Wind turbines face a unique set of operating conditions during their lifetime (Veers et al., 2023). These machines mostly operate inside the atmospheric boundary layer (ABL), often in a highly turbulent flow, and are subject to spatial and temporal variations in wind speed, wind direction, and temperature. Recent large-sized rotor blades may even partially operate between the ABL and the free atmosphere during events of strong atmospheric stability, facing inflow conditions unprecedented to date. Moreover, these machines are constantly exposed to atmospheric elements while being expected to operate reliably for many years with minimal levels of maintenance. This makes them prone to performance degradation due to soiling and erosion over time.

These challenges are reflected in the evolution of wind turbine blade shapes over the years (van Bussel, 2022). The core geometry of a wind turbine blade consists of radially stacked two-dimensional airfoil sections. Early wind turbine designs from the 1970s generally adopted pre-World War II standard airfoil shapes designed for aircraft wings from the National Advisory Committee for Aeronautics (NACA) (Abbott et al., 1945). These airfoils mostly featured relatively low thickness-to-chord ratios and narrow optimal operating windows. Such characteristics proved ill-suited to the highly unsteady conditions experienced by wind turbines and posed significant limitations on blade scaling. Moreover, they often suffered severe performance degradation due to soiling or small-scale damage, with wind turbine power production dropping nearly 30 % in some cases (Airfoils, Where the Turbine Meets the Wind, 2025). These considerations motivated researchers to develop airfoil shapes tailored specifically to wind turbine applications.

Development efforts were significantly accelerated by the introduction of panel methods, which have been studied since the early 1960s (Erickson, 1990). In particular, the coupling of inviscid panel methods with viscous boundary layer models in the early 1980s made low-speed aerodynamic analyses viable (Drela and Giles, 1987). During this decade the increase in computational power and the development of inverse design approaches, where designers specified the desired pressure distribution and where the panel method would compute an airfoil shape (Drela, 1989), significantly catalyzed the development of wind turbine airfoils. A hallmark of this evolution is that, since its beginning, it advanced in parallel within academia and wind turbine manufacturers. However, due to industrial confidentiality, commercial airfoil geometries were generally not disclosed, slowing down the validation of computational methods and the development of shared databases for wind turbine design.

From the start, airfoil designers developed families of airfoils, comprising thicker geometries which prioritized structural characteristics and more slender shapes to achieve high aerodynamic efficiency. An overview of the evolution of wind turbine airfoils in terms of aerodynamic efficiency and bending stiffness, which is assumed to be proportional to the sectional moment of inertia around the chord, is provided in Fig. 1. The figure is based on publicly available data gathered with different methods and in different testing conditions and which only partially cover the entire family in some cases. It should therefore be used to infer general trends in wind turbine airfoil design rather than as a comparison of the relative strengths of the airfoil families. One of the first airfoil families to be publicly issued is the S series, released by the National Laboratory of the Rockies (NLR; formerly known as NREL – the National Renewable Energy Laboratory) in the 1990s (Tangler and Somers, 1995). S-series airfoils feature multiple families with varying thickness-to-chord ratios and have been used in multiple commercial wind turbines (Thresher et al., 1994). However, as shown in Fig. 1, these airfoils are still relatively thin for the structural requirements of modern 100+ m long blades and are relatively sensitive to changes in roughness. In the same period (late 1970s), the FX-77 family of airfoils was developed at the University of Stuttgart (Wortmann, 1978). This family featured airfoils with relative thicknesses between 15.3 % and 50 %, with the thicker shapes derived from truncating a 34.3 %-thick design (flatback airfoil). To cater to the increased structural requirements of larger wind turbine blades, airfoil designers started developing thicker airfoil shapes, especially for use in inboard regions of the blade, as reflected by the lower efficiency and much higher sectional stiffness of root airfoils shown in Fig. 1. The University of Delft (the Netherlands) has developed several airfoils for wind turbine applications from 1990 onwards (Timmer and Rooij, 2003). They can be easily identified by their name: DU (Delft University) – W (wind) – followed by the year of development and the thickness of the design. Several airfoil families have also been developed by Risø in Denmark, namely the A1 (Bertagnolio et al., 2001; Dahl and Fuglsang, 1998), B1 (Fuglsang et al., 2004), and C2 families (Bak et al., 2008). These airfoils are relatively recent and have been designed with aero-structural considerations in mind (Bak et al., 2014). Risø airfoils, however, are used commercially and are not available open access. The Aeronautical Research Institute of Sweden developed a series of publicly available wind turbine airfoils for wind turbine applications in 1990 (Andres, 1990). They have become very popular among the academic community due to their use in several research reference turbines, such as the DTU 10 MW (Bak et al., 2013), IEA 10 and 3.4 MW (Bortolotti et al., 2019), IEA 15 MW (IEA Wind, 2020), and IEA 22 MW (Zahle et al., 2024). Moreover, they feature high thickness-to-chord ratios and good performance when soiled. Several airfoil families have been designed more recently by various institutions, such as ECN (Boorsma et al., 2015; Grasso, 2014), CENER (Méndez et al., 2014), and others (Cheng et al., 2014; Hansen, 2018), using a variety of design and optimization methods. These airfoil families are, however, not open access, and the shapes of the airfoils are not available publicly. On the other hand, SANDIA has recently developed a family of airfoils specifically tailored to offshore wind turbines (Karcher et al., 2025), i.e., for 100+ m long blades with high Reynolds (Re) and Mach (Ma) numbers. Indeed, the sheer size of modern wind turbine blades has increased the Reynolds numbers dramatically, especially offshore. Very limited experimental validation data exist at these Reynolds numbers due to experimental limitations, which makes the calibration and validation of simulation tools challenging.

