Stats NZ’s 2 October 2026 update, “Data to 2025,” reports long-run trends in an extreme-rainfall indicator at 30 climate sites across Aotearoa New Zealand. Between 1960 and 2025, the share of annual rainfall falling on very wet days very likely increased at 11 sites and very likely decreased at three. The source does not assign either of these two categories to the remaining sites. It would therefore be inaccurate to call the other 16 sites stable or unchanged.

The measure calculates what proportion of a year’s rainfall comes from very wet days; it does not measure a single daily record or total rainfall. The technical metric is R95pTOT: annual rainfall on days classed as very wet divided by annual rainfall. A very wet day is one when daily rainfall exceeds the 95th percentile of daily totals calculated over the 1991–2020 climate normal. The analysis applies a minimum daily total of 1 millimetre. This threshold identifies unusually wet days relative to the reference climate period.

A share metric describes how rainfall is distributed through the year. A rising share does not necessarily mean that annual rainfall increased: the share can rise while total rainfall stays similar, and total rainfall can rise without a change in the share. A trend at one station is also not equivalent to a single national-average rainfall regime. Stats NZ publishes total annual rainfall as a separate indicator in the same update. This article focuses only on the trend categories for the proportion contributed by very wet days.

The site network is not an evenly spaced grid covering every part of the country. The Ministry for the Environment’s technical annex says the atmosphere and climate indicators use observations from 30 NIWA climate stations. Sites are distributed to capture places where many people live; coastal and lower-elevation areas are more represented, while inland high elevations are less represented. This design matters when interpreting the geographic scope. Results from the network cannot automatically be extended to every catchment, town, or parcel of land.

The phrase “very likely” describes a confidence category for the direction of a statistical trend, not the magnitude of a change. Likelihood categories summarize uncertainty in trends observed at the sites. They do not say that a particular future rainfall event will certainly occur or that flooding will happen in a specified location. Likewise, sites not classified in the reported increase or decrease groups should not be assigned an unreported direction. The figures 11 and three are counts of classified outcomes across the 30-site network.

The trend period runs from 1960 through 2025, with 2025 as the latest data year rather than the starting point. The daily rainfall threshold uses the 1991–2020 normal, providing a common reference for identifying very wet days within the longer observational record. The trend period and threshold reference period therefore differ. The indicator is based on annual series derived from station observations and a trend assessment; it is not a short-term weather forecast, a projection for next year, or an estimate of climate-related damages.

The reported finding is that the very-wet-day share of annual rainfall very likely rose at 11 sites and very likely fell at three. That distribution does not support a claim of one uniform direction at every location in the country. Because a trend category does not report the amount of change, the 11 and three counts alone also cannot compare the size of changes between sites. The results inform monitoring of an extreme-rainfall indicator; they do not, by themselves, determine water-management decisions, infrastructure risk, or climate attribution.