Determining Forest Duff Water Content Using a Low-Cost ... - MDPI

sensors Article

Determining Forest Duff Water Content Using a Low-Cost Standing Wave Ratio Sensor Xiaofei Yan 1 , Yajie Zhao 1 , Qiang Cheng 2 , Xiaoliang Zheng 1 and Yandong Zhao 1, * 1 2

*

School of Technology, Beijing Forestry University, Beijing 100083, China; [email protected] (X.Y.); [email protected] (Y.Z.); [email protected] (X.Z.) College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, China; [email protected] Correspondence: [email protected]

Received: 8 January 2018; Accepted: 13 February 2018; Published: 22 February 2018

Abstract: Forest duff (fermentation and humus) water content is an important parameter for fire risk prediction and water resource management. However, accurate determination of forest duff water content is difficult due to its loose structure. This study evaluates the feasibility of a standing wave ratio (SWR) sensor to accurately determine the forest duff water content. The performance of this sensor was tested on fermentation and humus with eight different compaction levels. Meanwhile, a commercialized time domain reflectometry (TDR) was employed for comparison. Calibration results showed that there were strong linear relationships between the volumetric water content (θ V ) and the SWR sensor readings (VSWR ) at different compaction classes for both fermentation and humus samples. The sensor readings of both SWR and TDR underestimated the forest duff water content at low compacted levels, proving that the compaction of forest duff could significantly affect the measurement accuracy of both sensors. Experimental data also showed that the accuracy of the SWR sensor was higher than that of TDR according to the root mean square error (RMSE). Furthermore, low cost is another important advantage of the SWR sensor in comparison with TDR. This low-cost SWR sensor performs well in loose materials and is feasible for evaluating the water content of forest duff. In addition, the results indicate that decomposition of the forest duff should be taken into account for continuous and long-term water content measurement. Keywords: standing wave ratio; forest duff; volumetric water content; compaction

1. Introduction The forest floor plays a significant role in forest hydrological processes by affecting water and energy transfer between the sub-canopy atmosphere and the mineral soil [1–3]. It can be divided into the surface litter layer (Oi), with freshly fallen plant residue, the sub-surface fermentation layer (Oe), with partially decomposed but still recognizable organic material, and the underlying humus layer (Oa), with well decomposed organic matter [4,5]. Fermentation and humus are collectively known as forest duff. The retention of rainwater by forest duff not only influences the flow into the mineral soil but also provides an important source of water to vegetation [6,7]. Moreover, the moisture status of the forest duff is a critical determinant for fire ignition and spread [8,9], thus understanding the patterns of forest duff water content variation across time and space would be of great benefit to fire risk management and prevention [10]. In forest ecology, forest duff moisture strongly impacts leaf litter decomposition rates, which ultimately affects carbon cycling and microbial activity [11]. Although the importance of forest duff water content has been proved in many different fields, it has received less attention in forest ecosystem research compared with the moisture dynamics in mineral soil. The reason is that it is difficult to keep good contact between the sensor probe and the forest duff due to its loose structure and low density [12,13]. Sensors 2018, 18, 647; doi:10.3390/s18020647

