1. Introduction
Barren, severely disturbed sites lacking soil, such as mine sites and waste deposit sites, present severe challenges to ecological service restoration because of high temperatures, solar radiation, and wind speeds; extreme temperature changes; and low soil moisture and nutrient availability. Rapid erosion from these sites creates more problems by removing soil and increasing sedimentation in local waterways, negatively affecting aquatic environments [
1,
2]. Severely disturbed soils are often compacted and lack water holding capacity and soil nutrients [
3,
4,
5]. Thus, successional restoration should start with some development of soils and early-successional native species, which should allow sites to eventually return to natural successional processes [
6]. Returning nutrients to the soil and eventually ameliorating the coarse substrate into a forest supporting substrate could be accelerated by using early-successional hardwood species, as it has been shown that nutrient content and organics within hardwood forests are often greater than in conifer forests [
7,
8]. Reintroduction of original or new topsoil to barren sites can appear to be a promising method to ameliorate poor sites; however, there is a very high likelihood of more erosion [
9].
Seedlings in barren landscapes are often exposed to severe climatic conditions. Excessive heat and drought in the summer, lack of nutrients, frost heave in shallow soils, and cold damage in the winter are unfavorable conditions for plant survival. Shallow soils lack water holding capacity and dry rapidly, resulting in seedling mortality. In addition, shallow soil depth to bedrock and consequently a high water table increase susceptibility to frost heave from greater water concentrations near the soil surface [
10,
11,
12]. Frost heave has two main effects. First, young plants may be damaged or killed from root exposure. Second, soils that have been frost heaved are susceptible to erosion [
13,
14]. According to Graber [
15], four types of damage can occur to seedlings: (1) heaved out, with lethal exposure of root system, (2) partial heave, with exposed root damage commonly found, (3) girdling of plant stems via mechanical damage caused by substrate abrasion on the epidermis and cambium, and (4) seedling decapitation: separation of plant parts such as the roots from the aboveground plant parts.
Early-successional species are known for their ability to grow in post-disturbance landscapes [
16,
17]. With their need for full sunlight, they tend to grow rapidly, which can help capture sites quickly [
16,
18,
19]. We examined four such species: white birch (WB,
Betula papyrifera Marsh), gray birch (GB,
Betula populifolia Marsh), green alder (GA,
Alnus viridis Vill. ssp.
crispa Ait) and speckled alder (SA,
Alnus incana L. ssp.
rugosa Du Roi). All four species are classified as early-successional, relatively fast-growing species with relatively short lifespans [
20]. All four species have some histories in restoration efforts due to their ability to withstand and reclaim harsh sites [
16,
18,
19,
21,
22,
23]. White birch prefers deep, well-drained, rich soil, but can still compete in a wide range of soil types ranging from gravels to silts and organic bog to peat soils [
24]. Gray birch grows best on a moist, well-drained site, but is quite adaptable and can grow well on dry sandy or even gravelly soil [
25,
26]. Green alders prefer moist sites but can tolerate dry sites with soil types from gravelly to sandy and will regularly be found on upland glacial till sites [
27]. Speckled alders compete best on wet sites, such as beside waterways, bogs, and fens, but will also establish in mires with peaty soils [
28,
29]. In Minnesota, it was noted that speckled alder will grow in mesic to moist sites but was found more in moist sites [
30].
In this study, we examined a factorial of three site preparations: (1) site crusher to create a loose aggregate using a Meri-Crusher (MC) (MJ hydraulic 1.4 ST Meri-Crusher, Voukatti, Finland), (2) straw (S) coverage, and (3) coarse woody debris (CWD) coverage. The site crusher treatment involved mechanically ripping the top layer of aggregate to create a substrate that can be used to restore terrestrial ecosystems with little to no existing substrate available [
31,
32]. Breaking in situ soil beds is better than replacing with soil from elsewhere [
33], and benefits of ripping the soil include improved aeration, water infiltration, and root penetration [
34]. Straw coverage can affect the processes occurring in the top layer of the soil, such as reducing surface evaporation [
35], improving water infiltration [
36], and increasing moisture retention [
37,
38,
39], and can act as an insulator and regulate extreme temperatures in the winter to help prevent frost heave of planted seedlings [
40]. Coarse woody debris can create protected microsites and enhance biodiversity on a reclaimed site [
41]. The addition of CWD has been shown to decrease water loss and seedling desiccation by providing shelter from the elements [
42]. There is a preponderance towards larger CWD, as smaller pieces like mulch can draw what little nitrogen is available from the soil via different fungi to fuel decomposition [
43].