Despite research showing how these rotating blades feature strong three-dimensional flow characteristics, especially near the blade root (Guntur and Sørensen, 2015; Hand et al., 2001), modeling them as a series of bi-dimensional sections allows for several engineering simplifications and is the core approach used in most modeling tools. As such, all modern aero-elastic tools are based on polar data, and therefore accurate airfoil aerodynamic coefficients are essential to obtain reliable results. Because airfoil-based models are used in some capacity along the entire design chain of a wind turbine, accuracy impacts all phases of the process, from initial blade design to load verification and certification. In the absence of solid experimental reference data, the aerodynamic coefficients of a given airfoil can vary greatly based on the method used to compute the coefficients and the assumptions made. For instance, the aerodynamic coefficients used in the design of the DTU 10 MW, the IEA 15 MW, and the IEA 22 MW differ, despite all three designs featuring the same FFA-W3 family of airfoils. These uncertainties also make comparisons between different airfoil families difficult. Due to being developed in different times and using different tools, the aerodynamic coefficients used in the design of the aforementioned airfoil families cannot be directly compared in most cases. Moreover, even for those families that have been extensively tested experimentally, such as the DU series of airfoils (Timmer and Rooij, 2003), the FX-77 (Wortmann, 1978), some of the FFA-W3 airfoils (Fuglsang et al., 1998), or the S-series airfoils (Ohio State University Wind Tunnel Tests | Wind Research, 2026), making direct comparisons is challenging due to the coefficients being computed by different researchers in different wind tunnels, with different surface finishes, inflow turbulence, blockage ratios, and Reynolds and Mach numbers. In fact, the complex physics at play cause the performance of an aerodynamic shape to be strongly influenced by minor variations in external shape or surface finish and changes in inflow conditions.

Past efforts to gather coefficients for multiple airfoil families include the ECN (now TNO) Aerodynamic Table Generator, which utilizes wind tunnel measurements and empirical models to populate aerodynamic databases for industrial design codes. While historically significant, the tool is documented in Dutch and remains a proprietary institutional asset, limiting its availability to the broader research community (Bot, 2001). More recently, the viscous panel method RFoil (Koodly Ravishankara et al., 2025) was also used to provide coefficients for the FFA-W3, OSO, RISO, and DU families of airfoils in the work led by Sandia National Laboratories (Maniaci et al., 2025).

