Wind Resource Assessment: The Complete Guide for 2026

Wind resource assessment (WRA) is the process of measuring, modeling, and characterizing the wind resource at a specific location to estimate how much energy a wind farm would produce over its operating lifetime. It is the single most important technical input to every wind energy investment decision — the foundation on which turbine selection, layout optimization, energy yield estimates, and project finance are built.

A reliable WRA can make or break a project. Overestimate the wind resource by 10% and a wind farm's actual revenue will fall short of projections, potentially triggering loan covenant breaches. Underestimate it by 10% and a viable project may never get developed because the economics looked unattractive on paper.

This guide covers everything involved in a modern wind resource assessment: the data sources, measurement technologies, modeling approaches, and costs. It also explains how AI-powered tools are reshaping the early stages of the WRA process — compressing what used to take months into minutes.

What is Wind Resource Assessment?

Wind resource assessment is the systematic evaluation of wind conditions at a proposed wind farm site, including wind speed, direction, turbulence, shear, and variability over time. The output of a WRA is typically an energy yield assessment (EYA) that estimates the annual energy production (AEP) of a proposed wind farm, expressed as probability distributions (P50, P90, and other exceedance levels).

A complete WRA answers three fundamental questions:

1. How windy is this site? What are the mean wind speed, wind rose (directional distribution), and temporal patterns (diurnal, seasonal, inter-annual)?
2. How much energy will a wind farm produce here? Given a specific turbine model and layout, what is the expected net AEP after all losses?
3. How confident are we in that estimate? What are the uncertainty bounds, and what is the probability of underperformance?

The rigor and expense of a WRA scale with the decision it supports. A preliminary desktop screening for land acquisition requires days and costs hundreds of dollars. A bankable assessment for project finance requires 1-3 years and costs $150,000-$350,000 or more.

Why Wind Resource Assessment Matters

Three stakeholders depend on WRA quality, each for different reasons.

Developers

The WRA determines whether a project is worth pursuing. A site needs a minimum capacity factor — typically 25-30% for onshore wind in most markets — to support viable economics at current power purchase agreement (PPA) prices. If the WRA shows the site falls below that threshold, the developer saves years of time and millions in development costs by moving on. If the WRA confirms strong wind, the developer can proceed with confidence through permitting, grid interconnection, and financing.

Lenders and Investors

Wind resource uncertainty is the single largest risk factor in wind energy project finance. A 2024 study by DNV estimated that wind resource variability and modeling errors account for 50-70% of total energy yield uncertainty in a typical wind project. Lenders require an independent, bankable WRA before committing capital — and they size debt based on the P90 production estimate, not the P50. Higher-quality WRAs with lower uncertainty allow higher leverage, reducing the cost of capital.

Grid Operators and Policymakers

Accurate WRA data informs transmission planning, capacity credit calculations, and renewable energy target modeling. If wind resource assessments systematically overestimate or underestimate production across a portfolio, the consequences propagate through electricity markets and grid reliability assessments.

Traditional Approaches

Meteorological Mast Campaigns

The met mast has been the backbone of wind resource assessment for over three decades. A standard met mast is a lattice or tubular tower — typically 60 to 120 meters tall — instrumented with cup anemometers, wind vanes, temperature sensors, barometric pressure sensors, and humidity sensors at multiple heights.

How it works: The mast measures wind speed and direction continuously at high frequency (typically 1-second or 10-second samples, averaged into 10-minute intervals) over a minimum of 12 months. IEC 61400-12 standards specify sensor accuracy requirements, mounting configurations, and data quality thresholds. The resulting dataset captures the site's full seasonal cycle, diurnal patterns, wind shear profile, and turbulence characteristics.

Cost: $80,000-$250,000+ per mast, including equipment, permitting (FAA/aviation clearance can take 3-6 months alone), installation, ongoing maintenance, data retrieval, and decommissioning. Taller masts needed for modern 150m+ hub heights push costs toward the upper end.