Our goal was to quantify seedling mortality from the effects of three site preparation treatments in a site, no site preparation 2 × 2 × 2 factorial design resulting in eight separate treatments including control, testing four early-successional species. We hypothesized that the presence of a site preparation treatment will have a positive effect on the survival and establishment of the seedings. We also hypothesized that when site preparation treatments were used together, their effects might be additive. In addition, we hypothesized that there will be species differences in survival. Our objective was to quantify mortality (1) over the first summer, (2) over the following winter, and (3) mortality due to frost heave. In addition, we examined the combined full first-year mortality to assess the best practices and the best species to begin the restoration and forest succession process on highly disturbed sites.
2. Materials and Methods
2.1. Study Area
The study area is located within the Department of National Defence, CFB Gagetown, NB, Canada (45.68108, −66.50179). The study area had three sites located at an average altitude of 87 m above sea level and lies on a small western aspect. Average annual precipitation is 1376 mm year−1, with most of it falling in summer to winter months. Rainfall ranges from 94 mm in March to 133 mm in June. Snowfall during the winter months ranges from 41 cm in March to 72 cm in February. Average temperature for the hottest months (July and August) is around 20 °C and for the coldest months (January and February) −9 °C. Soil is predominantly parent material consisting of felsic pebble conglomerates and mixed igneous rocks. Surrounding forest is dominated by gray birch (Betula populifolia), red oak (Quercus rubra), and trembling aspen (Populus tremuloides).
2.2. Site Preparation Treatments
Three separate sites were selected at CFB Gagetown, NB, Canada (45.68108, −66.50179). A randomized site preparation design with site, no site preparation factorial of: S, NS; MC, NMC; and CWD, NCWD in a 2 × 2 × 2 factorial resulted in eight treatments, including a control. Note, a prefix of N on the site preparation treatment indicates no such treatment. The eight-treatment factorial was (1) NS × NMC × NCWD, the control, (2) S × NMC × NCWD, (3) NS × MC × NCWD, (4) NS × NMC × CWD, (5) S × MC × NCWD, (6) S × NMC × CWD, (7) NS × MC × CWD, and (8) S × MC × CWD. The eight treatments were replicated (blocked) four times per site; this equated to 32 plots per site, on three sites, resulting in 96 plots in the experiment. The MC treatment was created using a John Deere 200D excavator equipped with an MJ hydraulic 1.4 ST Meri-Crusher (Voukatti, Finland). The MC plots produced a 3.2 m by 3.2 m square of broken aggregate from loose soil, loose rock and solid rock substrate. Straw bales were used to distribute straw over approximately 18 m2 of area or roughly 2 experimental plots to a depth of 5 cm. The CWD was sourced from Department of National Defence (DND) near Gagetown, NB, using primarily GB. All CWD was processed into ~3 m length to match the MC plots. An average volume of 0.203 m3 ± 0.020 m3 of CWD was applied to 48 plots. The placement of CWD was performed last to hold the straw in place, and the CWD was distributed over each plot in a north to south orientation to provide shade at different times of the day.
2.3. Plant Materials
Seedlings from the four tree species were grown from seedlots bulked from seed collected from sources listed in
Table 1. WB, GB and GA were grown during 2020 at the Atlantic Forestry Center (AFC), Fredericton, NB (45.93501, −66.65745); SA were grown during 2020 at Scot and Stewart nurseries in Antigonish, Nova Scotia (45.53292, −61.86888). The seedlings grew in trays with root container volume of 110 mL. In January 2021, 1-year-old seedlings were cut to a height of roughly 10 cm to reduce the root-to-shoot ratio for these harsh sites and had a root-to-shoot ratio of 1:1 by dry weight. Seedlings were packed into boxes for frozen storage at −4 °C in AFC freezers. The seedlings were removed from frozen storage 3–4 days prior to planting to thaw out. On 7–8 May 2021, four individual replicates per species were randomly planted per plot, 32 plots per site, thus 128 individual species per site, and 384 individual species across the three sites. The seedlings were planted with narrow shovels at 0.5 × 0.5 m spacing. The experiment contained 1536 total planted seedlings across the three sites.
