People free a man from the rubble of a destroyed building after an earthquake hit Nepal, in Kathmandu. Photo / AP
In the aftermath of the Nepal earthquake comes another deadly threat - landslides.
The April 25 quake, which struck 80km northwest of Kathmandu, has left more than 5200 people dead, with at least 250 people still missing.
But the risk of landslides - capable of burying whole towns under tonnes of earth dislodged from the region's steep, quake-weakened countryside - will continue to pose a threat over the coming months.
And that risk would increase when monsoon rains arrived in summer, potentially overwhelming dams and sending catastrophic floods down valleys.
Dr Tom Robinson, from the University of Canterbury, has has produced preliminary maps of where landslides are likely to have occurred.
"They can also block or severely damage roads into affected regions, cutting those locations off completely at the time when they most need access."
As well as blocking roads landslides, could also block rivers, which presented probably one of the most severe hazards in Nepal currently, he said.
Once a river was blocked, it immediately began to pool behind the landslide, flooding upstream areas and any settlements or infrastructure.
"These dams often don't last long, and when they fail that can release a catastrophic flood wave down-valley.
"With monsoon season approaching this is a major hazard as dams can be overwhelmed by heavy rainfall and fail when the water behind has reached its peak height."
With satellite images being limited by flight paths and weather conditions, landslide modelling immediately after the event had enabled Dr Robinson and other scientists to rapidly identify areas where major landsliding is expected.
"This can help prioritise search and rescue zones on the ground as well as divert satellites to look at the regions of most concern first."
The models being used were based on previous events in USA, China, Taiwan, and New Zealand, and determined landslide hazard based on the slope angle and position, strength of ground shaking, and proximity to rivers and faults.
The models showed hazard on a relative scale with values closer to 1 having the highest likelihood of landsliding.
"Currently the models suggest the slopes between Kathmandu and the epicentre are likely to be the worst affected.
"This is an emerging field however, and this is the first event these models have been applied to in an emergency situation."
The University of Michigan has also modelled the risk, finding tens of thousands of high-risk spots for the coming weeks and months alongside locations where landslides likely occurred during the 7.8 earthquake.
"The majority of them, we expect, have already happened and came down all at once with the shaking on Saturday," said Dr Marin Clark, a geomorphologist at the university.
"But there will still be slopes that have not yet failed but were weakened. So there will be a continued risk during aftershocks and with the recent rainfall, and again when the monsoon rains arrive this summer."
Information from the modelling have been shared with agencies responding to the disaster, and the models were now being used to prioritise satellite observations and the analysis of data from those satellites.
"The satellites looked first at places where lots of people live, including Kathmandu and the foothills areas to the south," Dr Clark said.
"Those areas do not look significantly impacted by landsliding, but we're worried about the high country."
The university's modelling found the region at highest risk for landslides and mudslides was the mountainous area along the Nepal-Tibet border, north of Kathmandu and west of Mount Everest, directly above the fault rupture.
The highest-risk zone was at elevations above just under 2500 metres in a region that covered more than 45,400 square kilometres.
Key information:
•The model results show the relative landslide hazard for each 60mx60m cell on a scale from 0-1. A value of 1 represents cells that, compared to the rest of the affected area, are almost certain to produce landslides.
•A value of 0 represents a cell that, compared to the rest of the affected area, are almost certain not to produce a landslide. Values between 0and 1 show the relative likelihood for that cell.
•The model output shows landslide hazard. It is possible for a cell to have high hazard (value close to 1) and not produce a landslide in this event. Similarly a cell could have low hazard (value close to 0) and still produce a landslide.
•It is expected however that the majority of landslides will be located in the highest hazard values. In general, the areas with values >0.7 (yellow or warmer on the colour scale) are where the most intense landsliding would be expected.
•Affected area is estimated from the area covered by 20 per cent of highest hazard values, which in the test events has corresponded the region of most intense landsliding.