Timeline: 18-30 months from initial planning to completed dataset (3-6 months permitting and installation, 12-24 months of measurement).

Limitations: Met masts measure wind at a single point. A 200 MW wind farm spanning several kilometers may have only 1-3 masts, requiring spatial extrapolation. Masts shorter than the turbine's hub height require vertical extrapolation — a significant source of uncertainty, especially in regions with complex wind shear profiles. And in an industry where turbine hub heights now regularly exceed 120-150 meters, even the tallest practical mast (120m) may not reach the center of the rotor swept area.

LiDAR Campaigns

Ground-based LiDAR (Light Detection and Ranging) units have emerged as a critical complement to — and sometimes replacement for — met masts. A LiDAR device sits on the ground and uses laser pulses to measure wind speed and direction at heights from 40 to 300+ meters by tracking the movement of atmospheric aerosols.

Advantages over met masts: LiDAR can measure at the actual hub height of modern turbines (140-170m) without building an enormous tower. Units are portable and can be relocated during a campaign to test multiple positions. Permitting is simpler because there is no tall structure requiring aviation clearance. Deployment is faster — a LiDAR can be operational within days rather than months.

Cost: $60,000-$120,000 for a 6-12 month rental and deployment, including unit rental, installation, maintenance, and data processing.

Limitations: LiDAR measurements carry slightly higher uncertainty than well-calibrated met mast anemometers, particularly in complex terrain where the assumption of homogeneous flow over the measurement volume breaks down. Many lenders still require at least some met mast data alongside LiDAR for bankable assessments, though industry acceptance of LiDAR-only campaigns is increasing.

Modern Data Sources

Every wind resource assessment relies on reference datasets to extend short-term site measurements into long-term production estimates. Three data sources dominate.

ERA5

ERA5 is the fifth-generation atmospheric reanalysis dataset produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). It provides hourly estimates of atmospheric variables — including wind speed and direction at multiple pressure levels and surface heights — on a global grid at approximately 31 km resolution, from 1940 to near-present.

ERA5 is the current gold standard for long-term reference data in wind resource assessment. Its strengths include:

• Hourly temporal resolution (compared to 6-hourly for older reanalysis products)
• Consistent global methodology, enabling cross-site comparison
• Multi-decade record enabling robust measure-correlate-predict (MCP) analysis
• Continuous updates, with data typically available within 5 days of real time

The primary limitation is spatial resolution: a 31 km grid cell cannot capture local terrain effects that drive site-level wind resource variability.

MERRA-2

MERRA-2 (Modern-Era Retrospective analysis for Research and Applications, Version 2) is NASA's global atmospheric reanalysis dataset, providing hourly data at approximately 50 km resolution from 1980 to present. While coarser than ERA5, MERRA-2 uses a different atmospheric model and assimilation system, making it a valuable independent cross-reference.

Many WRA consultants use both ERA5 and MERRA-2 in parallel to assess long-term reference data sensitivity — if the long-term correction differs significantly between the two datasets, it flags potential uncertainty in the MCP process.

Satellite and Remote Sensing Data

High-resolution topographic data (SRTM at 30m resolution, ASTER at 30m) and land cover datasets (Copernicus Global Land Cover at 100m, ESA WorldCover at 10m) provide the terrain and surface roughness inputs critical for microscale wind flow modeling. Satellite-derived sea surface temperature and ice cover data are used for offshore assessments.

Increasingly, satellite-based scatterometer data (measuring ocean surface winds) and SAR (synthetic aperture radar) wind retrievals are being used for offshore wind resource assessment as independent validation of modeled wind speeds.

AI-Powered Wind Assessment

The most significant development in wind resource assessment in the past decade is the application of machine learning to predict wind farm performance directly from site characteristics — without requiring on-site measurement.

