Tree-Ring Reconstruction of Bark Beetle Disturbances in the Picea schrenkiana Fisch. et Mey. Forests of Southeast Kazakhstan

: Ips hauseri Reitter is the most important bark beetle on Picea schrenkiana in southeast Kazakhstan, but its biology, ecology, and outbreak dynamics are poorly known. We dendrochronologically reconstructed a 200-year history of disturbances in the Kazakh Tien Shan P. schrenkiana forests. Only localized, low-severity bark beetle events occurred during the reconstructed period, indicating that extensive high-severity bark beetle outbreaks have not occurred historically in the Tien Shan spruce forest, unlike bark beetle outbreaks in spruce forests in North America, Europe, and Russia. Disturbance frequency doubled after about 1965, probably due to warming climate. Results, combined with the failure of an outbreak to fully develop after blowdown events associated with hurricane-force windstorms in 2011, indicate that prolonged drought may be necessary to sustain I. hauseri outbreaks, or that year-to-year variation in the Tien Shan weather prevents outbreak development. I. hauseri is probably less aggressive than I. typographus , at least on their natural hosts within their natural ranges.


Introduction
High-severity wind storms with heavy rains caused significant blowdown in spruce forests in the Tien Shan Mountains of southeastern Kazakhstan in May and June 2011. Large-scale blowdown events provide extensive breeding material for bark beetles, and bark beetle outbreaks in spruce forests are often triggered by such events [1][2][3][4][5][6][7][8]. Bark beetle populations, especially Ips hauseri Reitter (Coleoptera: Curculionidae), began to increase in blown down Picea schrenkiana Fish. & C.A. Mey; subsp. tienshanica Rupr. (Schrenk's spruce) [9,10]. I. hauseri is known to exhibit particularly aggressive population dynamics that enable rapid population buildup [11,12]. However, the frequency, size, and severity of past I. hauseri outbreaks in Kazakhstan are unknown. Additionally, non-native bark beetle species I. duplicatus L. and I. sexdentatus (Boerner), which were established in the mid-to late-1960s [9,10], were also observed in the blowdown area. The biology, ecology, and impact of these species in the Tien Shan spruce forests are unknown.
Picea schrenkiana is native to southeast Kazakhstan, Kyrgyzstan, and western China [13]. It is one of the most important tree species in Kazakhstan, of which less than 5% is forested [14,15], and Kyrgyzstan, The Tien Shan spruce forest is strongly dissected by avalanche and landslide chutes in many places, while in others it is fairly contiguous. Avalanches, landslides, and glacial outwash floods are relatively frequent disturbance agents. Landslides and flash floods occur with spring snow melt, but are generally largest when spring-summer precipitation coincides with glacial ice melt [30] and with glacial lake outbursts [29,31,32]. Avalanches occur during winters with heavy snowfall, especially when followed by rapid warming. Earthquakes occur with some frequency [27,33]. Fires are infrequent in the subalpine forest, occur only during extraordinarily severe weather conditions, affect small portions of the forest, and are usually human caused [28,34]. Annual area burned is associated with drought conditions and has an increasing trend for the period of record 1954-1999. The frequency and extent that avalanches, earthquakes, and fire affect spruce forest ecology in Kazakhstan are unknown. The severity of the 2011 wind events is unprecedented in living memory.

