Post by Robin on Nov 17, 2017 19:10:45 GMT 8
I have chosen a particularly large elliptical galaxy with broad diffuse outer wings to test the effects of differing segmentation map sizes on the resulting optimised parameters of the fit with ProFit. Of course, as one increases the size of the segment, they are increasing the number of pixels pertaining to that galaxy and so improving the fit to the model at the larger radii.
Below are three ProFound segmentation plots of the same galaxy with differing levels of dilation being applied (ranging from "only profoundMakeSegExpand(skycut=0.0)" to an additional "profoundMakeSegDilate(size=51)" and lastly "profoundMakeSegDilate(size=101)") to illustrate that for elliptical galaxies in particular, the default profound() settings may not encapsulate all of the flux of the galaxy:

Fitting a single Sersic profile to this elliptical galaxy, I now plot the optimised effective radius and the total magnitude against the total number of pixels within the segmentation map below:

As can be seen, both the effective radius and the total flux (note. mag is plotted here) systematically increase as more pixels are added to segmentation map. These differences are at a level of a few percent but they are systematic up until a point where they begin to plateau as presumably, any further dilation of the segmentation is no longer adding any appreciable flux from the galaxy.
As an aside, bare in mind that as more pixels are added to the segmentation map, more computations must be performed to fit the model and so this comes at a computational cost. Below is a figure showing how the elapsed time of optimisation (MCMC CHARM routine w/ 10,000 iterations) against the num of pixels in the segmentation image:

Ensure that your segmentation maps are sufficiently dilated such that you aren't introducing any systematics when fitting large elliptical galaxies with diffuse extended wings.
Below are three ProFound segmentation plots of the same galaxy with differing levels of dilation being applied (ranging from "only profoundMakeSegExpand(skycut=0.0)" to an additional "profoundMakeSegDilate(size=51)" and lastly "profoundMakeSegDilate(size=101)") to illustrate that for elliptical galaxies in particular, the default profound() settings may not encapsulate all of the flux of the galaxy:

Fitting a single Sersic profile to this elliptical galaxy, I now plot the optimised effective radius and the total magnitude against the total number of pixels within the segmentation map below:

As can be seen, both the effective radius and the total flux (note. mag is plotted here) systematically increase as more pixels are added to segmentation map. These differences are at a level of a few percent but they are systematic up until a point where they begin to plateau as presumably, any further dilation of the segmentation is no longer adding any appreciable flux from the galaxy.
As an aside, bare in mind that as more pixels are added to the segmentation map, more computations must be performed to fit the model and so this comes at a computational cost. Below is a figure showing how the elapsed time of optimisation (MCMC CHARM routine w/ 10,000 iterations) against the num of pixels in the segmentation image:

Ensure that your segmentation maps are sufficiently dilated such that you aren't introducing any systematics when fitting large elliptical galaxies with diffuse extended wings.