Regional patterns of increasing Swiss needle cast impacts on Douglas‐fir growth with warming temperatures

Abstract The fungal pathogen, Phaeocryptopus gaeumannii, causing Swiss needle cast (SNC) occurs wherever Douglas‐fir is found but disease damage is believed to be limited in the U.S. Pacific Northwest (PNW) to the Coast Range of Oregon and Washington (Hansen et al., Plant Disease, 2000, 84, 773; Rosso & Hansen, Phytopathology, 2003, 93, 790; Shaw, et al., Journal of Forestry, 2011, 109, 109). However, knowledge remains limited on the history and spatial distribution of SNC impacts in the PNW. We reconstructed the history of SNC impacts on mature Douglas‐fir trees based on tree‐ring width chronologies from western Oregon. Our findings show that SNC impacts on growth occur wherever Douglas‐fir is found and is not limited to the coastal fog zone. The spatiotemporal patterns of growth impact from SNC disease were synchronous across the region, displayed periodicities of 12–40 years, and strongly correlated with winter and summer temperatures and summer precipitation. The primary climatic factor limiting pathogen dynamics varied spatially by location, topography, and elevation. SNC impacts were least severe in the first half of the 20th century when climatic conditions during the warm phase of the Pacific Decadal Oscillation (1924–1945) were less conducive to pathogen development. At low‐ to mid‐elevations, SNC impacts were most severe in 1984–1986 following several decades of warmer winters and cooler, wetter summers including a high summer precipitation anomaly in 1983. At high elevations on the west slope of the Cascade Range, SNC impacts peaked several years later and were the greatest in the 1990s, a period of warmer winter temperatures. Climate change is predicted to result in warmer winters and will likely continue to increase SNC severity at higher elevations, north along the coast from northern Oregon to British Columbia, and inland where low winter temperatures currently limit growth of the pathogen. Our findings indicate that SNC may become a significant forest health problem in areas of the PNW beyond the coastal fog zone.


glauca [Beissn.] Franco, and Pseudotsuga macrocarpa (Vasey) Mayr).
P. gaeumannii is indigenous in western North America and occurs wherever its host is found (Boyce, 1940). In its native distribution range, epidemic outbreaks of SNC are most severe within the coastal fog zone in Oregon, Washington, and British Columbia and have steadily increased in severity since 1980 (Black, Shaw, & Stone, 2010;Hansen et al., 2000;Omdal & Ramsey-Kroll, 2010;Shaw, Filip, Kanaskie, Maguire, & Littke, 2011). The affected area with visible SNC symptoms-chlorosis and premature needle loss-seen from annual aerial surveys of coastal Oregon has set new record highs each of the last 6 years (Ritóková et al., 2016). In the 20th century, P. gaeumannii has emerged in Douglas-fir plantations in many parts of the world including much of Europe, Chile, Australia, and New Zealand (Buchwald, 1940;Kimberley, Hood, & Knowles, 2011;Lanier, 1966;Marks & Pederick, 1976;Osorio, 2007;Peace, 1962). There is mounting concern that SNC is increasing in severity, frequency, and range in association with rising winter temperatures and spring/summer precipitation and will continue to intensify over the 21st century due to climate change (Watt, Stone, Hood, & Manning, 2011;Zhao, Mainwaring, Maguire, & Kanaskie, 2011).
While the epidemiology of SNC, genetic diversity and mechanisms of pathogenicity of P. gaeumannii on Douglas-fir have been well studied in young plantations, knowledge remains limited on the history and spatial distribution of SNC impacts on mature trees in the U.S. Pacific Northwest (PNW) and elsewhere. Phaeocryptopus gaeumannii has long believed to have been pervasive but innocuous in Douglas-fir forests prior to 1950 (Boyce, 1940;Hood, 1982;Peace, 1962). Increased severity since ~1950 is thought to be at least in part, climate-mediated because the causal fungus is sensitive to small differences in temperature and moisture (Black et al., 2010;Manter, Reeser, & Stone, 2005;Stone, Coop, & Manter, 2008). In coastal Oregon, a recent dendrochronological study indicates that SNC has affected periodically Douglas-fir growth at least back to 1592, which was the earliest of the available tree-ring records (Lee, Beedlow, Waschmann, Burdick, & Shaw, 2013). SNC impacts as measured by tree-ring width peaked in 1984-1986-thought to be a period when the fungal population reached epidemic levels following several decades of environmental conditions favorable to growth and reproduction of P. gaeumannii .
Growth reduction in Douglas-fir due to SNC in the PNW is symptomatic in the Coast Range of Oregon and Washington (Hansen et al., 2000;Rosso & Hansen, 2003). In the most heavily affected 10-to 30-year-old plantations in north coastal Oregon, height and basal area growth were reduced by ~25% and 35%, respectively, during a SNC outbreak in the mid-1990s (Maguire, Kanaskie, Voelker, Johnson, & Johnson, 2002). In a highly affected mature forest stand near Tillamook, Oregon, SNC reduced growth of 80-year-old Douglas-fir trees by 85% during severe SNC outbreaks with some trees showing 10 or more years having no observable growth (Black et al., 2010).
At less diseased sites in coastal Oregon, annual radial stem growth of mature Douglas-fir was reduced by 18%-28% on average during SNC outbreaks between 1590 and 2011 .