Field and Laboratory Methods
We collected dendrochronological material and tree inventory data in the summer of 2012 from 7 storm-damaged and 7 undamaged sites in valleys that experienced severe wind storms in 2011 (Butakovka Valley (BU), Chimbulak (CH), Medeu Ski Valley (ME)), and 3 undamaged sites in an area that did not incur storm damage (Bolshaya Almatinka (BA)) in southeast Kazakhstan (Figure 1).
We inventoried live and dead trees of 1-cm or greater DBH within a 0.04-ha circular plot for each site. Elevation, steepness (percent slope), and aspect were recorded at each plot. Stand density (trees per ha), percent mortality, average tree diameter (including live and dead trees), and quadratic mean The Tien Shan spruce forest is strongly dissected by avalanche and landslide chutes in many places, while in others it is fairly contiguous. Avalanches, landslides, and glacial outwash floods are relatively frequent disturbance agents. Landslides and flash floods occur with spring snow melt, but are generally largest when spring-summer precipitation coincides with glacial ice melt [30] and with glacial lake outbursts [29,31,32]. Avalanches occur during winters with heavy snowfall, especially when followed by rapid warming. Earthquakes occur with some frequency [27,33]. Fires are infrequent in the subalpine forest, occur only during extraordinarily severe weather conditions, affect small portions of the forest, and are usually human caused [28,34]. Annual area burned is associated with drought conditions and has an increasing trend for the period of record 1954-1999. The frequency and extent that avalanches, earthquakes, and fire affect spruce forest ecology in Kazakhstan are unknown. The severity of the 2011 wind events is unprecedented in living memory.

Field and Laboratory Methods
We collected dendrochronological material and tree inventory data in the summer of 2012 from 7 storm-damaged and 7 undamaged sites in valleys that experienced severe wind storms in 2011 (Butakovka Valley (BU), Chimbulak (CH), Medeu Ski Valley (ME)), and 3 undamaged sites in an area that did not incur storm damage (Bolshaya Almatinka (BA)) in southeast Kazakhstan ( Figure 1).
We inventoried live and dead trees of 1-cm or greater DBH within a 0.04-ha circular plot for each site. Elevation, steepness (percent slope), and aspect were recorded at each plot. Stand density (trees per ha), percent mortality, average tree diameter (including live and dead trees), and quadratic mean diameter (QMD) were computed for each plot. Variability in tree size was assessed using the coefficient of variation of the mean (CVM) of tree diameter. Increment cores were taken from the tallied trees, other than those that could not be sampled without killing them. Where possible, two increment cores were taken from opposite sides of each tree parallel to the slope at a height 20 cm or less above mineral soil. Trees in storm-damaged plots were usually uprooted, often in varied directions, and sometimes found removed from their original rooted location, so it was not always possible to ascertain growth position relative to the slope and sample from the desired bole aspect. Samples from trees with butt decay were taken above the decayed portion when possible. Increment cores were not taken from small trees (destructive sampling was prohibited), nor from shattered or twisted stems in the wind-damaged areas, but the trees are included in our summarized stand descriptive statistics.
Increment cores were mounted in wooden trays and sanded with progressively finer grits until individual cell structure was observable (usually to 15 µ). Samples were cross-dated using a combination of visual pattern matching and skeleton plots [35]. On samples that did not include pith but ring curvature indicated that pith was within 10 years of the innermost ring, concentric ring pith locators were used to estimate pith dates [36]. However, frequent decay in the lower bole made age determination of the majority of trees impossible. After crossdating, ring-width measurements were recorded with J2X software on a Velmex TA system ® with sensor accuracy of one µm [37]. Initial cross-dating of individual series was checked with COFECHA crossdating analysis software [38]. Samples that did not correlate with the plot-level chronology at a level of 0.3 or higher, or had ring anomalies such as branch traces, injury, or other deformation that precluded cross-dating or ring-width measurement were removed from the dendrochronology analyses [39]. When two satisfactory increment cores were obtained from the same tree, a mean ring-width was computed for each year.

