Weather Data Sources for Construction Planning

The quality of weather risk analysis depends entirely on the quality of the underlying weather data. ERA-5 reanalysis data, produced by the European Centre for Medium-Range Weather Forecasts, has emerged as the gold standard for construction weather analysis, offering 45+ years of consistent, hourly, global coverage across 80+ weather variables.

However, ERA-5 is not the only option. Weather stations, satellite observations, and numerical weather prediction models each have roles to play. Understanding the strengths and limitations of each data source is essential for selecting the right approach for your project and defending your weather analysis in contractual disputes.

Types of Weather Data

Reanalysis Data (ERA-5)

Reanalysis data is produced by running a modern weather prediction model backwards through time, assimilating all available historical observations to produce a best estimate of atmospheric conditions at every point on the globe for every hour since 1979. ERA-5, the fifth generation of ECMWF reanalysis, is the most advanced and widely used reanalysis dataset.

ERA-5 works by combining millions of observations (from weather stations, radiosondes, satellites, aircraft, and ships) with a sophisticated atmospheric model that enforces physical consistency. The model fills in gaps between observations and produces estimates for locations and times where no direct measurements exist. The result is a seamless, gap-free, physically consistent global weather record.

Key characteristics of ERA-5:

• Global coverage: data available for any location on Earth, including remote sites, offshore locations, and developing regions with sparse observation networks
• 45+ years of records: hourly data from 1979 to present (with a latency of approximately 5 days)
• 80+ weather variables: temperature, wind, rainfall, snow, humidity, visibility, radiation, soil conditions, and more
• Consistent methodology: the same model and assimilation system is applied across the entire period, eliminating the discontinuities that affect station records
• Spatial resolution of approximately 31km: suitable for regional-scale analysis, though local microclimate effects may not be fully captured

Weather Station Networks

Weather stations are physical instruments that directly measure atmospheric conditions at specific locations. National meteorological services operate networks of stations, with the densest coverage in developed countries and sparser networks in developing regions and remote areas.

Advantages: Weather stations provide the most accurate point measurements. A well-maintained, quality-controlled station record is the closest representation of actual conditions at that specific location. For projects located very near a weather station, station data can be highly valuable.

Limitations: Station networks are fundamentally sparse. Even in the UK, which has one of the denser networks globally, the average distance between synoptic weather stations is 40-60km. In remote regions, the nearest station may be hundreds of kilometres away. Station data also suffers from practical issues: equipment failures create data gaps, station relocations introduce discontinuities, changes in surrounding land use affect measurements over time, and individual stations may not measure all variables needed for construction analysis.

A common pitfall in construction weather analysis is using data from the nearest airport weather station for a project located 30km away in different terrain. Airport stations are typically located on flat, open ground at low elevation, while construction sites may be on hillsides, in valleys, near coasts, or at higher elevations. The weather at these locations can be materially different from the airport record.

Satellite Observations

Weather satellites provide global coverage of atmospheric conditions, with particular strengths in observing cloud cover, precipitation estimates, sea surface temperatures, and atmospheric moisture content. Satellite data is a key input to reanalysis datasets but can also be used directly for some applications.

Advantages: Global and consistent coverage, particularly valuable over oceans and remote regions where station data is unavailable. Geostationary satellites provide continuous monitoring of weather development, while polar-orbiting satellites provide higher-resolution snapshots.

Limitations: Satellites observe from above, which means surface-level measurements (wind speed, temperature, humidity at ground level) are inferred rather than directly measured. Precipitation estimates from satellites have significant uncertainty compared to ground-based measurements. The historical record of satellite data is shorter than station records, with comprehensive coverage only available from the 1980s onwards, and the record is fragmented by changes in satellite instruments and platforms.

Numerical Weather Prediction (Forecasting)

Numerical weather prediction (NWP) models use current atmospheric observations as initial conditions and simulate the atmosphere forward in time to produce weather forecasts. Global models (such as ECMWF's IFS or the American GFS) provide forecasts out to 10-15 days, while regional models provide higher-resolution short-range forecasts.

Advantages: NWP models provide the only source of future weather information. Forecast accuracy has improved dramatically over recent decades, with current 5-day forecasts being as accurate as 3-day forecasts were 20 years ago. Short-range forecasts (1-3 days) are highly reliable for most weather variables.

Limitations: Forecast skill decreases with lead time. Beyond 10-15 days, individual forecasts have limited value (though ensemble approaches can provide useful probability information). NWP forecasts cannot replace historical analysis for long-term planning: they tell you what weather to expect next week, not what the weather risk is for a project starting in six months.