www.mdpi.com/journal/sensors

Sensors 2018, 18, 647

2 of 11

Gravimetry is the most accurate method to determine water content, but the processes of destructive sampling and oven drying are very time-consuming and labor intense. The neutron scattering method can be used to measure water content in porous media with the advantages of being non-destructive and quite accurate [14]. However, this method may result in a hazard for human health [15]. Time domain reflectometry (TDR) is a promising method to determine the forest duff water content in a rapid and accurate way [16]. Börner et al. [12] placed one TDR probe at the interface of the forest floor layer and the mineral soil to measure their average water content, meanwhile the second TDR probe was employed to measure the water content of the mineral soil only. Then the forest floor water content could be calculated by the two-probe configuration. The results showed that this method would underestimate the water content at low moisture contents and overestimate it at high moisture contents. Canone et al. [17] designed a TDR probe with eight blades evenly spaced around the central rod, which would permit more homogeneous spatial distribution of energy and maintain good contact with the surrounding forest floor in the water content measurements. Experimental results indicated that the developed TDR probe performed better than the classic three-rod probe. In addition, a capacitance sensor is an alternative non-destructive technique to continuously measure the forest floor water content [18–20]. The natural processes in earthworm-free forests usually keep the forest floor loose and uncompacted. It is well known that forest duff, as a type of loose spongy material, has the characteristics of low density and high porosity, and the lower forest duff tends to have greater compaction and slower moisture loss rates [21]. The distinct spatial variability of forest floor compaction can significantly affect the contact between the sensor probe and the surrounding forest duff, which leads to a large deviation in the water content measurement. However, to the best of our knowledge, no study has been reported examining the effects of the compaction on forest duff water content measurements. The main objective of this study was to evaluate the performance of a standing wave ratio (SWR) sensor to determine the forest duff water content, and evaluate the accuracy of fermentation and humus moisture measurement at eight different compactions. The behavior of the SWR sensor was compared with the performance of a commercial TDR sensor. Fermentation, humus and mineral soil with different water contents were tested respectively under laboratory conditions. 2. Materials and Methods 2.1. Water Content Sensors The standing wave ratio (SWR) sensor, also known as an impedance sensor or amplitude domain reflectometry, consisted of a 100 MHz sinusoidal oscillator, a coaxial transmission line, and a stainless steel sensing probe (Figure 1). The high frequency measurement (higher than 30 MHz) can minimize the measurement uncertainty induced by the solutes in water [22,23]. The electromagnetic wave provided by the sinusoidal oscillator spreads along the coaxial transmission line into the probe, and if the impedance of the probe is different from that of the transmission line, a proportion of the incident wave would be reflected back along the line to the oscillator. As a result of the reflected wave interfering with the incident wave, a voltage standing wave would be set up on the transmission line. Therefore, voltage difference (∆U) can be measured at both ends of the transmission line, and it can be expressed as a function of the probe impedance [23]: ∆U = 2A

ZP − Z L ZP + Z L

(1)

where A is the electromagnetic wave amplitude (V), ZP is the probe impedance (Ω), and ZL is the transmission line impedance (Ω), which is dependent on its physical dimensions and dielectric constant of the insulating material. ZL is equal to 50 Ω in this research. Moreover, ZP is determined by the dielectric constant of the surrounding material of the probe because its geometry has been immobilized. In other words, ZP is decided by the water content when the fixed probe is inserted into some porous

Sensors 2018, Sensors 2018, 18,18, 647x FOR PEER REVIEW

3 of 3 of 11 11

immobilized. In other words, ZP is decided by the water content when the fixed probe is inserted medium. Thus, we can quantify the forest floor or mineral moisture measuring voltageby into some porous medium. Thus, we can quantify the soil forest floor orbymineral soil the moisture difference (∆U) of the transmission line. As shown in Figure 2a, the length and diameter of the measuring the voltage difference (ΔU) of the transmission line. As shown in Figure 2a, the SWR length sensor developed work weredeveloped 16 cm andin5 this cm, work respectively. hafand four5 stainless steel rodsItof and diameter of in thethis SWR sensor were 16Itcm cm, respectively. haf approximately mm diameter and 5 cm length. rods and were5 evenly disposed a cylinder four stainless5.8 steel rods of approximately 5.8 mmThree diameter cm length. Threealong rods were evenly and one in the center. Furthermore, distance the central anddistance the otherbetween three rods disposed along a cylinder and the onesame in the center.between Furthermore, therod same the was approximately 19other mm from to center. All rods were inserted in a brass support form a central rod and the threecenter rods was approximately 19 mm from center to center. Alltorods were coaxial probe. inserted in a brass support to form a coaxial probe.

Figure 1. Diagram of the standing wave ratio (SWR) sensor and measurement circuit. Ua and Ub, Figure 1. Diagram of the standing wave ratio (SWR) sensor and measurement circuit. Ua and Ub , output voltages of point a and point b on the coaxial transmission line, respectively; ΔU, differential output voltages of point a and point b on the coaxial transmission line, respectively; ∆U, differential output voltage of point a and point b. output voltage of point a and point b.

TDR can determine the dielectric constant of materials surrounding the probe to measure the TDR canwater determine theThe dielectric constant of materials the cable probetotothe measure the volumetric content. electromagnetic wave travelssurrounding down a coaxial TDR probe. volumetric water content. electromagnetic wave down coaxial cable the TDR probe. Part of the wave would The be reflected in the joint oftravels the cable anda the probe duetoto the impedance Part of the wave would be reflected in the joint of the cable and the probe due to the impedance difference, and the remainder of the wave would propagate through the probe and return to the difference, of the wave would propagate probe and return to the source source atand the the endremainder of the probe. The round-trip time of thethrough wave isthe described as [24]: at the end of the probe. The round-trip time of the wave is described as [24]:

