Hi!
I am trying to process Merian survey data with lsstpipe. I wonder what the best approach is to produce flats and biases? The data was taken over more than 4 years intermittently, so there are some variations in the flats and biases. Should I do nightly bias and flat or weekly, or just produce 1 flat and bias for everything? Thanks!
To add to this question, I think the key component here is “when should I split my calibration data into different epochs?”
The generation of calibration data for DECam is made possible via the pipelines already provided in cp_pipe (e.g., cpFlat.yaml, cpBias.yaml), but the decision as to when to split up the calibrations and certify the data into different date ranges is not so well documented (and, indeed, is more of a science question than one of data processing).
For LSST in early operations, new epochs are defined whenever the camera has been powered down. However, for long-term DECam observations, I imagine the choice as to when to split raw data into different epochs may have a different answer. It may be sufficient to split data into distinct observing runs (as has been done for HSC), or whenever a sufficient number of visits are available within a given epoch to reduce errors due to low number statistics.
Hi @zhouconghao, I am the Forum Moderator this week. This inquiry is actually more of a scientific question rather than one regarding LSST Science Pipelines data processing. Since @lskelvin ’s response suggested several viable approaches, I will mark that reply as the solution. Please feel free to open a new topic if you encounter issues or have questions regarding the LSST Science Pipelines. Thank you!