Scientists are getting closer to understanding the tsunami that struck Palu in Indonesia in September.

Damaging waves rushed ashore after a Magnitude 7.8 quake, but researchers said at the time that they were surprised by their size.

Now, a survey in the bay in front of the Sulawesi city shows significant subsidence of the seabed.

This likely contributed to the abrupt displacement of water that then crashed ashore.

More than 2,000 people lost their lives in the disaster. Preliminary results of various investigations are being reported here at the Fall Meeting of the American Geophysical Union - the largest annual gathering of Earth and space scientists.

The quake occurred on what is called a strike-slip fault, where the ground on one side of a rupture moves horizontally past the ground on the other side. It is not a configuration normally associated with very large tsunamis.

Nevertheless, this is what happened in the early evening of 28 September, around prayer time.

Two main surges of water were observed, the second being the biggest and pushing up to 400m inland.

Udrekh Al Hanif, from the Indonesian Agency for Assessment and Application of Technology (BPPT) in Jakarta, told the meeting that the generation source of the tsunami had to be very close to the city because of the short interval between the onset of the quake and the arrival of the high water - less than three minutes.

He and colleagues have sought answers in a depth (bathymetric) map of the long, narrow inlet that leads to Palu at its head. The team is still working through the results, but the data indicates the seafloor in much of the bay dropped down in the quake.

This, combined with a sharp movement of the crust northwards, could certainly have produced a tsunami, the Indonesian scientist said.

"When we overlap the bathymetric data from before and after, we can see that almost all of the area of the seafloor inside the bay subsides. And from this data, we can also observe [the movement] to the north. So, actually, we have a vertical and a horizontal displacement," Udrekh Al Hanif told BBC News.

Whether this behaviour was enough on its own to explain the tsunami's size is still open to question.

There is evidence of several underwater landslides in the data. These, too, could have been a factor.

Another possibility is an upwards thrust of the seabed in a zone some distance from Palu where the strike-slip fault splits into diverging tracks.

Movement on both tracks at the same time might have compressed the crust in between.

"This was a very unusual event but the tectonics tell us it could happen again," said Finn Løvholt from the Norwegian Geotechnical Institute.

"Indeed, this is not the first time an event has happened in Palu. It's probably the third or fourth event that has caused many fatalities. We had events in the 1960s and 1920s."

And this history is evidenced in the local culture where there are specific words to describe features of a tsunami and a quake. In the September event, Palu witnessed a lot liquefaction, where the structure of soils in the city was seen to collapse, to become fluidised and flow even on very low gradients.

Houses were consumed in the mud. Local people call this "Nalodo", which means something like "buried in black".

Hermann Fritz, from the Georgia Institute of Technology in the US, said Palu demonstrated the challenge confronting the local population. "This tsunami arrived very fast, within minutes," he stressed.

"That basically leaves no time for warnings. That's very different from Japan (in 2011) where there was an eternity of time - more than 30 minutes everywhere until the first person was killed by the tsunami. That's the challenge for these local tsunamis: people have to self-evacuate."

Widjo Kongko, also from the BPPT, spoke of some complacency following an emergency drill conducted in Palu in 2012. "It said go to high ground within 5-10 minutes. The people need to learn that the tsunami can come much, much quicker."