I'm not exactly sure by what standard you're distinguishing between "survey" and "empirical study," considering all of your cited studies also rely on surveys.

https://babel.hathitrust.org/cgi/pt?id=umn.31951003089863c

Not prepared to read through over 100 pages of unrelated stuff, perhaps you could add a page number? It sounds like this source is included only for a critique of the original study though, and I'll accept that that study isn't perfect.

http://scholarcommons.usf.edu/etd/1862

Ninety officers returned the surveys for a response rate of 36%.

This type of sampling comes with both weaknesses and strengths. One important weakness of using this convenience sample is that the results generated on the nature of the police sub-culture and the frequency of interpersonal violence on the part of police will not necessarily be generalizable. Although these results may not be generalizable, this sample is satisfactory for testing relationships among the variables—traditional police sub-culture, police domestic violence. This sample comes entirely from Central Florida, which further limits generalizability.

This paper is focused on a link between a domestic violence and a "traditional police sub-culture," it is not intended to be taken as a reliable, generalizable source of overall domestic violence.

http://webapp1.dlib.indiana.edu/virtual_disk_library/index.cgi/4951188/FID707/Root/New/030PG297.PDF

Did not investigate this one because I don't have the means to read floppy disk .iso images readily available.

https://digitalcommons.law.umaryland.edu/cgi/viewcontent.cgi?article=2519&context=fac_pubs

This one does reference the studies you mentioned, along with other studies showing much higher numbers. It then goes on to say:

The data on intimate partner abuse by police officers are both dated and potentially flawed, but in ways that make it more likely that abuse is being under—rather than over—reported.59 Most of the studies rely on self-reporting by police officers to establish prevalence of abuse. Self-reporting is a notoriously unreliable measure; as one study noted, “The issue of the reliability of self-reports data is problematic when considering any socially undesirable behavior.”60 Intimate partner abuse is frequently underreported,61 both by those who experience it and those who commit it. Underreporting is likely to be particularly prevalent among law enforcement officers “who fear, even when anonymity is assured, that admitting their own or their colleagues’ abusive behavior may jeopardize careers and livelihoods and break up families."