The numbers on this page aren’t typed in by hand. The weights, thresholds, worked example and dataset details are produced from the code and data behind the scores each time the site is built, so if the method changes, this page changes with it.
The month score
Every island gets a score out of 100 for each month: how good the weather and the sea usually are for a holiday spent outdoors. It’s built from long-term averages, not a forecast, so it tells you what that month is usually like.
The score has six parts. Each one earns between no marks and full marks from a single reading, and the marks are then weighted and added up.
- Warmth: the average daily high. Too cool and too hot both lose marks.
- Sea: the average sea temperature offshore. Lagoons and sheltered bays usually run a little warmer. An island with no sea reading takes its warmth mark here.
- Rain: the rain that falls in the month. The less, the better.
- Sunshine: the sun’s energy reaching the ground, from NASA’s POWER model.
- Storms: tropical storms expected in the month, from 30 seasons of storm tracks.
- Humidity: the average relative humidity. Drier air feels more comfortable.
There are two profiles, each with its own weights and thresholds: one for the 18 tropical islands and one for New Zealand’s eight islands. Between the reading that earns full marks and the one that earns none, the mark changes in a straight line.
| Part | Weight | Full marks | No marks |
|---|---|---|---|
| Warmth | 20% | 26–30.5 °C | 22 °C or below, or 34 °C or above |
| Sea | 20% | 25.5–29.5 °C | 22 °C or below, or 32 °C or above |
| Rain | 25% | 70 mm or less | 300 mm or more |
| Sunshine | 0% | Not counted | |
| Storms | 25% | None expected | 0.2 a season or more |
| Humidity | 10% | 74% or less | 84% or more |
| Part | Weight | Full marks | No marks |
|---|---|---|---|
| Warmth | 35% | 21–26 °C | 12 °C or below, or 32 °C or above |
| Sea | 10% | 18–24 °C | 12 °C or below, or 28 °C or above |
| Rain | 25% | 50 mm or less | 180 mm or more |
| Sunshine | 30% | 6.5 kWh/m² a day or more | 1.5 kWh/m² a day or less |
| Storms | 0% | Not counted | |
| Humidity | 0% | Not counted | |
Storm counts are smoothed before they’re scored: each month’s count becomes 25% of the month before, 50% of the month itself and 25% of the month after. A record of 30 seasons is short for something this rare, and smoothing stops one stray storm from making a single month look much worse than the months either side.
Ratings and the good season
Each month’s score, rounded to a whole number, gets a rating:
| Rating | Score |
|---|---|
| Excellent | 80 to 100 |
| Good | 65 to 79 |
| Fair | 50 to 64 |
| Poor | 0 to 49 |
An island’s good season is every month within 10 points of its best month. It’s the line over every island’s chart: for Rarotonga, “Best from May to November”.
A worked example: Rarotonga
Here are Rarotonga’s best and worst months, worked out from the same data as its chart. Each part’s mark is multiplied by its weight, and the points add up to the score.
Rarotonga in June 93 out of 100, excellent
| Part | Reading | Mark | Weight | Points |
|---|---|---|---|---|
| Warmth | 26.4 °C | 100% | 20% | 20.0 |
| Sea | 25.2 °C | 91% | 20% | 18.3 |
| Rain | 87 mm | 93% | 25% | 23.1 |
| Sunshine | 3.5 kWh/m² a day | 14% | 0% | 0.0 |
| Storms | None | 100% | 25% | 25.0 |
| Humidity | 78% | 62% | 10% | 6.2 |
| Score | 100% | 92.6 Rounds to 93 | ||
Rarotonga in February 64 out of 100, fair
| Part | Reading | Mark | Weight | Points |
|---|---|---|---|---|
| Warmth | 29.6 °C | 100% | 20% | 20.0 |
| Sea | 27.4 °C | 100% | 20% | 20.0 |
| Rain | 192 mm | 47% | 25% | 11.8 |
| Sunshine | 6.1 kWh/m² a day | 89% | 0% | 0.0 |
| Storms | 0.13 a season | 38% | 25% | 9.4 |
| Humidity | 81% | 29% | 10% | 2.9 |
| Score | 100% | 64.0 Rounds to 64 | ||
June beats February by 29 points. Most of the gap is storms (15.6 points) and rain (11.4 points), though February does better on sea. Sunshine earns June only 14% of its marks, but it carries no weight in the tropical profile. The next section explains why.