While SNC symptoms are often noted in plantations in Northern
Idaho and Western Montana (Hagle, Gibson, & Tunnock, 2003), there have been few broad-scale studies to quantify the impact of SNC on tree growth outside of the coastal fog zone in the PNW. A comprehensive study involving 59 young Douglas-fir stands (10-23 years) found no growth reductions in the Oregon Cascades during a SNC outbreak between 2001 and 2006 (Filip et al., 2007). This study assessed the SNC impact by observing natural outbreaks but this is difficult to do in mature forest stands where the history of SNC outbreaks is unknown and must be inferred. A previous dendrochronological study (Lee et al., 2016), using earlywood (EW) and latewood (LW) ring width chronologies, showed that the ubiquitous SNC affected Douglas-fir growth across a longitudinal transect from the west side of the Coast Range to the west slopes of the Cascade Range of Oregon. Air temperature and dewpoint deficit (DPD) were the most important climate factors affecting Douglas-fir growth, and soil moisture and SNC modified the growth response to these climate factors. Lee et al. (2016) examined the growth-climate relations for mature Douglas-fir by adjusting the tree-ring width series for SNC but did not reconstruct the local and regional history of SNC impacts on radial stem growth. We extend the findings of Lee et al. (2013Lee et al. ( , 2016 to determine the impact of SNC across a diversity of ecoregions in the PNW ranging from wet maritime in the Coast Range to dry Mediterranean in the Cascade Range. The key growth pattern in tree-ring records associated with SNC of coastal Douglas-fir is a sinusoidal cycle of anomalously low growth having a primary periodicity of ~20-30 years and a harmonic periodicity of ~4 years . The cyclical patterns of SNC impact on Douglas-fir growth occur throughout the life of the tree and because of the effects of synoptic seasonal weather patterns on fungal growth, are synchronous across coastal Oregon. Three major phases of the infection cycle of P. gaeumannii are relevant to the understanding of the climate-disease relation and history of SNC outbreak events (Manter et al., 2005): (1) P. gaeumannii reproduces by ascospores and pseudothecia (i.e., fruiting bodies) proliferate in winter from December to April; (2) sporulation occurs in synchrony with bud break and shoot elongation from May to July and only current-year needles are infected (Hood & Kershaw, 1975;Stone, Capitano, & Kerrigan, 2008); and (3) needle colonization by hyphal growth on the needle surface into the stomata occurs year round following initial infection. Because released spores can dessicate and lose viability within several days, leaf wetness at time of sporulation is important for ascospore germination (Manter et al., 2005;Stone, Coop, et al., 2008;Watt et al., 2011). Within the SNC impact zone along the coast of Oregon and Washington, disease severity is associated with a combination of mild winter temperatures and high leaf wetness in spring (Manter et al., 2005;Stone, Coop, et al., 2008) as well as mild summer temperatures ranging between the optimum temperatures for ascospore germination and germ tube growth at 18°C and 22°C, respectively (Capitano, 1999). These climatic conditions for disease severity have also been verified in New Zealand (Watt et al., 2011).
The maturation period for P. gaeumannii ranges from 1 to 2 years in some areas of the Coast Range to 4-7 years in the Cascade Range of Oregon and Washington as evidenced by pseudothecia on young and older needles, respectively (Stone, Capitano, et al., 2008). The low-to higher-frequency variations in the tree-ring width series of coastal Douglas-fir have been attributed to the slow buildup of fungal abundance over multiple generations of P. gaeumannii .
This study reconstructs the regional history of SNC impacts on Douglas-fir growth for nine sites in western Oregon differing in site conditions, elevation, topography, and proximity to the coast based on master chronologies of EW and LW ring widths of the host and nonhost tree species. Specific objectives of this research were to test the following hypotheses: (1) SNC is sensitive to winter and summer temperature, and summer precipitation, and so, spatial variability in SNC severity can be attributed to variations in site conditions, location, and elevation; (2) SNC impacts display primary periodicities of 20-30 years and secondary periodicities of 4-6 years as seen for the coast sites ; and (3) winter temperature is a more limiting factor of P. gaeumannii than summer conditions in the high Cascades and vice versa for low-to midelevation sites. We hypothesized that the impacts of SNC should be synchronous at low-to mid-elevations where summer conditions are more limiting to fungal dynamics than winter conditions but delayed at higher elevations in the Cascade Range where winter conditions are less favorable to P. gaeumannii.

| Research sites
Nine mature, closed-canopy forest stands dominated by Douglasfir were located on state and federal lands in western Oregon, USA ( Figure 1). The nine stands represent a range of climatic and edaphic conditions at varying elevations and proximal distances to the coast (Table 1) Basal area density is based on a complete plot survey of individual trees at these permanent field sites.
c Basal area density is based on multiple stand measurements using a wedge prism that has a basal area factor of either 20 or 30.

