How we turn data into understanding

An anomaly is the difference between an observed or modelled value and the climatological mean for the same location and the same time of year, calculated over a stated reference period.

Temperature anomalies are expressed in degrees (°C). Precipitation anomalies are usually expressed as a percentage of the normal amount, because rainfall varies by orders of magnitude between places.

An anomaly is meaningless without its baseline. We never show one without stating the reference period.

A percentile ranks a value against the distribution of values in the reference period for the same place and time of year. The 98th percentile means the value is higher than 98% of the reference values.

A return period is the average interval between events of at least a given magnitude. A '1-in-50-year' event has about a 2% chance of occurring in any given year; it can occur in consecutive years.

Percentiles and return periods depend on the length and quality of the record. Where a record is short, we say so.

A climatology is a long-term statistical description of the weather — commonly a 30-year average. The World Meteorological Organization's current standard climate normal is 1991–2020.

Long-term trends are sometimes shown relative to a pre-industrial baseline such as 1850–1900. The same data looks different against different baselines, so the baseline is always labelled.

A trend is always stated with its period, for example '°C per decade, 1991–2020'.

Observation time is when a measurement was taken. Forecast initialisation time is when a model run started. Forecast valid time is the moment a forecast describes. Dataset update time is when a provider published the data.

We store every timestamp in UTC and show UTC explicitly. Where useful we also show the viewer's local time, labelled as such.

Data is stored in canonical units: degrees Celsius (°C) for temperature, millimetres (mm) for precipitation, metres per second (m/s) for wind speed and hectopascals (hPa) for pressure.

Conversions to °F, inches, km/h, knots or mph happen only at the presentation layer. Temperature differences (anomalies) are scaled but never offset when converted.

Satellite precipitation (GPM IMERG) is an estimate merged from passive-microwave and infrared sensors and calibrated against gauges where they exist. It is excellent for seeing where heavy rain falls across oceans and data-sparse regions, but a single pixel is not a rain-gauge reading: light rain, orographic rain and very short bursts can be under- or over-estimated. The 30-minute Early Run arrives within hours; later runs are more accurate.

Infrared imagery (for example Himawari Band 13) measures brightness temperature — how warm the emitting surface looks at about 10.4 µm. Cold, bright cloud tops usually mean tall convective clouds, but brightness temperature is not air temperature, and thin high cloud can look colder than the air below it.

Land surface temperature (MODIS) is the skin temperature of the ground under clear skies. It can be far hotter than the 2 m air temperature in a forecast, and clouds leave gaps. True-colour imagery is a daily mosaic of daytime swaths, so gaps between orbits are normal.

Anomaly layers carry their provider's baseline. The GHRSST MUR sea surface temperature anomaly is relative to a 2003–2014 MUR climatology, which is shorter and more recent than the WMO 1991–2020 normal used elsewhere on this site, so the two are not directly comparable.

The ENSO value is NOAA CPC's Oceanic Niño Index (ONI): the 3-month running mean of sea surface temperature anomalies in the Niño-3.4 region (5°N–5°S, 170°W–120°W), from ERSSTv5. Anomalies are relative to centred 30-year base periods that CPC updates every 5 years, which removes the effect of long-term warming from the index.

We label a season 'El Niño conditions' when the ONI is at or above +0.5 °C and 'La Niña conditions' at or below −0.5 °C. An official El Niño or La Niña episode needs at least five consecutive overlapping seasons past the threshold; a single season is not an episode.

The IOD value is the Dipole Mode Index published by NOAA PSL from HadISST1.1: the SST anomaly of the western tropical Indian Ocean (10°S–10°N, 50°E–70°E) minus that of the south-eastern tropical Indian Ocean (10°S–0°, 90°E–110°E). It is monthly and arrives a few months after the fact.

We use the Bureau of Meteorology's ±0.4 °C thresholds to describe the value as positive, negative or neutral. The provider does not state the base period of its anomalies, so the value is shown with that caveat and marked 'partly verified' until the baseline is confirmed.

The MJO value is the Bureau of Meteorology's Real-time Multivariate MJO index (RMM; Wheeler & Hendon 2004), built from outgoing longwave radiation and 850/200 hPa winds. Its two components give a phase (1–8), which says where the enhanced convection sits, and an amplitude, which says how strong it is.

An amplitude of 1 or more is conventionally treated as an active MJO; below 1 it is described as weak. Phases 4–5 place the enhanced convection over the Maritime Continent, which matters for rainfall in Indonesia.

Global temperature comes from Copernicus Climate Pulse: the daily global mean of 2 m air temperature in the ERA5 reanalysis, with a daily anomaly relative to the 1991–2020 climatology of the same dataset. Values for the most recent days are preliminary and can change slightly when final ERA5 data arrive.

Annual anomalies are the mean of complete calendar years only. The monthly climatology band is built from the same series: for each month, the mean and the 10th–90th percentile range of the 30 monthly means in 1991–2020.

The percentile compares the mean anomaly of the latest 30 days with the mean anomaly of the same calendar window in each year from 1991 to 2020, and reports the share of those 30 years it exceeds. It is a rank within the baseline period, not a return period.

Live tropical cyclones come from NASA EONET, which aggregates advisories from the Joint Typhoon Warning Center and NOAA's hurricane centres. Each position, time and maximum sustained wind is shown as the agency published it; we do not smooth, interpolate or re-analyse tracks.

Winds are published in knots. We show knots and convert to m/s (× 0.5144) and km/h (× 1.852). Agencies use different averaging periods (for example 1-minute sustained winds at JTWC and NHC, 10-minute winds at most other agencies), so values from different agencies are not strictly comparable.

meteorologist.id is not a warning service. For warnings, follow the responsible national meteorological agency.

While data sources are being integrated, the interface is built with fixtures: placeholder values and synthetic series. Every fixture carries a FIXTURE badge, and synthetic charts are hatched and captioned 'not data'.

Values with no connected source show a dash and an UNAVAILABLE badge. We do not fill gaps with invented numbers, and we never show a 'Live' badge without a live source.

Provider attribution is part of the component architecture: charts, maps and metrics carry a source line, and maps keep their legally required attribution visible.

When a licence requires specific wording (for example Copernicus or BMKG), we use that wording. The full register is on the Sources page.