Field procedures

In this pilot study, we used detailed scale orthophoto maps (Institut Cartogràfic de Catalunya, 1:5000 black-and-white) to locate sampled sites and trees. These maps allowed good visualization of the zones and a better understanding of the processes by an exact localization of individuals in the area.

A total of 230 trees were sampled using different strategies at selected locations within the avalanche-affected area at Canal del Roc Roig.

1 Trees along two transects crossing the avalanche path were sampled to estimate the frequency of avalanche events in relation to the avalanche-path distance (this aspect is not dealt with in this paper). Each transect was 2 m wide, and all the trees were systematically sampled. One transect, 400 m long, was established in the upper part of the track at 2075 ma.s.l. measured at the centre of the gully; the second transect, 150m long, was located near the run-out zone at 1785 ma.s.l.

2 Selected old pines growing in the downslope forest right edge were sampled to document the occurrence of ancient avalanches (Reference Pelfini, Santilli, Caccianiga and GironiPelfini and others, 2001).

3 After inspecting the area, we sampled several zones where trees were more exposed to avalanche effects: at the arrival zone (see Reference Molina, Muntán, Andreu, Furdada, Oller, Gutiérrez, Martínez and VilaplanaMolina and others, 2004), at the end of the track where most trees were dead (some of them transported), and in the centre of the gully path, above and below the upper transect.

From trees that were dead or wounded, trunk cross-sections and wedges, respectively, were sampled (Fig. 3). From living trees cores were preferred to cross-sections. Cores were sampled from 0.30m height (basal cores) to 1.30m from the base of the tree trunk (dbh (diameter at breast height) cores). To reduce tree damage, an attempt was made to sample scars with increment borers (two to three cores per scarred tree) and this proved useful. From each tree the following data were recorded: dbh, height, tree morphology, health, position in relation to the centre of the gully, and age by counting trunk internodes in young trees. Only part of this information is used in the present paper. We extracted wood samples from 132 mountain pines.

Fig. 3. Types of signals in 1996 tree-rings: (a) wedge showing a scar, the dark coloration in the centre belonging to the secretion of resin and tannins; (b) compression wood in the last rings of the core, showing its dark coloration; (c) core with suppression growth in the last rings; (d) cross-section of a dead tree. Total number of sampled trees is 132 (109 trees were cored; cross-sections were taken from 18 dead trees, and wedges were sampled from 5 scarred trees).

Out of the avalanche zone, a selection of old trees that were apparently not affected by avalanches was sampled to build a master chronology. An ancient pine, 400 years old, was found, but most of the pines were less than 200 years old. The reliable period was 125 years according to an expressed population signal (EPS) value of 0.85 (Reference Cook and KairiukstisCook and Kairiukstis, 1990). This chronology was used as a reference series to compare with the series from trees that had been damaged by snow avalanches, to date the year of death of dead trees, and to verify the dates of tree-ring signals.

Laboratory procedures

Cores, wedges and cross-sections from the 132 sampled trees were prepared (mounted and sanded), visually dated and cross-dated according to standard dendrochronological methods (Reference Stokes and SmileyStokes and Smiley, 1968). Tree-ringsweremeasured and the cross-datingwas verified using the computer program COFECHA (Reference HolmesHolmes, 1983). Some samples could only be dated visually due to the anomalous tree-ring growth of trees affected by avalanches. All samples were carefully checked under a microscope, and details of single tree-ring characteristics were recorded to build the time-frequencydistributionof tree-ring signals related to avalanches.