| Dendrochronological and climate data
Tree core samples from 17 to 29 dominant and codominant Douglasfir trees and 4 to 16 western hemlock trees were taken at approximately breast height (1.4 m) using 5-mm diameter increment borers, collecting one or two cores per tree at each site (Lee et al., 2016) ( Table 2). The cores were air-dried, mounted, sanded, and digitized using a color flatbed scanner. EW and LW tree-ring widths were measured to the nearest 0.01 mm using WinDENDRO 2008g software (Regent Instruments Inc., Quebec, Canada). The individual ring width time series were visually cross-dated and verified using the program COFECHA (Holmes, 1983) to ensure the correct calendar year was assigned to each ring. The interseries correlation ranged from 0.46 to 0.55 for Douglas-fir and 0.12 to 0.39 for western hemlock ( Table 2).
The mean sensitivity ranged from 0.16 to 0.26 for Douglas-fir and 0.18 to 0.32 for western hemlock. The EW and LW time series were log transformed and detrended using either a cubic spline smoother with a 50% frequency response of 32 years or a horizontal line to remove the age-related trend in the initial ~70 years. Master chronologies of EW and LW ring widths for the nine sites were calculated as the median of the standardized time series for individual trees.
The Douglas-fir master chronologies by Lee et al. (2016) were used to infer the seasonal effects of temperature, water, and disturbances on intra-annual tree growth using climate data from several sources.
Divergences between the master chronologies of Douglas-fir and western hemlock were used as auxiliary information to detect outliers in the Douglas-fir chronologies associated with forest disturbance agents. The divergence between the observed and climate-based model prediction of Douglas-fir growth is the primary data source to infer the history of forest disturbances because the western hemlock chronologies are not pure climate proxies free of disturbance. In this study, we reconstruct the history of forest disturbances in Douglas-fir using growth-climate relations similar but more parsimonious to those reported by Lee et al. (2016) and the western hemlock chronologies.
T A B L E 2 Site and tree core sampling information and dendrochronology summary statistics for Douglas-fir (Lee et al., 2016) and western hemlock The average correlation between each detrended time series and the mean of all other detrended time series. c The mean absolute first difference of each detrended time series relative to its running mean value. The index ranges from 0 (i.e., no variability) to 2 (i.e., high variability with periodicity of 2 years). d The forest stand contained two different-aged cohorts of Douglas-fir growing in adjacent sections.
Monthly mean maximum, minimum, and dewpoint temperatures and total precipitation data for the period 1895-2012 were obtained from the PRISM Climate Group at Oregon State University at http://prism.oregonstate.edu. The divisional monthly Palmer Drought Severity Index (PDSI) for the period 1895-2012 was obtained from the National Oceanic and Atmospheric Administration's National Climate Data Center at http://www.ncdc.noaa.gov. Variables were obtained for the specific study site locations when spatially interpolated climate data were available and for the nearest location or division otherwise (Lee et al., 2016). DPD was calculated from monthly PRISM data as the difference between mean air temperature and the dewpoint temperature and is a measure of evaporative demand. The climate variables used in the time series model were daily maximum air temperature, total precipitation, DPD, and PDSI which were summarized as seasonal averages that corresponded best with intra-annual growth. Following Cook (1985, 1987, the master chronology (Y t ) has been described by Lee et al. (2016) as having a mean response function with components for climate (C t ) and disturbance (D t

| Time-series intervention analysis
The climate component, C t , represents the interactions of temperature, precipitation, soil moisture, and evapotranspiration demand on tree growth and is assumed to be the same for all trees within a stand (Fritts, 1976). The disturbance component, D t , has been described by Lee et al. (2016) as a set of pulse functions to detect outliers representing the outbreak events of one or more forest disturbance agents.
Maximum likelihood for a Gaussian time series was used for parameter estimation and hypothesis testing to reconstruct the forest disturbance history of SNC outbreaks by site for the years 1895-2011.
Given the explicit characterization of the growth-climate relation (C t ) of the host species (Lee et al., 2016), the disturbance events are unknown and must be inferred by comparing Y t and C t to detect growth divergences which cannot be attributed to the measured climate variables. Outlier detection tests are performed using the Wald test statis- where β i is the model parameter for the pulse disturbance event at time t = t i for i = 1, 2, … r. In general, the disturbance events t* = (t 1 , t 2 , …, t r ) are unknown and must be inferred.
Some multiple testing adjustments such as Bonferroni or Tukey and/or specification of a critical value A are appropriate for detection of multiple outliers because every point is a potential outlier. Chang and Tiao (1983) suggest values of A between 3 and 4. The forest disturbance events identified by Lee et al. (2016) were improved upon through iteration and perturbation based on the Wald test statistic using a greater critical value A of 4 for the reconstruction of SNC impacts by site. See Lee, Wickham, Beedlow, Waschmann, and Tingey (2017) for a more complete description of the time-series intervention methodology.