Why the two profiles differ
Why the tropical score leaves out sunshine
The sunshine reading is the sun’s energy reaching the ground, from NASA’s POWER model. In the tropics it follows the height of the sun more than the cloud, so it’s lowest in midwinter, and in the tropics midwinter is often the best time to go. On 17 of the 18 tropical islands, the month with the least solar energy falls in the good season. Counting it would mark the best months down. Cloud and wet weather still cost marks, through rain.
In New Zealand the sun’s energy peaks in midsummer, the best time to go, so sunshine carries 30% of the temperate score.
Why New Zealand’s islands leave out storms
The temperate profile is an outdoor-comfort score: is it warm, dry and bright enough to be outside? Storms carry no weight in it. Few reach these islands at storm strength: each had between 0 and 6 pass within 200 km in 30 seasons, against as many as 31 for a tropical island (Tongatapu). The rain they bring still counts, through the rain part.
How storms are counted
The storm figures come from IBTrACS, the international archive of tropical storm tracks, published by the US National Oceanic and Atmospheric Administration (NOAA).
We count every storm with winds of at least 34 knots (63 km/h) whose track passed within 200 km of an island’s centre between 1 January 1991 and 31 December 2020, from the South Pacific, East Pacific and South Indian basin records. Each storm counts once, in the month it came closest. Dividing each month’s count by 30 gives the storms expected in that month in a typical season.
Winds of 34 knots are tropical-storm strength. In the South Pacific a storm that strong is a tropical cyclone; around Hawaiʻi the count takes in tropical storms as well as hurricanes. It counts storms that came near, not the harm they did.
Storms near each tropical island
| Island | In 30 seasons | A season |
|---|---|---|
| Tongatapu | 31 | ≈ 1.0 |
| Efate | 27 | ≈ 0.9 |
| Vavaʻu | 20 | ≈ 0.7 |
| Mamanuca Islands | 16 | ≈ 0.5 |
| Yasawa Islands | 16 | ≈ 0.5 |
| Taveuni | 16 | ≈ 0.5 |
| Viti Levu | 15 | ≈ 0.5 |
| Rarotonga | 14 | ≈ 0.5 |
| Upolu | 14 | ≈ 0.5 |
| Savaiʻi | 14 | ≈ 0.5 |
| Niue | 14 | ≈ 0.5 |
| Aitutaki | 10 | ≈ 0.3 |
| Bora Bora | 6 | ≈ 0.2 |
| Maui | 6 | ≈ 0.2 |
| Oʻahu | 5 | ≈ 0.2 |
| Moʻorea | 4 | ≈ 0.1 |
| Tahiti | 2 | ≈ 0.1 |
| Bali | 0 | None |
Where temperatures are estimates
Temperatures and rain come from a nearby weather station’s long-term records wherever there’s one, through Meteostat. For 10 islands there’s no station near enough for temperatures: Aitutaki, Yasawa Islands, Taveuni, Upolu, Savaiʻi, Tongatapu, Vavaʻu, Niue, Aotea Great Barrier Island and Kawau Island.
For those, we estimate. NASA’s POWER model gives temperatures for every point on Earth, but over a small island its grid cell is mostly ocean, so its days come out too cool and its nights too warm. We take the model’s figures and correct them, month by month, by the average gap between the model and the stations on the islands that have one, worked out separately for tropical islands and for New Zealand’s islands:
| Islands | Daily high | Daily low | Stations used |
|---|---|---|---|
| Tropical islands | +2.7 to +3.0 °C | −2.8 to −2.3 °C | 10 |
| New Zealand’s islands | +0.7 to +1.7 °C | −2.8 to −1.2 °C | 6 |
The corrections, month by month
| Islands | Jan | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep | Oct | Nov | Dec |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Tropical islands, highs | +2.9 | +2.9 | +3.0 | +2.9 | +2.7 | +2.7 | +2.7 | +2.9 | +2.9 | +2.9 | +2.9 | +2.8 |
| Tropical islands, lows | −2.6 | −2.6 | −2.5 | −2.5 | −2.8 | −2.6 | −2.6 | −2.4 | −2.4 | −2.3 | −2.3 | −2.4 |
| New Zealand’s islands, highs | +1.7 | +1.7 | +1.5 | +1.1 | +1.0 | +0.7 | +1.0 | +1.3 | +1.7 | +1.6 | +1.3 | +1.5 |
| New Zealand’s islands, lows | −1.6 | −1.6 | −2.1 | −2.7 | −2.5 | −2.8 | −2.7 | −2.3 | −1.6 | −1.2 | −1.5 | −1.4 |
The estimated islands take their rain from the model too. Each island’s chart says where its numbers came from, and calls an estimate an estimate.
