Pyroclastic flows have often been considered to be by and large, unsurvivable, especially since the devastation in 1902 of St. Pierre on Martinique, many fatalities are due to hypoxia and severe thermal injuries (Hansell et al. 2006; Baxter et al. 1998).
According to the USGS webpage about pyroclastic flows, these events can occur very suddenly, without much warning at all (e.g. Mt. Ontake) and travel fast, and can be several hundreds of degrees Celsius. Pyroclastic flows also contain considerable amounts of toxic gases and rock fragments. The presence of water bodies in the path can result in lahars and mudflows.
As pyroclastic flows are variable in nature and can be sudden, there is no one-way of volcanologists protecting against its effects, except to get out of the way (if possible). The use of fire suits will limit mobility and still may not protect from the flow itself.
In the AGU blog So you want to be a volcanologist?, Jessica Bell, a volcanologist from the USGS makes the point that being prepared is crucial for greater chances of survival, emphasising the point of 'doing your homework' about the eruptive history of the volcano before going to the volcano. Gas masks, gloves, long sleeves and long pants are also crucial. Critically, a volcanologist must be:
very careful to avoid areas that could be dangerous – within the path of pyroclastic flows or mudflows or lava bombs, near gas vents or lava flows, or unstable ground are just a few.
It should be noted that several media reports mention that not all volcanoes have these kinds of shelters.
But where the shelters are present, they could provide an obstacle that could affect pyroclastic flow propogation but only when the shelter is distal to the eruption and if the flow does not last too long, related to modelling performed by Todesco et al. (2006) and earlier by Baxter et al. (1998) who assert that this situation is survivable.
Baxter et al. (1998) also state that through their models, the most effective shelters can shield against the dynamic pressure and heat transfer associated with a pyroclastic flow, especially if the flow is of short duration. This effectiveness obviously diminishes when the flow lasts a considerable amount of time, is large-scale or the shelter is too close to the eruptive source.
What these shelters won't protect from is fine ash dust, and not everyone sheltering will necessarily have a gas mask (as per the volcanologist above). Another problem with this, as noted after the Mt. Unzen eruption of 1991 is that even if shelters are spaced evenly apart - time is a major factor - the time to reach the shelter against the time it takes for the pyroclastic flow to reach that location.
A lot of shelters around some volcanoes that have had a past history of pyroclastic flows have 'exclusion zones' with the shelters in the distal regions of previous eruptions, this method of exclusion zones is claimed by Hansell et al. (2006) as being one of the only effective protective measures from proximal effects of an expected pyroclastic flow - which also describes, alongside the modelling by Baxter et al. (1998), the way that these shelters can have any effectiveness by being in the distal regions.
Baxter et al. 1998, Physical Modelling and Human Survival in Pyroclastic Flows, Natural Hazards
Hansell et al. 2006, The health hazards of volcanoes and geothermal areas, Occupational Environmental Medicine
Takahashi, 2007, Unzen-Fugendake Eruption Executive Summary 1990－1995, Unzen Restoration Project Office,
Kyushu Regional Construction Bureau,
Ministry of Land, Infrastructure and Transport
Todesco et al. 2006, Pyroclastic flow dynamics and hazard in a caldera setting: Application to Phlegrean Fields (Italy Geochemistry, Geophysics, Geosystems