.day.mean.{year}.nc
The app uses a conda environment where all the librairies are installed. To create the environment, run the following commands:
cd MHW-Detection
conda env create -f environment.yml
conda activate mhw-detection
Once the installation done you will first run the preprocess.py to create all the files needed to run the Dashboard.
python preprocess.py
panel serve app/interactive_map_panel.py --show
The dashboard has three tabs, each providing a different view of SST data.
A video of the changes of SST anomalies in the world across time. The diverging colormap (blue = cooler, red = warmer) is centered on zero and clipped to ±5 °C, making it easy to scan regional warming and cooling events through time.
A spatial-to-temporal click-to-inspect workflow:
-
Variability Map (left): displays the mean monthly standard deviation of SST anomalies across years, highlighting where the ocean fluctuates most. Clicking any location triggers the right panel.
Note: this map shows interannual variability, not heatwave intensity.
-
Interactive analyses (right): clicking a grid cell generates three stacked plots:
- OLS trend: long-term linear trend in °C/decade estimated by Ordinary Least Squares regression.
- Extreme events: time series of SST anomalies with points above the 95th-percentile threshold highlighted in red.
- Event count histogram + KDE: number of extreme events per year with a kernel-density curve to reveal multi-decadal shifts in frequency.
Below an example of the Anomaly Explorer Tab
Marine Heatwave (MHW) detection of 5 or more consecutive days where SST exceeds the local 90th-percentile climatological threshold.
- Metric selector: switch between days per year and events per year.
- Year slider: pan through annual MHW maps to inspect spatial patterns.
- Click-to-inspect: clicking the map plots a bar chart + KDE of the selected metric at that location across all available years.
Below an example of the Marine HeatWave Tab
(some code)
# to visualize the training
tensorboard --logdir forecast/runs
- Reynolds, R.W., N.A. Rayner, T.M. Smith, D.C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 1609-1625.
- Hobday, Alistair J., et al. "A hierarchical approach to defining marine heatwaves." Progress in oceanography 141 (2016): 227-238.