In addition, significant uncertainty is also present in the post-stall region. In fact, aerodynamic coefficients in the attached flow regime are typically obtained using viscous panel methods, computational fluid dynamics (CFD), or experimental measurements at moderate angles of attack. In contrast, post-stall extrapolation methods are widely used to empirically extend the coefficients to a -180 to +180° range. Once again, depending on the method and the assumptions used, estimations can vary greatly. For instance, the maximum drag value within the coefficients used in the NREL 5 MW RWT designs is approximately 1.3 before stall-delay correction, while this value is 1.3 to 1.5 for the DTU 10 MW, IEA 15 MW, and IEA 22 MW, depending on the airfoil thickness. As wind turbine blades have complex geometries and a finite aspect ratio, these values are often set empirically based on experience or previous simulations (Zahle et al., 2024). Indeed, wind tunnel experiments on two-dimensional semi-infinite sections, such as the ones presented in Timmer (2020), indicate maximum drag coefficients above 1.8 for a selection of wind turbine airfoils. Similar values are also recommended in some of the original post-extrapolation models, like that of Viterna and Janetzke (1982), which suggest a value of approximately 1.8 depending on the specific characteristics of the airfoil such as camber, relative thickness, and leading-edge radius. These differences cannot be imputed to Reynolds number alone (Battisti et al., 2020) but rather to the finite aspect ratio of wind turbine blades and their complex geometry, as noted in Burton et al. (2011), where mean drag coefficients measured on standstill blades are discussed. Overall, the values that should be used in engineering models remain highly uncertain. Although horizontal-axis wind turbines do not typically operate in the post-stall region, these conditions can be encountered when parked and can give rise to unsteady phenomena, such as vortex-induced vibration (VIV), which can drive ultimate loading. Even in the absence of unsteady loading on the blade, accurate sectional coefficients at high angles of attack can help reduce modeling errors in engineering tools.

Due to the issues described so far, computing reliable airfoil coefficients can be challenging for wind energy practitioners who wish to build their own blade design, as it requires significant expertise. It is, in fact, not uncommon to see coefficients computed using low-order viscous panel methods in recent research efforts (Gupta et al., 2024), introducing significant uncertainty in drag coefficient estimation and in the near-stall region (Ramanujam et al., 2016).

This work aims to address these issues by computing a dataset of high-Reynolds airfoil coefficients that can serve as a common research basis for modern wind turbine design. One of the key aspects of this effort is the fact that the airfoil polars have been computed using the same modeling choices and inflow conditions for the various families included in the dataset, thus ensuring comparability. Moreover, the coefficients included in this dataset take full advantage of CFD and are computed including compressibility, the effects of which are becoming increasingly relevant for modern wind turbines (Vitulano et al., 2025).

Multiple airfoil families are included in the dataset, allowing for fair comparisons between them. Coefficients are computed using consistent state-of-the-art numerical methods, which are described in the following sections of this study.

The study is organized as follows. Section 2 outlines the characteristics of the dataset, including the airfoils and simulated conditions. Section 3 describes the numerical methods used to compute the dataset, including the validation and verification of such methods. Section 4 presents a selection of results.

The dataset analyzed herein contains airfoil coefficients with varying Reynolds and Mach numbers of several open-access commonly used wind turbine airfoils. The dataset includes the FFA-W1, FFA-W2, FFA-W3, DU, FX77, and OSO airfoil families. One notable exception amongst open-access airfoils is represented by NREL S-series airfoils (Tangler and Somers, 1995), as they are relatively slender for modern blade designs. In fact, only airfoils with thickness-to-chord ratios greater than 21 % are included in the dataset. The NACA63(3)418 airfoil is included despite featuring a relative thickness of 18 % as it has been used in recent actual wind turbine blades (Braud et al., 2025). The full list of geometries is shown in Fig. 2 and is summarized in Table 1.

The Reynolds (Eq. 1) and Mach (Eq. 2) number are defined as 1Re=ρcUμ,2Ma=UγRT, where ρ is the air density, μ is the dynamic viscosity of the fluid, γ is the ratio of heat capacity at constant pressure and at constant volume, R is the gas constant, and T is the temperature of the fluid. Modern wind turbine blades, particularly those of offshore machines, are increasingly large. As such, chord-based Reynolds numbers can exceed 15×106 (Karcher et al., 2025). The mean values assumed in this study and their measurement units are summarized in Table 2, which also reports the reference conditions in terms of pressure P and turbulence intensity ti considered for the simulations. A turbulence intensity of 0.35 % is considered to match the conditions where the computational model is tuned and validated, as explained in detail in Sect. 3.1. In addition, this value is in line with the state-of-the-art. For instance, it is not uncommon to see aerodynamic coefficients used in blade design exercise being computed with Ncrit values in the range of 7 to 9 (Zahle et al., 2024), which correspond to low levels of turbulence intensity, as shown in Eq. (6) discussed later on in this study.