Dendrochronological Detection of Bark Beetle Outbreaks
Bark beetle outbreaks are inferred and dated in tree-ring chronologies by growth releases in surviving trees and death dates of attacked trees [7,8,[40][41][42][43][44][45][46][47][48]. Release of trees subjected to inter-tree competition indicate within-stand mortality or removals. Methods vary depending on whether analyses are based on site or individual tree chronologies, based on raw ring width measurements or standardized series, standardization method, whether or not the individual-tree series are scaled against its potential for release, and rule sets used to classify a growth release as outside the normal variability. Veblen et al. (1991) [40] compared different standardization approaches for detecting Dendroctonus rufipennis Kirby (Coleoptera: Curculionidae) outbreaks in P. engelmannii Parry ex Engelm. series and was able to detect growth releases better with non-standardized series. Removing trends related to tree age and geometry is more critical when dissimilar series are compared on a yearly basis (e.g., climate, defoliator host and non-host series, etc.), and where events of interest occur during juvenile tree growth, but has less utility for detecting growth releases, which by definition are substantial growth increases relative to growth in the recent past. We analyzed raw ring width series without standardization.
Within a given tree, growth releases related to stand-level disturbance are usually detected using the relative increase in the running mean of growth increment, usually constrained by limiting the minimum time between releases [40,45,48]. The rule set used to determine that a period of increased growth is a release outside of normal variation affects both the date assigned to inferred release initiation and the level of disturbance that the analysis can detect [48]. A disturbance is inferred when the forward 5-or 10-year running mean of the series exceeds the backward 5-or 10-year running mean by a given percent; changes on the order of 50 to 200% are commonly used. More conservative rule sets (requiring greater growth increases) detect dramatic release events that result from overstory removal, and less conservative rule sets detect stand-thinning events in addition to severe events [48,49]. Both the lengths of the running means and the required proportional growth increase affect the dates that inferred disturbances are associated with, requiring passage of several years before surging growth exceeds previous trends. Less conservative rule sets result in earlier release-initiation dates than conservative rule sets and more closely align with aerial and ground pest detection surveys [48]. Requiring a running mean of a given length can preclude automated (software-driven) detection of events at the beginning and end of the chronology.
Alternatively, releases are detected with a boundary line technique, which scales the percent growth change relative to site-and species-specific maximal growth change potential, using pre-determined boundary lines for individual species [42,43]. There are similar matters related to the magnitude and duration of growth changes required for releases to be inferred as disturbance events, and establishing the boundary lines requires large data sets collected from multiple sites, tree ages, and tree sizes. This method is applicable in mixed-species and mixed-age forests where different trees have drastically different release potential, i.e., canopy and sub-canopy trees may respond differently to stand-opening events, or some species are not responsive to such events. Our study design is not suited to establishing the boundary lines, our sample size is marginal for that purpose, and we were precluded from sampling small trees.
Site-level disturbances are usually inferred using thresholds such as at least 2 trees recording growth release over a five-year period. Widespread or regional outbreaks are usually inferred when 25% (or higher threshold) of sites record disturbances within a five-year period. These thresholds filter out release events at different levels.
Our criteria for growth releases was applied to individual spruce trees, and computed with Microsoft Excel 2013, as follows:

1.
A growth release was determined to have occurred when the 10-year running mean of ring widths increased by 50% or more and growth was sustained above the pre-release level for 10 or more years, with single-year deviations permitted.

2.
The initial year of the growth release was dated as the first year in which the growth increment exceeded the previous 10-year mean by 50%. Thus, each event was detected using the running mean, but dated with respect to the first year of the sustained growth increase.

3.
The final year of the growth release was defined as the year prior to the year in which growth returned to the pre-release rate (i.e., the last year of elevated growth). Two types of releases were included as dated events but excluded from release-duration computations: (a) Growth of released trees sometimes plateaued, without declining, at a rate higher than the pre-release average, which occurs when released suppressed trees reach better canopy positions, or when the disturbance occurs near the end of the chronology; (b) Individual trees occasionally incurred multiple releases, with later releases occurring before growth stabilized from the initial release.

4.
A minimum time of 15 years between initial release years was used to distinguish discrete events.
Tree-level growth releases were composited by site, and site-level events were identified as those in which 2 or more trees per site initiated a growth release within a 5-yr period. Landscape-scale events were identified as those in which 25% or more of sites recorded releases over a 5-yr period. We also analyzed all detected releases without filtering the analyses by the events detected on a specified number or proportion of trees or sites affected.