2.4. Soil Properties
For each plot, three depths to bedrock measurements were made using an iron rod hammered into the soil until it contacted the solid rock substrate. Soil samples were collected in October 2020 from 3 sites × 2 treatments MC, no MC (NMC) × 4 block replicates for a total of 24 total soil samples. Soil analysis included texture sand, silt, clay, and rock (%), and soil macronutrient analysis including organic matter, carbon (C), and nitrogen (N) using phosphorus (P) available potassium (K), calcium (Ca), magnesium (K), concentration, and pH by the Laboratory for Forest Soils and Environmental Quality at the University of New Brunswick using techniques as outlined in Major et al. 2022 [
44]. Soil bulk density assessment was performed from 16 plots per site, 8 meri-crushed plots and 8 non-meri-crushed plots, for a total of 48 samples in the fall of 2022.
Table 2 provides a summary of soil properties among sites. Soil moisture (%) was measured periodically using an ML3 Theta Probe soil moisture sensor (Cambridge, UK), by recording three readings per plot in two randomly selected blocks of each site, varying site order each time they were taken. Soil moisture readings were taken on 13 collection dates ranging from 3 March 2021 to 9 September 2021.
2.5. Mortality Assessment
Percent mortality per species per plot was determined from the four replicates per plot. Mortality assessments were performed at the end of August 2021 for summer-related mortality, and April 2022 for frost heave mortality. If a root plug lifted 25% or more and had died, it was considered frost heave mortality. A large majority were completely heaved from the ground. If the root plug lifted less than 25% and died, it was considered part of the winter mortality. Winter-related mortality was largely due to fungal presence found on expired seedlings. The difference between summer-related mortality 2021 and what was not due to spring frost heave mortality 2022 was considered winter-related mortality.
2.6. Statistical Analysis
The experiment was established as a randomized block design. Initial soil properties were subjected to analyses of variance (ANOVA); site and MC were fixed effects, and block was a random effect. The following ANOVA model was used:
where Y
ijk is the dependent soil property trait of the
ith block,
jth site, and the
kth MC treatment, and µ is the overall mean. B
i(S
j) is the random effect of block nested within site (
i = 1, 2, 3, 4), S
j is the effect of the
jth site (
j = 1, 2, 3), M
k is the effect of the
kth MC treatment (
k = 1, 2), SM
jk is the interaction effect between
jth site effect and
kth MC treatment effect, and
eijk is the random error component.
Available soil moisture in 2021 was subjected to analyses of variance (ANOVA); site, S, MC and CWD were fixed effects, and block was a random effect. The following ANOVA model was used:
where Y
ijklm is the dependent soil moisture availability of the
ith block, the
jth site,
kth S treatment,
lth MC treatment, and
mth CWD treatment, and µ is the overall mean. B
i(T
j) is the random effect of the
ith block nested in the
jth site (
i = 1, 2, 3, 4), T
j is the effect of the
jth site (
j = 1, 2, 3), S
k is the effect of
kth S treatment (
k = 1, 2), M
l is the effect of
lth MC treatment (
l = 1, 2), C
m is the effect of the
mth CWD treatment (
m = 1, 2), and
eijklm is the random error component. Interactions of the single effects are included up to three-way interactions; four- and five-way interactions are folded into the error term.
Summer-related, winter-related, frost heave, and total first year mortality were analyzed using analysis of variance (ANOVA). MC, S, CWD, species, and site were all considered as fixed effects. Block was a random effect. The following ANOVA model was used:
where Y
ijklun is the dependent mortality factor of the
ith block, within the
jth site, of the
kth species, in the
lth S treatment,
nth CWD treatment,
oth MC treatment, and µ is the overall mean. B
i(T
j) is the random effect of block nested in site (
i = 1, 2, 3, 4), T
j is the effect of the
jth site (
j = 1, 2, 3), P
k is the effect of the
kth species (
k = 1, 2, 3, 4), S
l is the effect of the
lth S treatment (
l = 1, 2), C
n is the effect of the
nth CWD treatment (
n = 1, 2), M
o is the effect of the
oth MC treatment (
o = 1, 2), and
eijklno is the random error component. Interactions of the single effects are included up to three-way interactions; four- and five-way interactions are folded into the error term.