Where each island’s numbers come from
| Island | Temperature | Rain |
|---|---|---|
| Rarotonga | Rarotonga weather station, 11 km away, 1991–2020 normals (Meteostat) | Same |
| Aitutaki | Estimated (no nearby station): NASA POWER model, adjusted by the gap between model and stations elsewhere | NASA POWER model, 2001–2025 (no nearby station) |
| Viti Levu | Nadi weather station, 2 km away, 1991–2020 normals (Meteostat) | Same |
| Mamanuca Islands | Nadi weather station, 32 km away, 1991–2020 normals (Meteostat) | Same |
| Yasawa Islands | Estimated (no nearby station): NASA POWER model, adjusted by the gap between model and stations elsewhere | NASA POWER model, 2001–2025 (no nearby station) |
| Taveuni | Estimated (no nearby station): NASA POWER model, adjusted by the gap between model and stations elsewhere | NASA POWER model, 2001–2025 (no nearby station) |
| Upolu | Estimated (no nearby station): NASA POWER model, adjusted by the gap between model and stations elsewhere | NASA POWER model, 2001–2025 (no nearby station) |
| Savaiʻi | Estimated (no nearby station): NASA POWER model, adjusted by the gap between model and stations elsewhere | NASA POWER model, 2001–2025 (no nearby station) |
| Tongatapu | Estimated (no nearby station): NASA POWER model, adjusted by the gap between model and stations elsewhere | NASA POWER model, 2001–2025 (no nearby station) |
| Vavaʻu | Estimated (no nearby station): NASA POWER model, adjusted by the gap between model and stations elsewhere | NASA POWER model, 2001–2025 (no nearby station) |
| Niue | Estimated (no nearby station): NASA POWER model, adjusted by the gap between model and stations elsewhere | NASA POWER model, 2001–2025 (no nearby station) |
| Tahiti | Tahiti Faʻaʻā Airport weather station, 2 km away, 1991–2020 normals (Meteostat) | Same |
| Moʻorea | Tahiti Faʻaʻā Airport weather station, 26 km away, 1991–2020 normals (Meteostat) | Same |
| Bora Bora | Bora Bora (Faanui) weather station, 6 km away, 1991–2020 normals (Meteostat) | Same |
| Oʻahu | Honolulu Airport weather station, 14 km away, 1991–2020 normals (Meteostat) | Same |
| Maui | Kahului Airport weather station, 16 km away, 1991–2020 normals (Meteostat) | Same |
| Bali | Denpasar Ngurah Rai Airport weather station, 7 km away, 1991–2020 normals (Meteostat) | Same |
| Efate | Port Vila Bauerfield Airport weather station, 5 km away, 1991–2020 normals (Meteostat) | Same |
| Waiheke Island | Auckland Airport weather station, 34 km away, 1991–2020 normals (Meteostat) | NASA POWER model, 2001–2025 (no nearby station) |
| Rangitoto Island | Auckland Airport weather station, 26 km away, 1991–2020 normals (Meteostat) | NASA POWER model, 2001–2025 (no nearby station) |
| Aotea Great Barrier Island | Estimated (no nearby station): NASA POWER model, adjusted by the gap between model and stations elsewhere | NASA POWER model, 2001–2025 (no nearby station) |
| Rakiura Stewart Island | Invercargill Airport weather station, 39 km away, 1991–2020 normals (Meteostat) | Same |
| Tiritiri Matangi | Auckland Airport weather station, 47 km away, 1991–2020 normals (Meteostat) | NASA POWER model, 2001–2025 (no nearby station) |
| Kawau Island | Estimated (no nearby station): NASA POWER model, adjusted by the gap between model and stations elsewhere | NASA POWER model, 2001–2025 (no nearby station) |
| Chatham Islands | Chatham Islands weather station, 1 km away, 1991–2020 normals (Meteostat) | Same |
| Kāpiti Island | Paraparaumu Airport weather station, 8 km away, 1991–2020 normals (Meteostat) | Same |
Who it suits
Each island guide scores its island from 0 to 5 for 12 kinds of trip: families, couples, honeymoons, adventure, beaches, snorkelling and diving, culture, food and wine, nature, rest, value and luxury.
