The TRACK program automatically and objectively identifies suitable features in time sequences of meteorological or oceanographic data. These features are then tracked through the time sequence to produce feature trajectories. These trajectories are then analysed to produce statistical diagnostic fields. These can then be used to aid the validation of General Circulation Models (GCMs) of the atmosphere and oceans which are used for forecasting and climate prediction. This approach can also be used to study features such as tropical cloud complexes in observational data (satellites) and to generate climatologies from both model and obseravational data.
A wide range of techniques have been used with many adapted to work directely in a spherical domain. To identify suitable features numerical methods and techniques from spatial data analysis and image processing have been used. To perform the tracking an existing technique from dynamic scene analysis has been adapted to work directely on the sphere and to perform the tracking adaptively. Finally to perform the statistical analysis, nonparameteric kernel estimators have been used with new, efficient spherical kernels. So the statistical fields can be determined directely on the sphere negating the systematic error usually introduced when using projections to estimate statistical quantities. Also, cross-validation and adaptive smoothing have been explored for the spherical domain to determine suitable values of the smoothing parameters of the spherical kernels. Both, density and regression estimators are used to produce the following statistical diagnostic fields:
2. Standard deviation of feature strength.
3. Mean feature speed.
4. Standard deviation of feature speed.
5. Feature density.
6. Genesis density.
7. Lysis density.
8. Track density.
9. Mean track lifetime.
10. Mean velocity.
11. Growth/Decay rates
12. Shape diagnostic, anisotropy, orientation.
2. "Feature Tracking On The Unit Sphere.", Monthly Weather Review, 1995, v123, 3458-3465.
3. "Spherical Non-Parametric Estimators Applied To The UGAMP Model Integration For AMIP",Monthly Weather Review, 1996, v124, 2914-2932.
4. Spherical kernels gif
