**3. Results**

#### *3.1. Soils and Weather and Soil Climate*

The soils were all medium textured, ranging from sandy loam to clay loam (Table 1). The similarity in textures, organic matter content, and classifications reflect their common glacial till origins. Soil organic matter was low, averaging 0.036 kg kg−<sup>1</sup> at the 5 cm depth and a third of that amount at 20 cm (Table 1). The sites are in the Upper and Lower Cordilleran Ecoregions of western Alberta [34]. The values of soil water content were obtained from each bulk density sample, and the soil water potential was measured with a handheld tensiometer next to each bulk density sample [11].

Differences in the depth of winter snowpack dominated the weather data over the 7 years that data were collected. The depth of snow cover was deeper the first three winters (1994–1997) than during the last four winters (1997–2001). Data to illustrate the differences between a heavy and light snowpack winter are shown in Figure 3. Although the sites were separated by up to 330 km, the confidence limits for snow depth within a year showed that the differences were small within the Upper and Lower Cordilleran Ecoregions of western Alberta [34].

**Figure 3.** Snow depth and daily minimum air temperatures for a winter with normal snowfall (1996–1997) and below normal snowfall (1997–1998). Values are the mean (solid lines) and the confidence limits (dashed lines) for the 9 sites each winter.

Periodic cycles of colder air temperatures had a more pronounced effect on soil temperatures when snow depth was low (Figure 3). A snow depth of about 10 cm was insufficient to prevent soil temperatures from dropping to near −4 ◦C at 5 cm and −2 ◦C at the 20 cm depth when daily minimum air temperatures were between −25 and −33 ◦C (3–14 January 1998). A snow cover of at least 40 cm prevented soils from dropping below −1 ◦C at the surface despite daily minimum air temperatures of −30 ◦C and colder, which occurred four times during the winter of 1996–1997. Whether the soil water froze either time will be an important part of our discussion.

Snow depth was measured over the undisturbed litter layer of the harvest-only treatment. Most of the forest floor was compacted or displaced on the adjacent seven-cycle treatment. Nevertheless, the differences in soil temperature between the control and trafficked soil were less than 1 ◦C (Figure 3). Thorud and Duncan [35] previously reported that the removal of the forest floor had less effect on soil freezing than did the depth of snow cover.

#### *3.2. Skidding Traffic Increases Bulk Density*

The mean values of bulk density and standard errors for each treatment, depth, site, and year are shown in Table 2. The only obvious trend in standard errors was values decreasing with depth—0.046, 0.039, and 0.033 Mg m<sup>−</sup><sup>3</sup> for depths of 5, 10, 20 cm, respectively.


*Forests* **2022**, *13*, 553

#### *Forests* **2022**, *13*, 553

**Table 2.** *Cont.*

The main effects model of the ANOVA show trafficking, depth, site, and year are all highly significant (*p* < 0.0000, Table 3). Soil depth is the dominant variable because bulk density increases rapidly with increasing depth (Table 4). The relationship is consistent for the entire period (depth and year interaction was not significant). These upland forest soils typically have a thin mineral topsoil with the B horizon beginning within 6 to 16 cm of the surface [36].

**Table 3.** Analyses of Variance of soil bulk density collected over a sampling period of 7 years (mixed ANOVA with depth and year as repeated factors).


\* Significance level, Pr < 0.05; \*\* Significance level, Pr < 0.01; \*\*\*\* Significance level, Pr < 0.0001.


**Table 4.** Summary of descriptive statistics for standard erros at the 5, 10, and 20 cm d.

Only the three cycles of skidding treatment caused a significant increase in bulk density (*p* < 0.0000, Table 3, Figure 4). An increase in bulk density between 3 and 7 cycles occurred but was not significant. A factor contributing to the sustained increase between 3 and 7 cycles on these wet soils was that samples for bulk density were collected from predetermined points within the 6 m wide skidding corridor [11]. Therefore, some of the increase in bulk density between 3 and 7 cycles is attributed to increased areal coverage of the skidding corridor. By seven cycles, the traffic pattern had stabilized into a well-defined skid trail along the 40 m transect. Bulk density increases faster when compaction is only measured in wheel tracks [27].

Site was a significant factor (*p* < 0.0000, Table 4) because increases in bulk density from three cycles of skidding were only significant at six of the nine sites. After 12 cycles of skidding, the average increase in bulk density when trafficking at the 5, 10, and 20 cm depths were 0.161, 0.106, and 0.084 Mg m<sup>−</sup>3, respectively. The percentage increases in bulk density with a soil depth were 14.2%, 8.2%, and 6.2%, respectively. The low increase in bulk density is partly attributed to the use of wider tires on the pressure harvesting machines (Table 1). Compaction of the wettest of these soils reduced their low air-filled porosity until the soils were essentially saturated at the higher skidding cycles [11]. Essentially saturated soils contained trapped air at a high degree of saturation. Trapped air prevents a further increase in bulk density because water is incompressible and trafficking does not allow time for the water in the soil to drain [11,37]. Sites 6 and 8 were not significantly compacted

and had soil water potentials lower than −25 kPa (Table 1). The increase in bulk density for these soils was less than 6% despite the soils having an average air-filled porosity of 0.32 m<sup>3</sup> m<sup>−</sup><sup>3</sup> after three cycles [11].