Tree-ring growth signals revealed a variety of responses to disturbances, which could be due to avalanche events: eccentricity, compression wood, abrupt growth changes (releases and suppressions), scars and tree death. Eccentricity is a response to trunk tilting, although trees can also show eccentric tree-rings for other reasons. When conifers are forced to incline or bend, rings develop special characteristics: in the upper side of the trunk, rings become narrower, and in the downslope side rings grow wider. The latter can also form compression wood, which is darker than in normal rings (Reference TimellTimell, 1986). Eccentricity may be detected by statistical analysis of the variation in ring width (Reference Braam, Weiss and BurroughBraam and others, 1987), but the simultaneous observation of narrow and wide rings, together with compression wood, is a better technique for obtaining exact dates of avalanches. However, when working with cores it is difficult to assess eccentricity. For this reason, in this study we consider suppression in the upslope samples, release in the downslope ones, and compression wood separately. Abrupt growth changes, release and suppression periods, are sequences of wider and narrower rings, respectively, than the previous ones. A release is usually caused by elimination of competition, while suppression can be due to (i) partial loss of tree crown (breakage of branches or tree trunk), and/or (ii) partial uprooting. Scars are produced by trees as a reaction to wounds from the impact of stones, flowing snow, and branches and trunks of dragged trees carried by the avalanche. Scars were only sampled in young trees. Tree death was dated by cross-dating of tree-ring series of dead trees with the master chronology. We have considered sudden tree death when the last rings did not present a reduction in growth rate. A table was constructed with the tree-ring features from 132 dated trees. It contained only the starting year for every signal, and not the complete period during which the signal was observed.

Tree age was estimated by counting whorls along the main axis in young trees. For older trees, ages were estimated counting the rings of the basal samples (usually cores); when the pith was not present, this age was corrected by adding the missing rings to the pith (Reference Bosch and GutiérrezBosch and Gutiérrez, 1999). Since basal cores were taken at 0.3 m from the base, we also added rings from the number of whorls counted in young trees to estimate the final tree age. When cores were sampled higher than 0.3 m from the base we also added the missing rings to the base. The way we estimated tree age was not accurate (Reference Gutsell and JohnsonGutsell and Johnson, 2002), but this was not significant since tree-age class intervals are larger (50 years) than the error in age estimation. We discarded 22 out of 132 dated trees because cores were rotten and age could not be estimated.

Validation of tree-ring signals from 1996 known avalanche season

Many of the 132 dated trees (40%) were affected by 1996 avalanches at Canal del Roc Roig as shown by the different signals in tree-rings. Here we discuss the signals detected from 1996 to 2000 (Fig. 4). External tree morphology observations are not considered in this paper, but some comparisons were made for single trees to better understand some tree-ring responses.

Fig. 4. Tree-ring signal frequency, 1996–2000. The different types of signals are shown for each year.

Growth suppression was the most frequent signal in 1996 (50%). In some of the samples, the narrowest ring was found to be not in 1996−the immediate growing season after the events−but in 1997. Some of the trees still presented a suppressed growth rate in 2002. However, other disturbances can cause loss of branches, for instance, snow accumulation on the tree crown (Reference SchweingruberSchweingruber, 1996).This can magnify the amount of signals during avalanche seasons. Compression wood was also very frequent in treerings (25%) after 1995 and lasted from 1 to several years. Scars appeared mainly on the upper side of the trunk, but we only sampled five scarred trees.

Near the run-out zone, the series of trunk cross-sections from dead trees showed a sudden tree death after the growing period of 1995. In the upper part of the track, however, the final year of growth (1996) of two dead trees was very narrow. In the case of one tree in the central part of the track, this growth behaviour was due to incomplete uprooting, and the tree survived another year. Even though data are scarce, it can be deduced that the impact of the snow was less severe in the upper part of the track, near the starting zone, than at the end of the track where most of the trees were completely uprooted or broken and sometimes transported.

Tree-age distribution differences exist between trees with and without 1996 signals (Fig. 5a). The proportion of trees older than 100 years that do not have signals is higher than those with signals. The fact that the avalanche did not affect old trees can be explained by (i) their larger dimensions, and (ii) the fact that they do not grow near the path, where snow-avalanche effects are more severe. Only 3 out of 22 trees older than 100 years recorded a response in the 1996 ring and it was suppression (Fig. 5b). Suppression was present in all age classes. Compression wood was more abundant in young individuals (0–50 years), but was also present in trees 50–100 years old (2 out of 44 trees), though not in trees older than 100 years. The age of dead trees ranged from 50 to 100 years, but this could be due to biased sampling; we specifically sampled dead trees near the end of the track, near the run-out zone.

Fig. 5. (a) Age-class distribution for trees with (n= 44) and without (n = 66) signal in the 1996 ring. (b) For trees with signals in the 1996 ring, age-class distribution for each type of signal (re, release; su, suppression; co, compression wood; wo, wounds (scars); de, dead trees). Tree-ring signals (n = 53) were redundant in some trees.

