What is a typical meteorological year (TMY)?

What is a typical meteorological year?

A TMY holds hourly values of solar radiation and meteorological variables for one year at one location. NREL's TMY3 user manual describes it as a reasonably sized annual data set that typifies conditions at a location over a longer period, such as 30 years. It keeps natural daily and seasonal variations, but it is not designed to contain extremes and should not be used to predict the weather of a particular period.

The year is built month by month. For each calendar month, every January in the record is compared with the long-term distribution of January weather and the most typical one is kept, then the same for February and so on. The 12 selected months are joined into one year, and the joins between months from different years are smoothed (for 6 hours either side in the Sandia method).

The first TMY data set was produced by Sandia National Laboratories in 1978 for 248 US locations. NREL updated it as TMY2 in 1994 and as TMY3 in 2008, which covers 1,020 locations in the United States and its territories.

How does the Sandia method pick a typical month?

The weights were chosen for solar energy systems and buildings. The TMY3 manual states that they are not appropriate for creating a typical year for wind energy.

• Nine daily indices are calculated: maximum, minimum and mean dry-bulb temperature; maximum, minimum and mean dew-point temperature; maximum and mean wind speed; and daily global horizontal radiation. TMY2 and TMY3 add direct normal radiation.
• For each index, the Finkelstein–Schafer statistic measures how far a candidate month's cumulative distribution lies from the long-term one.
• The statistics are combined with weights. In TMY2 and TMY3, global and direct radiation carry 5/20 each, dry-bulb and dew-point temperature 4/20 each, and wind speed 2/20.
• The five months with the lowest weighted sums are ranked by how close they are to the long-term mean and median.
• Persistence checks then exclude the month with the longest run, the month with the most runs and any month with no runs of unusually warm, cool or dull days, and the highest-ranked remaining month is selected.

How does the ISO 15927-4 method differ?

ISO 15927-4:2005, Hygrothermal performance of buildings — Calculation and presentation of climatic data — Part 4: Hourly data for assessing the annual energy use for heating and cooling, is the ISO standard for building a reference year of hourly data for heating and cooling energy calculations. It uses the same Finkelstein–Schafer statistic, but on the daily means of three primary variables: dry-bulb air temperature, global solar irradiance and humidity (water vapour pressure, or equivalents such as relative humidity). Each candidate month is ranked on each variable, the ranks are added up, and wind speed is used as a secondary check before the month is chosen. The standard suggests a record of at least ten years.

• Variables: Sandia uses daily maxima, minima and means with fixed weights; ISO 15927-4 uses daily means of three primary variables and sums their ranks.
• Solar emphasis: TMY3 gives half the total weight to solar radiation; ISO 15927-4 treats solar irradiance as one of three primary variables.
• Record length: TMY3 used 30 years (1976–2005) where available and 15 years (1991–2005) elsewhere; ISO 15927-4 suggests at least 10 years.
• Output: both produce a year of real months that matches the long-term statistics month by month, and both are only as good as the data behind them.

What are TMYx files?

TMYx is the name Climate.OneBuilding.Org gives to the typical years it builds from hourly data in NOAA's Integrated Surface Database, using the TMY and ISO 15927-4:2005 methods. The site lists 17,315 TMYx locations, and in its 2026 release the solar radiation for each site comes from the ERA5 reanalysis.

Many locations come in several versions: one built from the whole available record and others built from 15-year windows such as 2011–2025, 2009–2023 and 2007–2021. A recent window reflects current conditions, while a longer record gives a more stable selection. The TMY3 manual notes that the smaller the pool of years, the less likely the selection represents the climate.

TMY, AMY, XMY or design summer year: which for which job?

• Typical year (TMY, TMYx, test reference year): annual energy use and comparing design options. By construction it smooths out extremes, and the TMY3 manual states it is not suited to designing for worst-case conditions.
• Actual meteorological year (AMY): one real calendar year of weather. It lets you stress-test a building against a historical event, such as a heatwave, that typical years average out.
• Extreme meteorological year (XMY): proposed by Drury Crawley and Linda Lawrie in 2015. It starts from the same record as the TMY but deliberately selects more extreme months, and they recommend simulating a TMY together with two XMYs to capture the range of building performance.
• Design summer year (DSY): a hot year for overheating assessment. CIBSE's 2025 UK weather data, for example, provides three: one with a one-in-seven-year hot event, one with the most intense heat and one with the longest-lasting heat event.
• Future typical year: a typical year adjusted to a future climate projection, usually by morphing.

How C4B builds a typical year for your site

Every C4B search shows the nearest weather station's Climate.OneBuilding.Org file (usually TMYx), on every plan. In projects on paid plans, C4B also builds a site-specific typical year from gridded reanalysis at the project's coordinates, selecting months with the ISO 15927-4 (TMYx) method from the latest 15 complete years in the dataset (2006 to 2020 where CERRA is used).

On the Technical and Enterprise plans, a custom EPW builder lets you choose the dataset (ERA5-Land or CERRA), the period (30 years, 15 years or a custom window), the selection method (Sandia, TMYx/ISO 15927-4 or an entropy method) and presets for typical, hot, cold, seasonal-extreme or extreme years.