Automatic Detection of Bow Echoes from Weather Radar Images

Bow echo is a kind of convective storm that often associates with swaths of damaging straight-line winds and small tornadoes. When observed using weather radar, visual features of bow echoes includes rear inflow notches and bowing segments, which are useful predictors of damaging straight-line winds. In this paper a method is proposed for automatic detection of bow echoes from single radar reflectivity images. The method first detects quasi-linear convective systems (QLCSs) based on radon transform, then describes the QLCSs using their contour models, and finally identifies concave segments along the tailing edge of the contour models as indicator of bow echoes. These identified concave segments are also used to forecast damaging winds by extrapolating their tracks. The proposed method is demonstrated using two bow echo cases. The results show that bow echoes are successfully detected, and the corresponding damaging winds could be forecasted at least 15 min before the winds occur.