Broadcast signals on long wave, medium wave and short wave use amplitude modulation (AM). This means: there is a strong carrier at the center frequency. The audio is mirrored left and right of it (above and below). Here is a spectrum of the powerful Luxemburg signal on 1440 KHz. You can recognize an audio bandwidth of about 5.2 kHz. The carrier peaks out of the spectrum: it is more than 20 dB stronger (factor 100) than the audio. You can detect it long before you receive any audio. In this case the signal is strong enough for good reception: the audio is about 40 dB above the general noise level.
Now there are about 100 stations transmitting at the same time on 1440 khz. If the carriers all were exactly on one frequency it was boring. But they aren't. There are deviations of a few Hz. So it makes sense to watch the spectrum from 1439990 to 1440010 Hz a bit closer. Modulated audio usually starts about 50 Hz apart from the carrier- a great advantage: the spectrum around a carrier is clear even if a station dominates the channel. With software spectrum analyzers and a common PC you can detect up to 40 carriers on each MW channel at the same time in the night. Most carriers are stable on their "exclusive" frequency over years. Others have individual moving characteristics.
Here is an example of 820 kHz in the early morning here in Handewitt (19 Jan 2011). This is a frequency only used in North and South America in a distance of 5000 - 12000 km. On short wave overseas reception is nothing special. On MW it is. All these carriers are visible while audio remains the general noise level. The carriers use up most of the transmission power.
This screen shot was made with the excellent Spectrum Lab software in conjunction with the Perseus and ALA1530S magnetic loop active antenna. The difficulty is to identify these offsets. Remote receivers all around the world like the Perseus SDR network can help to measure some but the list is far from complete. Look at the signals stopping just after 4:00 UTC. They are probably from Venezuela and Argentina because these stations sign off at that time. From sunrise and sunset hours you can also draw some conclusions. By the way when you have multipath reception via different layers of the ionosphere lines get less sharp - typical phenomenon on shortwave.
The FMSCAN prediction uses the results of the identified signals to predict signal levels. These offsets are the best beacons of MW propagation.
...are getting rare with the digital switch over. Some years ago they were excellent for the demonstration of doppler effects caused by airplanes. The spectrum shows the Norwegian Greipstad signal on E2 (48.252 MHz) received here in Handewitt (Germany) over the distance of 395 km. Weather/DX conditions (tropo) were normal while receiving this. The dopplers are often stronger than the direct signal. This demonstrates that the reception of signals behind the horizon is significantly emphasized by airplanes.
Most airplanes travel with about 10 km altitude. From there the horizon is about 300 km away. The waves go from the transmitter up to the airplane and down again. That doubles the distance: 600 km if it was right in the middle. It corresponds with my own observations. With a 5 element antenna you can receive FM signals from powerful transmitters up to 500 km almost all the time. Airplane scatters are part of the FMSCAN predictions.
About 50 km around the international Hamburg airport you can receive more airplane scatters than here. Airplanes are lower, and they are more frequent there, resulting in stronger signals in distances up to 300 km coming in for about a minuter and then fading out again.