Depth of Quakes

I am wanting to put together ways to determine depth of quakes as part of the predictions. However i am not sure yet on how to go about that… I see a lot of quakes being reported as 10KM, is there something in how the depth is determined that forces an estimate the closer to the surface, and likewise becomes more accurate when the distance increases to all the different numbers we see in the Blot Echo Range… My first thought is that as (i assume) sound wave returns closer to the surface it flattens out more and more and thus it becomes harder to determine… in either case i will continue to learn about how and where these measurements are taken for USGS to post a depth and see if i can use a combination of average depth for area and maybe the landformations of the areas in question to be able to create a depth value in the prediction than can become better.