It must be noted that the inflow turbulence wind turbine airfoils experience during operation varies widely due to many factors. Depending on the turbulence level in the incoming wind field, the actual turbulence on the blade section scales based on the relative velocity, which depends on the local tip speed ratio of each section. Especially in the case of outboard blade sections, this can lower the turbulence intensity level significantly. The many factors at play make choosing a single value challenging. However, providing users of the dataset with coefficients computed with low inflow turbulence and a fully turbulent boundary layer allows for interpolation between the two, in order to fine-tune the coefficients to the application and the level of desired conservativeness, if used for blade design applications.

All the simulations performed in this study model the fluid as compressible. This is a key distinction compared to the aerodynamic coefficients used in common academic reference wind turbines (Bortolotti et al., 2019; IEA Wind, 2020; Zahle et al., 2024), where the flow is modeled as incompressible. This simplification is becoming increasingly limiting, as recent studies have shown how, due to high tip speeds and local flow acceleration, transonic flow is possible in certain conditions near the blade tip (De Tavernier and Terzi, 2022). This condition cannot be appropriately modeled if the flow is considered incompressible. On the other hand, Vitulano et al. (2025) have shown that unsteady Reynolds-averaged Navier–Stokes (URANS) simulations are a viable way of predicting airfoil performance in transonic conditions. In addition, compressibility is known to have an effect on the aerodynamic coefficients, particularly on the slope of the lift coefficient and the minimum drag coefficient in the pre-stall region, which tends to increase as the Mach number increases (Gudmundsson, 2014).

The second key improvement of this dataset is the fact that multiple Reynolds numbers are considered. In fact, the Reynolds number can vary greatly during operation in actual wind turbines, as the relative inflow velocity changes due to the varying wind speed and rotor speed. In order to select the most appropriate combination of Reynolds and Mach numbers for the various airfoils, the IEA 22 MW blade design was assumed as a reference, as it is the most recent and representative of the size and characteristics of upcoming offshore rotors. Individual airfoils are assigned to a representative span of the blade based on their thickness, as thicker airfoils are generally used in the inner part of the blade, while thinner ones are preferred in the outboard part to improve aerodynamic efficiency. Once the average chord of the blade section is computed, the inflow velocity for a given Reynolds number (Eq. 3) can be computed by inverting Eq. (1): 3U=Reμρc.

The chord length, Reynolds number, inflow velocity, and Mach number ranges assumed for each airfoil relative thickness are shown in Table 3. Five Reynolds numbers are simulated for each airfoil: 2.5×106, 5×106, 10×106, 15×106, and 20×106. As a consequence of the smaller chord of the tip airfoils and since the same range of Reynolds number was simulated for all relative thicknesses, a maximum velocity of 153.3 m s⁻¹ is considered for airfoils with relative thickness of 21 % and below. This velocity should rather be regarded as a conservative upper bound and is unlikely to be reached during steady-state wind turbine operation. However, its inclusion ensures that the dataset remains valid under transient scenarios, such as rotor overspeed events, extreme gusts, or rapid changes in operating conditions, which can temporarily lead to significantly increased local inflow velocities. The portions of blade span upon which the airfoils are assumed to insist are shown in Fig. 3.

All data included in this dataset are computed using two-dimensional CFD. More specifically, compressible URANS simulations are performed in Ansys Fluent 23R2. The adoption of the URANS approach is not necessarily intended to capture unsteady flow features. Rather, time-dependent simulations are preferred over a steady-state approach as the former have shown to be less prone to numerical instabilities and more accurate, especially when laminar to turbulent boundary layer transition is modeled. The numerical set-up is based on a previously validated approach (Balduzzi et al., 2021; Papi et al., 2021; Venturi et al., 2024). The working fluid is dry air at sea-level pressure and 0 °C temperature (Table 2) and is modeled as an ideal gas.

