Data Methodology
Where the weather data comes from, how it is processed, and what the numbers on the platform actually mean — explained for quantity surveyors, project managers and contract administrators, not meteorologists.
- Where the data comes from
- Observation, model and reanalysis data — what is the difference?
- How your site location is matched to a grid point
- Archive vs. forecast data
- What data fields are used
- Trade thresholds
- Tender planning averages and benchmark values
- Reproducibility and independent verification
- Limitations
- When a physical on-site station may be preferable
Where the data comes from
Construction Weather uses Open-Meteo as its primary weather data provider. Open-Meteo is an open-source weather API that aggregates data from national meteorological agencies — including NOAA (United States), DWD (Germany), Météo-France, ECMWF (European Centre for Medium-Range Weather Forecasts) and the UK Met Office — and exposes it as a unified, globally consistent dataset.
For historical data the platform uses the Open-Meteo Archive API, which draws on ERA5 reanalysis produced by ECMWF. ERA5 is one of the most widely cited global climate datasets in peer-reviewed research and is routinely used by meteorologists, climate scientists and engineering consultants worldwide.
Open-Meteo's API is free to use, its source code is publicly available, and its methodology is documented. This means the data underlying platform outputs can be independently examined and cross-referenced.
Who produces ERA5?
ERA5 is produced by the European Centre for Medium-Range Weather Forecasts (ECMWF), an intergovernmental organisation supported by 35 European states. It is published under the Copernicus Climate Change Service (C3S), funded by the European Union.
ERA5 covers the global atmosphere from 1940 to the present day. It is updated with a five-day lag and is considered one of the most accurate and complete historical weather datasets available. Open-Meteo exposes ERA5 data on a 0.1° × 0.1° land-surface grid (approximately 9 km × 9 km at mid-latitudes), optimised for the near-surface conditions relevant to construction operations.
Open-Meteo's documentation identifies ERA5 as the source for its Archive API. Any independent party can query the same endpoint for the same coordinates and date range to reproduce the underlying figures.
Observation, model and reanalysis data — what is the difference?
These three terms are used throughout this page and in the wider construction contract context. Here is what they mean in plain language.
Station observations
A physical instrument — an anemometer, rain gauge or thermometer — at a specific location measures conditions and records them at regular intervals. Station data is the most direct measurement of weather at a point, but coverage is patchy: there may be no station close to your site, and historic station records can contain gaps.
Construction Weather does not use raw station data directly. It uses model and reanalysis data that has been informed by station observations during calibration.
Numerical weather models
A weather model uses mathematical equations to simulate the atmosphere based on initial conditions (themselves derived from observations). Forecast models run forward in time to predict future conditions. Recent forecast data — the last 90 days — blends live model output with station observations to improve accuracy near the present date.
Forecast accuracy degrades with lead time. Days 1–3 ahead are generally reliable; days 8–14 are indicative only.
Reanalysis data
Reanalysis takes historical observations — from stations, weather balloons, satellite instruments and ocean buoys — and combines them with a weather model run over the past. The result is a globally consistent, gap-free historical record that is more accurate than any single station could provide and available everywhere on Earth.
Reanalysis data is fixed and stable: once a date is processed it does not change. This makes it suitable for contractual purposes where consistency of the record matters.
How your site location is matched to a grid point
ERA5-Land divides the Earth's surface into a regular grid with cells approximately 9 km × 9 km at mid-latitudes (0.1° × 0.1° in latitude and longitude). When you enter a site address or postcode, the platform converts it to a latitude and longitude coordinate and then retrieves data from the nearest ERA5-Land grid cell centre.
Open-Meteo handles this grid-cell selection. The API returns the exact latitude, longitude and elevation of the grid point used, which the platform records alongside every query. This means you can verify after the fact which grid point was used for a given report.
For most UK and European construction sites the nearest grid point will be within 5–7 km of the actual site location. In practice, weather conditions at the grid point are representative of the broader area, though local terrain effects can cause differences (see Limitations below).
Elevation handling
The elevation of the ERA5-Land grid cell is provided by Open-Meteo alongside the weather data. ERA5-Land applies an elevation correction to temperature estimates — a commonly used lapse rate of approximately 6.5 °C per 1,000 m — so that temperature values are representative of the grid-cell surface elevation rather than sea level.
If your site is at a significantly different elevation from the surrounding terrain (for example, a hilltop or valley bottom), the grid-cell elevation may not match the site precisely. In such cases the temperature values shown may over- or under-estimate actual on-site conditions by a degree or two.
Precipitation and wind are not elevation-adjusted in the same way. Orographic effects — where hills force moist air to rise, causing heavier rainfall on the windward side — are partially captured by the model but may be smoothed over at 9 km resolution.
Archive vs. forecast data
The platform distinguishes between two data regimes depending on the date range you request:
- Historical archive — dates more than 90 days in the past use the Open-Meteo Archive API, which draws on ERA5 reanalysis. This data is verified, stable and does not change after the fact.
