*2.1. Study Site*

#### 2.1.1. Location

The study was carried out at two sites in Kochi University Forest, which is located at the northwest boundary of Kami city, Kochi prefecture, Japan (133◦36- 32.0-- E, 33◦42- 12.8-- N). It has an area of 124 ha and is composed of eight compartments, which are divided into two blocks: the west and the east blocks (Figure 1). The altitude ranges from 660–1045 m. Coniferous planted forests and natural broad-leaved forests account for 60% and 40% of the vegetation, respectively, which is dominated by *Cryptomeria japonica* and *Chamaecyparis obtusa*, and *Quercus acuta* and *Q. serrata*, respectively [28]. All the forest roads have a width of 3.0 m, including shoulders, and are unpaved. The spur roads were constructed for forest management and logging operation performed by an 8-t class excavator-based machine (CAT 308D CR) and a 3-t class forwarder (IWAFUJI U3) [29]. The average annual precipitations from 2013 to 2021 at the two nearby meteorological stations, Motoyama town (6100 m to NNW) and Shigeto in Kami city (7900 m to ESE), were 3096 ± STD 637 mm and 3778 ± STD 757 mm, respectively [30].

Two routes were selected for the study: Sites 1 and 2. Site 1 was constructed in FY (fiscal year) 2013–2016 and Site 2 was constructed in FY 2019–2021. Field investigations of Sites 1 and 2 were carried out in 2017 and 2021, respectively. The coordinates of the routes were surveyed using a survey compass with a transit telescope for Site 1 and using a single band GNSS receiver [31] for Site 2.

The entire Kochi University Forest is in the Chichibu geology belt, which consists of an accretionary wedge sedimented from the Carboniferous period to the Jurassic period and is located north, adjacent to the Shimanto geology belt. The detailed geological features are as follows: geological period, Mesozoic–Early Cretaceous–Albian period to Cenozoic–Paleogene–Oligocene–Selandian period; lithology and muddy millstone or metamorphic chert; metamorphism, high P/T wide-area metamorphic rocks, and green mudstone belt [32]. Although many areas are prone to landslides in the Sanbagawa geology belt in Kochi prefecture, which is north adjacent to the Chichibu belt, there are few such areas in both Chichibu and Shimanto geology belts in Kochi prefecture [33]. From the terrain analysis of Yamasaki et al. [34], the average slope gradient is 25–30◦, the average height difference within a 500-m circle is 300–400 m, and the contour-round-about factor is moderate for the Kochi University Forest area, resulting in moderate to highly difficult terrain within the Kochi prefecture [34].

**Figure 1.** Study site. Site 1: A spur road constructed in FY 2013–2016 in the west block, Site 2: A spur road constructed in FY 2019–2021 in the east block of Kochi University Forest.

## 2.1.2. Spur Road Construction Method

The spur roads were constructed by three university forest officers using an 8-t class excavator (CAT 308D CR; bucket capacity, 0.28 m3). They had 24, 8, and 6 years of experience at the time of Site 1 spur road construction in 2013. They learned the Shimanto method in some training courses held outside the university and then commenced spur road construction within the university forest on their own around the 2010s.

Figure 2 shows the spur road construction in Kochi University Forest using the Shimanto method. The soil is dug over once with additional depth, then compacted with 20–30-cm-thick layers. An example of the cross-section of the middle-grade slope in Figure 2a is 20◦ of natural ground. On a steeper natural ground slope (Figure 2b, e.g., 30◦), the embankment volume and fill slope length tend to be larger. In such cases, the center line of the spur road is shifted to the uphill side in order to decrease the embankment volume (Figure 2c). The cut slope can be constructed as vertical if its maximum height is less than the limit height of 1.4 m.

Figure 3 shows a practical example of spur road construction in Kochi University Forest during a training course held for university forest officers of other universities in 2012. The excavator dug over the soil for extra depth (Figure 3a). Then, the loose soil was mixed using the bucket, spread over the designated road width (Figure 3b), and compacted by the crawler (Figure 3c). The compaction was carried out for a few layers with a thickness of 20–30 cm each up to the level of the existing road height (Figure 3d).

