Jack Murphy and several colleagues at SAIC have released a very interesting paper that uses a variety of geospatial and seismic data to locate the 2006 and 2009 nuclear explosions conducted by the DPRK, as well as estimate the yield and depth of burial. (Murphy et al, Exploitation of the IMS and Other Data For a Comprehensive, Advanced Analysis of the North Korean Nuclear Tests. You can also click on the image for the poster.)
Murphy et al conclude that the best fit is a pair of tests with yields of 0.9 kilotons at 200 meters (2006) and 4.6 kilotons at 550 meters (2009).
Murphy or one of the alii had a very clever idea — use the ratio of certain seismic data from tests in 2006 and 2009 to infer the yield and depth of burial for each event having determined the relative location of the two events as well as some other parameters. (Ok, it is a little more complicated than that — Murphy has been one of the developers of a method for using network averaged teleseismic P wave spectra in place of traditional Mb and Ms measurements. If you want to better understand the method, I would probably start with Murphy and Barker, Network-averaged teleseismic P-wave spectra for underground explosions, 1989. To be honest, I find this stuff pretty challenging even when I have the time to read slowly.)