| Swiss needle cast index of impact on stem growth
To reconstruct the site-specific history of SNC impact, divergences between the observed and climate-based model predictions were detected as outliers based on statistical fit of the TSIA model with pulse interventions. Growth anomalies that could not be attributed to climate were modeled by a pulse intervention event. For the diseased years specific to Douglas-fir, the SNC index of impact was equal to one minus the negative pulse intervention parameter backtransformed to the original scale (Lee et al., 2016). For example, a parameter value of −1 corresponds to a SNC index value of 1-10 −1 = 0.9 or a 90% reduction in growth. The SNC index of impact was defined to be 0 for the other years.

| Canonical correlation analysis
Once the history of forest disturbance was reconstructed for each site, we cross-correlated the SNC index with seasonal averages of temperature and precipitation . The PRISM monthly climate data were summarized as seasonal averages of temperature and precipitation that corresponded to the three major phases of the infection cycle (Manter et al., 2005). Winter temperature was calculated as average daily maximum temperature for one or more months between December and March. Spring/summer precipitation and temperature were calculated as total precipitation and mean daily maximum temperature for one or more months between May and July. Cross-correlation and canonical correlation analyses were used to identify the weighted moving averages of current and past seasonal temperature and precipitation values that optimize the correlation with SNC impact.

| Spectral analysis
For sites in coastal Oregon, the tree-ring width chronologies for mature Douglas-fir were synchronous and displayed cyclical patterns having primary periodicities of 20-30 years and secondary periodicities of 4-6 years that could not be attributed to climate . These concurrent sinusoidal patterns in Douglas-fir stem growth adjusted for climate were attributed to SNC within the coastal fog zone . This recent finding has led us to expand our scope to detect SNC impacts on Douglas-fir growth outside of the coastal fog zone. The episodic patterns of SNC impact are best examined using spectral or frequency domain analysis. Spectral analysis partitions the variance of a time series into a discrete and finite set of components, each of which is associated with a particular frequency band. The representation of a time series in the frequency domain is called the spectrum. The spectrum measures the strength of different frequency components, that is, peaks in the spectrum occur at the key frequencies. The primary frequency of a time series corresponds to the sinusoidal term that is most strongly associated with the time series, the secondary frequency with the next most strongly associated.
Cospectral analysis examines the synchronicity of SNC chronologies based on whether the spectrum at multiple sites has the same key frequencies (i.e., variations in each time series are associated with sinusoids having the same frequencies). Cospectral analysis was used to determine whether the aclimatic growth anomalies of inland Douglas-fir displayed a cyclical pattern having a primary periodicity of 20-30 years and a secondary periodicity of 4-6 years in synchrony with the SNC index for coastal Douglas-fir.

| RESULTS
All sampled stands experienced anomalously low radial stem growth that could not be accounted for by current and previous year seasonal climatic metrics used in this study. Regional synchronous patterns of growth suppression across the sites, attributable to forest disturbances, appeared in the master chronologies in 1918, 1951, 1959, and, most notably in 1984-1986

| Pseudoperiodicities in tree growth
The time series chronologies for EW and LW ring widths displayed a mixture of low, medium, and high frequencies indicative of variability in growth response to both climate and SNC ( Figure 3). One coast site, HCTL, and three Cascade sites, FC, SG, and TC, displayed strong sinusoidal patterns having a primary periodicity of 14-40 years and a secondary periodicity of 6-8 years as indicated in the frequency domain by peaks in its spectrum at a frequency of 0.25-0.07 cycles/year and 0.12-0.17 cycles/year, respectively. One low elevation coast site, CH, within the SNC impact zone displayed sinusoidal impacts on growth every 5-7 years. The pseudoperiodic patterns of growth were evident in both EW and LW width. For two inland sites, JP and MM, where evaporative demand was high during the annual summer drought, the spectrum had a primary peak at 0 frequency which was seen in the F I G U R E 2 Chronologies of earlywood and latewood widths of mature Douglas-fir across a network of sites from the Coast Range to the west side of the Cascade Range of Oregon (Lee et al., 2016). The vertical gray lines denote climatic pointer years of 1918, 1951, 1959, 1984, 1985, and 1986 time domain as a decreasing trend in growth rates in the most recent decades when temperatures were increasing. The 470-year-old Douglas-fir trees at SG displayed a slower declining growth trend since ~1905. The broad peaks in the spectrum indicated a mixture of climate and SNC effects that interacted and were confounded.

