Field infrared thermographic data from UAV-based measurements on an operating 1.5 MW wind turbine

This dataset supports the publication: Rackwitz, L., Poeck, N., Balaresque, N., von Freyberg, A., and Fischer, A.: UAV-based infrared thermography for laminar-turbulent transition detection on wind turbines in operation: quantifying motion-blur effects using blade image velocity, Wind Energy Science Discussions [preprint], https://doi.org/10.5194/wes-2026-79, in review, 2026. This archive contains the raw infrared-thermographic frames from the in-field measurements in the paper above, published so the results can be independently inspected. The evaluation code and the processed frames (e.g. motion-deblurred images) are not part of this release. This release covers the in-field portion of the study: infrared recordings of the rotor blades of an operating GE 1.5sl wind turbine (rated power 1.5 MW, rotor diameter 77 m, hub height 68 m), taken from an unmanned aerial vehicle (UAV). The complementary laboratory measurements on a scaled model rotor, also reported in the paper, are published separately at https://zenodo.org/records/20486757. The archive comprises: data/ -- 1500 infrared frames as TIFF (float32 radiometric temperature in Kelvin) from one measurement flight, organised by blade orientation in the image (Downstroke, Upstroke, Vertical) and by rotor blade (Blade_A, Blade_B, Blade_C). examples/ -- a small preview subset (2 raw frames as TIFF + false-colour JPG: a Vertical frame with clearly visible motion blur, and a Downstroke frame shown in Fig. 11 of the paper, with a turbulent wedge on the blade) plus one full metadata sample as JSON, so the format can be inspected without downloading the full dataset. README.md -- full documentation of the folder structure, file naming conventions, and metadata schema. Measurements were recorded with an InfraTec PIR uc 605 long-wave infrared camera using the IRBIS software suite, on 12 December 2025 between 12:39 and 12:54 local time (CET, UTC+1). Each raw recording was stored as a proprietary .irb multi-frame sequence; the frames used here were converted to TIFF for this release. The release does not contain the complete recordings: for every passage of a rotor blade through one of the three target orientations, it contains the frame in which the blade is best aligned with that orientation (PEAK) plus four neighbouring frames (pre1 ... pre4 before PEAK, or pre1 ... pre3 before and post1 after PEAK for Upstroke). Each TIFF holds one frame: Pixel data: a float32 array of radiometric temperatures in Kelvin (640 x 480 px). The conversion from the camera's native float64 SDK output is confirmed lossless (verified bit-exact for all 1500 frames) while roughly halving the file size. Values are radiometric temperatures from the camera's factory calibration (emissivity set to 1.0) with a one-point non-uniformity correction (NUC) applied before each run; per the associated publication, they are interpreted in a relative sense rather than as validated absolute surface temperatures. Embedded thumbnail: a reduced-resolution 8-bit grayscale preview stored as a SubIFD, together with display-range hints, so spec-aware viewers can show a quick preview. The full-resolution float32 data is always the primary image. Metadata: a JSON string in the TIFF ImageDescription tag, containing the source recording name, frame index and total frame count, the frame timestamp in local time and in UTC, image dimensions, the camera acquisition parameters embedded into the .irb file, and the blade orientation. The files can be read with standard TIFF readers that support 32-bit floating-point data, e.g. tifffile.imread(path) in Python. Visible-camera and LiDAR data recorded during the campaign are not included: they were not used for the reported results and may allow identification of the site. They are available from the authors upon reasonable request. Please cite both this dataset (using the DOI/citation this repository generates for the record) and the paper cited above. Contact: Lennart Rackwitz -- [e-mail]