What Construction Teams Need from Weather Data

Construction weather analysis has specific requirements that not all data sources can meet:

Hourly or sub-daily resolution: Many construction weather thresholds are defined on a daily or hourly basis. A day with 10mm of rainfall in a single hour is very different from 10mm spread over 24 hours, but daily data cannot distinguish between these scenarios. Hourly data (as provided by ERA-5 and most weather stations) is the minimum for robust threshold analysis.

Multi-variable coverage: Construction activities are affected by combinations of weather variables. A complete analysis needs simultaneous data for wind, rain, temperature, humidity, visibility, and snow at minimum. Data sources that provide only one or two variables (some satellite products, for instance) cannot support comprehensive threshold analysis.

Site-specific data: Weather risk is inherently local. Data must represent conditions at the project site, not at a location tens of kilometres away with different terrain, elevation, or exposure. This is where reanalysis data excels: it provides estimates at any coordinates, interpolated from the global grid.

Long historical record: Statistical analysis of weather risk requires a sufficiently long record to capture the range of conditions that may occur. A minimum of 20-30 years is generally recommended for robust probability estimates. ERA-5's 45+ year record comfortably exceeds this threshold.

Consistency: The data quality and methodology must be consistent across the entire analysis period. Discontinuities in station records (from relocations, equipment changes, or data gaps) can introduce systematic biases. ERA-5's consistent reanalysis methodology eliminates this issue.

Comparing Data Sources for Construction Use

For long-term weather risk analysis and contingency calculation, ERA-5 reanalysis is typically the most suitable primary data source. Its global coverage, 45+ year consistent record, hourly resolution, and multi-variable completeness address all the core requirements. The 31km spatial resolution is adequate for most construction applications, as weather patterns at the scale of individual construction sites are strongly influenced by the broader regional climate.

Weather station data serves as valuable supplementary information, particularly for validating reanalysis data and for capturing very localised effects at sites located close to a high-quality station. For projects near well-maintained stations with long records, a blended approach using both ERA-5 and station data can provide the most accurate analysis.

Satellite data is primarily useful as an input to reanalysis rather than as a direct construction planning resource. Its strengths in global coverage and precipitation monitoring are captured within ERA-5 through the data assimilation process.

Numerical weather prediction (forecasting) is essential for operational weather management during construction but is not suitable for long-term planning. Forecasts inform short-term decisions (schedule this week's concrete pour for Thursday when conditions are favourable) while historical analysis informs long-term planning (this project needs 45 days of weather contingency over the next 18 months).

Common Pitfalls in Construction Weather Data

Using Airport Data for Remote Sites

Perhaps the most common error is relying on data from the nearest airport weather station for a project site that is located in different terrain or at different elevation. Airport stations are valuable but they represent conditions at flat, open, low-elevation locations. A construction site on an exposed hillside, in a coastal location, or in a valley will experience different wind speeds, temperatures, and precipitation patterns.

Ignoring Microclimates

Local topography creates microclimates that can differ significantly from the regional pattern. Valley sites may experience temperature inversions and fog more frequently than surrounding areas. Hillside sites may be more exposed to wind. Urban sites may benefit from urban heat island effects but face different wind patterns due to surrounding buildings. While no data source perfectly captures microclimatic effects at 31km resolution, awareness of these factors is essential for interpreting any weather analysis.

Using Too Short a Record

Weather patterns exhibit significant year-to-year variability. An analysis based on five or ten years of data may not capture the full range of conditions that could occur. A decade that happened to be unusually dry would produce an underestimate of rainfall risk, while a decade with several severe winters would overestimate winter constraints. A minimum of 20-30 years of data is recommended for robust statistical analysis.

Ignoring Data Quality Issues

Not all data is of equal quality. Station records may contain errors, gaps, or discontinuities. Satellite-derived estimates have known biases for certain variables. Even reanalysis data has limitations in regions with sparse observational input. Understanding the quality characteristics of the data being used is essential for interpreting results and communicating confidence levels.

Confusing Climate Averages with Risk

Climate averages (mean monthly rainfall, average temperature) are not the same as weather risk assessments. A location with average rainfall of 80mm in November could experience anything from 20mm to 200mm in any given year. Weather risk analysis must work with distributions of possible outcomes, not just averages. This requires access to the underlying hourly or daily data, not just monthly summary statistics.