| Outlier detection using time-series intervention analysis
To reconstruct the site-specific history of forest disturbance, the growth anomalies that could not be attributed to climate were detected as outliers based on time series analysis with pulse interventions for the disturbance years (Lee et al., , 2016). An outlier is detected when the pulse intervention parameter, which represents the divergence between observed growth and the climate-based model prediction, is statistically less than zero based on a one-sided Wald test at an overall 0.05 level of significance with a Bonferroni correction. In the period 1895-2011, the number of disturbance years ranged from 4 (3%) to 20 (17%) for EW and 3 (3%) to 12 (10%) for LW (Table A1).
Many of these growth anomalies represented a divergence between the Douglas-fir and western hemlock chronologies, most notably in 1984-1986 ( Figure 4). Several low growth anomalies were detected as outliers in both the Douglas-fir and the western hemlock chronologies, most notably in 1917, 1918, 1937, and 1991 and consequently were not attributed to a species-specific forest disturbance such as SNC (Table A1). Two low elevation coast sites, CH and HCTL, within the SNC impact zone experienced the most disturbance outbreak events, while a hot, dry valley site, JP, and a high Cascade site, SG, outside of the coastal fog zone experienced the fewest number of forest disturbances.
Examination of the negative outlier years indicated that high DPD in the previous-year (pDPD) reduced growth more in a disturbance year than in years when disturbance was absent ( Figure 5). This interaction of pDPD and disturbance was alternatively modeled as a change in slope of pDPD and resulted in a more parsimonious model having fewer pulse intervention terms and a better statistical fit based on a lower Akaike information criterion. The slope of pDPD for the disturbance outbreak years was most negative for the coastal sites where Douglas-fir forests were more heavily infected by SNC and least negative for montane sites on the west slope of the Cascade Range of Oregon where SNC was less severe ( Figure 6).

| Reconstruction of SNC outbreaks for the period 1895-2011
The outbreak reconstructions based on the SNC index indicated that reductions in Douglas-fir stem growth attributable to SNC have occurred periodically in western Oregon for the period of study from 1895 to 2011 (Figure 7). Synchronous outbreaks in 1959Synchronous outbreaks in and 1984Synchronous outbreaks in -1986 were inferred for four sites. Two coast sites, HCTL and HCTU, recorded synchronous outbreaks since 1895, whereas CH outbreaks were coherent after 1950. The 1984-1986 outbreak was exceptionally intense resulting in estimated reductions in EW growth of 34%-61% for CH and 31%-44% for FC. These episodic growth anomalies that could not be attributable to climate have varied in frequency and intensity but were highly synchronous and averaged 1-2 years in duration across the region from the coast to the mid-Cascades (Table 3).
The average return interval-which is the number of years between the start of two consecutive outbreaks-ranged from 7.9 to 15.0 years F I G U R E 3 A comparison of the spectrum of (a) earlywood and (b) latewood ring width chronologies for the years 1895-2011 indicates a wide range of tree variability between growth periods within a site and between sites. The master chronologies for two coast sites, Cascade Head and Horse Creek Trail Lower, have a primary periodicity of ~40 years and a secondary periodicity of 6.2-8 years. The master chronologies for two high elevation sites, Soapgrass Mountain and Toad Creek, have a primary periodicity of ~20 years which also is indicative of a Swiss needle cast disease cycle within the SNC impact zone and 14.6 to 48 years outside the SNC impact zone based on EW growth series (Table 3) and FC, and a higher-elevation site, TC, were coherent at several key frequencies but were not in phase ( Figures A3 and A4). In general, F I G U R E 4 Observed and predicted earlywood width chronologies of Douglas-fir in comparison with observed earlywood chronology of Western hemlock for the nine study sites based on time-series intervention model. Vertical red bars represent anomalously low growth associated with Swiss needle cast (SNC) and not accounted for by the effects of climatic factors. These anomalous growth years are identified as statistically significant divergences between the host chronology (black line) and the predicted growth response to seasonal climatic factors (gray line) at the 0.05 overall level of significance. Based on the interaction of previous-year dewpoint deficit and SNC, each year was classified into one of two disease states: (1) growth suppression due to SNC (•) and (2) no growth suppression (▲). The master chronology for Cascade Head was plotted on a different scale because the interannual variability was greater at this coast site SNC outbreaks for SG and TC lagged the outbreaks for lower elevation sites by several years. The synchronization of SNC impact on Douglas-fir across the landscape indicated that there were climate factors, which favored disease conditions for low-to midelevation sites in western Oregon (Figure 8). Peak SNC outbreaks were sometimes delayed by several years in the high Cascades where freezing winter temperatures slowed the development of pathogen population from reaching epidemic levels.

| Seasonal climate factors
Cross-correlation and canonical correlation analyses were used to determine the climatic factors associated with the SNC index for each site, focusing on EW growth for two coast (CH and HCTL) and two Cascade sites (FC and SG) where the magnitude and frequency of disease impacts and the statistical power of the test for correlation were greater. Disease impact related to winter temperature, summer precipitation, and temperature consistently across all study sites. The  T A B L E 3 Growth anomalies associated with a forest disturbance specific to Douglas-fir were detected using time-series intervention analysis and represented divergences between the observed and climate-based model predictions that were statistically significant at the 0.05 level. The model parameters for the pulse interventions were used to infer the magnitude of the growth response to forest disturbance that could not be attributed to climate and diverged from the western hemlock chronologies. The frequency, duration, and magnitude of the tree-ring-based reconstructions of forest disturbance history are reported by growth period for each site   reduces the growth of Douglas-fir by restricting gas and water exchange via stomatal occlusion and early needle abscission (Manter, Kelsey, & Stone, 2001). Unlike most forest pests and diseases specific to Douglas-fir in the PNW, P. gaeumannii is found wherever its host is found (Boyce, 1940)  Both laminated root rot (Thies & Sturrock, 1995) and Armillaria spp. (Shaw & Kile, 1991) are mortality agents of Douglas-fir and reduce the growth of conifers. The mature Douglas-fir trees for SG showed a slow decreasing century-long growth trend similar to that of mountain pine (Pinus mugo Turra) trees in the Swiss National Park which were in response to annual changes in climate and the abundance of the causal fungus (Lee et al., , 2016 in 1984in , 1996in , and 2004in (Black et al., 2010. Within the SNC impact zone of Oregon, reductions in volume growth of young Douglas-fir by SNC were estimated to range between 23% and 50% Maguire et al., 2002).

