Concur with an3_bolt.
A strong inversion layer can cause EM energy (including radio/radar/etc) to "bounce" between the layer and the ground multiple times causing unusually long ranges. You will find that it is unreliable as there are gaps between the bounces. You may hear the broadcast at 480 NM and 400 NM but not at 440 NM for example. Can be frustrating for those who spend a bit of time at low level over the ocean. You hear VHF loud and clear and think you might not have to struggle with HF for the next position report but the VHF coverage is gone as quick as it came.
More
here
Sudden changes in the atmosphere's vertical moisture content and temperature profiles can on random occasions make microwave and UHF & VHF signals propagate hundreds of kilometers up to about 2,000 kilometers (1,300 mi)—and for ducting mode even farther—beyond the normal radio-horizon. The inversion layer is mostly observed over high pressure regions, but there are several tropospheric weather conditions which create these randomly occurring propagation modes. Inversion layer's altitude for non-ducting is typically found between 100 meters (300 ft) to about 1 kilometer (3,000 ft) and for ducting about 500 meters to 3 kilometers (1,600 to 10,000 ft), and the duration of the events are typically from several hours up to several days. Higher frequencies experience the most dramatic increase of signal strengths, while on low-VHF and HF the effect is negligible.
And more again
here.