Start where the cascade starts. … of the … usable Pacific series in the SPC emissions file sit below 4 tonnes of CO₂-equivalent per person per year. Kiribati is at …, the Solomon Islands at …, Tuvalu at …. New Caledonia is the one exception at …, and that number is nickel smelting. Territorial accounting books an export industry’s output to the place it happens, so the figure is describing a smelter rather than a way of living.
There are … further series in the same file that I have kept out of that count, and the reason is the bookkeeping rather than the places. Palau appears at … tonnes per person in … and still … by …, and at its peak it reaches …, about … times the highest value any other series in this file ever reaches. That reads as fuel bunkering or a population denominator in the accounting rather than a footprint. The other five, Nauru, the Marshall Islands, Guam, American Samoa and the Northern Marianas, sit between … and … tonnes across … years and take at most … distinct values each; … never change at all. That is a placeholder holding a row open rather than a record of anything. All six stay on the chart in grey and say so on hover. A gap you can see seemed more useful than a country that quietly disappears.
GHG Emissions · Pacific territories · tonnes CO₂-eq per person
Surface temperature anomaly · Pacific territories
Each territory carries its own surface-temperature anomaly for the selected year. Double-click one to zoom in and open its record beside the map.
Hover a territory for its reading. Double-click to zoom in.
The SPC sea-surface temperature file runs … across … countries and territories. Averaging all of them, the ten warmest years in that … record are …. I did not choose that list. It is the top ten as the file ranks them, and not one twentieth-century year is in it.
Through the 1961 to 1990 climatological baseline the regional mean sat at … °C. Across 1995 to 2024 it sits at … °C. Against the 1850 to 1900 pre-industrial window, the last ten complete years average … °C. Three different baselines, one direction.
The sawtooth riding on the trend is ENSO, and it is worth shading in rather than leaving a reader to trip over it. El Niño years lift regional sea-surface temperature by several tenths of a degree and La Niña years pull it back; 2016 and 2024 are both El Niño peaks. Without those bands the same series looks noisier than it is.
The ocean takes up roughly 91% of the excess energy in the climate system (IPCC AR6 WGI Ch. 9). That is where the cascade splits in two, and the next two sections follow one branch each.
Sea-surface temperature anomaly · SST_ANOM · every territory · every year
One row per territory, one cell per year, on the same scale as the map above. …
A regional signal, built from national series too coarse to read one at a time
The SPC sea-level file reports to 0.1 metres, so every observation in it is one of five values: −0.2, −0.1, 0.0, 0.1 or 0.2 metres. That is the resolution of the file, not a property of the places in it. Kiribati reads 0.0 for its first … years because the true value rounded to zero, and I would rather say that plainly than draw a flat line and let it look like a finding.
Averaged across all … territories, the rounding cancels out. The regional mean rises at … mm per year across …, which is the same range published satellite-altimetry estimates give for the tropical Pacific over those years. So here is what I think this file can honestly show: not which island is most exposed, but that the Pacific is rising.
Sea-level rise · Pacific regional mean · 1993 to 2023
The same trend means different things depending on what sits underneath it
Rainfall anomaly is not moving one way across the Pacific. Fitting a linear trend to each of the … national series in RAIN_ANOM (…) splits the region almost evenly: … territories trending drier, … wetter. On its own that is close to a coin toss, and it would be easy to stop there.
What caught my attention is which ones. The four steepest drying trends are …. Kiribati and Tuvalu are atoll states and Nauru is a single raised coral island, and not one of the three has a river, a lake or a highland catchment. Their fresh water sits as a thin lens above seawater underground and is recharged by rain and by nothing else. Identical millimetres mean something quite different there than they do in New Caledonia, which is the other steep one and does have rivers.
… are trending wetter, and I would not read that as the good half of the split. Drainage, cropping calendars and catchment sizing are all built around a distribution, and moving in either direction moves away from what they were built for.
