1,117,692 daily observations across 9 Philippine grid cells, 1940 to 2024. Two reanalyses of the same atmosphere put Manila's warming at 0.072 and 0.122 °C a decade — so this page reports ranges. Both agree 8 of the 9 cities set their hottest day on record since 2020.
Two reanalyses of the same atmosphere disagree by 1.09 °C on Manila's mean temperature and by a factor of 1.9 on how fast it is warming. That is not a rounding detail, so no single absolute temperature appears on this page and every trend is a range.
ERA5 and ERA5-Land are both ECMWF reanalyses of the same period, the second at higher resolution over land. Where they disagree is the honest measure of how well any of this is known, and they disagree more than most charts of warming would suggest.
Mean absolute difference on Manila's annual mean: 27.69 against 26.6 °C. Neither is quoted on this page as the temperature.
ERA5 warms faster in every city checked. For Manila, 0.122 against 0.072 °C a decade.
Cities where both models show warming. A check asserts it, because the page's whole framing depends on the two disagreeing about how much and not about whether.
Trend per decade for the 6 cities fetched from both models, computed over the 1950–2024 years they both cover. Comparing ERA5's 1940 start against ERA5-Land's 1950 one would be partly a comparison of periods, and a check asserts the shared window is used.
°C per decade across the 6 cities. The more conservative of the two.
The same cities, the same years. The upper bound is 0.25.
Last ten years of the window against the first ten, on the conservative model and then the other. Even the low end is a shift a person would notice over a lifetime.
The hottest single day in each city's 85-year record. This is the most robust thing on the page, because a maximum does not care about a model's mean bias in the way a trend line does.
Cities whose hottest day on record falls in 2020 or later, out of 9.
In Laoag on 2022-04-29. Every one of the nine records but one is from this decade.
Whose record still stands from that year. It is the one city in this set at altitude, 8.3 °C cooler than Manila on average.
A threshold count rather than a mean, by decade. The threshold is a stated constant, and the caveat matters: because the two reanalyses sit more than a degree apart on absolute temperature, the number of days crossing any fixed line is model-dependent. The direction here is robust; treat the magnitude as indicative.
Days a year above 35 °C, 1950s to 2020s — from 0.3 to 30.99.
6.8 times as many, reaching 20.39 a year.
Not ten. Every decadal figure on this page carries its year count, and a check fails if a partial decade is not labelled as one.
Mean temperature per decade for all 9 cities. Two things are visible: the 1970s are the coolest decade in most of them, and the curve is flat until roughly 2000 and then is not.
The 1970s. Cooler than the 1940s, which is why a straight line through the whole record understates the recent part.
The 2010s, and the partial 2020s are running warmer still.
Cities whose warmest complete decade is the 2010s — the most recent complete one available.
Mean annual rainfall and the shape of the year. Temperature varies modestly across the archipelago; rainfall varies by more than a factor of two, and the timing varies more than the total.
A year, on average. The wettest of the nine.
The driest, a ratio of 2.05 between the two — much wider than any temperature difference here.
Wettest month against driest, 28 times apart. The sharpest monsoon split in the set; an annual total hides it completely.
This page is built on a model, and that changes what its numbers mean.
ERA5 is a physical model of the atmosphere run backwards, constrained by whatever observations existed at the time. Its 1940s values rest on far fewer measurements than its 2020s ones, which is part of why the two reanalyses disagree more about early decades than recent ones. PAGASA holds the Philippine station record and is not reachable from a script here.
Roughly 25 km for ERA5 and 10 km for ERA5-Land. That cannot separate Manila from the land around it, so nothing here measures urban heat island effects, and "Manila" means a cell centred near it rather than the city.
Daily rainfall totals do not identify a typhoon, and wind is not in this dataset at all. The typhoon page uses IBTrACS track data for that, which is a different archive entirely.
Reporting the midpoint of two models that differ by a factor of 1.9 would be a false precision. Where this page gives a single trend it names the model it came from, and a check fails if the two ever converge to within 0.02 °C a decade, because then this whole framing should be revisited.
One fetcher, nine cities, eight CSVs, and a cache.
The API's default silently resolves to one of them. A page that quotes "the" temperature without saying which model produced it is quoting a coin flip, so era5 and era5_land are named in the request and carried through as separate rows.
ERA5 begins in 1940 and ERA5-Land in 1950. Comparing a 1940-2024 slope against a 1950-2024 one is partly a comparison of periods, and the early decades are exactly the ones whose inclusion changes the answer. Each model's own full-span trend is kept in a separate row labelled with its span.
Partial years are excluded from every mean and trend. A missing January would otherwise drag a mean down and read as a cold year, and a check asserts no annual row falls below the threshold.
The 2020s hold 5 years. They appear on the charts because leaving them off would hide the warmest data in the record, and they carry their year count because plotting five years beside ten as though both were decades would be worse.
Open-Meteo's free tier counts an 85-year request as many calls and returns 429 well before 27 of them finish. Six cities from Luzon to Mindanao establish the disagreement as well as nine would; all nine still get the full daily series. The cities chosen are named in the fetcher rather than left implicit.
About 3 MB of JSON per city per model. Cached so that changing an aggregate does not re-fetch 85 years from a free API eighteen times over, and not committed.
Every figure on this page traces to one of these, through a CSV in
data/ph-weather/. Each is checked against its source query on every
build.
Daily maximum, minimum and mean temperature and precipitation for nine points across the Philippines, 31,047 days each from 1 January 1940 to 31 December 2024 — Reanalysis, not observation: a physical model constrained by whatever measurements existed, on a grid of roughly 25 km. A city here is a grid cell, not a weather station.
The same variables from the higher-resolution land-surface reanalysis, used as an independent check on ERA5 — Fetched through Open-Meteo. It disagrees with ERA5 by 1.09 °C on Manila's mean temperature and by roughly a factor of two on the warming rate, which is why this page reports a range rather than a number.
The Philippine authority for station observations and climate normals — Cited because it is what should be used and is not. pagasa.dost.gov.ph is not reachable from a script here, and station records are a different measurement from reanalysis.
If you have raw data and want to transform it into meaningful insights like this analysis, I'd love to help! Whether it's business data, research datasets, or any information you want to visualize and understand better.
Have a dataset you'd like analyzed or need a mini AI project as a starter guide? Send me your suggestions!
I wrote a companion post — “I Asked How Much Hotter It Has Got And Got Two Different Answers” — in simple, everyday words.