Climate change-associated trends in biomass dynamics are consistent across soil drainage classes in western boreal forests of Canada

Although positive net biomass change of the world’s forests is critical to offsetting global anthropogenic CO₂ emission, recent studies have demonstrated that the rate of positive aboveground biomass change has decreased with calendar year due to high biomass loss from mortality coupled with growth insufficient to offset these losses (Brienen et al. 2015; Chen and Luo 2015; Chen et al. 2016). The increased tree mortality in tropical forests has been attributed to shortened tree longevity associated with greater climate variability and feedbacks of faster growth on mortality (Brienen et al. 2015). In boreal and temperate forests, increases in tree mortality (van Mantgem et al. 2009; Michaelian et al. 2011; Luo and Chen 2013; Hember et al. 2017) and decreases in aboveground biomass accumulation (Ma et al. 2012; Chen and Luo 2015; Hogg et al. 2017) have been hypothesized to be the result of warming-induced increased demand for evapotranspiration exceeding water availability (McDowell et al. 2008).

The response of net aboveground biomass change and its mortality and growth components appear strongly spatially variable (Brienen et al. 2015; Girardin et al. 2016). The variable responses to climate change could result from spatially dependent tree species composition and local site conditions. The response of aboveground biomass accumulation and wood production to climate change has been previously shown to differ with stand species composition (Coomes et al. 2014; Chen and Luo 2015) and tree mortality increases more dramatically in late-successional species than drought-tolerant early-successional species in regions with temporally decreasing water availability (van Mantgem et al. 2009; Luo and Chen 2013). If the hypothesis of local evapotranspiration exceeding local soil moisture contents is the cause for increased tree mortality and decreased net aboveground biomass change, sites with rapid soil drainage should be more sensitive to climate change as these sites have less buffering capacity than sites with reduced drainage and better water retention. Recent studies have begun to support this theory. For example, across North America, more arid sites have higher tree mortality during drought conditions than more mesic and wet sites (Hember et al. 2017). Alternatively, tree species growing in drier sites have adapted to water deficit, and may be more resistant to droughts. It remains poorly understood whether the response of net aboveground biomass change and its components to climate change differ with local site conditions such as site drainage or moisture regime. A better understanding is needed to fully appreciate the impacts of climate change on forests with heterogonous local site conditions.

The central and western boreal forest presents a unique opportunity to examine this question. Not only is it topographically diverse, but it has also been shown to be strongly affected by climate change. Studies generally demonstrate a loss in net above ground biomass change (Ma et al. 2012; Chen and Luo 2015; Chen et al. 2016) and increases in mortality either due to declining water availability (Peng et al. 2011; Hember et al. 2017). The effect of climate change on tree growth in the region is more complex, with studies demonstrating: increases in young stands but no change, or even decreases, in old stands (Chen et al. 2016); a slight increase early successional conifer forest types but a decrease in late-successional and mixed forest types (Chen and Luo 2015); and, strongly spatially variable trends dominated by heat and drought stress (Girardin et al. 2016).

Here, we use a network of 1279 permanent sample plots across western boreal forests of Canada to specifically test whether stands on well drained sites had greater temporal decreases in growth (here defined as biomass added through both growth of surviving stems and recruitment of new stems), and temporal increases in mortality than stands on less drained sites during the past half-century. Since aboveground biomass change and its components are strongly regulated by forest age in the boreal forest (Oliver and Larson 1990), and long-term temporal trends are forest age-dependent (Chen et al. 2016), we modelled age as a continuous function, similar to previous studies (Chen and Luo 2015; Chen et al. 2016).