Percent mortality was arcsine, square root transformed to fulfill normality. All effects were considered significant at the p = 0.05 level, though all p-values are presented for individual consideration. Assumptions of independence of data, normality of distribution, and normality of variance were met prior to assessing each statistical analysis (ANOVA). The general linear model from Systat Software Inc. (Chicago, IL, USA) was used for analysis. If significant, site and species were post hoc tested using Tukey mean separation test (p = 0.05). Significant interactions are referred to as either rank change interaction, or magnitude effect interaction. A rank change interaction occurs when one effect has a greater mean under one scenario and lower mean under another scenario. A magnitude effect would refer to when the effect mean is always greater than under both scenarios, but of different magnitudes.
3. Results
There was little or no difference in soil nutrients or texture between MC and NMC plots, so they were folded into the site-only analysis. Additionally, soil bulk density showed no significant difference between sites or MC and NMC plots. Soil analysis in fall 2020 showed greater total N in site 1 compared to sites 2 and 3 (
Table 2). Site 1 had a significantly lower Ca content and pH than sites 2 and 3. Two soil texture properties were significantly different: silt, which was greater on site 1 (43.8%) than on sites 2 and 3, and sand, which was greater on site 3 (58.2%) than on sites 1 and 2.
Soil moisture content corresponded with the precipitation graph (
Figure 1A). When the greatest rainfall occurred (1 July onwards), the average soil moisture content increased. Dry conditions from 3 June to 30 June resulted in significantly lower soil moisture content for NS than for S treatments. As the soil accumulated more rain, the effect of straw was not significant.
All summer-related mortality main effects excluding CWD were significant (
Table 3). Straw was highly significant and accounted for the greatest amount of variance (9.2%) in summer-related mortality and appeared in most other significant interactions that were magnitude effects.
The significant MC × S × site (
p = 0.003) interaction (
Table 3) for summer-related mortality was a result of a single site rank change on only NS plots. Site 3 had the greatest summer-related mortality on NS, NMC treatments (25.8%), whereas it had low summer-related mortality on NS, MC treatment plots (7.0%) compared to site 2 (21.0%) (
Figure 2A). The S × site × species interaction in summer-related mortality was largely due to a magnitude effect under NS. Site 2 NS had the greatest mortality particularly for SA (45.3%), which is also seen in
Figure 2B, whereas sites 2 and 3 were somewhat equal in summer-related mortality for the other species.
The significant MC × S × CWD interaction in summer-related mortality was a result of a single rank change again under NS (
Figure 3). Under the NS, CWD treatment, NMC plots had greater summer-related mortality than MC plots, with 20.7% and 7.8%, respectively. Summer-related mortality was reversed to a lesser degree under the NS, NCWD plots, with 11 and 14%, respectively.
Overall, S and NS had 2.7% and 13.2% summer-related mortality, respectively. Site 1 mortality of 3.3% was significantly lower than that of sites 2 and 3 with 11.1% and 9.4%, respectively.
Winter-related mortality had the smallest mortality impact, resulting in the loss of only 36 seedlings, affecting only 1.9% of the total seedlings. Winter-related mortality had only one three-way interaction, site × S × CWD interaction (
Table 4), and it was a small rank and largely a magnitude effect driven by S on site 2 (
Figure 4).
S, NCWD on site 2 had the greatest mortality, with 12% winter-related mortality (
Figure 4). This was also driving the two significant two-way interactions: site × CWD and S × CWD (
Figure 5A,B). Overall, S and NS resulted in winter-related mortality of 1.7% and 4%, respectively. Overall, CWD, NCWD had 1.6% and 3.6% winter-related mortality, respectively. MC, NMC had a winter-related mortality of 3.5% and 1.7%, respectively.
