The Bor Forest Island Fire Experiment Fire Research Campaign Asia-North (FIRESCAN)
III. Vegetation and Fuels
1. Prefire Vegetation
Stand structure and vegetation composition were evaluated on a 10 m wide by 320 m long transect crossing Bor Forest Island in an east-west direction. Along the transect we determined stand composition, average tree height and diameter (DBH, 1.3 m), stem basal area, average tree age, and standing volume. DBH and tree condition (healthy, declining, dead) were determined for all 525 trees in the transect. Trees were categorized into 4-cm diameter classes (0-4cm, >4-8cm, >8-12cm, >12-16cm, >16-20cm, >20-24cm, >24-28cm). We measured tree height and height to the bottom of the crown for 5 randomly selected trees in each diameter class (a total of 45 trees). Stand age was determined by increment cores taken from 20 trees. Downed wood volume was determined by measuring diameters and lengths of fallen trees along a 100m transect. Downed trees were classified by stage of wood decomposition as either intact or losing the shape of the tree. Living ground cover was described with regard to the vegetation structure and composition on 40 1m2 plots evenly distributed across the transect. Projected cover and abundance of species in the grass/low brush layer were estimated visually according to the Drude Scale (Schenikov 1964):
Vegetation was typical of central taiga forests of Western Siberia. Before the experimental fire, Bor Forest Island supported a pure stand of Scots pine (Pinus sylvestris L.), well-stocked and in even-aged patches (trees on most of the island were 130 years old). The canopy cover was relatively high (0.6-0.7) and average density was 1470 trees ha-1. Average diameter and height of living trees were 18 cm and 17 m, respectively. There were also about 170 stems ha-1 of standing dead trees, heavily concentrated in the smaller size classes (Figure 5). The volumes of standing living trees, snags and downed wood were 248, 14.6, 17.3 m3 ha-1, respectively. Downed wood was in various stages of decomposition.
Fig.5. Diameter class distributions of dead and living trees in a 320 m by 10 m sample plot, one year before the fire. Size classes indicated are upper limits of 4 cm diameter classes.
Advance regeneration was sparse, but included both Pinus sibirica Ledeb. and Pinus sylvestris L. The understory was of low density, with scattered patches of Rosa acicularis and Salix caprea dominating in some areas. The edge of the island and depressions were covered by Spiraea salicifolia. The ground cover was of mosaic character, but a dense mat of Cladonia lichens mixed with mosses dominated over much of the area. Lichens were represented by Cladonia stellaris, C. sylvatica, C. rangiferina, and C. uncialis. Green moss (Pleurozium schreberi) and low shrub plants (Vaccinium vitis-idaea, V. myrtillus, V. uliginosum and Ledum palustre) were found in shallow depressions (Table 3). The heights of the lichen, moss, and low shrub layers were 5-7, 3-5 and 12-35 cm respectively. The total biomass of living ground cover was 15.9 t ha-1. Litter and forest floor organic layer load was 17.6 t ha-1.
Tab.3. Characteristics of pre-fire living ground cover in a pine-lichen forest, Bor Forest Island experimental site.
2. First-year Vegetation Recovery and Stand Conditions
Photo points: Ten permanent photo points were established on the island before the fire on sites selected to represent the range of prefire stand conditions in terms of tree size distribution, occurrence of snags, regeneration, and canopy closure. Both plot centers and base points for the camera were permanently marked. A standard 35 mm lens was used for photographs. Figures 6a, 6b, and 6c illustrates typical stand conditions at one of these photo points before, immediately after, and one year after the fire. Note the degree of crown drying between 1993 and 1994, the lack of obvious regeneration or herbaceous vegetation in the year following the fire, and the loss of bark from dead trees.
Fig.6. View from photo point 5 of typical stand condition on the Island. a. Before the fire, b. Immediately after the fire, c. July 1994, one year after fire.