Tree growth responses can show a delay, and tree-ring features may not appear until 1 or more years later (Reference JohnsonJohnson, 1987). The problem of dating the 1996 avalanche was solved because most of the signals were detected during the immediate growing period of 1996 (40.15% of tree-ring signals in relation to 132 sampled trees with and without signals), and a smaller proportion during subsequent years (Fig. 4). We found that scars provide a decisive criterion for establishing the exact year, because they are produced by an immediate tree reaction to wounds. For this reason, a higher number of samples should be gathered from scarred trees in future sampling campaigns.

It is important to point out the significance of finding all these signals together because tree-rings can show similar responses to different environmental disturbances. Consequently, only the combination (redundancy) and a high number of these responses reliably confirms past events. Here the evidence of signals from past avalanche events is discussed. Tilting of tree trunks causes eccentricity and provides redundant signals: abrupt growth changes (suppression and release), compression wood or all together. Therefore, the exact year of tilting can be dated, providing reliable information on past events. One of the clearest signals in tree-rings is sudden tree death, but wood goes rotten and tree-rings become decayed after 50–70 years (Reference Bosch and GutiérrezBosch and Gutiérrez, 2001). Near the run-out zone, it was observed that the sapwood of some of the dead trees from 1996 was already rotting, so this source of information about the date and age of dead trees may disappear faster.

Past avalanche events at Canal del Roc Roig

At present, when snow avalanches reach the railway of Núria tourist resort and damage the line, the train crew documents them, as was done during the winters of 1972 and 1996. These two avalanche seasons were detected after examination of the dendrochronological samples. As explained above, two snow avalanches were documented in 1996. However, it is not possible to distinguish them in treering signals, because tree signals are not different for the two events, as trees only provide annual resolution for avalanche events.

Besides 1996 and 1972, three other feasible avalanche years were detected by a high frequency of signals: 1991, 1986 and 1930 (Fig. 6). There is no historical evidence for these possible events. These should be checked in future campaigns to verify their occurrence. Other possible avalanche events before 1930 could not be corroborated because samples were scarce and results not reliable, even though tree-ring signals were registered up to the mid-18th century (Fig. 6).

Fig. 6. Yearly frequency distribution of signals detected in tree-ring series susceptible of being caused by avalanche events in Canal del Roc Roig and sample depth (number of trees), 1750–2002.

The amount of signals dated for 1930 and their spatial distribution (Fig. 2) point to a large avalanche. Pines affected during winter 1930 occupy a broader area in the starting zone than those affected during 1996. Thiswouldmean a larger amount of snow collapsing downslope than during the 1996 events. The starting zone was also different: signals belonging to 1996 are not detected on the hydrographical right side of the path, while signals linked to 1930 are abundant on this side. The hypothesis that a large avalanche occurred in 1930 is quite probable, as these forests have not been managed and few of the sampled dead trees were 70 years old. Snow avalanches do not always run through the full extent of their tracks and flow out over the run-out zone. Small avalanche events are more frequent (Reference CarraraCarrara, 1979). It is interesting to note the variety of trajectories that snow avalanches can have along the same path. In Figure 2, dead trees about 70 years old caused by the February 1996 events were abundant on the righthand side in a forested area near the run-out zone. This place is some metres above the central part of the track section, and trees had not been destroyed during the 1972 event. The establishment of trees in this area probably took place mainly after 1930; some of them were small and so they could have survived this event. Therefore, the hypothesis could be that the 1930 avalanche destroyed trees growing there, and this avalanche could have had a similar trajectory to that of 1996. Further research must be undertaken on this topic.

This study was also useful in corroborating the mapped avalanche limits edited in the avalanche-paths map (ICC, 2002) by the location of all the trees with signals from different past avalanche events (Fig. 2). For instance, the boundaries described in the map in the run-out zone on the other side of the railway track precisely coincide with those found by dendrochronological means (see Reference Molina, Muntán, Andreu, Furdada, Oller, Gutiérrez, Martínez and VilaplanaMolina and others, 2004). But in other areas, as in the starting zone, some discrepancies exist. So from the avalanche signals in tree-rings, cartography can be improved.