**Figure 2.** Spur road construction in Kochi University Forest using the Shimanto method. The soil was dug over once with additional depth, and then compacted with layers of 20–30 cm thickness. (**a**) Example of the cross-section of the middle-grade slope (e.g., 20◦) natural ground. (**b**) On a steeper natural ground slope (e.g., 30◦), the embankment volume and fill slope length tend to be larger. (**c**) In such cases, the center line of the spur road is shifted to the uphill side in order to decrease the embankment volume. The cut slope can be constructed as vertical if its maximum height is less than the limit height of 1.4 m.

**Figure 3.** Example of spur road construction in Kochi University Forest during a training course. The excavator dug over the soil for extra depth (**a**). Then, the loose soil was mixed using the bucket, spread over the designated road width (**b**), and compacted by the crawler (**c**). The compaction was carried out for a few layers with a thickness of 20–30 cm each up to the level of the existing road height (**d**).

#### *2.2. Field Survey*

2.2.1. Measurement of Roadbed Strength

The roadbed soil strength was tested using a simple dynamic cone penetrometer. There are some types of dynamic cone penetrometers for in situ soil strength measurement in Japan [35], varying in terms of the cone size, hammer mass, and height of a single strike with a hammer. Using a dynamic cone penetrometer, the strength of the soil is expressed as the number of hammer strikes required to penetrate a certain depth of the soil. Umeda et al. [36] standardized the strength values of three penetrometers. Yoshinaga et al. [37] concluded that the in situ density of underground soil can be estimated with the measured strength of a simple dynamic cone penetrometer called the Doken-type penetrometer. Gotou et al. [38] verified the roadbed strength of spur roads and forest roads using a Doken-type simple dynamic cone penetrometer. In the present study, a Doken-type simple dynamic cone penetrometer was used for the measurement of the roadbed strength. The Doken-type penetrometer was developed for in situ measurement on steep slopes with a lighter hammer (5 kg) for convenience rather than the other types [39]. Its measuring procedure has been standardized by the Japan Geotechnical Society, Committee for Standard of Geotechnical Survey, (Japanese Industrial Standards)-1433-2012 [39].

Figure 4 shows the standardized design of the Doken-type dynamic penetrometer (Figure 4a) [39]. The soil strength is measured by striking the hammer at a height of 500 mm (Figure 4b,c) with the road kept vertical. By measuring the penetration depth of each strike or a few strikes, the penetrated depth is converted into an *N*<sup>d</sup> (or *N*c) value, which represents the number of strikes required for a penetration depth of 10 cm.

**Figure 4.** Doken-type dynamic penetrometer. Standardized design of the Doken-type dynamic penetrometer (**a**) (the Japan Geotechnical Society, Committee for Standard of Geotechnical Survey 2014 [36]). The soil strength is measured by striking the hammer at a height of 500 mm (**b**–**d**) with the road kept vertical. By measuring the penetration depth by each strike or a few strikes, the penetrated depth is converted into an *N*<sup>d</sup> value, which represents the number of strikes required for a depth of penetration of 10 cm.

The *N*<sup>d</sup> value is defined as follows:

$$N\_{\rm d} = 10 \text{ (N/}\Delta\text{h)},\tag{1}$$

where Δ*h* is the penetrating depth of *N* strikes. In general, Δ*h* is 10 cm; however, for softer soil, Δ*h* was set to one strike. and for harder soil, Δ*h* was set to less than 10 cm. The maximum penetrating depth was set to 100 cm because some research reports have pointed out that the friction of the rod cannot be ignored when the penetrating depth exceeds 100 cm [38–40]. In some cases, the *N*<sup>d</sup> value was too large, implying that there was a large rock just under the penetrating point. In such cases, we moved the penetrating point 20–0 cm apart from the original point.