Absolute and relative mortality and net biomass change differed significantly among drainage classes while absolute growth differed marginally between drainage classes (Tables 1 and 2). On average over the sampling period, growth was 2.46 ± 0.09 (mean ± bootstrapped 95% confident intervals), 2.55 ± 0.03, and 2.37 ± 0.19 Mg ha⁻¹ year⁻¹ on well, moderately, and poorly drained soils, respectively. Mortality was 1.17 ± 0.16, 1.58 ± 0.07, and 1.43 ± 0.35 Mg ha⁻¹ year⁻¹ on well, moderately, and poorly drained soils, respectively. Net biomass change was 1.30 ± 0.15, 0.97 ± 0.06, and 0.96 ± 0.35 Mg ha⁻¹ year⁻¹ on well, moderately, and poorly drained soils, respectively (Fig. 1a). Relative growth was 2.06 ± 0.13, 2.06 ± 0.05, and 1.70 ± 0.30% year⁻¹; relative mortality was 0.77 ± 0.11, 0.96 ± 0.05 and 1.02 ± 0.24% year⁻¹; and relative net biomass change was 1.29 ± 0.13, 0.85 ± 0.05 and 0.72 ± 0.31% year⁻¹ on well, moderately, and poorly drained soils, respectively (Fig. 1b).

Fixed effects	Sum of squares	df	F	P
Biomass gain from growth (Mg ha −1  year −1 )
D	0.80	2, 1466.1	2.41	0.090
f(A)	12.33	1, 1441.8	73.97	<0.001
Y	0.96	1, 4019.7	5.75	0.017
D × f(A)	0.01	2, 1546.8	0.02	0.976
D × Y	0.86	2, 4029.4	2.58	0.076
f(A) × Y	0.49	1, 3444.6	2.96	0.085
Biomass loss from mortality (Mg ha −1  year −1 )
D	14.77	2, 1713.4	6.39	0.002
f(A)	43.72	1, 1615.1	37.83	<0.001
Y	21.45	1, 3898.2	18.56	<0.001
D × f(A)	20.86	2, 1716.9	9.02	<0.001
D × Y	3.61	2, 3933.1	1.56	0.210
f(A) × Y	0.66	1, 3588.3	0.57	0.450
Net biomass change (Mg ha −1  year −1 )
D	11.56	2, 1804.6	4.06	0.017
f(A)	143.50	1, 1643.9	100.89	<0.001
Y	34.16	1, 3854.9	24.02	<0.001
D × f(A)	30.33	2, 1729.9	10.66	<0.001
D × Y	8.38	2, 3901.1	2.95	0.053
f(A) × Y	0.19	1, 3646	0.13	0.714

Forest age is logarithmically transformed (lnFA) and calendar year is represented by Year. Linear mixed-effects model fit tests used Satterthwaite approximations of degrees of freedom

Fixed effects	Sum of squares	df	F	P
Biomass gain from growth (% year −1 )
D	1.30	2, 1226.5	4.03	0.018
f(A)	27.54	1, 1404.2	170.78	<0.001
Y	4.16	1, 4200.9	25.78	<0.001
D × f(A)	3.57	2, 1558	11.07	<0.001
D × Y	9.45	2, 4183.5	29.31	<0.001
f(A) × Y	24.79	1, 3871.3	153.68	<0.001
Biomass loss from mortality (% year −1 )
D	3.69	2, 1717.4	3.35	0.035
f(A)	7.48	1, 1658.6	13.57	<0.001
Y	7.21	1, 3898.4	13.08	<0.001
D × f(A)	5.61	2, 1814.4	5.09	0.006
D × Y	0.61	2, 3899.3	0.55	0.578
f(A) × Y	2.65	1, 3729.5	4.80	0.029
Net biomass change (% year −1 )
D	17.35	2, 1393.2	11.02	<0.001
f(A)	136.93	1, 1341.1	173.96	<0.001
Y	27.00	1, 3794	34.30	<0.001
D × f(A)	16.51	2, 1484.8	10.49	<0.001
D × Y	17.84	2, 3794.3	11.33	<0.001
f(A) × Y	7.00	1, 3571.9	8.89	0.003

Forest age is the inverse of logarithmically transformed age (lnFA ⁻¹ ) and calendar year is represented by Year. Linear mixed-effects model fit tests used Satterthwaite approximations of degrees of freedom

After accounting for the effects of forest ageing, both absolute and relative rates of growth, mortality, and net biomass change changed significantly with calendar year (Tables 1 and 2). Across all drainage classes, absolute growth declined at −0.001 ± 0.002 Mg ha⁻¹ year⁻¹ with calendar year, while absolute mortality increased at 0.015 ± 0.004 Mg ha⁻¹ year⁻², resulting in a decline of absolute net biomass change of 0.016 ± 0.004 Mg ha⁻¹ year⁻² (Table 1; Fig. 2a). The temporal trends in growth, mortality, and net biomass change did not differ significantly among drainage classes, as indicated by insignificant interactions of calendar year and drainage class (Table 1; Fig. 1b).