Postfire vegetation sampling: When we returned to study forest regeneration processes on the experimental site, one year after the fire, we did not resample prefire plots to determine vegetation cover since living ground cover was virtually nonexistent. We observed only isolated individual pine seedlings in the interior of the island and newly emerged small sedge and wild rosemary (Ledum) sinusia where the fire burned into the edges of the bog. Otherwise there was no visible regeneration of either herbaceous or woody species. Plots will be resampled in 1995 and in following years to document vegetation development.
3. Fuel Loading and Surface Fuel Consumption
To determine ground fuel load and structure, the Russian team measured litter, forest floor organic matter, and living ground cover, which was represented mainly by lichens. Along the east-west transect, we established 10 1 m2 plots (one every 10 m) for determining prefire lichen, litter, and small branch loads. Forest floor organic matter was measured in 0.5 m by 0.5 m subplots. Samples were oven dried to constant weight and then weighed. Before sampling, we measured depths of lichens and litter and of the forest floor organic layer. After the fire we established 10 additional 20 by 25 cm plots next to the prefire plots to measure the organic matter remaining after fire. Twenty by 25 cm plots were also established next to these plots the year after the fire to determine post-fire litter volume. The Canadian team also measured preburn and immediate postburn fuel loads at 12 sites throughout the burned area, primarily to quantify fuel consumption in the fire. Their methods are described in Section III.
Pine-lichen stands are remarkable for rapid development of flammability after a fire and their high flammability over a long period during the fire season. A considerable load of ground fuels on the forest island (Table 4) promoted the occurrence of high fire intensity, even in the absence of significant ladder fuels. The prefire surface fuel loads of 33.6 t ha-1 were decreased by 50 percent a year following the fire. The Canadian team estimated prefire ground and small (<3 cm) surface fuel loads of 44.9 t ha-1, about 60% of which was consumed by the fire. We will continue to explore reasons for these differences. The prefire duff layer averaged 28.7 t ha-1, with 13.9 t ha-1 of living lichens over the duff. A higher proportion of surface fuel than ground fuel (lichen and duff) was consumed. The Canadian group reported 48 % reduction in the lichen and duff layer by weight and a 88% reduction in depth (all the lichen layer plus an average of 1.3 cm of duff). Ivanova (Table 4) reported 79% reduction in the loose surface litter and 21% reduction in the forest floor (duff) layer one year after fire. Except around the bases of trees, there were few places where the fire burned to mineral soil. Essentially all of the loose litter measured in 1994 had accumulated since the fire. This litter consisted primarily of needles (77.6 percent) and bark (8.6 percent) shed from injured and dying trees following the fire. The prefire research allowed us to estimate prefire fuel loads and biomass consumption, and to investigate postfire processes of forest restoration in this area.
Tab.4. Russian measurements of prefire and postfire (1 yr. after fire) structure and load of fuels in a pine-lichen forest (dry weights) at Bor Forest Island. Miscellaneous litter includes lichens (prefire), needles, bark, etc. Small branches are branches up to 2 cm in diameter. The Forest floor is the compacted forest floor organic layer.
To measure the effects of the fire on the carbon and nutrient contents of the soil organic layer, and to monitor postfire accumulation of soil organic matter, two 30m x 30m plots were established. Prefire and postfire samples were collected on 5 and 7 July, 1993. The chemistry of the humus layers before and after the fire is described in Table 5. Fire intensity on Plot 2 was higher because of a greater amount of downed woody material.
Tab.5. Element content and standard deviations (kg ha-1) of the humus layer of Bor Forest sland before and after fire.
4. Tree Mortality
Stand-level mortality: The extent of crown fire was estimated from low-level aerial photographs by determining the percentage of the area of the island that experienced complete canopy removal (all foliage and small twigs were combusted). Because the fire burned the entire surface of the island, the remaining area was assumed to have burned in surface fire. Areas of crown scorch were determined based on foliage color (light green or brown in areas where crowns were scorched, dark green where they were not). A 30 m by 30 m plot was established in July 1994 to evaluate stand structure, tree mortality, and insect damage. The plot contained 203 trees 6 cm or greater in diameter. Tree diameter (DBH) was measured at 1.3 m from the ground surface. Maximum scorch height on the bole and the percent of crown with dried foliage were recorded for each tree. Trees were classified as alive, dead, or dying. The plot was located on the northwest corner of the island in an area that had burned in surface fire only. It extended from the edge of the island up to the higher ground in the interior. In 1994 and 1995, mortality was also evaluated for the 525 trees on the vegetation transect described above.