Thresholds of the *N*<sup>d</sup> values were determined for judging the strength requirements as follows. Okimura and Tanaka [41] predicted the potential sliding layer depth in forested areas of sandy soil caused by heavy rain as *N*<sup>10</sup> < 12. The *N*<sup>10</sup> value is measured using a handy dynamic cone penetrometer termed the Kobe University type, which provides nearly the same value as the Doken-type penetrometer. Sugiyama et al. [42] reported that the average *N*<sup>d</sup> value of a collapsed railway embankment was 5.4. Ogawa [43] studied the relationship between slope surface failure and water content variation on the forested slope and found that most of the sliding depth was 0.5–1.5 m, of which the *N*<sup>d</sup> values ranged from 5–10. Hiramatsu and Bitoh [44] analyzed a landslide area on the Chichibu geology belt in Yusuhara town, Kochi prefecture, and proposed a threshold *N*<sup>d</sup> value of 9 to estimate the slippage depth of the soil layer. Koyama [45] conducted a series of field investigations on spur roads using a simpler dynamic penetrometer called the Tottori FK-type penetrometer and found that cracks appeared at the road shoulders if the *N*<sup>d</sup> equivalent values were less than 1.4 with a penetrating depth of up to 25 cm. For these preceding achievements, we set the following three thresholds: *N*<sup>d</sup> = 1.4 as "very weak" for clacks on shoulders, *N*<sup>d</sup> = 5.4 as "weak" for the collapse of the embankment, and *N*<sup>d</sup> = 10.0 as "moderate" for natural slope failure.

#### 2.2.2. Measurement of Bearing Capacity

The California bearing ratio (CBR) is a widely used parameter for assessing the bearing capacity of a road subsurface. While proper CBR values are obtained by testing standardized sample soils using appropriate instruments in a laboratory, in situ CBR values are obtained using a purpose-built instrument. Therefore, although the CBR value was originally for the construction of public roads, it is often employed for assessing the bearing capacity of unpaved forest roads and spur roads using in situ CBR testers [46–49].

Kobayashi and Fukuda [46] measured in situ CBR values of spur roads using a sphere-shaped rammer type in situ CBR tester and obtained the average value of 18.9%. Kobayashi et al. [47] investigated two spur road routes constructed using the Ohashi method with an in situ CBR tester called CASPFOL, and obtained average values of 14.7% and 23.4%. Suzuki et al. [48] evaluated the total strength of a spur road constructed using the Shimanto method in a volcanic ash soil area. For the bearing ratio, they used a CASPFOL tester and obtained in situ CBR values of 2.5–9.8% for ruts on the cut slope sides and 8.0–12.5% for ruts on the fill slope sides. Sawaguchi et al. [49] investigated the age change of the bearing capacity for spur roads in Neogene period gravelly soils for the Tohoku area, northern Japan, using a CASPFOL tester. They found that the CBR values increased from 1.4% to 4.0% at the shoulder, from 1.7 to 5.6% at the center, and from 2.7 to 7.8% at the ruts over 10 years, even if there was nearly no traffic during the gap years [49].

In the present study, we used an in situ CBR tester called CASPFOL (Figure 5a) [50,51]. The soil strength is measured by striking the rummer at a height of 500 mm with the road kept vertical through the built-in acceleration meter (Figure 5b,c). The acceleration of the rummer is measured with the instrument when the rummer falls and strikes the soil. The standard measuring method requires five measurements around a designated point with a radius of 20 cm (Figure 5d).

**Figure 5.** In-situ CBR measuring instrument (CASPFOL). Schematic diagram of CASPFOL (**a**) [51]. The acceleration of the rummer is measured by the instrument when the rummer falls and strikes the soil (**b**,**c**). The soil strength is measured by striking the rummer at a height of 500 mm with the road kept vertical through the built-in acceleration meter. The standard measuring method requires five measurements around a designated point with a 20 cm radius (**d**).

