November science update

Hello world,

Edit: Over the past week, a few papers were published on WTF. Jason Wright blogged about the first two, you can view that discussion here (including links to the original papers). The third paper has only been 'submitted' to the journal, meaning that it has not gone through the peer review process necessary for publication. We will discuss the contents on this once it has been accepted. Be sure to visit the sub-reddit (which now has over 4,000 subscribers) if you want to discuss more with the community.

Updates to the Observations

As discussed in the last report, we have modified our observing strategies to optimize data quality. This was a concern because the image was saturating because the defocus command not properly executing. When this happened, the scatter increased in the light curve, an effect we want to minimize. To remedy this, we have now selected settings that will work whether the defocus command executes or not. Just after this change in the schedule request we were reviewing the data and saw something.

We saw the data points trending downwards - like the start of a dip. But this trend was not downwards enough to be a sure thing (it was too soon), especially if we considered the measurement errors (which is a must!). Furthermore, we only saw the downwards trend in *one* of the three filters at only at *one* of the observatory sites (what is plotted in the figure above). OK, so nothing significant, right? Right? Well, we weren't so confident about whether or not it was real because we couldn't explain what was causing the trend we observed. And if we couldn't explain the data, we couldn't let our guard down. So here we are frantically checking the LCO scheduler to see when new data was expected to be taken. If the dip trend continued at the same rate, it was the next observations that would confirm it.

At the same time we are asking ourselves hundreds of questions. Why would a trend like this appear only in one filter at one observing site? Was is astrophysical or instrumental? Did a thin layer of clouds roll in that affected the conditions? Did the image get contaminated by scattered moonlight? Did we just not understand our errors well enough? Is there some unknown source of correlated noise in the data?

And then Tyler triumphantly announced - this comparison star is BAD! It was one of the comparison stars that we have been using all along. But after the configuration change to address saturation, the conditions were just barely right (or wrong) enough to affect the one star enough to make it look like there was a dip in the data. Removing the bad comparison star fixed everything, and the data now lines up with the rest of the curve to reveal nothing but a flatline. I guess that is good. For now. We remain patient. This highlights the odd nature of astronomy (and science as a whole), occasionally things just don't work right even though they should.

On the management side of things, we are now set up with a new computer which we have named Toph. Tyler chose this name because we wanted a theme which is expandable if we ever need future computers and the computer generally won't have a monitor attached. Toph is from Avatar: The Last Airbender and is a blind, but fierce fighter. Blind, no monitor. Look, we're scientists first and good at naming things somewhere further down the list.

This is where all the data will be stored locally. Each image is about 7 megabytes in size, but throughout this campaign we will take thousands (possibly over ten thousand images). At the moment we almost have 20 GB of images! From these images we will have Toph automatically extract and produce the light curves and then email them to us and send alerts if a dipping event occurs. This is largely possible with the photometry code developed by Rachel Street and the other astronomers at LCO.

Happy holidays to you all,

~the entire WTF team

Thank you once again for your support!


Featured Posts
Recent Posts