Approximately 57 percent of the area of the island burned as a crown fire (see Figure 15). All of the trees in these parts of the stand died. In another 25 percent of the area, the surface fire was severe enough to scorch most of the crowns. Most of these trees were also dead within a year. In the plot established to evaluate mortality and insect infestation in the underburned area, trees had an average diameter of 15 cm. Of the 203 trees sampled, 17 percent were 6-11 cm in diameter, 53 percent 12-17 cm DBH, and 30 percent 18-23 cm DBH (Figure 7). Fifty-five percent of the trees in the plot had maximum char heights of greater than 2 meters (Figure 8). Except for the smallest size class, there was generally a positive correlation between tree diameter and char height. A similar relationship has been observed by Tsvetkov (1993) for Larix. Tree height was related to the proportion of dry foliage in the crowns, and nearly all trees less than 12 cm DBH were dead and had 100% dry foliage. At the time of sampling in July 1994, 75 percent of the trees were dead or dying (Figure 7). Most trees 20-24 cm DBH are expected to survive. We anticipate additional mortality as insect damage increases and trees of low vigor continue to die. On the east-west vegetation transect, 94.5% of the 525 trees alive at the time of the fire were dead in 1994; this had increased to 98.5% by July 1995, at which time the only living trees remaining were in scattered pockets on the slope along the margin of the island.
Fig.7. Diameter class distribution of live, dead, and dying trees one year after the fire on the 30 by 30 m sample plot.
Fig.8. Distribution of scorch height by diameter class on the 30 by 30 m sample plot.
Modelling: We also measured fire injury to individual trees for the purpose of beginning to develop models for predicting mortality of P. sylvestris following wildfires. Immediately following the fire, we established 5 plots of twenty trees each. Plot locations were randomly selected but were linked to the transects established for fuel sampling. Twenty trees were tagged in an approximately circular pattern around the plot center. Plotradius varied with tree density, so that each plot included 20 trees. An attempt was made to distribute plots in areas of varying fire severity. However, this initial sampling resulted in a very high proportion of sample trees in medium to high fire severity areas, and most of these trees had died by 1994. In July 1994 we established 5 additional plots. Plot centers were again randomly located, but sampling was restricted to areas of low to moderate fire severity to ensure an adequate sample of surviving trees for modelling purposes. In both years, all trees in the plots were tagged and numbered. Measurements included maximum and minimum scorch height on the bole, tree diameter, tree height, height to bottom of crown, and depth of residual forest floor organic matter at the base of the tree (as a possible indication of impacts of fire on shallow roots). In 1993, we also did visual estimates of percent crown scorch. This was not possible in 1994 because dead needles resulting from postfire insect damage and decreases in tree vigor could not be distinguished from scorch, and many needles had already been shed. In 1994 we also noted whether trees had visible evidence of insect infestation on the boles (as indicated by exit holes in the bark and by visible insect frass).
Data were analyzed following procedures described in Regelbrugge and Conard (1993). Logistic regression analysis (Walker and Duncan 1967) was used to model the probability of postfire tree mortality as a function of tree size and fire damage variables. The model used is of the form:
where P(m) is the probability of postfire mortality, X1 through Xk are independent variables and ß1 through ßk are model coefficients estimated from the data. We used DBH, maximum and average ((maximum – minimum)/2) stem bark char height, relative char height (height of bark char/tree height), depth of forest floor organic matter, and percent of canopy volume scorched as independent variables to predict fire-induced mortality. The SAS LOGISTIC procedure was used to obtain maximum likelihood estimates of the model coefficients and model fit was evaluated using the Homer and Lemeshow goodness of fit statistic (SAS Institute 1989;Saveland and Neuenschwander 1990).