| DISCUSSION
Defoliation due to SNC is the cause of the synchronous cyclical patterns of aclimatic growth anomalies within the coastal fog zone (Black et al., 2010;Lee et al., 2013) and is the most likely cause of the concurrent patterns of cyclical growth anomalies across western Oregon. While there are several similarities between SNC and forest insects, SNC can be differentiated from other forest disturbance agents based on differences in their spatial extent, frequency, duration, and magnitude of impact, and potential to cause tree mortality.  (Flower et al., 2014;Furniss, 2014;Shaw, Oester, & Filip, 2009;Swetnam & Lynch, 1989, 1993, while SNC severity generally increases after several years of warm winters followed by wet summers . Some forest insects of Douglas-fir (e.g., Douglas-fir beetle) are ubiquitous in the PNW, but as a disturbance agent, their effects are sporadic and are associated with endogenous (e.g., population dynamics, host tree vigor, and susceptibility) and exogenous factors (e.g., climate) (Bentz et al., 2010;Raffa et al., 2008;Swetnam & Betancourt, 1998). Forest insects, alone or in combination with root diseases and drought stress, are very important causes of tree mortality and, consequently, are unlikely to have caused the cyclical patterns of aclimatic growth anomalies with no concurrent increase in the mortality rate.
Prolonged periods of high VPD conditions during the summer drought reduce stomatal conductance and can lead to carbon starvation and ultimately reduced annual stem growth, more so during a period of increased SNC severity when carbon uptake is further reduced by stomatal occlusion and early needle abscission. For five study sites (CH, FC, MM, SG, and TC), isotopic measurements of δ 13 C and δ 18 O discrimination in tree rings indicated that the growth anomalies in disturbed years could not be attributed to a physiological response to temperature and water stress but were consistent with a reduction in photosynthetic capacity by a loss of functioning stomata (data will be presented in a future manuscript).
Our tree-ring chronologies of SNC outbreaks represent population cycles of the causal pathogen that are strongly associated with winter and summer temperatures and summer precipitation. Close correlations have been found between these seasonal climatic factors and epidemiological measurements of needle retention and P. gaeumannii abundance as well as growth impacts in the PNW (Hansen et al., 2000;Lee et al., 2013;Maguire et al., 2002;Manter et al., 2001;Winton, Manter, Stone, & Hansen, 2003;Winton, Stone, Watrud, & Hansen, 2002) and New Zealand (Watt, Stone, Hood, & Palmer, 2010). Lowfrequency variability in SNC impacts on Douglas-fir growth is caused by the intensification of P. gaeumannii abundance to epidemic levels over several decades . The onset of a slow-developing SNC outbreak occurs when the pathogen population reaches epidemic levels. High-frequency variability in SNC impacts is the result of a delayed growth response to the infection of newly emerged needles each year and ensuing pathogen colonization in each needle age class.

