Well, it depends on the technique you decide to use. IW is an adquisition mode, so you could have both GRD and SLC in IW mode. If you are going to estimate the glacier velocity by using SAR interferometry techniques (InSAR, DInSAR, PSI, etc), you should use SLC, because according with the documentation of sentinel-1 the phase is missing in the GRD product. So, for this reason I strongly recommend you to use SLC in order to apply InSAR or DInSAR approaches.
the authors recommend the GRD mode for glaciar motion. Please look at it, if you don´t have access to the paper, let me know and send me your email, and I will send you.
I am wondering if someone explains about specific need of polarization of images i.e. VV, VH, HH or HV to estimate the glacier velocity. In my study site, only data available for VV polarization. Can I use those datasets for glacier veloctiy estimation. Thank you very much in advance