Of the 201 trees sampled for development of mortality models, 57 percent were dead by a year after the fire. Ninety-nine percent of dead trees sampled, as well as 85 percent of the living trees, were infested with bark beetles. Although infestation levels tended to be low in living trees, at least another year will be required to ensure that all mortality has occurred before developing final models. Only 14 of 100 trees were intermediate in canopy scorch estimates. Forty-three percent had no crown scorch and 43 percent had 90 to 100% crown scorch. Of those with no crown scorch, 56% were dead a year after the fire. Of those with 90 to 100% scorch, 97.6% had died.
General population characteristics for trees used to develop mortality models are in Table 6. These trees were selected to provide a well-distributed set of characteristics for model development, and not to describe the stand in general. Diameters were slightly larger than those measured in the 30 by 30m plot and in the vegetation transect, primarily because sampling focused on areas of incomplete mortality along the fringes of the island, where average tree size tended to be larger that in the interior.
Tab.6. Population characteristics of trees used in developing mortality models.
Preliminary mortality models were developed both for the trees measured only in 1993 (where we could incorporate canopy scorch) and for all trees. For 1993 trees, a model incorporating canopy scorch, DBH, and crown length seemed to provide excellent fit, with 97 percent concordant pairs, and a Hosmer and Lemeshow goodness of fit statistic with p=0.9937 (Table 7). We also attained good fit with a model incorporating crown scorch and DBH(94.7 percent concordant pairs, goodness of fit p=0.9518; Table 7). The best model for all trees together incorporated relative average char height and DBH2. This model had 83.3% concordant pairs, with p=0.7702 for the goodness of fit statistic (Table 7). Any of these models might be sufficient for estimating stand-level mortality, but the ones based on crown scorch will be more reliable for predicting mortality of specific trees. Unfortunately, models based on crown scorch have the problem that model parameter estimates are very dependent on how soon the measurements are made after the fire. If these measurements are not made immediately, increasing crown drying in stressed trees may cause browning of needles not directly related to crown scorch. This problem was evident in data from the mortality plot (not shown) where crown drying was not nearly as reliable a predictor of mortality as was the crown scorch we measured immediately after fire. However, all models presented here are preliminary, as we expect continuing mortality.
Tab.7. Parameter estimates for the best fitting mortality models for 1993 and 1994 data sets.
Insect damage: The year following the fire, insect damage was evaluated in early July on trees in the 30 m by 30 m plot established for characterizing forest structure. All trees in the plot were visually inspected for signs of insect damage (emergence holes in bark and insect frass). In addition, a representative test area was established in the northeastern corner of the experimental plot. Trees were categorized as living, dying, or dead. A bark pallet of approximately 0.2 m2 was removed from each of five trees, and counts were made of the number of living larvae under the bark, the number of mother passages in the bark, and the number of larval passages in the wood. We also recorded the percentage of the crown that had dried needles. Living larvae were brought back to the laboratory for identification.
Visual inspection showed that all trees in the 30m by 30m plot were infested with bark insects. The bark of many trees had already begun to peel off even a year after the fire, and piles of bark at the foot of the trunks were common. Insect activity was so intense that one could hear the sounds of bark insects in the tree stems throughout the forest. All of the five sample trees showed evidence of heavy insect infestation in the bark plates sampled (Table 8).
Tab.8. Insect populations in the sample trees, in July 1994 at Bor Forest Island
Although there was considerable variation in the level of insect infestation from tree to tree, the levels were in general quite high, both in dead trees and in those that were still living. All of the living larvae taken from the sample trees were in the genus Monochamus, but species identification was not attempted. The adult beetles observed were mainly grey capricorn beetles (Acanthocinus edulis L.), brown capricorn beetles (Criocephalus vresticees L.), and big pine weevils (Hylobius abietis L.). The small plot area makes it likely that important species were not observed.