2L 0.5 t 2Lε0.5 c t=

(2) (2)

c where t is the round-trip time (s); L is the TDR probe length (m); ε is the dielectric constant of the where t is the time (s); is the TDR probe length (m);space ε is the constant medium, andround-trip c is the velocity of Lelectromagnetic wave in free (3 ×dielectric 108 m s−1). Due to of thethe fact 8 − 1 medium, and c is the velocity of electromagnetic wave in free space (3 × 10 m s ). Due to the fact that the dielectric constant of free water (≈80) is significantly greater than that of dry solid wood that the dielectric constant free water ≈80) is significantly thanofthat of dry solid (2~5), (2~5), dry soil (3~8), andofair (≈1), the (changes in dielectricgreater constants forest floor andwood mineral soil dry soil (3~8), and air ( ≈ 1), the changes in dielectric constants of forest floor and mineral soil may may be predominantly attributed to water content [25–27]. Consequently, the round-trip time bewould predominantly attributed towater watercontent contentof [25–27]. Consequently, the round-trip timeInwould be be proportional to the the material surrounding the probe [28]. this work, proportional to the water content of the material surrounding the probe [28]. In this work, a portable a portable TDR probe (TRIME-PICO 64, IMKO, Ettlingen, Germany) equipped with a data-reader TDR probe (TRIME-PICO IMKO, Germany) a data-reader (HD2,As IMKO, (HD2, IMKO, Germany)64, was used Ettlingen, to measure the forestequipped floor andwith mineral soil moisture. shown Germany) usedTDR to measure thetwo forest floor and moisture. shown in Figure in Figurewas 2b, The probe has stainless steelmineral rods of 3soil mm diameterAs and 114 mm length.2b, The thespace TDRof probe has is two steel rodstoofcenter. 3 mm Water diameter and 114 mm length. The of theby rod the rod 20 stainless mm from center content measurements werespace recorded the is data-reader 20 mm fromin center to center. Water content measurements were recorded by the data-reader in the 3 −3 the unit of volumetric water content (cm water × cm forest floor or mineral soil). unit of volumetric water content (cm3 water × cm− 3 forest floor or mineral soil).

Sensors 2018, 18, 647

4 of 11

Sensors 2018, 18, x FOR PEER REVIEW

4 of 11

Figure 2. Photograph of the probes used: (a) the SWR probe; (b) the commercial time domain Figure 2. Photograph of the probes used: (a) the SWR probe; (b) the commercial time domain reflectometry (TDR) probe. reflectometry (TDR) probe.

2.2. Potential Uncertainty Due to Temperature Effects on Dielectric Sensor Readings 2.2. Potential Uncertainty Due to Temperature Effects on Dielectric Sensor Readings Temperature effects on forest duff water content measurements using dielectric sensors Temperature effects on forest duff water content measurements using dielectric sensors stemming stemming from the temperature-dependent response of the sensor circuit itself and the temperature from the temperature-dependent response of the sensor circuit itself and the temperature dependence dependence of the dielectric properties of the medium (e.g., the water in the forest duff). The former of the dielectric properties of the medium (e.g., the water in the forest duff). The former has been has been characterized by Saito et al. (2009) [29]. In this study, a potential uncertainty due to characterized by Saito et al. (2009) [29]. In this study, a potential uncertainty due to temperature temperature effects on duff permittivity was considered by combining a permittivity model effects on duff permittivity was considered by combining a permittivity model developed for porous developed for porous materials [30]: materials [30]: α εα = ϕ (3) llεll(TT)  +ϕssεs sα+ ϕaaεaaα (3) ϕl+ϕs +ϕa = =11 l

s

a

(4) (4)

with a temperature-dependent equation describing the permittivity of free water [31]: with a temperature-dependent equation describing the permittivity of free water [31]: 2 −8 3 3 55 ( T − 25)2 − 2.80 × 10 ε l (T )T =78.54 [1 − 4.58 × 10− ( TT−2525 )+  78.54[1  4.58 1.19× 10 2525  1.19 T  25   2.80  108 T(T −  ]) ] l

(5) (5)