How It Works

AI-based wind assessment models are trained on paired datasets: site characteristics (inputs) and actual wind farm production data (outputs). The site characteristics typically include:

• Reanalysis-derived features: Wind speed, direction, shear, veer, turbulence, and atmospheric stability at multiple heights, extracted from ERA5 or MERRA-2 for the grid cell containing the site
• Terrain features: Elevation, slope, aspect, terrain complexity indices (ruggedness, terrain variation), and surface roughness derived from high-resolution topographic and land cover data
• Climatological features: Seasonal and diurnal wind patterns, inter-annual variability, air density, temperature extremes, and precipitation frequency

The model learns the complex, nonlinear relationships between these input features and real-world capacity factor from hundreds of operational wind farms. Once trained, it can predict capacity factor for any new location where the same input features can be extracted — which is essentially anywhere on Earth with ERA5 coverage.

Advantages

• Speed: Results in minutes instead of months or years
• Cost: Orders of magnitude cheaper than traditional approaches
• Consistency: The same model and methodology applied to every site, enabling true apples-to-apples portfolio comparison
• Scale: Practical to screen hundreds of candidate sites — impossible with traditional methods at any reasonable budget

Limitations

• Cannot replace bankable assessments for project finance — lenders require on-site measurement
• Accuracy depends on the quality and geographic diversity of the training dataset
• Does not produce a full wind regime characterization (wind rose, turbulence intensity, directional shear)
• Performance may degrade in regions not represented in training data

AI-powered assessment is not replacing traditional WRA — it is filling a critical gap at the beginning of the development pipeline where speed and cost matter more than bankable precision.

Step-by-Step WRA Process

A complete wind resource assessment, from initial screening to bankable energy yield, follows these steps:

1. Site Screening and Pre-Feasibility: Use wind resource maps, reanalysis data, or AI-powered tools to identify and rank candidate sites based on estimated wind resource, proximity to grid infrastructure, land availability, and environmental constraints. This stage eliminates clearly unsuitable sites before any capital is committed.

2. Measurement Campaign Design: Select met mast and/or LiDAR positions based on preliminary wind flow modeling, terrain analysis, and the proposed turbine layout. Determine the number of measurement points, sensor configurations, and campaign duration.

3. On-Site Data Collection: Deploy met masts and/or LiDAR units and collect 12-24 months of continuous wind data. Maintain instruments, retrieve data, and monitor data quality throughout the campaign.

4. Data Quality Control: Screen raw data for instrument errors, icing events, tower shadow effects, and other anomalies. Apply standard industry QC procedures (IEA Wind Task 32 guidelines, IEC 61400-12 requirements). Remove or flag suspect data periods.

5. Measure-Correlate-Predict (MCP): Correlate the short-term on-site measurements with long-term reference datasets (ERA5, MERRA-2) to derive a long-term wind resource estimate. The MCP process accounts for the fact that your 12-24 month measurement period may have been windier or calmer than the long-term average.

6. Wind Flow Modeling: Apply microscale models — either linear (WAsP) or computational fluid dynamics (CFD) — to extrapolate the measured wind resource from mast/LiDAR locations across the entire wind farm site. This accounts for terrain-driven speed-up, roughness changes, and elevation effects.

7. Energy Yield Assessment: Calculate gross energy production using the modeled wind resource and the selected turbine power curve. Apply loss factors (wake losses, electrical losses, availability, curtailment, environmental shutdowns, degradation) to determine net AEP. Perform uncertainty analysis to generate P50, P75, P90, and P99 estimates.

8. Independent Review: For bankable assessments, an independent engineer (typically DNV, UL Renewables, or K2 Management) reviews the entire WRA and EYA, often conducting parallel analysis to validate results. This independent review is a standard lender requirement.

Cost Comparison

The cost and timeline of wind resource assessment varies enormously depending on the approach and the decision stage.

The cost difference between screening 20 sites illustrates why AI-powered tools have found rapid adoption:

• 20 sites via WindAI: ~$1,000, completed in one afternoon
• 20 sites via consultant desktop studies: $100,000-$400,000, delivered in 2-3 months
• 20 sites via met mast campaigns: $1.6M-$5M, requiring 2+ years

No rational developer installs 20 met masts. The question is what you use to narrow 20 candidate sites down to 3-5 finalists that merit measurement investment. AI-powered screening makes that filtering process dramatically faster and cheaper.

