Corrected event dataset of FY-4A LMI, 2019–2023

Abstract. The Lightning Mapping Imager (LMI) aboard the Fengyun-4A (FY-4A) satellite, once one of the only two geostationary lightning detection payloads operating in space, has accumulated a substantial volume of observational data. Extensive efforts have been made to correct lightning geolocation deviations, including payload misalignment correction, cloud-top height parallax correction, and thermal deformation correction. These measures have substantially improved the geolocation accuracy of LMI observations. However, individual correction schemes are not necessarily applicable across the entire LMI field of view; furthermore, a comprehensive, unified correction dataset has yet to be established, which has limited the wider utilization of LMI data. To address the remaining systematic geolocation deviations in current LMI lightning products, we propose a new correction method based on reference data from the World Wide Lightning Location Network (WWLLN). Using ground-based lightning observations as a benchmark, the LMI field of view is subdivided into 400 subregions arranged in a 20 × 20 grid. Within each subregion, sensitivity experiments are conducted to match spaceborne LMI detections with ground-based lightning events, thereby quantifying the systematic deviation for each subregion. A weighted curve-fitting approach is then applied to the coordinate deviations derived from the matched events to obtain a correction curve for each subregion. These subregional correction curves are subsequently mapped back to the image coordinates, enabling an analysis of the spatiotemporal variability of lightning geolocation deviations across the full LMI coverage. Finally, the fitted curve values are applied as correction terms to the original data, resulting in the construction of a new, refined correction dataset. Building upon the existing Level-2 lightning products, this method significantly enhances the geolocation accuracy of LMI observations. The coordinate deviations between LMI detections and ground-based lightning network observations exhibit pronounced convergence in both the zonal and meridional components, indicating a substantial improvement in geolocation performance. In addition, a domain-wide assessment of geolocation accuracy reveals that, except for regions such as Xinjiang and Mongolia where lightning occurrence is too sparse to support robust curve fitting, the geolocation accuracy across most of the LMI field of view is relatively stable, with an average error of approximately 15 km (about 1.5 pixels), achieving high practical accuracy. The corrected dataset is publicly available at https://doi.org/10.11888/Atmos.tpdc.303312 (Zhang et al., 2026).