| Conceptual SNC disease cycle
We combined our dendroecological findings with the epidemiology of SNC to develop a conceptual model of the disease cycle driven by needle retention and fungal fruiting body abundance which have routinely been used as indices of disease severity (Hansen et al., 2000;Hood, 1982;Manter et al., 2005;Michaels & Chastagner, 1984). SNC reduces assimilation of carbon and tree diameter by stomatal occlusion and early needle abscission (Hansen et al., 2000;Manter, Bond, Kavanagh, Rosso, & Filip, 2000). Consequently, yearly changes in SNC impacts depend upon inoculum abundance, ascospore germination, and pathogen colonization in association with climatic conditions, which affect the proportion of stomata occluded and needle retention (Manter et al., 2005). Douglas-fir trees on the coast typically retain up to 4 years of needles but may only have current and 1-year-old foliage due to premature needle abscission in severely affected plantations (Hansen et al., 2000;Maguire et al., 2002;Zhao et al., 2011). In our conceptual model, the disease cycle begins when pathogen abundance is at epidemic levels, resulting in loss of 2-year-old and older needles and a significant reduction in stem growth ( Figure 12). The pathogen population will be reduced due to premature needle abscission resulting in fewer infected needles and a reduction in inoculum.
Peak SNC outbreaks reduce tree growth for several consecutive years because photosynthetic capacity is restored to normal only after all needle classes have formed (Saffell et al., 2014). A delay of several years between inoculation and growth of the fungus and tree growth reduction is expected because the pathogen infects only the newly emerged needles (Hood & Kershaw, 1975;Stone, Capitano, et al., 2008). This lagged growth response to SNC is represented by a 4-year periodicity in disease impacts ( Figure 12). The slow buildup of pathogen abundance from endemic to epidemic levels over several generations is represented by a 20-year periodicity. The conceptual disease model is illustrated as having a dominant periodicity of 20 years but, in reality, the primary periodicity varies by site and is as low as 6 years at Tillamook, Oregon, where more favorable climatic conditions allow the fungus to develop faster (Black et al., 2010;Lee et al., 2013;Stone, Coop, et al., 2008). Pseudothecia can be commonly found on 4-to 7-year-old needles in the Cascade Range of Oregon and Washington, and on 1-to 2-year-old needles in some areas of the Coast Range where pathogen dynamics are enhanced by more favorable climatic conditions (Stone, Coop, et al., 2008). Pathogen abundance is not reset to endemic levels by abscission of 2-year-old and older needles in areas where disease is constantly severe as indicated by a <10-year disease cycle and the presence of pseudothecia on 1-to 2-year-old needles.
We represent five replications of the conceptual disease cycle in the time and frequency domains ( Figure 13). The dominant pattern in the disease cycle is a peak impact occurring every 20 years ( Figure 13a) which is represented in the frequency domain by a peak in its spectrum at a frequency of 0.05 cycles/year (i.e., periodicity of 20 years) (Figure 13b). The spectrum also has a secondary periodicity of 4 years (frequency = 0.25 cycles/year) which is seen in the time domain as a periodic impact on growth every 4 years. The other local peaks in the spectrum occur at the harmonic periodicities, for example, twice and thrice the secondary periodicity. The spectrum in Figure 13b is the signature of SNC impact that is unique and can be used to identify a SNC disease cycle and separate the confounding effects of climate, SNC, and other forest disturbances at a site.

| Spatiotemporal variability in SNC impacts
Detectable SNC impacts occur wherever Douglas-fir is found in western Oregon and are greater and more frequent nearer the coast, at lower elevations on south-facing aspects in the coastal fog zone (Manter, Winton, Filip, & Stone, 2003;Manter et al., 2005;Rosso & Hansen, 2003). Areas where environmental conditions are less favorable to the development of P. gaeumannii experience F I G U R E 1 2 Conceptual model of Swiss needle cast impact on tree growth in association with the abundance of Phaeocryptopus gaeumannii and number of needle classes retained . The number of needle classes retained varies from one (when the tree is heavily infected) to four (least infected). Pathogen abundance increases from endemic (when 2-year-old and older needles are abscised) to epidemic levels (when tree is heavily infected) over several decades. The disease cycle begins anew with a peak reduction in growth when pathogen abundance reaches epidemic levels and is then reset to endemic levels following the early abscission of two-year-old and older needles. Growth reductions display 4-and ~20-year periodicities because P. gaeumannii infects only the newly emerged needles at time of sporulation and has a 4-year life cycle F I G U R E 1 3 The sinusoidal pattern of Swiss needle cast (SNC) impact on growth of Douglas-fir is represented by five repetitions of the 20-year disease cycle. The SNC index has a primary periodicity of ~20 years and secondary periodicity of ~4 years as seen in the (a) time and (b) frequency domain. The total area under the spectrum is equal to the variance of the time series less frequent and less intense SNC outbreaks. The least severely impacted sites are found in warm, dry environments at low-to midelevations in the Willamette Valley and Cascade Range and in cool environments in the high Cascades. Spatial variability in SNC severity depends upon site conditions, location, proximity to coast, and elevation (Rosso & Hansen, 2003;Stone, Coop, et al., 2008) and whether winter or summer conditions are more limiting (Zhao et al., 2011).
SNC reduces Douglas-fir growth relative to other tree species and therefore can affect competitive outcomes between tree species. This is very well illustrated by the Sitka spruce (Picea sitchensis) vegetation zone of the Coast Range of Oregon and Washington where P. gaeumannii is most abundant and Douglas-fir is a minor species compared to western hemlock, even though Douglas-fir grow much more rapidly than hemlock where environmental conditions for P. gaeumannii are less favorable. The composition of the Sitka spruce vegetation zone is distinct and shaped not only by climate favorable to growth of Sitka spruce and western hemlock but also to P. gaeumannii and its negative effect on the competitive ability of Douglas-fir.

