22 Pacific Island countries and territories. More than 170 years of climate records. A region facing rising temperatures, rising seas and growing climate pressure — with far fewer resources to absorb the impact. Here's what the data shows.
The recent picture is anything but simple. In 2025, the wider Pacific typhoon basin recorded 27 named storms, including 13 typhoons, linked to 653 deaths and about US$10.8 billion in damage. Yet the 2025–26 South Pacific cyclone season, covering countries such as Fiji, Tonga, Vanuatu and Samoa, produced only 2 cyclones — a record low, tied with 1990–91.
Storm statistics: 2025 Pacific typhoon season and 2025–26 South Pacific cyclone season records. Elevation exposure: small island developing states (SIDS) literature. See references in the footer.
Rather than throwing every indicator at the story, we let the data pick: a technique called forward selection adds one predictor at a time, keeping whichever most improves the model, and stops once the rest stop pulling their weight.
Sea-level anomalies were selected first. Adding greenhouse-gas emissions, rainfall anomalies and crop yield produced a final model with an adjusted R² of 0.64. These four indicators showed the strongest combined relationship with regional temperature in the selected model.
This is an exploratory, correlational model based on 28 annual observations. It identifies patterns that move together; it does not prove that one indicator causes another.
Across the regional series, temperatures have risen substantially since the 1850s. Year-to-year values still move up and down, but the long-term direction is clear.
Sea level is one of the clearest and most consistent signals in the dataset. 21 of 22 territories have sea-level data, and all 21 show a rising trend (Pitcairn has no sea-level records), with most estimates falling within a relatively narrow range.
The direction is remarkably consistent. That makes elevation and exposure especially important: low-lying atoll communities have less physical room to absorb the same change in sea level.
The relationships are not equally clear across every indicator. Crop yield showed no statistically significant regional trend in this analysis (p=0.40). That is important too: climate signals are not uniform, and the absence of a clear trend should not be turned into a stronger claim than the data supports.
For the forward-looking view, I compared a straight-line trend with a second-degree polynomial. Sea level was better represented by the linear model, while temperature fitted the quadratic model more closely. Both are shown here because extrapolation becomes more uncertain the further it moves beyond the observed data.
Across climate policy, Pacific resilience and renewable-energy planning, many of these priorities are already being pursued. The challenge is turning them into action at the scale and speed the data suggests is needed.