There was only one frost heaved mortality three-way interaction (
Table 5); it was largely a magnitude effect among the three site treatments and was driven by the large frost heaved mortality under NS, NCWD with NMC, MC having 25% and 47% mortality, respectively (
Figure 6).
Frost heaved mortality with all other treatment combinations was 5% or less, with S, CWD having zero frost heaved mortality. The three two-way interactions S × MC, CWD × MC and CWD × S were largely magnitude effects found in the three-way interaction figure (
Figure 6). The one other two-way interaction site × S was also a magnitude effect, with sites 2 and 3 having the largest magnitude with site 1 lesser (
Figure 7).
ANOVA for frost heaved mortality showed that all main effects were significant (
Table 3). Straw accounted for the most variance in the model (19%), and CWD had the second largest variance (12.9%). Overall, S vs. NS had 1.1% and 20.7% frost heaved mortality, CWD, NCWD had 2.5% and 19.4%, and MC, NMC had 13.4% and 8.5%, respectively. The species that was affected the most was SA, with 14.8% mortality, then green alder at 11.7%, and both birches at 8.6%. Site was significant because of the reduced mortality on site 1 (6.8%) while sites 2 and 3 had increased mortalities of 12.9% and 13.1%, respectively.
There was also only one total first-year mortality three-way interaction (
Table 6), and it was largely a magnitude effect, with some rank change, and it was again among the three site treatments and driven by the large frost heave mortality findings.
Again, under NS, NCWD with NMC, MC had 39 and 60.9% mortality, respectively (
Figure 6). The three-way interaction MC × S × CWD (
p < 0.001) (
Figure 8) showed trends very similar to that discussed in the frost heaved model, with MC alone having the greatest mortality (60.9%). Since spring 2022 mortality was the sum of the past three mortalities, it followed similar trends. MC presence was largely negative by itself without other treatments; however, once CWD or S was added, MC had a positive effect.
The species × straw interaction (
p < 0.001) (
Figure 9A) was a magnitude effect. When S was present, all species mortality stayed under 10% (
Figure 9A), but when S was not present, there was an average 21.3% increase in mortality for WB, GB, and GA, and the greatest increase of 45.4% mortality in SA, which resulted in the greatest mortality through the year of around 56.3% (
Figure 9A). Straw × site interaction (
p < 0.001) was also due to a magnitude effect. The lack of S on site 1 (16.8%) had a more muted effect when compared to site 2 and 3 at 46.9% and 41%, respectively (
Figure 9B).
After the first full year, the overall survival of the seedlings was significantly affected by all main effects, excluding MC (
Table 3). However, MC was significantly present in two-way interactions with CWD (
p < 0.001) and site (
p = 0.007) and also a three-way interaction with S and CWD (
p = 0.001). Like the other models, S had the highest variance component (19.1%), and it was present in four of the six significant interactions. Overall, S and NS had 7.7% and 34.9% total mortality, respectively. CWD and NCWD had 11.8% and 30.7% total mortality. Sites 1, 2 and 3 overall had 11.3%, 27.9%, and 24.6% total mortality, respectively, by spring 2022.
Total mortality was greatest in SA, with 33.6% (
Figure 10A). WB, GB, and GA all had under 25% mortality, with GA being at exactly 20% mortality through the first year. Frost heave was the largest contributor to mortality regardless of species, with WB, GB, GA, and SA experiencing 8.6%, 8.6%, 11.7%, and 14.8% mortality, respectively, across all sites. The effect of winter-related mortality was relatively minor, with only 36 cases. Of all species, 1.9% were affected, with most mortality occurring in SA, with 4.4% of all SA dying (
Figure 10A). When broken down by treatments in
Figure 10B, the addition of S greatly limited the high level of frost heaved mortality, as did the addition of CWD but to a lesser extent. The treatment of MC alone has the greatest mortality, with 60.9% of all MC plots succumbing to 13.5% summer-related mortality and 46.9% frost heave mortality. An additive effect could be seen where the more treatments being added reduced the mortality. The treatment of MCSCWD has the lowest mortality rate, with only 3.1% of the species dying within that plot type (
Figure 10B).