Pretty sketchy at best. The key figure, the surface heat flux across the time interval of the one earthquake presented as an example, is the size of a postage stamp such that the axis is nearly unreadable. If I decoded the numbers right, the "surface latent heat flux" varied between 0 and 200 W/M-squared in a few days in early June, according to their curve. As heat flow is usually roughly 100 milliwatts per meter squared, these numbers are unusual. The "3-sigma" anomaly prior to the earthquake comes from some kind of wavelet transform, apparently with an amplitude of 50 W/M-squared.

Wild and unlikely stuff. The other earthquake prediction papers are from Bulgaria, India, Russia, and Mexico.