- Recent and near-term — dates within the last 90 days and up to one day ahead use the Open-Meteo Forecast API, which blends recent station observations with model output. The dedicated Forecast Lookahead feature extends this window further forward. Accuracy decreases with lead time.
Where a requested date range spans both regimes, the platform merges the two datasets and aligns them by date so your output is seamless regardless of the period covered.
Why 90 days?
ERA5 reanalysis is published with a lag of approximately five days. For dates within the most recent 90 days, reanalysis data is either not yet available or only partially complete. The platform switches to the Open-Meteo Forecast API for this period because it incorporates the most recent available observations and model runs.
For contractual purposes — for example, compiling an EOT evidence report — the recommended practice is to use dates that fall comfortably within the archive period (more than 90 days ago) so the underlying ERA5 data is fully settled. More recent dates may be included but should be noted as drawing on forecast-model data rather than fully processed reanalysis.
What data fields are used
The platform retrieves and processes the following parameters from Open-Meteo for each day in a requested range.
Precipitation
Total precipitation (mm), rainfall (mm), snowfall (cm) and the number of hours with measurable precipitation. Used for rain, snow and general wet-weather thresholds across all trades.
Temperature
Maximum, minimum and mean air temperature at 2 m, plus apparent (feels-like) temperature. Used for frost, freeze, heat-stress and concrete-pour thresholds.
Wind
Maximum sustained wind speed and maximum wind gust speed at 10 m, with dominant wind direction. Used for elevated-work, crane, roofing and open-area trade thresholds.
Humidity & dew point
Mean relative humidity and mean dew point temperature derived from hourly data. Used for painting, coating, waterproofing and finishing-trade thresholds.
Sunshine & UV
Total daily sunshine duration (seconds) and maximum UV index. Used as supporting indicators for worker exposure and scheduling context.
Weather code
WMO weather interpretation code summarising the dominant condition for the day (clear, cloudy, rain, snow, thunderstorm, etc). Used for headline day classifications.
Trade thresholds
Each construction operation on the platform has a set of weather thresholds that define an "unsuitable" day for that trade. These thresholds are derived from a combination of:
- Published British Standards (BS) and industry guidance notes where applicable
- NEC and JCT contract clause conventions for adverse weather events
- CIOB and RICS guidance on weather risk and delay analysis
- Practical site experience for operations not covered by formal standards
Thresholds are applied independently per parameter. A day is flagged as unsuitable if any single threshold is breached — for example, a day with acceptable rain but excessive wind would still be classified as unsuitable for crane operations.
How thresholds are applied
Each day's weather data is compared against the threshold set for the selected trade. The comparison uses daily aggregates — maximum wind, total precipitation, minimum temperature — rather than hourly data, which reflects the way adverse weather clauses are typically assessed in standard construction contracts.
The resulting unsuitable-day count is a binary classification: each day either breaches one or more thresholds (unsuitable) or does not (suitable). The platform does not apply partial or proportional counting.
Threshold values are visible within the platform for each trade. If you believe a threshold is not appropriate for a particular contract or operation, you should verify the applicable threshold with the contract administrator before relying on the output.
→ See how trade thresholds power Construction Weather Advisories
→ See how unsuitable-day counts support EOT Evidence reports
Tender planning averages
The tender planning tool calculates average unsuitable days per month by applying trade thresholds to ten years of historical archive data retrieved from the Open-Meteo Archive API (ERA5 reanalysis) for the requested location.
The result is a statistical average: the expected number of unsuitable days in a given calendar month, based on the decade of historical data. Individual years will vary. The averages are intended to inform programme allowances and contract risk registers, not to predict the specific conditions in a future project month.
Ten years of ERA5 data represents a robust climatological baseline for most locations, balancing data stability against the need to capture recent climate trends. A 10-year average is consistent with the approach recommended by the CIOB for weather-risk assessment in project programmes.
How benchmark values are calculated
For each month in the 10-year window, the platform counts the number of days on which the trade's threshold conditions were breached. These annual monthly counts are then averaged to produce a single figure — for example, "2.3 unsuitable days in March" for a particular trade at a particular location.
The calculation is deterministic: given the same location, trade and date window, the result is consistent. There is no randomness or sampling involved. The underlying ERA5 data is accessible via the Open-Meteo Archive API, so an independent party with access to the same API could reproduce the calculation for the same coordinates and date range.
The tender planning tool queries the Open-Meteo Archive API directly for the full historical range — the archive endpoint covers settled historical data rather than provisional forecast output, making it appropriate for climatological averaging.
Reproducibility and independent verification
For contractual and dispute-resolution purposes, it matters whether data can be verified independently. Here is how Construction Weather data can be checked.
Open data source
ERA5 reanalysis data is publicly available via the Open-Meteo Archive API. Any party with internet access can query the same endpoint for the same coordinates and date range to retrieve the same underlying figures — no subscription or special access is required.