An open-field, bullet-shaped domain is used as shown in Fig. 4. This geometry allows for the use of a single inlet boundary, where inflow velocity, turbulence intensity, and length scale are specified, and a single outlet boundary, where the free-stream pressure is imposed. The domain boundaries are placed 600 chords away to avoid interference with the pressure distribution on the airfoil, as discussed in more detail in Sect. 3.1, while also allowing the airfoil's wake to fully develop. The angle of attack is changed by rotating the airfoil inside the domain via a sliding interface. While not presently included in this dataset, this approach also allows for dynamic simulations to be performed. The k-ω SST turbulence closure (Menter, 1994) is used in fully turbulent simulations. The numerical decay of turbulence from the inlet to the airfoil is prevented using ad hoc source terms in the k and ω transport equations: this avoids artificially large length scales and consequent contamination of the solution in non-turbulent regions due to excessive eddy viscosity (Spalart and Rumsey, 2007). Boundary layer transition in the free transition simulations is modeled using the four-equation γ-Reθ transition model (Langtry and Menter, 2009; Menter et al., 2004). In order to fine-tune the characteristics of the transition model, the correlation developed by Drela (1998) for the transition-critical Reynolds number is used, as explained in more detail in Sect. 3.2.

The simulations are initialized from a steady Reynolds-averaged Navier–Stokes (RANS) simulation with the same inflow conditions and model set-up as the following URANS runs. RANS simulations are run for 2000 iterations or until all residuals drop below 1×10-6. After initialization, the simulations are run for 50 chord-based through-flow times or until numerical convergence, where numerical convergence is achieved if the percentage variation over the first and last values used to compute the mean aerodynamic coefficients is less than 0.5 %.

Experimental and numerical data generated during the AVATAR project (Schepers et al., 2018) serve as the benchmark for model validation and calibration in this study. During the project, experiments were conducted on the DU-00-W-212 airfoil at Reynolds numbers ranging from 3×106 to 15×106. Such high-Reynolds-number data are exceptionally rare and were made possible by the DNW High Pressure Wind Tunnel (HDG) in Göttingen. By pressurizing the tunnel to increase air density, flight-scale Reynolds numbers were achieved on a model with a chord of only 15 cm. The resulting dataset includes aerodynamic coefficients from positive to negative stall, alongside detailed surface pressure distributions. Furthermore, these benchmarks were used by project members to cross-verify various computational tools (Sørensen et al., 2016). The following sections detail the numerical framework: Sect. 3.1 discusses grid and domain sensitivity; Sect. 3.2 provides a verification of the fully turbulent set-up against established CFD results; Sect. 3.3 explores boundary layer transition modeling sensitivity; and Sect. 3.4 addresses the methodologies for post-stall extrapolation to high angles of attack.

The operating condition of an airfoil, often expressed in terms of Reynolds and Mach numbers, greatly influences its performance. While this is commonly acknowledged and is not a novelty in itself, the variety of conditions of the examined dataset, which have been tailored specifically to multi-megawatt wind turbine rotors, nevertheless enables the identification of trends directly relevant to rotor aerodynamic design and optimization.

As a first example, lift and drag coefficients, as well as the aerodynamic efficiency, are shown as a function of the angle of attack for the FFA-W3-241 airfoil in Fig. 14. As expected, the Reynolds number has a strong influence on all quantities of interest. For both the lift and the drag coefficients, fully turbulent simulations are more sensitive than transitional ones to the changes in inflow conditions, as a general decrease in drag coefficient with increasing Reynolds number can be noted. The stall limit (static stall angle and maximum lift coefficient before stall) increases instead. These changes are reflected in airfoil efficiency, which grows with the Reynolds number in fully turbulent conditions (Fig. 14c).