Statistical Comparisons
Temporal patterns of growth releases were compared within and between plots for all sampled sites as well as between wind-and non-wind-damaged sites. Release frequency was computed for each site for 1930-1965 and 1966-2012 (an a posteriori classification that we discuss later), covering the period common to all sites. Mean site, stand, and growth release frequency of wind-damaged and non-damaged groups were compared using two-tailed t-tests for independent samples, and associations between growth release frequency and site and plot characteristics were measured using Pearson's correlation coefficient. Growth release frequencies for different time periods were compared with paired t-tests. Comparisons were completed using SPSS Release 18.0.0 [50].
Debris-flow, avalanches, and earthquake disturbances manifest in tree-ring series of surviving trees as abrupt, usually short, declines in radial growth [29,33] and are unlikely to be confused with bark beetle-related disturbance. We did not find evidence of such events in our samples, presumably because we minimized sampling near forest edges.

Results
Usable dendrochronology samples were collected from 219 P. schrenkiana. Samples from Crataegus, Sorbus, and Salix were not used in the analysis because they were either very young or did not correlate well with the spruce series. Average chronology length was 151 years, varying from 96 to 262 years ( Table 1). The chronology lengths underestimate stand age (or the age of the oldest tree in mixed-age stands), often substantially, because of the prevalence of stem decay. Absent rings were observed on 8.45% of cores, or 0.153% of measured rings. Site mean inter-series correlations were 0.537 ± 0.079. Trees greater than 120 years old were present at all sites but one ( Table 1). The age at which susceptibility to I. hauseri increases is unknown, but 120 is the age at which susceptibility to I. typographus increases considerably in Norway spruce, P. abies (L.) Karst [2].

Disturbance Events
Growth releases were detected on all sites, but incidence was low and varied considerably (Table 1). We detected 112 growth releases, averaging 0.49 per tree, though chronology length varied considerably by tree. Overall, 0.44% of measured rings were determined to be the initial year of a growth release (1 in 227 rings).
Events for which 2 or more trees on the same plot incurred a growth release within a 5-yr period were asynchronous and exceptionally infrequent (Figure 2). Only 17 events were detected, and the 25% filter indicating a large-scale event was never met, excluding an early event when sample size was low. Almost half of the events (47%) occurred in the last 60 years.

