Wind meteorology for power forecasting

Wind is a highly complex and extremely valuable meteorological variable in short-term power forecasting. In this article, we explain the basic science behind its formation, how it behaves across different atmospheric layers, and how it’s represented in weather models.

The physical forces at play

Wind is air in motion, driven by physical forces acting on the atmosphere. To understand wind, we first need to understand the forces at play:

• Pressure gradient force: Air moves from high to low pressure. You can think of this force by imagining squeezing a bottle: the air inside, pushed by higher pressure, moves toward the lower-pressure outside. This principle holds if the bottle contains only air; it’s the difference in pressure that drives motion. The greater the difference in pressure, the stronger the resulting force, and thus the wind.
• Coriolis force: This is an apparent force, i.e., it arises not from a direct physical interaction, but from observing motion in a rotating frame of reference (in this case, Earth). It causes moving air to deflect to the right in the Northern Hemisphere and to the left in the Southern Hemisphere. For example, if you stood at the equator and threw a ball toward the pole, it would appear to curve sideways rather than travel in a straight line. This force increases with latitude and with wind speed.
• Centrifugal force: This force arises from curved motion, pushing air outward from the center of rotation, like the pull you feel when spinning a rock tied to a string. In atmospheric flows, it becomes particularly important in circular systems, such as cyclones or anticyclones, where air rotates around a pressure center.
• Friction (or viscosity): This force acts near the Earth’s surface, where air interacts with terrain, buildings, and vegetation. In fluid dynamics, we refer to this as viscosity – the internal friction within the air itself.

How forces shape wind at different heights

Together, the physical forces described above determine how the atmosphere moves. For example, in the free troposphere (i.e., above 1–2 km of height), mainly the pressure gradient and Coriolis forces dominate. The result is a near-parallel alignment between wind and pressure lines, which meteorologists call geostrophic wind.

Near the ground – thus, also at wind turbine height – the picture becomes more complex. This is the Planetary Boundary Layer (PBL): the lowest part of the atmosphere, where terrain, surface roughness, solar heating, and ground friction introduce chaotic variability.

The PBL can range from 100 meters at night to over 1,000 meters during the day, depending on weather conditions. During daytime, solar heating energizes the surface and lower atmosphere, leading to vertical mixing of air and the formation of a convective mixed layer. This mixing helps distribute wind speed and direction more evenly through the layer. At night, the surface cools and the energy driving turbulence weakens, causing the boundary layer to shrink and become more stratified.

As the sun rises again, the cycle restarts: warming leads to renewed vertical mixing and expansion of the PBL.

In this layer, wind is no longer smooth or laminar. It is dominated by turbulence: chaotic air motion characterized by eddies (irregular swirls of air), gusts, and rapid shifts in speed and direction, especially near obstacles like hills, forests, or buildings. Forecasting wind here requires not just physics, but also, as we’ll show, smart approximations for everything we can’t directly model.

Simulating wind in NWP models

As we’ve seen above, forecasting wind means accounting for a dynamic interplay of forces, all of which behave differently depending on altitude, terrain, and time of day. Our most accurate wind forecasts today rely on sophisticated tools that can represent these interactions across space and time. Chief among them are Numerical Weather Prediction (NWP) models.

NWP models simulate the atmosphere by dividing it into a 3D grid of cells, each representing a small volume defined by latitude, longitude, and altitude. The finer the grid (i.e., the higher the resolution), the better the model can capture small-scale features, such as terrain effects or turbulence, but at a higher computational cost.

Within each cell, the model solves differential equations that describe how atmospheric variables evolve over time, including:

• Momentum equations for wind acceleration
• Thermodynamic equations for temperature and heat transfer
• Continuity equations for mass conservation
• State equations linking pressure, temperature, and density

These equations describe large-scale atmospheric behavior, but many small-scale processes near the Earth’s surface are too fine-grained to resolve directly. To address this, models use parametrizations – simplified sub-models based on physics and empirical data. In wind forecasting, these are essential for representing friction, turbulence, and terrain-driven flows, all of which affect turbine-level wind speeds.

A central challenge is modeling turbulence in the lower atmosphere. To do this, NWP models apply Reynolds averaging, which separates wind into a mean (predictable) and a fluctuating (turbulent) component. This introduces new momentum terms representing the transfer of energy by turbulent eddies – terms that must be estimated using turbulence parametrizations.

A key output of this process is Turbulent Kinetic Energy (TKE), a measure of the turbulent motion within each grid cell. High TKE indicates strong, chaotic air motion. It plays an important role in turbulence modeling and is increasingly available in high-resolution models, with broader integration underway.

Ultimately, all these modeling techniques serve one practical goal in power forecasting: predicting wind at the exact height of wind turbines.

Forecasting wind at turbine height

Wind turbines typically operate at hub heights between 80 and 150 meters, a region within the PBL where, as explained, atmospheric behavior is complex, variable, and heavily influenced by local conditions. However, NWP models provide wind forecasts at fixed vertical levels that rarely correspond exactly to turbine height.

To fill this gap, forecasters often interpolate between model layers or use simplified theoretical formulas, such as the logarithmic wind profile – an idealized formula that assumes smooth vertical wind shear – to estimate wind speed at hub height.

In practice, however, real-world wind profiles deviate from this ideal due to several interacting factors:

• Atmospheric stability: Stable layers (common at night) reduce turbulence and cause wind shear.
• Terrain and roughness: Hills, forests, and buildings modify wind speeds and directions locally.
• Diurnal cycles: Daytime heating expands the boundary layer and enhances vertical mixing, while nighttime cooling suppresses it, concentrating turbulence closer to the surface.

Because the PBL is dominated by turbulence and fine-scale variability, accurate wind forecasts at hub height depend heavily on how well these effects are modeled. That’s where TKE, mentioned above, becomes relevant as a quantifiable indicator of turbulence intensity.

From physics to forecasts

Wind may seem simple – just moving air. But behind every gust lies an interplay of forces, atmospheric layers, and parameterized physics. By understanding which dynamics dominate at each level, we turn physics into accurate, turbine-level power forecasts.