This dataset contains the Version 1.2 CYGNSS Level 2 Climate Data Record which provides the time-tagged and geolocated average wind speed (m/s) and mean square slope (MSS) with 25x25 kilometer resolution from the Delay Doppler Mapping Instrument aboard the CYGNSS satellite constellation. The reported sample locations are determined by the specular points corresponding to the Delay Doppler Maps (DDMs). A subset of DDM data used in the direct processing of the average wind speed and MSS is co-located inside of the Level 2 data files. Only one netCDF data file is produced each day (each file containing data from up to 8 unique CYGNSS spacecraft) with a latency of approximately 1 to 2 months from the last recorded measurement time. The Version 1.2 CDR represents is a collection of reanalysis products derived from the SDR v3.1 Level 1 data (https://doi.org/10.5067/CYGNS-L1X31 ). Calibration accuracy and long term stability are improved relative to SDR v3.1 (https://doi.org/10.5067/CYGNS-L2X31 ) using the same trackwise correction algorithm as was used by CDR v1.1 (https://doi.org/10.5067/CYGNS-L2C11 ), which was derived from SDR v2.1 Level 1 data (https://doi.org/10.5067/CYGNS-L1X21 ). Details of the algorithm are provided in the Trackwise Corrected CDR Algorithm Theoretical Basis Document. CDR Level 2 and 3 products (ocean surface wind speed, mean square slope, and latent and sensible heat flux) are generated from the CDR L1 data using the v3.1 SDR data processing algorithms. These products also exhibit improved calibration accuracy and stability over SDR v3.0. Trackwise correction is applied to the two primary CYGNSS L1 science data products, the normalized bistatic radar cross section (NBRCS) and the leading edge slope of the Doppler-integrated delay waveform (LES). The correction compensates for small errors in the Level 1 calibration, due e.g. to uncertainties in the GPS transmitting antenna gain patterns and the CYGNSS receiving antenna gain patterns. It should be noted that the trackwise correction algorithm cannot be successfully applied to all SDR v3.1 L1 data so there is also some loss of samples that were present in SDR v3.1.