Disturbance Events
Growth releases were detected on all sites, but incidence was low and varied considerably ( Table  1). We detected 112 growth releases, averaging 0.49 per tree, though chronology length varied considerably by tree. Overall, 0.44% of measured rings were determined to be the initial year of a growth release (1 in 227 rings).
Events for which 2 or more trees on the same plot incurred a growth release within a 5-yr period were asynchronous and exceptionally infrequent (Figure 2). Only 17 events were detected, and the 25% filter indicating a large-scale event was never met, excluding an early event when sample size was low. Almost half of the events (47%) occurred in the last 60 years. Unfiltered growth releases (single trees per plot) were also infrequent ( Figure 3, Table 1). Tree events were also asynchronous, seldom affecting 25% or more of sites. Growth releases occurred on most sites in the late 1920s, and in the 1930s if not in the 1920s, and again in most sites in the 1980s, especially the early 1980s (Figures 2 and 3). Though releases occurred more frequently during these periods, they were not synchronous. Unfiltered growth releases (single trees per plot) were also infrequent ( Figure 3, Table 1). Tree events were also asynchronous, seldom affecting 25% or more of sites. Growth releases occurred on most sites in the late 1920s, and in the 1930s if not in the 1920s, and again in most sites in the 1980s, especially the early 1980s (Figures 2 and 3). Though releases occurred more frequently during these periods, they were not synchronous.
There was a trend of increasing single-tree growth release frequency with time ( Figure 3, Table 1). This trend was consistent throughout the reconstruction, though there were insufficient numbers of trees in the chronologies to draw conclusions prior to 1910. Release frequency increased beginning approximately the mid-1960s, almost doubling in the 35 years after 1965 compared to the 35-year period ending in 1965, from one release in 19.2 years to one in 9.9 years (Table 1). Please note that these are tree-and plot-level releases, not extensive disturbance events.
Within sites, individual-tree growth releases occurred in single years (occasionally 2 adjacent years) rather than over several contiguous years, with little or no temporal clustering at the site level. Occasionally, different trees on the same site incurred releases over several years (e.g., CH4 and BA2 in Figure 3). This may result from emerging bark beetles attacking nearby trees in consecutive years such that mortality at a given location occurred over several years, typical of spot expansion observed with other bark beetle species [51].  There was a trend of increasing single-tree growth release frequency with time ( Figure 3, Table  1). This trend was consistent throughout the reconstruction, though there were insufficient numbers of trees in the chronologies to draw conclusions prior to 1910. Release frequency increased beginning approximately the mid-1960s, almost doubling in the 35 years after 1965 compared to the 35-year period ending in 1965, from one release in 19.2 years to one in 9.9 years (Table 1). Please note that these are tree-and plot-level releases, not extensive disturbance events.
Within sites, individual-tree growth releases occurred in single years (occasionally 2 adjacent years) rather than over several contiguous years, with little or no temporal clustering at the site level. Occasionally, different trees on the same site incurred releases over several years (e.g., CH4 and BA2 in Figure 3). This may result from emerging bark beetles attacking nearby trees in consecutive years such that mortality at a given location occurred over several years, typical of spot expansion observed with other bark beetle species [51].
Computed as an average of all detected releases, regardless of tree or site, releases lasted 27.4 ± 15.1 years, based on 60 tree-level events for which duration could be estimated. The magnitude of change in growth rate for individual releases was usually between 200 and 400%, rarely below 100%, and sometimes as high as 1200%.
Single-tree growth release frequency was associated with greater stand density and less steepness both before and after 1965 ( Table 2) (stand density and steepness were not significantly correlated with each other (r = −0.32, sig. = 0.25)). Comparisons of release frequency and temporal patterns by stand age are precluded by the incidence stem decay, which produces short series from older trees, and other sampling challenges. Computed as an average of all detected releases, regardless of tree or site, releases lasted 27.4 ± 15.1 years, based on 60 tree-level events for which duration could be estimated. The magnitude of change in growth rate for individual releases was usually between 200 and 400%, rarely below 100%, and sometimes as high as 1200%.
Single-tree growth release frequency was associated with greater stand density and less steepness both before and after 1965 ( Table 2) (stand density and steepness were not significantly correlated with each other (r = −0.32, sig. = 0.25)). Comparisons of release frequency and temporal patterns by stand age are precluded by the incidence stem decay, which produces short series from older trees, and other sampling challenges.  Though not a study objective, we found no evidence of defoliator outbreaks in the spruce series. Defoliation leaves a signature of suppressed radial growth in tree-ring chronologies that can be used to dendrochronologically reconstruct defoliator outbreak regimes [52], but no such evidence was found in our series at any scale.

Storm Damage
Wind-damaged and non-damaged sites are similar with respect to average tree size, elevation, steepness, and aspect (Table 1). On average, stand density was lower and tree diameter more uniform on wind-damaged sites compared with trees on non-damaged sites. Though wind-damaged and nondamaged sites had, on average, similar numbers of disturbances since 1965, two damaged sites in Medeu Valley had incurred unusually low numbers of disturbances prior to 2011 (Figures 2 and 3). Releases were not detected in the years just prior to the storms, indicating that storm damage was not influenced by prior mortality events. Mortality in 2012 was 65% on wind-damaged sites, but this underestimates the complete impact of the storms, as damaged trees were still dying from physical storm damage (partially pushed over, broken, or missing significant portions of the crown) at the Though not a study objective, we found no evidence of defoliator outbreaks in the spruce series. Defoliation leaves a signature of suppressed radial growth in tree-ring chronologies that can be used to dendrochronologically reconstruct defoliator outbreak regimes [52], but no such evidence was found in our series at any scale.

Storm Damage
Wind-damaged and non-damaged sites are similar with respect to average tree size, elevation, steepness, and aspect (Table 1). On average, stand density was lower and tree diameter more uniform on wind-damaged sites compared with trees on non-damaged sites. Though wind-damaged and non-damaged sites had, on average, similar numbers of disturbances since 1965, two damaged sites in Medeu Valley had incurred unusually low numbers of disturbances prior to 2011 (Figures 2 and 3). Releases were not detected in the years just prior to the storms, indicating that storm damage was not influenced by prior mortality events. Mortality in 2012 was 65% on wind-damaged sites, but this underestimates the complete impact of the storms, as damaged trees were still dying from physical storm damage (partially pushed over, broken, or missing significant portions of the crown) at the time of sampling.