The responses of relative growth rate and net biomass change to calendar year differed significantly among drainage classes, while calendar year trends related to mortality did not differ significantly (Table 2). Responses of relative growth to calendar year were −0.017 ± 0.006, 0.005 ± 0.002 and −0.021 ± 0.008% year⁻²; and relative net biomass change was −0.036 ± 0.009, −0.006 ± 0.003 and −0.030 ± 0.014% year⁻² on well, moderately, and poorly drained soils, respectively (Fig. 2b).

Both absolute and relative growth, mortality, and net biomass change responded significantly to forest ageing (Tables 1 and 2). Growth declined with forest ageing on all soil types, while mortality increased more significantly on moderately drained soils than on poorly or well drained soils, leading to a significantly larger decline in net biomass change on moderately drained soils than on poorly or well drained soils with forest ageing (Table 1; Fig. 3a). Relative growth rates and net biomass change had an initial sharp decline with forest ageing, with the largest declines on well drained followed by moderately drained soils (Table 2 & Fig. 3b). Relative mortality had a sharp initial increase with forest ageing, highest on moderately drained soils, followed by a relatively consistent increase (Table 2 & Fig. 3b).

As atmospheric CO₂ had a tight correlation to calendar year (r ² = 0.97, Additional file 1: Figure S2), trends in both absolute and relative growth, mortality, and net biomass change responses to atmospheric CO₂ largely followed those of calendar year although poorly drained sites had significantly lower growth than moderately drained or well drained sites (Table 3, Figs. 4a and 5a). Absolute growth did not respond to increasing ATA, but absolute mortality increased, resulting in a similar decrease in absolute net biomass change across all soil drainage classes (Table 3, Fig. 4b). Trends in absolute mortality related to temperature were also age-dependent, with older ages having increased rates of mortality with increasing temperatures (Table 3). The response of relative growth to increasing temperature was comparable to temporal trends, with increases on moderately drained sites and decreases on well drained and poorly drained sites (Table 3, Fig. 5b). With decreasing SPEI, absolute growth decreased, while loss from mortality increased, leading to a decline in absolute net biomass change with no significant difference among drainage classes (Table 3, Fig. 4). Trends in absolute growth and mortality responses to SPEI were also age dependent, with growth being less affected in older stands than in younger stands, but a large increase in mortality in older stands (Table 3). With decreasing SPEI, relative growth differed significantly between drainage classes while relative mortality increased, and relative net biomass change decreased with no significant differences among drainage classes (Table 3, Fig. 5c).

Fixed effects	Biomass gain from growth (Mg ha−1 year−1)	Biomass loss from mortality (Mg ha−1 year−1)	Net biomass change (Mg ha−1 year−1)	Biomass gain from growth (% year−1)	Biomass loss from mortality (% year−1)	Net biomass change (% year−1)
Atmospheric carbon dioxide concentration
D	0.094	0.002	0.017	0.017	0.035	<0.001
f(A)	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001
CO 2	0.008	<0.001	<0.001	<0.001	<0.001	<0.001
D × f(A)	0.976	<0.001	<0.001	<0.001	0.005	<0.001
D × CO 2	0.066	0.158	0.032	<0.001	0.490	<0.001
f(A) × CO 2	0.046	0.683	0.974	<0.001	0.055	0.002
Annual temperature anomaly
D	0.122	0.003	0.030	0.073	0.051	<0.001
f(A)	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001
ATA	0.373	0.002	0.002	0.006	0.009	<0.001
D × f(A)	0.889	<0.001	<0.001	<0.001	0.015	<0.001
D × ATA	0.336	0.882	0.877	<0.001	0.747	0.100
f(A) × ATA	0.086	0.036	0.207	<0.001	0.001	0.867
Standardized precipitation-evapotranspiration index
D	0.138	0.004	0.027	0.044	0.046	<0.001
f(A)	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001
SPEI	0.114	0.021	0.006	0.159	0.005	0.004
D × f(A)	0.517	<0.001	<0.001	0.001	0.002	<0.001
D × SPEI	0.179	0.650	0.618	0.041	0.582	0.065
f(A) × SPEI	0.337	0.010	0.026	<0.001	0.048	0.020

f(A) represents the natural logarithm of forest age for absolute rates and the inverse of the natural logarithm of forest age for relative rates

