The wildfire site is situated on the southwest exposed rock walls of a glacially over-steepened Alpine trough valley. The valley is characterized by almost vertical rock walls several hundred metres high, which are mainly made of Mesozoic limestone (Dachstein Formation). The limestone is characterized by predominantly thick bedding, sudden changes in the joint mass structures and the presence of dominant fault systems. In the wildfire-affected part of the rock wall, the bedding has a predominantly medium steep dipping of 35 to 45∘ in the direction of the rock wall (from northeast to northwest) (Melzner, 2015, 2018). The bedding planes form locations preferable for trees to grow and are usually covered with a thin layer of debris (Fig. 1). A fixed-rope climbing tour is installed in the rock wall, which is frequently used by numerous climbing tourists. The talus slope below the rock wall is relatively short and has an inclination between 30 and 40∘. The soil type consists of scree (Ø ≲10 cm) or medium compact soil with small rock fragments and some larger blocks. The scree is covered by a very thin layer of soil and organic matter which can be classified as brown rendzina and brown earth (Fig. 2). The vegetation is characterized by coniferous trees, mainly spruces and broad-leaved trees such as beech and larches. The pre-fire vegetation was composed of medium old forest and an understory of sparse bushes on the rockfall talus slope. The forest on the talus slope beneath the rock wall is designated as a protection forest for the houses in the valley floor. The annual precipitation is about 1743 mm. The highest 1 d precipitation amount since 1901 was measured on 12 August 1959 to be 118 mm, and the maximum annual precipitation measured was 2085 mm (1954). There are 20 to 30 convective summer thunderstorm days per year. Precipitation as snow occurs normally between November to April during which snow cover can reach thicknesses up to a few metres.

On 21 August 2018 at 09:30 LT a wildfire was presumably initiated by a carelessly discarded cigarette or the reflection of a broken glass bottle at the foot of the rock wall. At that time there were three groups of about 20 climbers on the via ferrata. Since the fire could only be extinguished from the air by helicopters, the via ferrata had to be evacuated to protect the climbers from falling rocks and branches caused by the downwind of the helicopters while fighting the fire. The fire rapidly spread up the rock wall (area size of about 3 ha) affecting the trees growing mainly on the bedding planes of the limestone (Fig. 1). The protection forest beneath the rock wall was not affected by the fire (Fig. 2). During the night of 21 to 22 August the first evacuations of the houses beneath the rock wall took place as burned trunks, rootstocks and rock blocks were falling down the rock wall, the latter approaching two houses. Sixteen mapped rockfall boulders which reached the settlement area had volumes smaller than 0.3 m3 (Fig. 3). With the exception of minor damage to one building, no severe damage to buildings occurred, and local inhabitants were not injured. In the following days the firefighter brigades tried to extinguish the fire from above the rock wall with fire hoses and from the air with helicopters carrying water containers and buckets. In total, four police and military helicopters were flying during the days of the fire, refilling the buckets with water from the nearby Lake Hallstatt every two minutes. During the 4 d of the firefighting operation, up to 100 people (firefighters, police officers, military personnel and mountain rescue team members) were on duty every day. Unusual low-wind conditions and rainfall (starting on 24 August 2018) prevented the spread of the fire towards the village of Hallstatt. The official end of the firefighting mission was on 28 August 2018. A rockfall hazard and risk assessment conducted by the Geological Survey of Austria (Melzner, 2015) formed an important part of the wildfire emergency response. Preventive rockfall hazard actions by the Austrian Torrent and Avalanche Control (WLV) after the wildfire included the (i) establishment of temporary rockfall protection measures (embarkments and simple rockfall fences) in order to be able to clear the wildfire area, (ii) clearance of the wildfire area (removal of loose stones, boulders, trees at risk of falling, etc.), (iii) repair of pre-existing rockfall protective structures damaged by the rockfall, and (iv) sowing of seeds in the wildfire-affected scree and soil.

