JAM: Joint Assessment of Models

Objectives

The International Energy Agency (IEA) Wind Technology Collaboration Programme (TCP) Task 57, Joint Assessment of Models (JAM), aims to coordinate international efforts to assess wind energy models used across applications—from resource assessment to turbine control. Running from February 2024 through January 2027, the Task brings together researchers from national laboratories, universities, and industry to ensure findings are grounded in real-world wind energy decisions.

At its core, Task 57 seeks to answer questions critical to the wind energy community: How accurate are today's wind energy models—and under what conditions do they fail? How much uncertainty should practitioners expect for a given model and application? And which models are best suited for specific decisions, whether siting a wind plant, designing a turbine, or optimizing farm control?

These questions have historically been addressed in a fragmented way, with results that are difficult to compare or generalize across groups. Task 57 addresses this by standardizing and coordinating validation activities internationally, so that results are reproducible and meaningful to a broad community of model users and developers—including turbine manufacturers, plant operators, and energy consultants.

The primary vehicle for achieving these objectives is the planning and execution of standardized benchmarks spanning a range of physical scales and modeling challenges, with particular emphasis on concurrent inflow measurements and uncertainty quantification.

Participation

The institutions participating in JAM are listed in Table 1. The email list consists of 156 participants. About 60 people have attended the yearly general meeting (held online in June 2025).

Progress, Results, and Impact in 2025

In 2025, Task 57 saw the completion of two major benchmarks: one using the U.S. Department of Energy's (DOE) American Wake Experiment (AWAKEN) dataset, which offers utility-scale wind farm wake measurements. A second benchmark used measurements from the U.S. DOE Rotor Aerodynamic, Aeroelastic, and Wake (RAAW) field campaign, focused on rotor inflow and single turbine characterization.

During 2025, JAM also released its third benchmark, a complex terrain inflow data benchmark, based on the Wind Science & Engineering Test Site in Complex Terrain (WINSENT) campaign. The fourth and last benchmark to be executed under the current phase of the Task is the WiValdi benchmark, focused on turbine-turbine interaction.

Stakeholder engagement through surveys and elicitation interviews has guided performance criteria and prioritization. The Task benefits from strong industry involvement through direct participation in benchmarking. Stakeholders include turbine manufacturers, wind farm developers, and consultants. In-kind contributions from national research labs, universities, and industry bolster efforts.

Some details of each benchmark is given below:

• First benchmark, AWAKEN wind farm wake: focused on inter-farm wake effects, brought together 16 international research groups to evaluate a wide range of wind farm models—from engineering tools to high-fidelity LES—against a detailed observational dataset from a 24 August 2023 case during the AWAKEN field campaign in Oklahoma (see Figure 1). Structured in three phases with progressively richer data, the benchmark showed that accurate inflow characterization is the dominant factor in model performance, often outweighing model complexity, and that even modest terrain can strongly influence turbine output. Results showed that higher-fidelity models do not consistently outperform simpler ones, particularly when inflow conditions are poorly captured. It also showed that additional observational data can significantly reduce prediction errors. Overall, the benchmark established a new standard for model evaluation, emphasizing the importance of inflow accuracy, terrain representation, phased benchmarking, and high-quality observational datasets for improving wind farm modeling.
• Second benchmark, rotor inflow reconstruction using data from the RAAW field campaign: evaluated how accurately modern techniques can reconstruct full inflow conditions from limited measurements—an essential challenge in wind turbine design validation. The study brought together eight inflow models and tested them against nine 10-minute reference cases derived from both a real-world 2.8 MW turbine experiment and a synthetic field campaign. Results showed that reconstruction accuracy varied significantly across models – see Figure 2. A key finding was that error spikes often occurred simultaneously across models, indicating the presence of coherent atmospheric structures not captured by hub-height measurements alone. Overall, the benchmark highlights the inherent uncertainty in inflow reconstruction, underscores the importance of accurately capturing atmospheric variability, and establishes a foundation for future work linking inflow reconstruction errors to turbine load validation accuracy.
• Third benchmark, complex terrain based on the WINSENT campaign: led by the Centre for Solar Energy and Hydrogen Research Baden-Württemberg (ZSW). A working group is meeting regularly towards case definition and model goals.
• The fourth benchmark, based on the WiValdi campaign: led by Germany's DLR, participants of the Task, it is focused on turbine-turbine interaction. The Wivaldi is still in early stages of gathering interest and problem definition.

The results of the AWAKEN benchmark have been documented in two peer-reviewed journal papers currently under review [1,2], one describing the observation dataset, and the benchmark case study selection; another presenting the comparative modeling results across all participants and benchmark phases, and presenting conclusions and lessons learned. The results of the rotor inflow benchmark based on the RAAW campaign are documented in [3].

In terms of impact, the following quote was received from a turbine manufacturer company regarding the AWAKEN benchmark: "This is really great work considering how this benchmark has come together to make some nice new conclusions on farm modeling. It is awesome that so many contributed to build out this study for these learnings."

Next Steps

For the remainder of the phase I of JAM, the next steps are:

• Finalize the selection of target conditions for the WINSENT benchmark and carry out regular working meetings with participants;
• Finalize gathering interest and start the activities related to the WiValdi benchmark;
• Work on incorporating uncertainty quantification metrics on the latest benchmarks.

References

1. Bodini, N., Abraham, A., Doubrawa, P., Letizia, S., Lundquist, J. K., Moriarty, P., and Scott, R.: The AWAKEN wind farm benchmark, Part 1: Observed conditions. Wind Energ. Sci. Discuss. [preprint], https://doi.org/10.5194/wes-2026-33, in review, 2026.
2. Bodini, N., Moriarty, P., Thedin, R., Doubrawa, P., Archer, C., Blaylock, M., Bottasso, C., Carmo, B., Cheung, L., Dubreuil, C., Floors, R., Herges, T., Houck, D., Kanjari, A., Kaul, C. M., Kelley, C., LI, R., Lundquist, J. K., Major, D., Nguyen, A. K., Optis, M., Parada, L. R. C., Peña, A., Quick, J., Ricarte, D., Radünz, W. C., Rai, R. K., Garcia Santiago, O., Schulte, J., Seim, K. S., van der Laan, M. P., Vimalakanthan, K., and Wise, A.: The AWAKEN wind farm benchmark, Part 2: Modeling results. Wind Energ. Sci. Discuss. [preprint], https://doi.org/10.5194/wes-2026-34, in review, 2026.
3. Rybchuk, A., Asmuth, H., Haghshenas, A., Hannesdóttir, A., Friedrich, J., Liew, J., Rinker, J., Houck, D., Thedin, R., and Moriarty P.: The IEA Wind Task 57 inflow reconstruction benchmark for a single turbine in simple terrain: real-world and synthetic case studies, Wind Energ. Sci. Discuss. [preprint], https://doi.org/10.5194/wes-2026-77, in review, 2026.