The scores are relative to the other islands on Island, not absolute: 5 means one of the best choices in the region for that need, and 0 means unsuitable. Each comes with a one-line reason drawn from the guide’s sourced facts. They are our judgement, not a measurement.
To keep them consistent, the scores are calibrated across all the guides at once: read side by side, one kind of trip at a time, against anchor islands that set the top of the scale. A score is changed only where an island is clearly out of line with its peers. The latest calibration was on .
| Kind of trip | Islands |
|---|---|
| Families | Rarotonga and Viti Levu |
| Couples | Aitutaki, Bali and Bora Bora |
| Honeymoons | Aitutaki and Bora Bora |
| Adventure | Taveuni |
| Beaches | Aitutaki and Mamanuca Islands |
| Snorkelling and diving | Aitutaki, Moʻorea, Niue and Taveuni |
| Culture | Bali, Savaiʻi, Tahiti, Tongatapu and Upolu |
| Food and wine | Waiheke Island |
| Nature | Chatham Islands, Kāpiti Island, Rakiura Stewart Island, Taveuni and Tiritiri Matangi |
| Rest | Aitutaki, Aotea Great Barrier Island and Savaiʻi |
| Value for money | Bali |
| Luxury | Bora Bora |
How the finder ranks islands
Find your island scores each island out of 100 for your trip, from four weighted parts:
| Part | Weight | How it’s worked out |
|---|---|---|
| Fit | 40% | How well the island suits who’s travelling (40% of fit) and the things you picked (60%), on the 0 to 5 “who it suits” scores. Pick nothing and that part counts as 3 out of 5. |
| Conditions | 30% | The island’s month score for the month you travel. |
| Budget | 15% | Full marks at your budget. Each cost tier cheaper than your budget costs 10% of this part; each tier dearer costs 35%, down to nothing. |
| Ease | 15% | Shorter journeys from your city score higher, falling in a straight line to nothing at 12 hours or more. |
It ranks on the unrounded score, so two islands that both show 70 keep their true order. Before scoring, it leaves an island out when:
- There’s no route from your city in that month.
- The journey is longer than your limit.
- You asked for a direct flight and there isn’t one.
- You asked for no visa and New Zealanders need one.
Who’s travelling
Each kind of traveller is matched to one or more of the “who it suits” scores, in equal shares:
| Traveller | Scores used |
|---|---|
| Couple | Couples |
| Honeymoon | Honeymoons · Luxury |
| Family with under-12s | Families |
| Family with teens | Families · Adventure |
| Friends | Adventure · Food and wine · Value for money |
| Solo | Culture · Nature · Value for money |
Reasons and caveats
Each result lists what’s in its favour:
- Something you picked that the island scores 4 or more out of 5 (two at most).
- A month score of 65 or more, which is good or better.
- A direct route from your city.
- A cost tier within your budget.
And the catches to weigh up:
- Storms: 0.05 or more expected in the month, or one in 20 seasons.
- Rain: 220 mm or more in the month.
- A cool sea: below 24 °C, when you’ve picked beaches or snorkelling and diving.
- No direct route: the journey time includes a change.
- Over your budget, and by how many tiers.
Commissions never change a ranking
Some links on Island go to partners, and if you book through one we may earn a commission. That money has no way into the numbers.
- The month score is worked out from climate, sea and storm records only.
- The finder ranks islands on what it knows about them, and this is all it knows:
- its page address
- its name
- its twelve “who it suits” scores
- its cost tier, from 1 to 4
- whether New Zealanders need a visa
- its journey time from each New Zealand city in each month, and whether it’s direct
- its twelve month scores
- the storms expected in each month
- the rain in each month
- the sea temperature in each month
- Partner links are added by the page templates, never written into a guide, and each one is labelled where it appears.
How each guide is checked
Every island guide follows a house style guide, and its first rule is that every specific fact traces to a source listed on the page: a price, a rule, a time, a distance, the name of a business. Government sources come first, then official tourism boards, then operators’ own websites, then reputable media. Forums and AI-generated sites are never used as sources.
Once a guide is drafted, a separate check goes through it claim by claim against those sources. Each claim is marked verified, corrected or removed: one that can’t be sourced comes out, and one that has changed is brought up to date. Where two official sources disagree, the page says so.
