Cyclone Tracking in Different Spatial and Temporal Resolutions

Abstract

The quality of cyclone tracks associated with model output of various resolutions is determined using a high-resolution dataset (1.125° ? 1.125°, 2 h) mapped to different spatial (triangular truncations, T21, T42, T63, T84) and temporal resolutions (4 h, 6 h, 12 h, 18 h, 24 h). Three sets of comparisons are performed to study the impact of increasing (i) temporal, (ii) spatial, and (iii) spatiotemporal resolutions (T21/24 h, T42/12 h, T63/6 h, T84/4 h). The different ?test? resolutions are compared with a ?perfect? reference dataset of optimal resolution. A general method to quantitatively compare two sets of tracks is developed. For a given test dataset, this method yields the ratio of perfect tracks (which are also identified in the reference set) and the ratio of missing tracks (present in the reference set, but missing in the test set). In 24-h data (T42) only about 45% of the cyclones can be identified with cyclones in 2-h data (73% in 12 h, 85% in 6 h).