Fig 1.
Location of 88 sites inventoried for aspen canker. For analyses, sites within the Yukon-Old Crow Basin were grouped with those from the North Ogilvie Mountains.
Fig 2.
Relationships between stand structural characteristics across the 88 study sites.
Mean aspen DBH vs. percentage of trees that were aspen (= relative aspen density) across stands where aspen grew with white spruce, black spruce, Alaskan paper birch, or a combination (A) and aspen basal area vs. mean aspen DBH (B) across study sites.
Table 1.
Aspen structural characteristics of 88 stands averaged by ecoregion.
Fig 3.
Canker incidence and canker mortality across ecoregions.
Box plots of (A) canker incidence (% aspen trees infected) by ecoregion, and (B) % of trees infected with canker that are dead. Ecoregions sharing different letters are significantly different at P < 0.05 (ANOVA model, F5,82 = 5.10, P < 0.001). Means ± 1 STD error are listed across the bottom of the graph.
Fig 4.
Incidence of canker infection by DBH class for each ecoregion and as an average across all 88 sites (last column).
For the last column, size classes sharing similar letters are not different at P < 0.05 (ANOVA, F5,291 = 5.42, P < 0.0001). CK = Cook Inlet, CR = Copper River, RM = Ray Mountains, TKL = Tanana Kuskokwim Lowlands, NOM = North Ogilvie Mountains, YTU = Yukon Tanana Uplands. The absence of size classes within ecoregions is because no trees were found within that size class, or that none of the trees within that size class were infected.
Fig 5.
Proportion of DBH size classes at 72 sites with canker (columns 1–8) and at 16 sites without canker (columns 9–10).
For sites with canker, the first 6 columns show data for only trees with canker categorized by ecoregion, and column 7 shows the mean across all sites for trees with canker. Column 8 shows the size distribution of all trees (with and without canker) at sites where canker was present. Values beside size classes in Column 8 refer to the % increase in proportion of infected trees vs. the population as a whole for each size class (*P < 0.10). For example, there are 6% more trees infected with canker in DBH class 1 than there are in the population (NS), while there are 35% fewer trees infected with canker in 15–20 cm size class than there are in the population as a whole (P < 0.10). Columns 9 and 10 show data for sites without canker, broken down into young sites (n = 9) and older sites (n = 7). For sites without canker, young sites (column 9) were located within recent burn perimeters scattered across the Yukon Tanana Uplands, Ray Mountains, and North Ogilvie Mountains ecoregions, while 6 of 7 old sites (column 10) were located south of the Alaska Range within the Cook Inlet and Copper River ecoregions. CK = Cook Inlet, CR = Copper River, RM = Ray Mountains, TKL = Tanana Kuskokwim Lowlands, NOM = North Ogilvie Mountains, YTU = Yukon Tanana Uplands.
Fig 6.
Modeling canker incidence based on stand parameters.
Odds ratios for stand structural parameters from the mixed-effects logistic model: Logit (π) = log (π/(1- π)) = –0.761–0.177*DBH– 0.023*Relative Aspen Density + 0.092*Aspen Basal Area + 0.106*Mean Aspen DBH– 1.413*CK– 3.263*CR—0.358*RM + 0.692*TKL– 0.736*NOM. Ecoregion β values are computed relative to YTU which does not appear in the model and has a log (odds) = the intercept when all other factors are 0. Variables with odds ratios significantly greater than 1.0 indicate a significant increase the odds of canker infection while those below 1.0 indicate a significant reduction in the probability of canker infection. For example, an odds ratio of 0.838 (red dot) means that for every 1 cm increase in DBH of a given aspen tree, the odds of having canker is reduced by 16.2% (= (1–0.839)*100), while an odds ratio of 1.112 (blue dot) means that every 1 cm increase in average stand-level DBH (a surrogate for stand age) increased the odds of having canker by 11.2% ((1.112–1)*100). Ecoregion abbreviations follow Fig 4.
Fig 7.
Modeling canker incidence among ecoregions.
Probability of canker infection for different ecoregions as a function of DBH of an aspen tree (A), mean stand-level aspen DBH, a surrogate for stand age (B), and aspen basal area (C). Probabilities were calculated by fixing all other variables as mean values using the following equation derived from a mixed-effects logistic regression model incorporating site as a random effect: Logit (π) = log (π/(1- π)) = –0.761–0.177*DBH– 0.023*Relative Aspen Density + 0.092*Aspen Basal Area + 0.106*Mean Aspen DBH– 1.413*CK– 3.263*CR– 0.358*RM + 0.692*TKL– 0.736*NOM. Ecoregion β values are computed relative to YTU which does not appear in the model and has a log(odds) = the intercept when all other factors are 0. Contrasts derived from logistic models indicated the following differences in canker incidence among ecoregions; TKLa, YTUab, RMbc, NOMabc, CKcd, CRd, where ecoregions with different letters are significantly different at P < 0.05. Ecoregion abbreviations follow Fig 4.
Fig 8.
Modeling canker responses to variations in stand structure and climate.
Odds ratios for canker incidence response to 0.25 hPa increases in VPD during May, June, July and August derived from the mixed-effects logistic equation model: Logit (π) = log (π/(1- π)) = α + β1 *DBH + β2 *Relative Aspen Density + β3 *Aspen Basal Area + β4 *Mean Aspen DBH + β5 *VPDmonth. Model results for stand structural variables and monthly VPD are show in S1 Table. Unit increases in variables with odds ratios above 1.0 significantly increase the probability of canker infection while those below 1.0 reduce the probability of canker infection. Thus, an odds ratio of 1.177 for VPD in May indicates that for every 0.25 hPa increase in VPD the odds of having canker increases by 17.7%.
Fig 9.
Changes in disease incidence at 3 assessments across multiple sites within 2 regions.
Box plots showing % of all aspen trees dead (2013, 2015, and 2018) and % of all aspen trees with canker (2015 and 2018) for sites within the Big Denver Fire perimeter (A, n = 4 sites, Ray Mts. ecoregion) and Gerstle River Fire perimeter (B, n = 6 sites, Tanana Kuskokwim Lowlands ecoregion).
Fig 10.
Box plots showing % of all aspen trees dead at select CAFI sites in 2000 and every 5 years thereafter.
When plot inventory began in 1995, dead trees were not inventoried, but all live trees were tagged and those that subsequently died were inventoried on subsequent dates. The CAFI program included other sites in the inventory program from 1996–2013, but the 12 sites here represent the cohort of sites with the longest sampling history (other than 4 sites first inventoried in 1994 but not included here). Also shown are the proportion of trees with canker at these same sites that we inventoried in 2018.