Story

I have a set of dispersed observations and while I see spectra in these, I also see that there is a non uniform background level in each of them. I want to remove that background before I do anything with the data because I know that this dispersed light can be thought as a simple additive term to my observations. I am using WFC3/G102 and I was told that the background can vary during the course of an observation and that there are several sources of background light. Since the dispersing process is one where different components can be thought as simply dispersed and added up, I know that I can solve for the various amount of light originating from each background component. This is important because each of them has a different 2D shape. I am also told  that some of these components vary during the course of an orbit and that on-the-ramp fitting assumes that count rates are constant during the observations, so I want to remove the variable background component from the data before running on-the-ramp fitting. I want to be able to run on-the-ramp fitting because it is essential to maximize the signal-to-noise of my observations and to remove cosmic ray impacts. In the specific case of WFC3, I was given three component images: Zodi, HeI, and Scattered light. I am told the first one should be constant but its level is time of the year and sky position dependent. The levels of the other two components are expected to vary but they do so in a pretty stochastic manner and we need to empirically determine their levels. See ISR WFC3 2020-04

Inputs

• A set of FLT files taken in the same visit
• The Zodi, HeI, Scattered light model images

Outputs

• A set of background subtracted FLT files
• An estimate of the Zodi, HeI, and Scattered light level as a function of time during each observation

Computations

• We create a mask of each of the FLT files (N files for example) so that we can avoid any part of the observations where we can see, or expect, an object's dispersed spectrum
• We revert back to the multiple read format of the same observations and use the IMA files (M reads each for example) so that we have separate observations (IMSET in the parlance of WFC3)   which are taken during the course of the observations
• Using the ensemble of N*M IMSET we have, we compute a unique level of Zodi light, N*M levels of HeI light and N*M levels of Scatter lights so that we minimize the residuals in the masked N*M IMSETs.
• We subtract our scaled estimated 2D models of the HeI and Scattered from each of the N*M IMSETs within our individual IMA files.
• We run on-the-ramp fitting on the N (now HeI and Scatered light subtracted) IMA files to produce FLT files.
• We re-estimate and subtract the amount of Zodi light from each of the N observations.

Drawbacks

• The method is reliant on a good set of background models which can be thought of as an orthogonal basis to represent the actual background. Any error in these models will lower our ability to properly compute and subtract the background from observations
• For WFC3, and since on-the-ramp fitting needs to be avoided, we need to run CALWF3 multiple times.