Indicators of fire severity (Figs. 1 and 2) are the colour of the trees and the decomposition degree of the leaves and needles. Unaffected trees have a green and unaltered colour, whereas burned or heated trees are easily recognizable by their brown colour. Varying degrees of consumption of the needles and leaves and organic matter can be related to different classes of fire severity. According the classification of Keeley (2009), the trees in the affected area show moderate or severe surface burn. This is visible in that most of the burned trees still have needles, but all understorey plants and pre-fire soil organic layer (besides a post-wildfire needle cover) were consumed. In the transition area between the burned and not-burned areas, the vegetation shows indicators for light fire severity, expressed by green needles, although the stems may be scorched, and the understory plants and soil organic layer are largely intact. At the foot of the rock wall we observed a burning tree falling down the rock wall carrying a large rock, which burst into various rockfall boulders during the first impact with the ground.

The vertical relief of the rock walls, the anabatic winds and patchy vegetation pattern caused an upward jumping of the fire, resulting in a spotty fire pattern (Fig. 1). Thus, the residence time of the fire and the heating duration were reduced, leading to a less direct influence of the high temperatures on the rock mass structure. Fire-induced rock surface alteration and cracking due to thermal shock are typical rock-weathering processes occurring during a wildfire (Dorn, 2003; Shtober-Zisu et al., 2015). Thermal shock takes place when the thermally induced stress event is of sufficient magnitude to make the material unable to adjust quickly enough to accommodate the required deformation and accordingly fail (Hall, 1999). As a result, the surface failure takes the form of cracking or exfoliation due to the compression and the shear stress it induces (Yatsu, 1988). Moreover, rocks composed of several minerals, each with different coefficients of thermal expansion, may experience stresses resulting from the minerals' differential thermal response to the heating and cooling cycles (McFadden et al., 2005).

Spalling or the formation of exfoliation fissures (caused by insolation weathering) may be less severe in such exposed-terrain conditions compared to more gentle slopes (Blackwelder, 1927; Zimmerman et al., 1994; Shakesby and Doerr, 2006; Shtober-Zisu et al., 2015). In the course of the wildfire, abundant small rock fragments had come to rest directly at the base of the rock wall. The rockfall boulders which were detached from the rock wall during the wildfire could be easily identified in the field, as they usually have at least one black (scorched) side (Fig. 2). Some smaller rockfall boulders with volumes of <0.3 m3 have reached the valley floor (Fig. 3).

The slope in the uppermost part of the rock wall is covered by gravel, stones and blocks with a matrix composed of fine clastic material and ash. It could be mobilized in the form of a debris slide or flow in a heavy precipitation event. Such an event has not been documented thus far in this area. As the organic material mantling the scree slope in the upper part of the rock wall was consumed completely (Figs. 1 and 2), we observed that the ash covers the surface. Ash has a kind of “sealing effect” which reduces the infiltration, accelerates the splashing effect and increases the surface runoff (Brook and Wittenberg, 2016). It can be assumed that future frost and thaw cycles will further weaken the rock or that the loose slope debris in the upper-rock-wall area will be remobilized by heavy precipitation events. In forests, wildfire usually generates a mosaic of different levels of burn severity (Neary et al., 2005). In sites affected by fire of light-to-moderate severity, needle cast occurs when leaves from the scorched trees fall down and blanket the surface, thus protecting the soil from further erosion (Cerdà and Doerr, 2008; Robichaud et al., 2013). There are numerous studies addressing the effect of ash deposits on runoff and erosion processes, rates, and quality (Bodí et al., 2011). Results, however, are inconclusive; while many suggest that ash temporarily reduces infiltration, either by clogging soil pores or by forming a surface crust (Onda et al., 2008), others indicate that ash and specifically the black char produced during light-to-moderate fires might increase infiltration by storing rainfall and protecting the underlying soil from sealing (Wittenberg, 2012). The ash layers may also protect the burned soil against raindrop impact and related splash erosion, and its leachates may reduce soil erodibility by promoting flocculation of the dispersed clays (Woods and Balfour, 2008). Ash particles penetrate, accumulate and shelter under the rock spalls formed during the fire, even for several decades (Shtober-Zisu et al., 2018).