50 likes | 146 Views
Methods for automatic processing. Requirements Accurate enough Useful for doing significant science Accepted by majority of community Robust Useful on most of the solar disk Errors limited Fast enough Real time Post facto Others?. Considerations. No manual intervention by definition
E N D
Methods for automatic processing • Requirements • Accurate enough • Useful for doing significant science • Accepted by majority of community • Robust • Useful on most of the solar disk • Errors limited • Fast enough • Real time • Post facto • Others?
Considerations • No manual intervention by definition • Excludes AZAM? • Semel: “locations with poor solutions are interesting” • Test auto methods against gold standard • Is AZAM the gold standard? • Any auto method will leave some discontinuities (some may be real, some due to noise or algorithm problems)
Considerations • May not need to disambiguate every image in a time series (but, flare changes) • Photospheric magnetic field is intermittent, not continuous, so prefer a method that minimizes dependence on continuity
Meeting the requirements • Only c2 and c3 seem promising enough (pending noise sensitivity tests) • Need to accelerate these by a large factor (but seems possible by coding in C) • Is there hope of improving other methods? • Is there some new approach that is better?
Personal opinion • The more good additional information that can be added to the c2/3 error metrics, the better (e.g. height gradient of |B| or div BHorizontal) • A hybrid approach may be helpful (one method for network and another for active regions) • Are we fooling ourselves with non-linear spatial averaging of real solar fields? • Should work harder on getting physically useful information that does not depend upon disambiguation