#### 2.2.3. Soil Property

The grain size distribution has significant effects on soil properties, especially mechanical features, such as compatibility [21]. Soil samples were obtained from a cut slope at two points each for the two sites. The samples were tested for grain size analysis by following the procedure of JIS-A-1202 [52]. Soil property as engineering materials is classified as one of 24 engineering soil types through a standard classification system of the Japanese Geotechnical Society using the parameters derived by the grain size distribution and accompanying tests for the grain size analysis procedures. An additional property, i.e., wideness of size distribution, is assessed using two parameters, the uniformity coefficient *U*c and coefficient of curvature *U*c - , which are defined as follows:

$$
\Delta I\_{\rm c} = D\_{60} / D\_{10\prime} \text{ and} \tag{2}
$$

$$\left(\text{II}\_{\text{c}}\right)' = (\text{D}\_{\text{30}})^2 / (\text{D}\_{\text{10}} \times \text{D}\_{\text{60}}),\tag{3}$$

where *D*10, *D*30, and *D*<sup>60</sup> are the grain sizes of the accumulated grain mass percentages of 10%, 30%, and 60%, respectively [52]. If *U*<sup>c</sup> ≥ 10 and 1 < *U*<sup>c</sup> - < 3, then the soil is supposed to have wide grain size distribution, which means that it is suitable for compaction [52].

#### 2.2.4. Experimental Design and Analysis

The first series of the investigation was conducted for Site 1 for the measurement of the roadbed strength. To check the equality of the compacted roadbed soil strength around the road width, Factor A was set as a relative position within the road width (Figure 6) with five levels: A1, downhill side apart from the shoulder at least 30 cm; A2, rut of downhill side; A3, center; A4, rut of uphill side; and A5, uphill side apart from the toe of the cut slope at least 30 cm. Factor B was set to the fiscal year of construction in order to check the age-related change in the strength. The spur road of Site 2 was constructed from 2013 to 2016. Because the constructed length of FY 2013 was nearly twice that of the other years, the FY 2013 constructed section was separated at its mid-point into two sections. Therefore, the levels of Factor B are B1, FY2013-1; B2, FY2013-2; B3, FY2014; B4, FY2015; and B5, FY2016. The spur roads were constructed from fall to winter, and Site 1 was investigated in the

winter of FY 2017. Hence, the number of years elapsed were 4.2, 4, 3, 2, and 1 for B1, B2, B3, B4, and B5, respectively. It should be noted that the number of years elapsed could not be treated as the effect of time on each road segment in a precise sense. However, in the analysis, it was assumed that the number of years elapsed had a meaning of elapsed time by assuming that each road segment has a homogeneous property. Nevertheless, the effect of the number of years elapsed should be carefully evaluated in this regard.

**Figure 6.** Measuring points on the spur road at each sample point of Site 1.

The number of repetitions of the measurement was set to five so that the distance between every two points would be the same. However, because of the slightly shorter section length of A3, the number of repetitions of A3 was four. Therefore, 24 points were investigated for Site 1 in total.

The obtained data were analyzed by two-way ANOVA [53,54]. The dependent variable was the *N*<sup>d</sup> value up to a depth of 100 cm average weighted by Δ*h*. Two independent variables were Factors A and B. However, in some points, the soil was too hard to reach a depth of 100 cm. When *N*<sup>d</sup> exceeded 50 or more, the penetration was stopped, and the *N*<sup>d</sup> value was averaged up to the maximum depth. One dummy data was added to the A3 level as the mean of four observations so that the number of repetitions is equal to the other levels of A, i.e., five. In the ANOVA table, the degree of freedom was adjusted for Factor A and the error concerning the dummy observations [53].

The second series of the investigation was carried out in Site 2 for the roadbed strength and bearing capacity. As for the roadbed strength, to compare the strength difference between natural soil and compacted soil as well as that between the downhill side and the uphill side, Factor A was set to the distance from the road center and the Factor B direction (Figure 7). The levels of Factor A were as follows: A1, 0.5 m from the center; A2, 1.5 m from the center; and A3, 2.5 m from the center. The points of A1 were nearly on the ruts. The distance between A1 and A2 was, in some cases, more than 1.0 m so that A2 was not on the fill slope. The distance between A2 and A3 was 1.0 m. The factor levels are B1, downhill side, and B2, uphill side.