Equations (3)–(5) were coupled to estimate the potential uncertainty caused by the temperature dependence of the permittivity, ϕ is of the the dielectric dielectricproperties propertiesofofthe themedium. medium.InInthese theseequations, equations,ε is ε is the permittivity, ϕ of is the volumetric fraction of each component in the duff,duff, and the subscripts l, s and of the volumetric fraction of each component in forest the forest andassociated the associated subscripts l, as refer water, component and air, respectively. Besides, α isBesides, an empirical calibration exponent [32] and atorefer tosolid water, solid component and air, respectively. α is an empirical calibration and can be[32] determined bydetermined a method proposed by Roth et al. (1990) [33]. In this study, that exponent and can be by a method proposed by Roth et al. (1990) [33].we In found this study, α = found 0.85 fitted to Equation (5)to with R2 = 0.963. we thatwell α = 0.85 fitted well Equation (5) with R2 = 0.963. 2.3. Field Sampling Sampling 2.3. Field The samples of of fermentation, fermentation,humus humusand andmineral mineralsoil soilwere werecollected collected from a forest ecosystem The samples from a forest ecosystem in ◦ 540 N, 116◦ 280 E), Beijing, China. This park is the property in Jiufeng National Forest Park (39 Jiufeng National Forest Park (39°54′ N, 116°28′ E), Beijing, China. This park is the property of of Beijing Forestry University is used for teaching and scientific research. The climate is a Beijing Forestry University andand is used for teaching and scientific research. The climate is a typical ◦ C and mean typical warm-temperate continental climate with a mean annual temperature of 11.6 warm-temperate continental climate with a mean annual temperature of 11.6 °C and mean annual annual precipitation of 600ofmm, of which 80%during falls during the period from to June to September [34,35]. precipitation of 600 mm, which 80% falls the period from June September [34,35]. The The sampling site is a homogeneous forest where the dominant species is Quercus variabilis and sampling site is a homogeneous forest where the dominant species is Quercus variabilis and the the forest is approximately 12thick cm thick (theOe Oi,and Oe Oa andlayers Oa layers are about 2, 46 and 6 cm forest floorfloor is approximately 12 cm (the Oi, are about 2, 4 and cm thick, thick, respectively). The boundary of the litter, fermentation, humus and mineral soil are flat and respectively). The boundary of the litter, fermentation, humus and mineral soil are flat and abrupt. abrupt. For sampling of the fermentation, the surface mineral soil cm), (0~10the cm), the horizon For sampling of the fermentation, humushumus and theand surface mineral soil (0~10 horizon to be to be sampled was first uncovered with the help of a shovel and knife. All of the samples were sampled was first uncovered with the help of a shovel and knife. All of the samples were collected collected vertically.vertically.

Sensors FOR PEER REVIEW Sensors 2018, 2018, 18, 18, x 647

55 of of 11 11

2.4. Experimental Procedure 2.4. Experimental Procedure The water contents of fermentation and humus were tested using SWR and TDR at eight The water contentsThe of fermentation and humus were using SWR and TDRatat65 eight different different compactions. samples of fermentation andtested humus were oven-dried °C for 48 h, compactions. The was samples of fermentation andfor humus oven-dried at 65 ◦samples C for 48 h, anddifferent mineral and mineral soil oven-dried at 105 °C 24 h.were In order to prepare with ◦ C for 24 h. In order to prepare samples with different compactions, soil was oven-dried at 105similar compactions, a procedure to the Proctor compaction test described by Ayers and Perumpral a procedure similar to[36]. the A Proctor compaction described by Ayers and Perumpral (1982) was (1982) was conducted polyvinyl chloridetest (PVC) cylindrical mold (diameter = 15 cm, height = conducted [36]. A polyvinyl chloride (PVC) cylindrical mold (diameter = 15 cm, height = 21 cm) was 21 cm) was used to pack the sample with a PVC rod (diameter = 5 cm, length = 48 cm, weight = 1 kg) used packhammer. the sample with a PVCwas rodcarried (diameter 5 cm, length the = 48hammer cm, weight 1 kg) asheight the drop as thetodrop Densification out=by dropping at a =constant of hammer. Densification was carried outthe by mold dropping the hammer at a different constant height of 20 cm when 20 cm when sample was placed into in layers. The eight compaction levels of sample was placed into the mold in layers. The eight different compaction levels of fermentation and fermentation and humus samples were achieved by zero, one, three, six, eight, 10, 15 and 20 blows humus samples were achieved by zero, one, three, six, eight, 10, 15 and 20 blows per layer, respectively. per layer, respectively. The experimental procedure was performed as follows. First the samples were oven-dried and Then the the mold mold was was packed packed by by placing placing samples samples in in layers layers with with a the empty mold was weighed. Then predetermined number number of blows blows per layer. The mold with sample was weighed and the volume of predetermined the packed sample was calculated to determine the volumetric After that, the volumetric water water content content (θ (θVV). After prepared sample was measured using a using hand-held digital SC-900 Soil SC-900 Compaction compaction ofofthethe prepared sample was measured a hand-held digital Soil Meter (Spectrum Inc., Plainfield, USA). Moreover, shown in as Figure 3, the water Compaction MeterTechnologies, (Spectrum Technologies, Inc.,IL,Plainfield, IL, USA).asMoreover, shown in Figure content of this sample was sample measured both TDR tests were repeated 3, the water content of this wasby measured by and bothSWR TDR sensors. and SWRAll sensors. All tests were twice on the prepared at the sameat volumetric contentwater and compaction. repeated twice on thesample prepared sample the samewater volumetric content and Furthermore, compaction. the compactionthe and water content conducted were in theconducted same way on Furthermore, compaction andmeasurements water contentwere measurements in samples the sameprepared way on at seven other compactions the same gravimetric water content (θ G ). After thecontent twenty-four samples prepared at seven keeping other compactions keeping the same gravimetric water (θG). tests atthe eight different compactions finished, the gravimetric content of the sample was After twenty-four tests at eightwere different compactions were water finished, the gravimetric water increased by adding a certain amount of deionized water (100 g) in a fine mist form and mixing content of the sample was increased by adding a certain amount of deionized water (100 g) in a fine thoroughly. Then the same experimental to test the compaction water were mist form and mixing thoroughly. Then procedure the same experimental procedureand to test thecontent compaction conducted the eight different compactions. After that, deionized water After was added 100 g intervals and water at content were conducted at the eight different compactions. that, at deionized water until added the sample reached saturation, measurements of compaction content were was at 100 g intervals untiland thethe sample reached saturation, andand thewater measurements of carried out as mentioned above. compaction and water content were carried out as mentioned above.