One more thing worth putting beside the trend rather than instead of it: Nauru’s year-to-year standard deviation is … mm, about … times New Caledonia’s … mm and the widest of any territory in the file. The line runs downward and the swing around it is that wide, so the average year is a weaker guide there than almost anywhere else on this page.
Rainfall anomaly · mm/decade trend · Pacific territories · 1979 to 2025
Two measured things, and no line drawn between them
The SDG 6.1.1 series tracks the share of each population using a safely managed drinking-water service. Kiribati, which carries one of the four steepest drying trends above, moved from …% in 2000 to …% in 2022. That is real progress over twenty-two years, and the national figure also hides a second number I think is more useful: coverage in Kiribati’s urban areas runs … percentage points ahead of its rural areas. An urban-rural distance like that turns up across the region, so it reads to me as a feature of dispersed geography rather than of any one country.
The widest gap in the file is …’s, at … points on national coverage of …%. Vanuatu is at …, Fiji at …. … territories report lower coverage in 2022 than in 2000: …. Reported definitions and survey methods shift over two decades, so I would treat those three as something to look into rather than as a verdict on anyone.
Rainfall is the recharge term for a rain-fed island’s freshwater lens, and in several of these places it is the only one. That is the mechanism, and it is exactly why I have drawn this link dashed instead of solid. The rainfall trend is measured. The water coverage is measured. Nothing in either dataset connects one to the other, and I am not going to draw a line the data has not earned.
Urban vs rural safely managed water access · sorted by gap
National coverage 2000 → 2022
Two million transects, no model
Digital Earth Pacific’s rates-of-change layer fits a trend to every
shoreline transect across the region and reports metres gained or lost per year.
… of them fall inside one of the twenty-two
Pacific territories here, and … of those carry
Digital Earth Pacific’s own good quality flag. Those are the only
ones I have counted, using their threshold rather than one of mine. Red retreats, blue
builds.
The pattern is not uniform retreat, and the literature agrees. Duvat’s 2019 synthesis of 709 islands and Kench’s 2018 work on Tuvalu both find most atoll islands stable or gaining area as reef-derived sediment moves around. Area and habitability are not the same measurement, though, and I want to be careful not to let one stand in for the other. An island can gain area on its ocean side and lose the village on its lagoon side in the same decade, and this layer would record that as a gain.
So this section measures an outcome. It does not divide that outcome between the links above, and neither will I.
Rate of shoreline change · median and interquartile range of each territory’s good-quality transects
The bar is the middle half of transects in that territory; the dot is the median. Nearly every territory straddles zero, and that is the observation I would pull out of this chart: the same coastline is building in one place and retreating in another, so a single national number for shoreline change hides more than it tells you. The tails run wider than the bars and are in each tooltip. … has the fewest transects surviving the quality flag of anywhere here, and with so few left its top decile reaches … m/yr, so I have drawn quartiles rather than deciles to keep every territory readable on one axis.
Shoreline balance · share of transects significantly retreating or accreting · sorted by net
Counted with Digital Earth Pacific’s own threshold, taken from
their published map project: sig_time ≤ 0.01 and
|rate_time| ≥ 0.3 m/yr. Transects between those bounds are neither
retreating nor accreting, and I have left them out of both bars rather than rounding them
into one. The grey remainder is the rest of the coastline.
The cascade is measured everywhere except where it matters most
The heat is in the record: the ten warmest years, every one of them this
century.
The sea anomalies are in the record, marking a Pacific that keeps getting less
settled.
The rise is in the record once the noise is averaged away, at
… millimetres a year.
The rainfall shift is in the record, uneven, and three of its four steepest drying trends sit
on islands with no rivers.
The shoreline change is in the record, traced across two million transects.
And the gas is in the record, accumulating upstream of all of it.
Between the gas and the shoreline sits a chain of measurable shifts: gas → heat → sea anomalies → sea-level rise → rainfall → shoreline.
Every solid link on this page exists because somebody decided that quantity was worth counting. Every dashed one is where I ran out of measurements, not where the connection stops.
What I cannot show you is what happens in between. That is not a gap in the Pacific. It is a gap in what has been chosen for measuring.