Both the research and the checking are done with the help of AI tools, as our about page explains. The tools read the sources themselves, and the numbers on the charts come straight from the datasets, not from a model’s memory. We don’t claim to have stayed anywhere we write about.
Prices appear only when a source gives them, as ranges with the month they were seen. Rules that change, such as entry requirements, link to the authority that sets them. Every guide shows the date it was last checked.
The datasets
These are the datasets behind the charts and scores, and how each one becomes what the site shows. Their licences, and the credit each one asks for, are on the credits page.
Climate
Meteostat station climate normals, with NASA Langley Research Center POWER Project model data
How it was made
Temperatures come from Meteostat weather stations within 60 km of where visitors stay, skipping stations above 300 m: the nearest that publishes complete long-term averages (1991–2020, or else 1981–2010) or, where none does, the nearest with at least 10 years of monthly records for every month, averaged. Where the nearest station is clearly unlike where visitors stay, a better-placed one is chosen by hand. Rain comes from the same station when it covers all twelve months, and sunshine hours only from published averages; otherwise rain is the NASA POWER model’s figure and there are no sunshine hours. Where no station qualifies, temperatures are NASA POWER model values, corrected month by month by the average gap between the model and the stations on the other tropical islands, or the other New Zealand islands. Solar energy, humidity and cloud always come from NASA POWER, averaged over 2001–2025.
Sea temperature
NOAA OISST v2.1 long-term monthly mean 1991–2020 (NOAA PSL)
How it was made
Long-term monthly averages of sea surface temperature for 1991–2020, read from the 0.25° grid square at a sea point chosen for each island or, when that square is land, from the nearest ocean square within 0.75°. They are open-ocean averages: lagoons and sheltered bays run warmer.
Cyclone tracks
IBTrACS v04r01, South Pacific, East Pacific and South Indian basins (NOAA NCEI)
How it was made
Every storm with winds of at least 34 knots (on the highest reading among the agencies that tracked it) whose track passed within 200 km of an island’s centre between 1 January 1991 and 31 December 2020. The tracks come from the South Pacific, East Pacific (which takes in Hawaiʻi) and South Indian records. Each storm counts once, in the month it came closest, and each month’s count is divided by the 30 years to give the storms expected in a typical season.
Island outlines
OpenStreetMap, via the Overpass API
How it was made
Each island’s coastline is drawn from OpenStreetMap: the one mapped island chosen for it or, for an island group, every island and islet whose centre lies inside a box drawn around the group. Fragments under 0.05 km² are dropped, overlapping parts are merged and the outline is simplified so it draws quickly. The island’s area is measured from that simplified outline.
Land on the Pacific charts
Natural Earth 1:50m physical vectors (land) and 1:10m physical vectors (minor islands)
How it was made
Natural Earth’s land and minor-island outlines, clipped to the Pacific between 95°E and 120°W and between 55°S and 25°N, joined across the 180° line so that shapes cut there meet again, then simplified.
Island facts
How it was made
Each island’s area, population (with the year it was counted) and highest point come from its Wikidata entry. Each entry was chosen by hand and checked against OpenStreetMap’s link to it or, where there is none, against the island’s coordinates and area. The Chatham Islands’ figures come from the entry for Chatham Island, the main island, and Niue’s from the entry for the country. Population is the figure Wikidata marks as preferred, or else the latest dated one. A figure Wikidata doesn’t give is left blank, and so is Kawau Island’s area, because Wikidata’s figure is ten times too small.
Exchange rates
Frankfurter (exchange rates from the European Central Bank and other central banks)
How it was made
The mid-market rate for one New Zealand dollar in each currency the islands use, from Frankfurter, which passes on rates from the European Central Bank and other central banks. Each currency’s rate has its own date, and the set is dated by the oldest. It’s a snapshot from when the rates were fetched, not a live rate.
Ferry timetables
How it was made
Auckland Transport’s published timetable for the ferry routes that call at Waiheke, Rangitoto, Rakino or Tiritiri Matangi, kept for the 60 days after it was fetched, with each route’s usual crossing time. It’s the planned timetable, so it doesn’t know about cancellations or delays.
Found a mistake?
Email hello@island.nz with the page, what’s wrong and, if you can, a source. We check it, fix it and update the page’s date.