Distance moduli from the web interface at low redshift

In the web interface (the one that returns models instantly), interpolation is used on a pre-calculated grid to calculate distance moduli. This is a problem at low redshift, where the redshift-distance moduli relationship is highly non-linear. The difference between the interpolated distance modulus and the actual one can be as large as 0.1 mags for z < 0.001, but falls to less than 0.01 mags for z > 0.05. If this is a problem you can either use EzGal yourself from python or use the model server, which emails results to you. Both of these methods will calculate the distance moduli precisely, instead of using interpolation.

Model calculations near z=zf

EzGal uses interpolation when calculating magnitude and mass values. This can be problematic as the output redshift approaches the formation redshift, because very close to the formation redshift the properites of the stellar population are changing very rapidly. This cannot be properly represented by interpolation, and so the properties returned by EzGal can be wrong. Unfortunately this is more of a model limitation than an EzGal limitation - at very young ages you can only reliably extract information about a stellar population for the ages output by the stellar model. Any finer time sampling will require some interpolation scheme which will introduce errors. At the very least for star formation histories with short bursts consider any redshifts which are within the burst to be suspect.

EzGal returns magnitudes past the useful regions of the SPS spectra

All models only cover a finite region of wavelength space, and often with differing resolutions. At the moment EzGal does not do a very good job of detecting the end of an SPS model. As a result of this, EzGal might return a magnitude (leading you to believe it is reasonable) even if the model ends halfway thorugh the filter, or if the resolution of the model is so low that only one or two points from the spectra go through the filter. In either case, the returned magnitude is pretty much junk, but there is no way to know this. At some point I will fix this. So be aware of where your models end. If you are using a very blue filter, do some sanity checks before trusting them at higher redshifts. Fortunately galaxies are not visible through blue filters at high redshift, so chances are this will never be a problem. Also make sure the models are reliable at very long wavelengths before you trust them. I don't have any filters redder than about IRAC ch2 on the web interface because I found many of the models had too low of a resolution to be useful past that point.

Improper calculations for CSPs with discontinuities in the star formation history

Discontinuties are always a source of trouble, and at the moment it is not clear whether or not EzGal is properly handling them when calculating CSPs. In fact, it seems not to be. If you do model a star formation history with discontinuities then expect EzGal to hit the maximum iteration limit, and expect your model to take a very long amount of time to calculate. Moreover, it still might not be correct. Test your results carefully. This is why (at the moment) I don't distribute any CSPs on the web interface that are a short burst of star formation.