**Figure 7.** Measuring points on the spur road at each sample point of Site 2.

Factor C was set to the fiscal year of construction. The Site 2 spur road was constructed in 2019, 2020, and 2021. The section distance of 2021 was nearly twice that of the others. Therefore, section FY2021 was separated into two sections at the mid-point. To check the strength difference between the upper and lower parts, Factor D was set to the penetrating depth, i.e., the level D1 was 0–50 cm, and D2 was 50–100 cm. The *N*<sup>d</sup> values were averaged for 0–50 cm and 50–100 cm, respectively. The number of repetitions was two. In Site 2, two points each were investigated at each section of Factor A, at a mid-point of the section and separated by 2 m, because the section length was not as long as that of Site 1. The total number of roadbed strength measurement points was 3 × 2 × 4 × 2 × 2 = 96. The obtained data were analyzed via four-way ANVOVA [53,54]. The dependent variable was the *N*<sup>d</sup> value averaged weighted by Δ*h*. Four independent variables were Factors A, B, C and D.

As for the bearing capacity, in situ CBR measurement, three factors were set. The first two factors, Factors B and C, were set to be the same as those of the roadbed strength. The third factor, Factor R, was set to repetition, having two levels of R1 and R2. Although repetition was not treated as a factor in general, it was included to check for the homogeneity of the measurement. The total number of in situ CBR measurement data was 2 × 4 × 2 × 5 = 80 because five measurements each were required for the standard measurement of the in situ CBR tester. The obtained data were analyzed via three-way ANVOVA [53,54]. The dependent variable was the in situ value. The three independent variables were Factors B, C, and R.

The multi-way ANOVA was analyzed with SAS software using the super-computing service of the Institute for Information Management and Communication, Kyoto University. The two-way ANOVA and the following Tukey test for multiple comparisons [54] were carried out using Microsoft Excel software.

#### 2.2.5. Date of Investigation

The investigation of Site 1 was carried out from November 2017 to January 2018. Surveying of the spur road route and road section measurement using a surveying pole with a precision of 0.1 m were conducted simultaneously.

The investigation of Site 2 was carried out from November 2021 to January 2022. The surveying of the spur road route and road section measurement with a surveying pole with 0.1 m precision were conducted simultaneously.

The soil samples were collected at the sections FY2013 and FY2015 of Site 1, the boundary of sections FY2019 and FY2020 (FY2019–2020), and the first half of FY2021 of Site 2. The FY2015 and the FY2019–2020 samples were collected in December 2020 and analyzed from December 2020 to February 2021. The FY2013 and FY2021 samples were collected in December 2021 and analyzed from December 2021 to January 2022.

#### **3. Results and Discussion**

#### *3.1. Overview of the Sites*

The total spur road lengths of the sites were 626 and 250 m for Sites 1 and 2, respectively (Tables 1 and 2). The route of Site 1 had a steep gradient in the section of FY2015, whereas the other sections were nearly flat or had a gentle gradient. In Site 2, the route extended upward; hence, its gradient was relatively high. The average road widths were nearly the same for both sites while the Site 1 route was slightly wider (3.0 m) than the Site 2 route (2.8 m).

**Table 1.** Descriptive statistics of Study Site 1: a spur road in the west block of Kochi University Forest.


Fiscal year of survey: 2017 (November 2017–January 2018).

**Table 2.** Descriptive statistics of Study Site 2: a spur road in the east block of Kochi University Forest.


Fiscal year of survey: 2021 (November 2021–January 2022).

The slope gradients of adjacent natural ground were estimated by the connecting fill slope toes and top ends of cut slopes. The average slope gradients of the adjacent natural ground were 26.0 ± STD 6.5◦ for Site 1 and 21.7 ± STD 2.6◦ for Site 2. Such a steep natural ground slope of Site 1 resulted in a higher cut slope height (1.3 ± STD 0.4 m) compared to that of Site 2 (0.9 ± STD 0.3 m). This implies that the excavation volume as well as the dig over depth of Site 1 would be larger than those of Site 2 (Figure 2).