Figure 3. Photograph of the experimental setup for water content measurement by TDR and SWR. Figure 3. Photograph of the experimental setup for water content measurement by TDR and SWR.

3. Results and Discussion 3. Results and Discussion 3.1. Effect of Compaction on Sensor Readings for Water Content Measurement 3.1. Effect of Compaction on Sensor Readings for Water Content Measurement In order to keep better linear relationships between volumetric water content determined In order to keep better linear relationships between volumetric water content determined gravimetrically (θV) and SWR output voltage (VSWR), the eight different compaction levels of gravimetrically (θ V ) and SWR output voltage (VSWR ), the eight different compaction levels of fermentation and humus samples were divided into three compaction classes: incompact, slightly fermentation and humus samples were divided into three compaction classes: incompact, slightly compacted and strongly compacted. The photograph in Figure 4 presented the same quality of humus samples prepared with incompact, slightly and strongly compacted classes. The range of

Sensors 2018, 18, 647

6 of 11

Sensors 2018, 18, x FOR PEER REVIEW

6 of 11

compacted and strongly compacted. The photograph in Figure 4 presented the same quality of humus samples prepared with incompact, slightly classes.and Thehumus range of compaction compaction measured by SC-900 and the and bulkstrongly densitycompacted of fermentation are shown in measured by SC-900 and the bulk density of fermentation and humus are shown in Table 1. Table 1.

Figure 4. The same quality of humus samples with three compaction classes. The left, middle and Figure 4. The same quality of humus samples with three compaction classes. The left, middle and right right samples were incompact with zero blows per layer, slightly compacted with three blows per samples were incompact with zero blows per layer, slightly compacted with three blows per layer and layer and strongly compacted with 15 blows per layer, respectively. strongly compacted with 15 blows per layer, respectively.

