Strong Lens
Time Delay Challenge

Testing accuracy on thousands of simulated lenses - blind.

Downloads Read the Paper How to Submit

Welcome to the Strong Lens Time Delay Challenge!

Strong lens time delays have been demonstrated to be a powerful tool for cosmology. At present, the power of this method of measuring distances in the Universe is limited by the small number of lenses with measured time delays. However, things are about to change: wide-field surveys this decade will increase the numbers of lensed quasars by an order of magnitude; LSST will find even more, and monitor them all. Each one of the thousands of LSST time delay lenses will have a multi-filter lightcurve of up to a thousand photometric measurements, extracted from observations spaced on average by a few days or so, over seasons of several months in length. We will see the 2 or 4 images of the AGN or supernova source brightening and fading, with time delays between the images that we will need to infer to a mean accuracy of nearly one part in a thousand, if we are to contribute a useful measurement of the accelerating expansion of the Universe. However, we will also see the effects of microlensing and calibration errors - and the data is sparse. How well can we really measure the time delays?

We are trying to answer this question by setting a series of blind data analysis challenges: the LSST DESC SL ``Evil Team'' is producing thousands of realistic simulated lightcurves, and the righteous Good Teams of the astronomy community are trying to infer the time delays that went into them. Gateway datasets provide the first ladder to climb: they contain enough simulated lightcurves to check that your code runs on the challenge data, and to let you know if your algorithms provide enough accuracy to make it worth your while attempting the main challenge. In the first challenge (TDC1) we had 46 gateway entries from 16 independent analysis teams, and 7 graduated to the main challenge.

On July 1, 2014, the TDC1 challenge came to a close. We analyzed the TDC1 submissions, and together wrote a paper (Liao et al 2015) on the community's results. As well as assessing time delay measurement methods, we were able to provide feedback to the LSST project (via the living observing strategy white paper) on how our accuracy depends on the survey observing strategy, which is still to be fully optimized.

We are planning a second time delay challenge, to take place in Summer 2016, with data release in the spring. If you are interested in taking part in TDC2, please consider joining the Good Teams' email list so we can keep you up to date.

The TDC1 data (and ground truth time delays) are now available on the downloads page to help with algorithm development and testing. The TDC1 light curves are only single filter, but the variability you see in the images is typical.

Downloads Read the Paper How to Submit