Comparisons of the Traditional Feature Matching and Optical Flow Algorithms for Atmospheric Motion Vector Estimation
Atmospheric motion vector (AMV) estimation from atmospheric variables is important in the studies of convective phenomena. AMVs are derived by tracking features (e.g., cloud top properties, radiances, or water vapor) in sequences of geostationary (GEO) or high latitude low Earth orbit (LEO) images spaced at regular intervals in time. However, estimation of AMVs from satellite data is a challenging problem. Current operational feature tracking methods for AMVs often generate vector fields with large errors. AMV performance can be improved by employing an optical flow method that produces winds with higher spatial and temporal resolution and that is less sensitive to gaps within individual images. We use a robust and efficient variational optical flow method that enforces the conservation of pixel brightness across a pair of images with a regularization constraint. In contrast to the traditional feature matching algorithm, which uses three images to generate a retrieval, the optical flow algorithm uses two images. Our approach generates a dense vector field for every pixel in a pair of images.
We test the performance of the optical flow method by generating the data of various convective weather phenomena using the numerical weather prediction models, derive AMVs from a sequence of water vapor fields, and compare them to nature run wind vector fields and to the vector fields generated by the traditional feature matching algorithm. We consider different configurations including infrared and microwave sounders in GEO and LEO. In particular, we show that the root mean square vector differences (RMSVDs) of optical flow retrievals are considerably smaller than those of traditional feature tracking retrievals. We also show that the traditional feature matching algorithm has a strong dependence on the time interval. However, the optical flow results do not depend on the time interval, which is an important detail allowing more flexibility when designing future missions.
