The improvements described here were implemented and retrospectively
hmi.Mharp_720s from inception (2010.05) to 2013.01.
The improvements have not yet made it fully into production code,
so some HARPs from 2013.01 to date use the old matching algorithm.
This is being corrected (April 2013).
Several users pointed out that the matching of HARPs to NOAA ARs
(embodied in keywords in
was not working reliably, especially for large HARPs which
contain several NOAA ARs.
This turned out to be due to the way the matching procedure worked,
which was bound to fail for large HARPs. We changed the algorithm,
and have gotten better results since. This is work in progress.
The match information is in the
NOAA_ARS keywords, which
are often not very useful when the NOAA AR happens to
coincide with a large HARP. The results here are for a new algorithm,
which is understood and implemented. It needs to be deployed
in production, and the old and less-helpful keywords need to
be updated for past data.
I compiled a plain-text table of all HARPs and the matching NOAA ARs (here), and the inverse table listing the HARP or HARPs corresponding to each NOAA AR (here). They cover 2010.05-2012.12.
We ran the new algorithm and tabulated monthly results. The tables below were originally made to identify cases where the match algorithm could be improved; now they mainly serve to turn up cases of un-physical NOAA AR movements. It is useful to have a new quick-look movie up while looking at these tables.
The NOAA ARs have artifacts, such as retrograde movement and large latitude fluctuations. Most remaining issues in these associations are due to these artifacts.
The main issue the new match procedure solves is to get a correct match when a given NOAA AR is contained in two different but overlapping HARP bounding boxes. You see such multi-overlaps on the match-summaries linked to below in the first table (NOAA-to-HARP), whenever there is a "+" character in the leftmost column. This means one NOAA AR matched two HARPs, which in general we do not want. An example of this type of situation is NOAA AR number 11218, which overlaps both HARP 605 and HARP 610, although the latter is the correct answer. As mentioned, this issue is solved.
If the indicated HARP after the "+" is given as Ø, this is not so bad (especially if it happens for only a few T_RECs, i.e. if the "Count" column is small), because it indicates there was no matching HARP at all. Most of these seem to be old NOAA ARs, etc.