Dice Question Streamline Icon: https://streamlinehq.com

Generalization of Zheng et al. (2019) AMV-based center-fixing to typical geostationary satellites

Determine whether the atmospheric-motion-vector-based center-fixing method of Zheng et al. (2019)—which locates a tropical cyclone center by minimizing the magnitude of the mean of direction vectors (MMDV) computed from AMVs derived from Gaofen-4 imagery—generalizes to typical geostationary satellites such as GOES ABI and Himawari AHI that have coarser spatial (approximately 0.5–2.0 km) and temporal (~10 min) resolution, by rigorously evaluating its center-location accuracy under these sensor characteristics.

Information Square Streamline Icon: https://streamlinehq.com

Background

Zheng et al. (2019) compute atmospheric motion vectors from geostationary infrared imagery (Gaofen-4) and identify the cyclone center as the minimum of the magnitude of the mean of direction vectors (MMDV). The success of this approach may depend critically on the very high spatial and temporal resolution of Gaofen-4, which is atypical of most operational geostationary platforms.

Typical geostationary satellites used operationally for tropical cyclone analysis (e.g., GOES ABI, Himawari AHI) have substantially coarser resolution (10-minute cadence and 0.5–2.0 km pixels), raising uncertainty about whether AMV quality and the derived MMDV field will suffice for accurate center-fixing. Establishing generalizability would determine whether this method can be adopted more broadly across basins and satellites.

References

However, the computation of high-quality AMVs was made possible by the extremely high spatial and temporal resolution of the Gaofen-4 satellite. It is unclear how well the approach of Zheng et al. (2019) would generalize to typical GEO satellites, which have much coarser resolution (10 min and 0.5-2.0 km).

Center-fixing of tropical cyclones using uncertainty-aware deep learning applied to high-temporal-resolution geostationary satellite imagery (2409.16507 - Lagerquist et al., 24 Sep 2024) in Section 1c (Approaches using cloud-derived or measured wind)