How WindAI Works

WindAI is a machine learning platform purpose-built for wind resource assessment screening. Here is what happens when you enter a set of coordinates:

Feature extraction. WindAI extracts 400+ engineered features for the target location from ERA5 reanalysis data, high-resolution terrain datasets, and atmospheric boundary layer physics models. These features span wind speed and direction at multiple heights, wind shear and veer profiles, atmospheric stability indicators (Richardson number, Obukhov length proxies), terrain complexity metrics (ruggedness index, elevation variance), surface roughness parameters, and seasonal/diurnal variability patterns.

Model prediction. The extracted features are fed into a machine learning model trained on 10 million+ hourly production observations from 300+ operational wind farms across 8 countries. The model has learned how these 400+ site characteristics map to real-world net capacity factor — capturing complex, nonlinear relationships that traditional linear models and rules of thumb cannot represent.

Validation. WindAI's predictions have been validated against held-out operational wind farm data, achieving an RMSE of 0.147, R-squared of 0.777, and annual prediction errors ranging from 2.1% to 7.8%. These accuracy metrics are comparable to preliminary consultant desktop studies at a fraction of the cost and time.

Output. You receive a predicted net capacity factor with uncertainty bounds, delivered in 2-5 minutes. This is sufficient for site ranking, preliminary economic screening, and go/no-go decisions on whether a site merits further investment in measurement campaigns or consultant studies.

WindAI does not replace bankable energy yield assessments — those require on-site measurement and independent engineering review. What it replaces is the expensive, slow process of screening sites before you know which ones deserve that investment. Your first 5 assessments are free at windai.tech, and each additional assessment costs $49.99.

For a deeper look at the model architecture, training data, and validation results, visit the research page.

Frequently Asked Questions

How long does a wind resource assessment take?

The timeline depends on the level of rigor required. AI-powered screening tools deliver results in minutes. Consultant desktop studies take 2-6 weeks. On-site measurement campaigns require 12-24 months of data collection, plus 3-6 months for permitting, installation, and analysis. A full bankable WRA from initial planning through independent review typically spans 1-3 years. The wind industry increasingly uses a staged approach — fast, inexpensive screening first, followed by progressively more detailed (and expensive) assessment for sites that pass each gate.

What is the minimum data requirement for a bankable wind resource assessment?

Most lenders require a minimum of 12 months of IEC-compliant on-site wind measurement data, though 24 months is preferred and reduces inter-annual variability uncertainty. The measurement must capture the full seasonal cycle and be correlated with at least 10-20 years of long-term reference data (ERA5 or MERRA-2). An independent engineer must review the complete analysis. Some progressive lenders are beginning to accept LiDAR-only campaigns (without met masts) if the LiDAR data meets IEC 61400-12-1 verification requirements.

How accurate are AI-based wind resource assessments compared to traditional methods?

AI-based tools like WindAI achieve screening-level accuracy comparable to preliminary consultant desktop studies — annual prediction errors of 2-8% against operational data. Traditional bankable assessments, which combine 12-24 months of on-site measurement with professional flow modeling and uncertainty analysis, achieve higher accuracy with formal uncertainty quantification (total uncertainty typically 8-12% of P50 AEP). The key distinction is that AI tools are designed for early-stage screening where speed and cost matter most, not for replacing the final bankable assessment.

Can I skip the met mast and go straight to project finance?

Not yet, in most cases. While the industry is moving toward greater acceptance of remote sensing and model-based assessments, the vast majority of project finance lenders still require on-site measurement data (met mast and/or LiDAR) as the foundation of a bankable energy yield assessment. The met mast or LiDAR campaign validates the modeled wind resource with direct physical measurement — a level of empirical grounding that no purely model-based approach can currently replace for a lender writing a $100 million check. Use AI-powered tools and desktop studies to identify the right sites, then invest in measurement campaigns for the ones that make the cut.