Jump1 and Jump2 for flux dips?

Do the (or perhaps could the) Jump1 and Jump2 parameters return negative values for flux dips? That would be very helpful for detecting (for example) dipper stars and giant eclipses.

Based on the provided notebook that defines the Jump parameters, there seems to be an Absolute Magnitude that removes the sign of the jump.

1 Like

That is a good point. Would a feature to directly access whether the latest diaSourceId was positive or negative helpful as well?

That would definitely be helpful, as long as negative “jumps” are being reported (I.e. not as zeros) all we’d need is a bit that indicates the sign

Obviously the Rubin alerts program is already a jump detector – every alert comes from 5-sigma jump in flux. Therefore we see strange objects like this anti-supernova, that was very bright when the templates were acquired, but now its gone. so the difference flux is -100000 nJ.

The jump detector features are looking for changes in difference flux. You could do this yourself with a collection of six SQL clauses, one for each filter band, like this:
abs(objects.g_psfFlux - objects.g_psfFluxMean) / objects.g_psfFluxMeanErr > 3

The jump features were designed as a way to find outbursts in AGN. Characteristically, the flux jumps up and down so they will generate long lightcurves in LSST, but interesting science is when they suddenly brighten over a few days. The advantages of the jump features are:

  • Its much simpler than the long SQL shown above.
  • The jump features use the mean and standard deviation from a specific interval from 70 days before now to 10 days before now. The ten-day standoff is to maximise the discernment of AGN flares.
  • There is also the jump2 statistic – brightening in a different filter band – that we would not expect from a measurement error.
  • Note that if you have a filter using the jump features, you can fetch the results with the “lite” lightcurve in the Kafka and thus do your own analysis of the flux jumps.

Hope this helps