**Figure 4.** Bulk density measured immediately after harvesting and skidding (year 0) as a function of the number of skidding cycles.

#### *3.3. Postharvest Changes in Bulk Density*

Changes in bulk density occurred in the 7 years postharvest (Figure 5). The first change was a significant increase in bulk density after 1 year. Year of sampling was a significant source of variation (*p* < 0.0000, Table 3) and was consistent across all levels of trafficking, including the harvest-only control. The interactions of traffic cycles and site, and trafficking cycles and year were not significant.

**Figure 5.** Mean bulk density of harvest-only (control) and 3, 7, and 12 skid cycles of soil trafficking over a period of 7 years. Each value of bulk density is the mean for the 5, 10, and 20 cm depths measured at nine sites (*n* = 144; Table 2).

The significant increase in bulk density after 1 year was approximately 0.03 Mg m<sup>−</sup><sup>3</sup> (Figure 5). This is a 2.4% average increase in bulk density, which is about 25% of the original increase as a percentage. The postharvest increase in bulk density of trafficked soil persisted for three winters when the snowpacks were deep (Figure 3). The bulk density samples for year 4 were collected in the fall of 1997 prior to the first of four winters when the snowpack was thin. By year 7, the bulk density of the three trafficked treatments had decreased to values similar to those measured immediately following trafficking. During the intervening time period, these sites had potentially undergone a minor amount of freezing (Figure 3). However, the bulk density of the harvest-only soil had not returned to the values measured immediately after harvesting.

The site and year interaction (*p* < 0.0000) and the depth and year interaction (*p* < 0.0027) were both significant (Table 3). Sites 1 and 8 had the greatest year-to-year variation (Table 2). The year-to-year variation in soil wetness was greater on these sites than the other sites. The sustained wetness on some sites also caused changes in soil morphology and decreased soil drainage class within the first 3 to 4 years [14], the time period when two-thirds of the data were collected. Wet or dry soils can sometimes make it more difficult to consistently collect high-quality soil cores with a standardized protocol [28,30]. Field staff also reported that wet soil made the collection of core samples on some sites more difficult.

The interactions that are not statistically significant are also relevant to understanding the dynamics of natural recovery processes. Traffic cycles and year (*p* < 0.1976), and traffic cycles, depth, and year interactions (*p* < 0.7903) are most notable (Table 4). Their lack of significance indicates that soil to a sampling depth of 22 cm behaved as a homogeneous unit over the entire 7-year period. This is a further indication that the soil freezing, if it occurred, was ineffective, and postharvest soil biological processes were not sufficiently effective to loosen the soil compacted by skidding at any depth.

#### *3.4. Measuring Bulk Density*

The mean of the standard errors for bulk density for the three depths in Table 2 is 0.039 Mg m<sup>−</sup><sup>3</sup> with a standard deviation of 0.0149 Mg m<sup>−</sup><sup>3</sup> (Table 3). The median value is 0.037 Mg m<sup>−</sup>3. The only obvious tread in standard deviation was values decreasing with depth (0.046, 0.039, and 0.033 Mg m<sup>−</sup><sup>3</sup> for depths 5, 10, 20 cm, respectively). The inclusive value for kurtosis is 13.39, and the data are positively skewed to the right (2.54) for all depths. Hence, the variability of the values of bulk density is not described by a bell-shaped distribution curve, but by the widening of the lower portion of the curve [38] (Figure 6). Standard errors > 0.065 Mg m<sup>−</sup><sup>3</sup> represent 3.6% of the samples. Half of these values, six of the seven values > 0.095 Mg m<sup>−</sup>3, and 75% of all values occurred on Site 9 in year 3 (Table 2). Site 9 was the only site where a small amount of gravel was encountered [11]. Nevertheless, the mean standard error at Site 9 for year 0 was 0.042 Mg m<sup>−</sup>3, including a maximum value of 0.069 Mg m<sup>−</sup>3. At that time, a small amount of gravel was not regarded as a problem or likely to have a measurable effect on the value of bulk density. The sampling protocol required that sampling rings be pressed into the soil by body weight. Roots were normally the dominant factor causing a high discard rate of potential bulk density sampling points. At Site 9, small gravel caused more sample locations to be abandoned or cores discarded during trimming. Nevertheless, a temporary relaxation of the commitment to the sampling protocol on a difficult-to-sample site is most likely responsible for this one-time deviation. This deviation reinforces the need for a strict adherence to a sampling protocol.

**Figure 6.** Histogram of all the standard errors reported in Table 2. Seven values between 0.095 and 0.150 g m<sup>−</sup><sup>3</sup> are not shown. Each value of standard error is based on 16 individual soil cores (Table 2).