Strong linear relationships between θV and VSWR in the three compaction classes for both the Strong linear relationships between θ V andAs VSWR in the three compaction classes both the fermentation and humus samples were found. shown in Figure 5, the variation of for compaction fermentation and humus samples found. curves, As shown Figure 5, the variation of compaction resulted in noticeable changes in were calibration andinthe responses of VSWR to compaction resulted in noticeable changes in calibration curves, and the responses of V to compaction variation SWR variation showed similar patterns for fermentation and humus in water content measurements. The showed similar patterns for fermentation and humus in water content measurements. The more more the sample was compacted, the lower the volumetric water content was determined at the the sample was compacted, the lower the volumetric water content determined the same readingthe of same reading of SWR. Furthermore, similar results were alsowas observed usingatTDR to measure SWR. Furthermore, similar results were also observed using TDR to measure the water content of the water content of the fermentation and humus in the above compaction classes (Figure 6). Reports in fermentation and humus in the above compaction classes (Figure soil 6). Reports in some of the literature some of the literature indicated that the compaction of mineral is commonly assessed through indicated that the compaction of mineral soil is commonly assessed through soil bulk density only and soil bulk density and is also influenced by gravimetric soil water content, and compaction is also influenced by gravimetric soil water content, and compaction only slightly affects the water slightly affects the water content measurements in mineral soil using dielectric sensors such as SWR content measurements in mineral soilused usingtodielectric sensors such as SWR TDR, because and TDR, because these sensors, determine volumetric waterand content, includethese the sensors, used to determine volumetric water content, include the information on the bulk and information on the bulk density and gravimetric water content of mineral soils [24,37].density However, gravimetric water content of mineral soilsthat [24,37]. our experimental demonstrated our experimental results demonstrated thereHowever, is an obvious dependenceresults on compaction for that there is an obvious dependence on compaction for volumetric water content measurements in volumetric water content measurements in fermentation and humus. Compared to mineral soil, fermentation and humus. Compared to mineral soil, fermentation and humus have a loose spongy fermentation and humus have a loose spongy structure with higher porosity and greater variation structure with higher porosity and variation in bulk density, is a critical factor in bulk density, so compaction is greater a critical factor and should not so becompaction neglected when measuring and should not be neglected when measuring forest duff water content. forest duff water content. The square error (RMSE) of theofregression values was calculated to evaluatetocalibration The root rootmean mean square error (RMSE) the regression values was calculated evaluate accuracy (Table 2). When the samples’ water content was determined using the SWR sensor, the RMSEs calibration accuracy (Table 2). When the samples’ water content was determined using the SWR 3 ·cm− 3 for fermentation and humus, respectively) of the incompact samples (0.058 and 0.041 cm sensor, the RMSEs of the incompact samples (0.058 and 0.041 cm3·cm−3 for fermentation and humus, 3 for were distinctly higher than those of the samples (0.036samples and 0.025 cm3 ·and cm−0.025 respectively) were distinctly higher thanstrongly those ofcompacted the strongly compacted (0.036 fermentation humus, respectively). moisture measurements of the incompact samples cm3·cm−3 for and fermentation and humus,Similarly, respectively). Similarly, moisture measurements of the 3 ·cm− 3 for fermentation and humus, using TDR also showed larger errors (RMSEs of 0.083 and 0.050 cm incompact samples using TDR also showed larger errors (RMSEs of 0.083 and 0.050 cm3·cm−3 for − 3 for fermentation respectively) of the strongly compacted samples and 0.039 cm3 ·cmsamples fermentationthan and those humus, respectively) than those of the(0.043 strongly compacted (0.043 and and respectively). The resultsand indicated that the performanceThe of both SWRindicated and TDR was −3 for fermentation 0.039humus, cm3·cm humus, respectively). results thatbetter the in measuring the water content of fermentation and humus with higher compaction. This might be performance of both SWR and TDR was better in measuring the water content of fermentation and due to the fact higher that loose materials such fermentation with lower compaction humus with compaction. Thisasmight be dueand to humus the fact that loose materialscould such not as be prepared as uniformly as the samples with higher compaction, and the sensor probes are inserted fermentation and humus with lower compaction could not be prepared as uniformly as the samples in large porecompaction, spaces randomly. with higher and the sensor probes are inserted in large pore spaces randomly.

Sensors 2018, 18, 647 Sensors 2018, 18, x FOR PEER REVIEW Sensors 2018, 18, x FOR PEER REVIEW

7 of 11 7 of 11 7 of 11

Table 1. Classification of blows per layer in the Proctor compaction test used for the characterization of Tablecompaction. 1. Classification of blows per layer in the Proctor compaction test used for the characterization sample Table 1. Classification of blows per layer in the Proctor compaction test used for the characterization of sample compaction. of sample compaction. Blows Compaction Measured Measured by by Bulk Compaction Sample Blows perper Compaction BulkDensity Density Compaction Sample Type 3) Sample Blows per Compaction by Bulk Density Compaction Layer SC-900Measured (kPa) (g/cm Classes 3 Type Layer SC-900 (kPa) (g/cm3) Classes Type Layer SC-900 (kPa) (g/cm ) Classes Fermentation 0.54~0.63 Fermentation 0, 10, 1