More general conclusions can be drawn from Figs. 15 and 16, where the design angle of attack, here taken as the point where the airfoil presents the maximum aerodynamic efficiency E=Cl/Cd, is shown together with the corresponding aerodynamic efficiency and lift coefficient as a function of Reynolds and Mach number. The same trends are reported for the static stall angle. These are considered the key parameters for assessing the performance of an aerodynamic section used in a wind turbine rotor. Lift-to-drag ratio is of paramount importance to minimize the penalty on torque (drag) per unit of useful aerodynamic force (lift). In a horizontal-axis wind turbine, this is especially important in the outboard sections of the blade, where, due to the blade's high tangential velocity, lift is directed mostly out of plane, thereby increasing unwanted thrust force rather than the driving torque, and drag is directed mostly in-plane, accentuating its detrimental effect on torque. The maximum airfoil efficiency for airfoils with 30 % or lower relative thickness in fully turbulent conditions is shown in Fig. 15 as a function of Reynolds number (c) and Mach number (g), respectively. For all airfoil families, lower thickness ratios generally lead to higher aerodynamic efficiency. For all airfoils, efficiency tends to increase with the Reynolds number. However, particularly for more outboard, thinner airfoils, efficiency plateaus with very little, if any, increase beyond 15×106 Reynolds number. From an inspection of Fig. 15g, compressibility appears to influence this effect. In fact, while the Reynolds number is the same for all airfoils and ranges from 2.5×106 to 2×107, the Mach number is not, as thinner airfoils are simulated with higher inflow velocities. If maximum efficiency is shown as a function of the Mach number (Fig. 15g), with the only exception of the DU91-W2-250 and FX77-258 airfoils, which underperform compared to the rest, all airfoils follow a very similar trend. When compressibility effects are low and the efficiency curves remain below Ma=0.2 (as is the case for the FFA-W3-301 airfoil for example), efficiency increases as the Reynolds (and Mach) number increase. Above Ma=0.25 however, the maximum lift-to-drag ratio starts to plateau for all the tested airfoils, with the FFA-W3-211 airfoil being a clear example of this trend. Both the Reynolds number and compressibility have a marked effect on stall performance. In fact, both the stall angle (Fig. 15e) and the design angle (Fig. 15f) tend to initially increase at lower Reynolds numbers, due to the increased resistance to stall as the airfoil boundary layer becomes increasingly thinner, but tend to decrease at higher Mach numbers, highlighting the effect of compressibility.

The variation in the design lift coefficient as a function of the Reynolds and Mach number is also significant, with maximum differences exceeding 20 % in some cases. The differences are stronger for more inboard airfoils, which are less affected by flow compressibility. These variations may have a significant impact on design, as blade sections can experience significant variations in Reynolds and Mach number depending on the rotational speed.

The same analysis is repeated in free transition conditions. The stall angle, the design angle of attack, the lift coefficient at the design angle of attack, and the maximum airfoil efficiency are shown as a function of the Reynolds and Mach numbers for various airfoil families in Fig. 16. While the effect of compressibility on the stall angle, which generally tends to be anticipated with increasing Mach number, is similar to what was noted in fully turbulent conditions, the variations in design angle of attack, maximum efficiency, and design lift coefficient follow different trends. When the boundary layer is allowed to transition, the model predicts greater resistance to stall, as demonstrated by the higher stall angles, especially at lower angles of attack. Resistance to stall is less influenced by the Reynolds number and appears to be more airfoil dependent. In addition, aerodynamic efficiency follows a different trend as a function of Reynolds number with respect to what was noted in fully turbulent conditions (Fig. 15c) and generally decreases as Reynolds number increases. This trend is also confirmed by experiments, although inflow characteristics, and particularly inflow turbulence, varied slightly in the AVATAR test campaign (Pires et al., 2016). The reason for the decrease in maximum efficiency can be explained by examining Fig. 16, where the pressure (Cp) and skin friction (Cf) coefficients for the FFA-W3-241 airfoil in both free transition and fully turbulent conditions are shown. In both cases, pressure loading in the front region of the airfoil suction side (x/c<0.4) tends to increase with Reynolds number due to a reduction in boundary layer thickness. In fully turbulent conditions, this increase is met by a corresponding decrease in skin friction, explaining the increased maximum efficiency. When the boundary layer is allowed to transition from laminar to turbulent however, the higher the Reynolds number, the earlier the flow tends to transition, as shown by the increases in skin friction in Fig. 17b. The earlier laminar-to-turbulent transition reduces the low-friction area of the airfoil operating with a laminar boundary layer. The anticipated transition point is balanced by the decrease skin friction in both the laminar and the turbulent regimes at higher Reynolds numbers (Fig. 17b), ultimately leading to the trends shown in Fig. 16c.

A similar trend in terms of the variation in maximum efficiency and design angle of attack as a function of the Reynolds number is also noted in the experiments performed by Ozlem et al. (2017) on the DU-00-W-212 airfoil. The experimental results are compared in Fig. 18 to the same set of airfoils shown in Fig. 16. In the experiments performed on the DU-00-W-212 airfoils, efficiency peaks at Re=9×106 (Fig. 18b), while the angle of attack at which the maximum lift-to-drag ratio is located (Fig. 18a) decreases as the Reynolds number increases before starting to increase at Re=1.5×107. The simulations are consistent with this trend, with some differences from airfoil to airfoil and depending on the inflow conditions. The Mach number is also much higher in the simulations, as it ranges between 0.01 and 0.46, while it does not exceed 0.1 in the experimental reference.