Discussion
We reconstructed a nearly 200-year history of disturbances in the Tien Shan spruce forest and found no evidence of either extensive high-severity bark beetle outbreaks or synchronous stand thinning events within that time frame. Growth releases were local in scale, and only weakly temporally coincidental on several plots during two periods.
Collectively, the evidence indicates that the primary biological disturbance agent in the Tien Shan spruce forest has historically functioned at a local scale. There is no evidence of extensive bark beetle outbreaks in our 200-year reconstruction. Furthermore, the evidence indicates that extensive high-severity bark beetle outbreaks such as occur in spruce forests in western Europe, Russia, and  [47] detected three extensive high-severity I. typographus and/or wind events plus an ongoing event in the last 300 years in P. abies forests in the Bohemian Forest. D. rufipennis incurs extensive outbreaks on average 1-4 times per century across much of its range [45,46,48,53,54]. None of the 17 plot-level events in our data occurred concurrently. Extensive, severe outbreaks by native bark beetle species are also not reported from spruce forests in central and western Asia [55][56][57].
Even plot-level events ( Figure 3) occur considerably less frequently than was observed with I. typographus and D. rufipennis [7,40,41,[44][45][46]48]. Čada et al. (2013) [7] reported extensive but less severe events every 10-50 years for I. typographus. Counting prolonged periods as single events, in the most recently reconstructed 100-yr period, 60 D. rufipennis events occurred on 23 sites on the Kenai Peninsula, Alaska [45], and almost 600 on 34 sites in Arizona [48]. Our data show only 17 events on 17 sites, and never came close to the 25% threshold needed to detect an extensive outbreak. If individual years were counted, rather than counting prolonged period as single events, the difference would increase by at least one order of magnitude.
In our study, disturbance frequency is associated with greater stand density and less steepness both before and after 1965, and weakly after 1965 with greater uniformity in tree size. High stand density is a common risk factor for coniferous bark beetles, and uniform tree size may reflect even-aged stands reaching a susceptible stocking or tree size threshold, though our data are not suited to teasing out interacting effects of tree size, stocking, age structure, and topographic factors as they relate to bark beetle susceptibility.

Doubling of Disturbance Frequency ca. 1965
Though disturbance frequency is low throughout the record and is related to local (stand-level) rather than landscape-scale outbreaks, it doubled after about the mid-1960s (Table 1, Figures 2  and 3). Factors that may have contributed to increased bark beetle activity after about 1965 include establishment of non-native bark beetle species and warmer temperatures. Tree species composition is about 95% spruce, so variability in host dominance and species composition are not relevant factors in the study area.

Non-Native Bark Beetles
Exotic bark beetles, including I. duplicatus and I. sexdentatus, established locally in the 1960s when raw logs were imported from Siberia and northern Kazakhstan during construction of railroads and the Medeu Mudflow Control Dam, and/or when European-grown spruce and pine were planted in the wildland-urban interface after completion of the dam [10,58]. The control dam protects the city of Almaty and the Medeu winter sports complex from avalanches and mud and debris flows. At 1750 m a.s.l., the control dam is located just within the lower elevation zone for P. schrenkiana, and close to P. sylvestris plantations. Areas below the dam became attractive for urban development, and European-grown trees of a variety of species are commonly planted in the new wildland-urban interface with little regard to the importation of exotic pests [10,58]. However, importation of raw bark-on logs used in construction is the more likely source of exotic bark beetle species because urban planting stock tends to be smaller than sizes typically attacked by these bark beetles.
Pheromone-trap monitoring shows that I. duplicatus and I. sexdentatus are established in the study area [9,10], but the overwhelming majority of Ips are hauseri [9,10]. Sagitov et al. (2016) [10] caught significant numbers of I. hauseri and small numbers of I. duplicatus in monitoring traps baited for I. typographus (synthetic lures for I. typographus are known to be attractive to I. duplicatus [59,60] and vice-versa [61,62]). The hauseri beetles were originally reported as typographus, but were later identified by M.Yu. Mandelshtam as hauseri. Later sampling by this group also caught I. sexdentatus. Kazenas et al. (2016) [58] caught I. hauseri and I. sexdentatus (pheromone lures were not specified). Increased disturbance frequency, even at a small local scale, might be related to establishment of exotic species. However, because few duplicatus or sexdentatus have been observed in the blowdown area we do not think that exotics are responsible for the post-1960 increase in disturbance frequency.