We found that temporal trends in both absolute and relative growth, mortality, and net biomass change associated with climate change were largely consistent across three drainage classes. Although we predicted that well drained sites would have more negative responses to climate change (i.e., decreased growth, increased biomass loss from mortality) than moderately or poorly drained sites, there are three major reasons we may not have observed this trend. First, it is possible that well drained sites foster a species community that is already resistant to lower water availability while sites with higher water retention benefit from having a soil water buffer. Previous studies in the boreal forest have emphasized the role of local site suitability in determining species compositions (Chen and Taylor 2012) and have demonstrated that these forests are moving towards more drought resistant early-successional conifers in response to climate change (Searle and Chen 2017a). These abiotic sorting mechanisms, and background compositional shifts, could be enhancing stand resistance to lowered water availability on well drained sites. Second, it is possible that communities, irrespective of composition, have adapted to local water availability conditions. Thus, responses of growth, mortality and net biomass change to temporal declines in water availability may be less dependent on local site conditions (to which communities are adapted) and more dependent on the magnitude of temporal decline. Third, we only examined long-term temporal and water availability trends. It is possible that well drained sites are more susceptible to singular drought events than less drained sites although examination of this would require more evenly sampled plots or dendrochronological techniques (e.g., Hogg et al. 2017). However, if this were the case, we might still expect to find a long-term difference as trees that experience droughts often have long-term growth reductions (Anderegg et al. 2015).

While absolute and relative measures of biomass dynamics were generally in agreement, relative growth increased over our study period on moderately drained sites but decreased significantly on well and poorly drained sites, whereas absolute growth consistently declined across all drainage classes. The discrepancy in direction between relative and absolute growth on moderately drained sites is likely an artifact of the calculation of relative growth rates. We calculated relative growth rates as the absolute growth rate divided by the mean standing biomass - if standing biomass declined more than absolute growth rates due to temporally increased absolute biomass loss from mortality, then relative growth rates would increase. The discrepancy in temporal trends in relative growth rates drove the differences in net biomass change; although net change declined across all drainage classes since relative mortality declined more than relative growth increased on moderately drained sites. It is also important to note that our study focusses on large trees (i.e., trees ≥9.2 cm in diameter at breast height). While this may bias our estimates of age effects, estimates of temporal trends have been shown to be consistent across diameter thresholds within boreal forests (Searle and Chen 2017b).

Our sensitivity analysis using climactic drivers further emphasizes that forests growing on different drainage classes are being affected similarly in the western boreal forest. There was no evidence that increasing temperatures or decreasing water availability adversely affected well drained sites more than moderately or poorly drained sites. However, we did find an overall temporal decline in net biomass change driven by higher loss from mortality over the past half-century in response, consistent with other studies in the region (Ma et al. 2012; Chen and Luo 2015; Chen et al. 2016) highlighting the negative impact of reduced water availability.

It is interesting that previously reported increased tree mortality rates to drought on more arid sites (Hember et al. 2017) were not reflected in our stand-level analysis. However, it is important to note that we used explicit measurements of drainage class as our measure of site water availability rather than the approach of Hember et al. (2017) who used average long-term climactic water availability as measured by aridity index. A recent tree ring analysis from across the boreal forest of Canada found a positive effect of a long-term soil moisture index (calculated using precipitation and evapotranspiration rates) and growth (Girardin et al. 2016). Together, these results suggest that long-term regional climates have pronounced effects on forest growth, mortality and net biomass change, irrespective of local site condition.

We found no evidence that drainage class influences estimates of temporal trends in growth, mortality, and net biomass change. Sensitivity analysis further confirmed that drainage class does not alter forest response to increases in atmospheric carbon dioxide and temperature, and decreases in water availability. Our results suggest that climate change acts as top-down control on forest growth, mortality and net biomass change, irrespective of local site condition.