The absolute differences between the fully turbulent and free transition cases in terms of stall and design angle of attack, aerodynamic efficiency, and design lift coefficient are shown in Fig. 19. Small differences between the performance in fully turbulent and in free transition conditions are generally desirable. These two extreme conditions lie at either end of the operational windows of a wind turbine blade and correspond to a dirty, worn-down blade or a new, clean blade, respectively. The larger the difference in terms of efficiency between these two configurations, the larger the performance degradation during the turbine lifetime will be. As shown in Fig. 19, there are generally larger differences between the free transition and fully turbulent scenarios at lower Reynolds numbers. This trend is most noticeable when examining maximum efficiency, as the gap between the free transition and fully turbulent cases closes as the fully turbulent efficiency increases.

The operating range of some of the airfoils in the database is compared in Fig. 20 by showing the ranges of angles of attack where efficiency exceeds 85 % of the maximum value at each Reynolds number. A broad high-efficiency operating range is desirable in wind turbine applications as these machines operate in highly turbulent inflow and face highly variable inflow conditions, which cause the angle of attack to vary during operation. In addition, modern wind turbine blades tend to deform as rotor load increases due to combined effects of aerodynamic loading and aeroelastic phenomena such as bend–twist and shear–twist coupling. As such, a broader operating range can increase design options and the optimal operational window. When fully turbulent and free transition conditions are compared (top and bottom rows of Fig. 20, respectively), the range of angles of attack with good performance remains similar, but the stall margin, i.e., the distance of this optimal operating range from the static stall point, is higher in the case of a clean blade. In free transition conditions, however, some airfoils, such as the DU08-W-210, OSO-21, and DU91-W2-250, suffer from reductions in the optimal operating window at high Reynolds number due to the effects of compressibility, which leads to early stall and performance degradation. FFA-W3 airfoils generally show good performance in this regard.

This paper describes a database of wind turbine airfoil aerodynamic coefficients. The aerodynamic coefficients are computed using a state-of-the-art CFD methodology, which has been fine-tuned and validated based on available experimental data. Results include the effects of compressibility, and computations are performed with and without laminar-to-turbulent transition. The total computational resources required to compute the database exceed 100 000 CPU hours on modern hardware (dual AMD-EPYC 7413 server). This dataset expands on measurement-based resources such as the Aerodynamic Table Generator developed by TNO, which is constrained by limited Reynolds number coverage, and provides a high-fidelity complement to similar recent datasets (Karcher et al., 2025) generated via viscous panel methods. The quantity and fidelity of the generated data allow for the effects of Reynolds and Mach number variation to be investigated across different families of wind turbine airfoils featuring relative thicknesses between 21 % and 36 %. In particular, these effects are more pronounced when considering fully turbulent computations rather than in the free transition cases. In fact, when not considering boundary layer transition, peak efficiency increases as the Reynolds number increases, but benefits are greater at lower Mach numbers. Above a free-stream Mach number of 0.2, efficiency gains appear to be limited by the effects of compressibility. In free transition conditions, the variation in the transition point due to the change in inflow conditions can compensate for the effects of Reynolds and Mach number variation, and the variation in airfoil efficiency as a function of these non-dimensional parameters is much more case dependent. Because of these trends, the difference in efficiency between the free transition case, representative of a clean blade, and the fully turbulent case, representative of a soiled blade, decreases as the Reynolds number increases. Based on this evidence, the benefits of operating a clean blade are more strongly felt at lower wind speeds, where the Reynolds number is lower, rather than at higher wind speeds, where the Reynolds number increases due to the higher rotational speed. Finally, the more recent airfoils included in this database, namely the OSO family, outperform other academic open-access families, such as the FFA-W3 and DU airfoils, over a large range of relative thicknesses, proving that modern design tools can indeed improve blade sectional performance. On the other hand, the variations in design angle of attack as a function of the Reynolds number and boundary layer modeling (free transition compared to fully turbulent) are more significant than the OSO family of airfoils, especially for the geometries with higher relative thickness. Finally, it is worth remarking that the presented dataset is to be intended as a “living” dataset, and new geometries may be added to it in the future based on the wind energy community's requests.