Warming Climate
There has been a generally increasing trend in temperature and a decreasing trend in precipitation across Kazakhstan since 1895 [28,29,63]. The annual frequency of frost days has decreased about 20 days since 1946 in Almaty Oblast, and heat waves have become more frequent, longer, and more intense [64]. Changes in tree growth response to climate occurred in the next lower vegetation zone ca. 1970, with increasing sensitivity to spring precipitation [65]. I. hauseri outbreaks are known to be associated with drought [11,12]. Outbreaks of other spruce-attacking bark beetles, especially I. typographus and D. rufipennis, are associated primarily with warmer-than-usual periods, with drought being a secondary factor [2,45,48,66,67]. Warmer temperatures can change beetle development rates, voltinism, overwintering survivorship, and emergence synchrony, and enable emergence when tree defenses are low, with different factors coming into play for different insect-host systems, seasons, and thresholds. Any of these factors can alter bark beetle-related disturbance frequency.
Beetle trap catches after the windstorms peaked in 2015 and declined after 2015, possibly due to exceptionally wet conditions in 2016. Spruce mortality from I. typographus peaks in the 3rd year after storm disturbance [1,5,68]. Spring 2016 precipitation in Almaty Oblast was in the highest 10th percentile since 1940 [64]. Spring moisture shortens the montane growing season in the Ile Alatau due to snow insulation and/or cold rains cooling the soil [65,69]. Drought-responsive multi-voltine bark beetles, particularly Ips, often respond rapidly to drought, and terminate abruptly with a shift to wetter conditions [2].

Outbreak Potential
A relevant question is: why did the storms not trigger an I. hauseri outbreak? I. hauseri and I. typographus supposedly fill similar niches, and I. hauseri is thought to be as aggressive as I. typographus [12], weather conditions were suitable, and considerable host material (damaged and not damaged) was available, so an outbreak might be expected to occur. Managers undertook aggressive salvage operations, in part to remediate the risk of a bark beetle outbreak, but much of the storm-damaged forest was inaccessible. Bark beetles attacked trees within and adjacent to the storm-damaged area, with some spot growth, but an extensive outbreak did not develop. Weather conditions after the windstorm were conducive to increasing bark beetle populations, but suitable conditions did not persist. The 2012 spring was within 10% of the driest seasons on record, with precipitation at 40-60% of normal, and spring and summer were exceptionally warm [70]. However, spring precipitation in 2016 was in the highest 10th percentile since 1940 [64], and the bark beetle populations declined immediately. Results of our study indicate that extensive severe outbreaks have not occurred in the last 200 years. Collectively, our results and the failure of a post-storm outbreak to continue after moisture returned to normal suggest that persistent drought is necessary to sustain I. hauseri outbreaks. Additionally, either I. hauseri is not as aggressive as I. typographus, or year-to-year variation in the Tien Shan weather prevents outbreak development, as there is no evidence of extensive stand-thinning or high-severity outbreaks such as occur with I. typographus in its natural range.