| SNC outbreaks are influenced by broad-scale forcing factors
The intensity and duration of SNC outbreaks are influenced by multi- Long-term PNW temperature records show an accelerated warming trend since 1970, most notably in the winter months (December-February) resulting in a longer freeze-free season and warmer minimum temperatures in winter (Abatzoglou, Rupp, & Mote, 2014). PNW precipitation records show a downward trend during a strong warm phase of the PDO (1925PDO ( -1946 followed by a heterogeneous but positive trend in spring precipitation during and after a strong cool phase of the PDO (1947PDO ( -1978 (Abatzoglou & Redmond, 2007).
These multidecadal trends in seasonal temperature and precipitation are reflected in the lower intensity of SNC outbreaks prior to 1950 relative to outbreaks after 1950.
Swiss needle cast outbreaks tend to be more severe during periods of relatively warm winter and cool, wet summer conditions . Synchrony in SNC impacts is attributed to a single or several consecutive extreme climate events toward the end of a disease cycle when the amount of inoculum is high. The wettest 25-year period was 1960-1984 with most years above the historical water year average precipitation (Abatzoglou et al., 2014). The warm winters and cool, wet summers in this period allowed P. gaeumannii to develop rapidly, resulting in a widespread SNC outbreak in 1984SNC outbreak in -1986. Following the SNC outbreak initiated in 1983, there was a delay of 1 year before the reduction in photosynthate production due to stomatal occlusion and early needle abscission was fully expressed in stem growth. The anomalously cool, wet summer in 1983 occurred during an extra strong El Niño event. However, broad-scale forcing factors such as ENSO and PNA modulate seasonal temperature patterns but not precipitation patterns (Abatzoglou et al., 2014). Long-term climate records show an increase in spring precipitation and a decrease in summer and autumn precipitation, indicating that summer conditions may become more of a limiting factor for P. gaeumannii in warm, dry environments with a changing climate.
Rapid warming in the PNW during December-February and March-May since ~1950 has been attributed to ENSO and PNA (Abatzoglou & Redmond, 2007;Abatzoglou et al., 2014). The warmest 10-year period was 1998-2007 with most years above the historical annual mean temperature. With warmer winters, SNC impacts are increasing in mature closed-canopy Douglas-fir stands on the east slopes of the Coast Range and in the high Cascades as SNC impacts on Douglas-fir growth were greatest in the 1990s for these sites.
Warming winters are expected to intensify SNC outbreaks more at high elevations and higher latitudes because the proliferation of pseudothecia is highly sensitive to small increases in winter temperature above the temperature threshold of ~4°C for fungal growth (Stone, Coop, et al., 2008). A temperature threshold of 4°C is consistent with the spatial shift in the association between SNC impact and winter temperature in December-February for the Coast Range to February-March in the high Cascades. Long-term observations show an increase in the length of the freeze-free season since 1970 by an average rate of 0.5 weeks per decade (Abatzoglou et al., 2014). This increase in winter temperature and freeze-free days is expected to continue in the 21st century, resulting in a proliferation of pseudothecia by extending the growing season and shifting temperatures above the threshold.
Swiss needle cast has been and will continue to be an import-  (Hood, 1982) and decrease elsewhere where the summer precipitation threshold is seldom exceeded.
Because the greatest warming due to climate change is predicted to occur in the winter and summer (Mote, Abatzoglou, & Kunkel, 2013), SNC in the PNW is expected to intensify in frequency and magnitude at higher elevations and/or higher latitudes along the coast and inland where current winter temperatures are a primary limiting factor to fungal growth.

APPENDIX Time Series Models with Pulse Interventions
The site-specific growth-climate-disturbance models reported by Lee et al. (2016) were modified to simplify the earlywood and latewood growth response to temperature and water stress and to better separate the confounding effects of climate and forest disturbances on radial stem growth. Supplementary dual isotope discrimination data and nonhost chronologies that were not previously available were used in this study to detect growth anomalies that represented divergences between the host chronology and the climate proxies including the nonhost chronology and isotopic measurements of δ 13 C and δ 18 O discrimination in tree rings. Years were classified into two categories (no suppression and suppression) based on outlier detection tests using time-series intervention analysis. This resulted in more parsimonious models than the previous time series models that classified years into three categories (no suppression, suppression, release) (Lee et al., 2016). Furthermore, we used a more stringent test for outlier detection and supplemental data to reconstruct the history of forest disturbance impacts on Douglas-fir growth at the stand level.
Several growth anomalies were classified as disturbance years based on observed divergences between the host and climate proxies when the Wald test for outliers was not statistically significant based on the Bonferroni adjustment and an experiment-wise error rate of 0.05. The growth-climate-disturbance models by site and growth period are presented in Table A1.

COSPECTRAL ANALYSIS OF SNC IMPACT
Swiss needle cast indices of impact for CH and FC display pseudoperiodic patterns having a primary periodicity of ~30 years and a secondary periodicity of ~6.7 years ( Figure A1b). The SNC chronologies for CH and FC are coherent and in phase at the key frequencies ( Figure A1c&d). To a lesser extent, SNC impacts for HCTL and FC are coherent and in phase ( Figure A2). SNC impacts for CH and TC are coherent at several key frequencies but out of phase ( Figure A3). SNC impacts for TC lag those for CH by several years. Similarly, SNC impacts for FC and TC are coherent at the key frequencies but SNC impacts for TC lag those for FC by several years ( Figure A4). SNC impacts for SG and TC are coherent and in phase at the primary frequency of ~0.054 cycles/year ( Figure A5). Low growth anomaly for Douglas-fir was concurrent with low growth anomaly for western hemlock and was not attributed to Swiss needle cast. The low growth anomaly in 1918 was likely caused by an exceptionally hot dry summer as indicated by a record high dewpoint deficit in June and PDSI < −2.4 (i.e., moderate-to-severe drought) for June-September.