Location and source in exports
Downloaded exports — including the Excel weather report and the single-day certificate — record the location name and the coordinates entered when the site was queried. These are the user-entered coordinates, not the exact grid-cell centre that Open-Meteo resolves internally. An independent party can query the Open-Meteo Archive API with the same coordinates and date range to retrieve comparable figures for the same area.
Deterministic thresholds
The unsuitable-day thresholds applied to each trade are fixed values, not estimates or probabilities. Given the same raw weather data and the same threshold set, the unsuitable-day count will be consistent. There is no model uncertainty or confidence interval involved in the classification step.
Stable historical record
ERA5 reanalysis data for settled historical dates is stable: the same query for the same coordinates and date range will return consistent results. This consistency — tied to the ERA5 dataset version served by Open-Meteo at the time of retrieval — is important when building an audit trail for contractual claims.
Third-party verification
A meteorological expert retained by either party in a dispute can access the same Open-Meteo API independently, apply their own threshold criteria, and produce a comparable analysis. Platform outputs should be treated as a starting point for expert review rather than a substitute for it in high-value disputes.
EOT claim documents
The EOT evidence Word document records the weather parameters and threshold values applied to each qualifying day, supporting disclosure to the other party. For very high-value claims, an independent meteorological expert should be instructed to review and, where appropriate, supplement the platform-generated data with their own analysis.
Limitations
We believe transparency about what the platform cannot do is as important as explaining what it can. The following limitations are inherent to the approach and should be understood before relying on outputs for contractual purposes.
Gridded vs. point measurements
ERA5-Land grid cells cover approximately 9 km × 9 km. The value assigned to a cell is a spatial average across that area. Conditions at a specific construction site may differ from the cell average, particularly in hilly, coastal or dense urban areas where local terrain creates its own microclimate.
Elevation and microclimate effects
Valley frost pockets, ridge exposure, sea-breeze effects and urban heat islands are not captured at 9 km resolution. A site in a valley may experience ground frost when the grid-cell data shows above-zero temperatures. A site on an exposed ridge may experience higher wind speeds than the grid-cell average.
Forecast uncertainty
Forecast accuracy declines with lead time. Days 1–3 ahead are generally reliable; days 8–14 are indicative only and should not be relied upon for precise operational decisions. Historical archive data is stable and does not change after the fact.
Missing or sparse data
In rare cases archive data may contain gaps or null values for specific parameters and dates — for example, in very remote regions or for early dates in the ERA5 record. The platform displays null values as absent rather than substituting estimates, to avoid misleading outputs in contractual contexts.
Threshold applicability
Unsuitable-day thresholds are general guides aligned to published standards and common contract practice. The applicable threshold for any specific contract event will depend on the exact wording of the contract, the nature of the work, and the judgment of the contract administrator. Always verify threshold applicability against your contract documents.
Not a substitute for expert advice
Platform outputs — including EOT evidence reports and Early Warning Notice drafts — are provided to assist review, not to replace it. Contractual and legal decisions should be made with qualified professional advice. In high-value disputes a meteorological expert witness should be instructed to review and, where appropriate, supplement platform-generated data.
When a physical on-site station may be preferable
Construction Weather uses gridded reanalysis and forecast data rather than a physical instrument on your site. For the majority of commercial, infrastructure and residential projects, this approach provides data that is fit for purpose — accurate enough to identify adverse weather days and to support contractual claims. There are, however, situations where an on-site station delivers material advantages.
Consider a physical station when:
- Your site is in a location with strong local microclimate effects — a narrow coastal valley, a high-altitude plateau, an area prone to fog or orographic rainfall — that are unlikely to be captured at 9 km resolution.
- The contract or project specification explicitly requires weather data from a named local meteorological station or from an instrument installed on site.
- The project is of very high value and weather-related claims are anticipated from the outset, making a contemporaneous on-site record a worthwhile investment in evidence quality.
- The relevant weather parameter is highly localised — for example, precise wind gust readings at height on a tall structure, or ground temperature at the exact location of a concrete pour.
Gridded data vs. an on-site station — a fair comparison
A physical station installed on a specific site produces the most direct measurement of conditions at that point. If the instrument is well-maintained and properly positioned, its readings will be more accurate for that precise location than any gridded product.
However, on-site stations also have limitations that gridded reanalysis does not:
- Coverage gaps: station data only exists from when the instrument was installed. ERA5 reanalysis covers the full project period, including dates before the station was commissioned.
- Maintenance: instruments require calibration and maintenance. A poorly positioned or uncalibrated instrument can produce less accurate data than reanalysis.
- Retrospective analysis: for EOT claims based on conditions that occurred in the past, an on-site station provides no advantage unless it was in place and recording during the period in question.
In practice, the most robust approach for high-value disputes is to use both: ERA5-based platform data to establish the overall weather picture, and on-site or nearby station records (where available) to cross-check or supplement for localised conditions.