The LOLland offshore Lidar EXperiment, or LOLLEX, ran from September 2022 to August 2023 in and around the Rødsand II wind farm, a 90-turbine array in the shallow waters of the Baltic Sea just south of the island of Lolland. Led by Shokoufeh Malekmohammadi of the University of Bergen, with colleagues from Bergen, the Technical University of Denmark and RWE, the campaign was part of the EU-funded Train2Wind training network. Its central innovation was deceptively simple: instead of bolting instruments to fixed platforms or buoys, the team mounted two commercial Doppler wind lidars on a crew transfer vessel that commuted nearly every day between Rødby harbour and the wind farm, turning routine maintenance traffic into a mobile atmospheric observatory.

The instrumentation combined complementary strengths. A WindCubeV2 lidar wind profiler measured wind speed and direction at eleven heights from 40 to 290 metres above the sea, sampling at 0.25 hertz and relying on an integrated inertial measurement unit and satellite positioning to track the vessel’s every pitch, roll and heave. Alongside it, a WindCube100S scanning lidar alternated on a 30-minute cycle between five-beam wind profiling and a vertical staring mode that sampled the vertical wind velocity at 1 hertz with 10-metre along-beam resolution, reaching as high as about 2.5 kilometres. Later in the campaign, a long-range Halo Photonics StreamLine XR+ scanning lidar was installed on the wind farm’s transformer platform to the north, firing range-height-indicator scans across the array every 16 seconds to capture turbine wakes in a fixed vertical plane.

Measuring wind from a moving ship is far harder than it sounds. Vessel motion corrupts the lidar’s line-of-sight velocity with spurious contributions from surge, sway, heave, roll, pitch and yaw, and while 10-minute averages on buoys are only marginally affected, instantaneous turbulence measurements can be badly distorted. The team applied a two-step motion correction that projects the platform’s translational velocity onto each beam and then uses time-resolved attitude data to build a rotation matrix, retrieving the true wind vector by least squares. The payoff was substantial: for the WindCubeV2, correction improved the coefficient of determination against the NORA3 reanalysis from 0.797 to 0.892 and cut the root-mean-square error from 1.9 to 1.5 metres per second, bringing ship-based performance close to that of fixed lidars.

The scanning lidar posed a subtler problem. In vertical stare mode only one equation is available for three unknown wind components, so dynamic motion correction is impossible. Instead, the researchers developed a static tilt correction, estimating that the instrument sat at a mean inclination of about 2.7 degrees and removing the resulting bias in the vertical velocity. Without correction the scanning lidar showed a median vertical velocity bias of 0.28 metres per second; a naive zero-mean assumption over-corrected to minus 0.20, while the new tilt method reduced the bias to just minus 0.05 metres per second, the smallest error and the tightest distribution of the three approaches.

With the corrections validated, the dataset, several thousand hours of vertical scans and more than 2,100 hours of platform-based RHI data, became a window into processes rarely seen above the open sea. The most striking case study captured Kelvin-Helmholtz billows on 22 February 2023, wave-like instabilities that form when vertical wind shear across a stable atmospheric layer exceeds a critical threshold. Between 550 and 750 metres altitude, the lidar recorded billows that grew, overturned and dissipated over roughly eleven minutes, their passage visibly broadening and lowering a layer of accumulated aerosol beneath the capping inversion. The accompanying surge in vertical velocity variance within a 400-metre-thick slab suggests exactly the kind of enhanced downward mixing that could, if billows overlap with turbine wakes, accelerate wake recovery under stable conditions.

A second case the same morning revealed a different flavour of entrainment. Around sunrise, with the vessel still in harbour, turbulence intensified near 600 metres and spread downward through a 200-to-800-metre layer, consistent with buoyancy-driven mixing initiated at the top of the boundary layer, possibly by cloud-top radiative cooling. Unlike the textbook Kelvin-Helmholtz event, this episode lacked narrow-band spectral peaks and organised periodic structures, pointing instead to intermittent, vertically distributed entrainment. That evening, a third case study detected internal atmospheric waves with a two-minute period riding near the nose of a 15-metre-per-second low-level jet, shear-driven oscillations spanning roughly 200 metres in height that may amplify mixing and wake meandering when they occur above wind farms.

The fourth case study turned to wakes themselves. During easterly flow on 13 June 2023, the platform-mounted lidar’s RHI scans resolved multiple turbine wakes stretching across the farm, which the team compared against virtual scans generated with the open-source PyWake framework driven by NORA3 inflow data. Most analytical wake deficit models, including the classic Jensen and several Gaussian variants, underestimated the observed wake losses, predicting wind speeds higher than those measured. Only the TurboGaussian model, closely related to Ørsted’s TurbOPark concept, produced longer wakes and larger velocity deficits in good agreement with the lidar, echoing earlier findings that common models overestimate wake recovery between Rødsand II and the neighbouring Nysted farm.

The campaign’s limitations are as instructive as its successes. Safety rules kept the vessel ashore when significant wave heights exceeded two metres or winds surpassed 12 metres per second, and in-farm observations were confined to daylight working hours. The mechanically steered WindCube100S proved sensitive to vessel translation, with 65 percent of its vertical stare scans flagged incomplete, while the optically switched WindCubeV2 maintained roughly 3.5 times higher data availability. Missing temperature and humidity profiles prevented a definitive separation of buoyancy- and shear-driven processes. Even so, the researchers argue that crew transfer vessels offer a cost-effective, mobile platform for shallow coastal seas, and that future deployments with gyroscopic stabilisation and co-located thermodynamic profiling could extend the approach to harsher environments, turning the everyday commute of offshore service boats into a routine probe of the atmosphere that governs how much power the world’s wind farms actually deliver.

Subject of Research: Ship-based Doppler lidar observations of wind farm flow, wake recovery and vertical momentum entrainment in the marine atmospheric boundary layer

Article Title: The LOLland offshore Lidar EXperiment (LOLLEX): a novel observational approach for the study of wind farm flow and entrainment

Article References: Malekmohammadi, S., Cheynet, E., Reuder, J., Linnemann, C., Sjöholm, M., Mann, J., & Giebel, G. (2026). The LOLland offshore Lidar EXperiment (LOLLEX): a novel observational approach for the study of wind farm flow and entrainment. Atmospheric Measurement Techniques, 19(19), 6267-6292. https://doi.org/10.5194/amt-19-6267-2026

Image Credits: AI Generated

DOI: 10.5194/amt-19-6267-2026

Keywords: offshore wind energy, Doppler wind lidar, atmospheric boundary layer, wake recovery, momentum entrainment, Kelvin-Helmholtz instability, turbulence, Rødsand II wind farm, crew transfer vessel, remote sensing, wind profiler, atmospheric waves