Reconstruction Limitations
The growth releases detected in this study cannot be unequivocally attributed to bark beetle-related tree mortality. It is standard practice in dendroentomology to confirm that some inferred outbreaks reflect documented events [71]. That is not possible here, because the detected activity is so minor and there are few records. Our reconstruction does show a disturbance in Chimbulak (at CH4) corresponding to a drought-associated outbreak in 1945 and 1946 in the same general area reported by Pruntenskii and Romaneko (1954) [9] (citing Zverev (1947) [25]), though they did not indicate either extent or severity. The reverse does appear to be true-neither our reconstruction nor local knowledge indicate that outbreaks occurred in the recent past. There are few records documenting forest pest activity in the Kazakh Tien Shan Mountains, especially any that quantify extent and severity. Though quantitative records are lacking, it is logical to assume that Ismukhambetov (1976) [26] would have noted recent extensive severe outbreaks. Likewise, local experts, including co-authors and forest managers, recall no extensive or severe outbreaks. Therefore, our reconstruction is consistent with local knowledge of bark beetle activity. Furthermore, our data indicate low levels of disturbance, so any erroneous attribution of disturbance agent would indicate an even lower level of bark beetle disturbance than what we infer. The reconstruction reflects small spatially and temporally isolated disturbances, with only two periods of increased disturbance levels across multiple sites, and very few sustained periods of disturbance at the site level.
Additionally, during release, only 1 to a few trees show synchronous releases, but radial growth increased considerably (200-400%). This reflects small-scale disturbances that killed individual large overstory trees, creating canopy gaps that provide growing space and resources for newly released understory trees, rather than watershed-or landscape-scale mortality events. The most reasonable explanation is that of bark beetles occasionally killing large, declining, injured, or diseased trees, thereby releasing understory trees from suppression. It is likely that the majority of the growth releases observed over the 200-year period reflect endemic levels of bark beetle activity that occasionally reach incipient outbreak levels. Chronology lengths from our plots are not a good indicator of stand age because of the high incidence of stem decay-some of the shortest series were from the oldest-appearing, but hollow trees. Our oldest dated inner ring was 1750, but Panyushkina et al. (2018) [72] and Zhang et al. (2017) [73] sampled trees dating in the early 1700s near the upper tree-line and the lower spruce zone, respectively. During work to support Panyushkina et al. (2018) [72], we sampled several trees dating to between 1702 and the 1750s, and one to 1579. Both upper and lower forest zones were influenced by agriculture, grazing, and harvesting, but the majority of the spruce zone is too steep to have been utilized for those purposes in the past. More extensive sampling would provide information about forest age, demographics and successional patterns. Our study does not provide information regarding the role that stem and root pathogens may play in these patterns, but we think that role is significant.

Conclusions
Ips hauseri is the most important bark beetle on P. schrenkiana in southeast Kazakhstan, but its biology, ecology, and outbreak dynamics are poorly known. Seasonal life history depends on temperature and is univoltine at high elevations but bivoltine in lower, warmer environments. The ability of I. hauseri to cause extensive tree mortality has been compared to I. typographus, which has similar niche requirements and phenology [11,12,22]. Outbreaks are associated with drought or triggered by blowdown events [11,26]. However, frequency, extent, and severity of outbreaks are unknown.
We dendrochronologically reconstructed a nearly 200-year history of forest disturbances in the Kazakh Tien Shan Mountains and found no evidence of extensive high-severity bark beetle outbreaks or even of synchronous stand-thinning events within that time frame. Growth releases were local in scale, consistent with small-scale disturbances that killed individual large overstory trees. These results, combined with failure of an outbreak to fully develop after blowdown events in 2011 indicate that prolonged drought may be necessary to sustain I. hauseri outbreaks, or that year-to-year variation in the Tien Shan weather prevents outbreak development. I. hauseri is probably less aggressive than I. typographus, at least on their natural hosts within their natural ranges. The evidence indicates that the primary biological disturbance agent in the Tien Shan spruce forest has historically functioned at a local scale and that extensive high-severity bark beetle outbreaks such as occur in spruce forests in western Europe, Russia, and North America, primarily by I. typographus and D. rufipennis, are not typical in the Kazakh portion of the Tien Shan. Though disturbance frequency was very low throughout the reconstructed period, it doubled after about 1964, probably due to warming climate.
Tree-ring reconstructions from spruce forests [7,40,[45][46][47][48]53,54] are typically longer than our Tien Shan reconstruction. Rather than reflecting a relatively young forest age, our reconstruction is short due to the prevalence of stem decay in older trees, which results in short chronologies from old trees.