1 Introduction

In 2018, during the West Japan torrential rain disaster, heavy rains caused extensive landslides and floods, primarily in the southern part of Hiroshima Prefecture. A total of 87 people died owing to sediment-related disasters (primarily debris flows), accounting for approximately 80% of the deaths (Hashimoto et al. 2020). In August 2014, 74 people lost their lives during the Hiroshima City landslide disaster (Tsuchida et al. 2019). Heavy rain disasters that claim many lives have become more frequent in recent years, and reducing the human damage caused by landslides and other sediment disasters has become an urgent problem. Many survivors of sediment-related disasters are residents living in sediment-related disaster hazard areas; thus, effective measures to encourage early evacuation of people living in areas with high disaster risks are required.

Some precursors can be observed when a landslide disaster occurs. Table 1 summarizes the precursors of landslide disasters according to a report of Sediment Disaster Precursor Information Review Committee for Warning and Evacuation compiled by the Ministry of Land, Infrastructure, Transport and Tourism (MLITT) (Sediment Disaster Precursor Information Review Committee for Warning and Evacuation, 2006). In some cases, these premonitory phenomena were used to make evacuation decisions for residents, and they have saved their lives. Therefore, the MLITT and local governments have called for the use of these precursory phenomena, in addition to regular rainfall information, in warning and evacuation actions.

Table 1 Precursors of debris flow disasters (MLITT,2006)
Full size table

Premonitory phenomena, such as boulders, rumbling of the ground, spring water from a bedrock, and river turbidity, which can be perceived visually and audibly, indicate that heavy rains have caused slope collapse or slip failure in the ground and an increase in groundwater level. In other words, the relationship between the precursors and landslide disasters can be easily explained. However, “the smell of rotten soil” has been cited as an omen, but the relationship between odors and landslides has not been fully clarified. The fact that odors are precursors is based on the experiences of many survivors in the past. Several survivors of the 2018 West Japan torrential rain disaster also reported that they smelled a strange odor before and after the landslide disaster, some of which are shown in Table 2 (Kaibori and Yanagisako 2019). It was reported that the odor was an “earthly smell” or “earthy green smell,” which occurred immediately before or 30 min before the debris flow. The odor occurred in the wide space around the survivors’ residences. In the report by MLITT, the odor was considered to be caused by the release of odors from the ground when a slope failure occurs upstream in a debris flow or by ignition owing to the friction of boulders and fallen trees when the earth and sand flow down (Sediment Disaster Precursor Information Review Committee for Warning and Evacuation, 2006). However, considering the reports that survivors smelled a strange odor several tens of minutes before the disaster occurred, the mechanism of odor generation during sediment disasters must be investigated in more detail.

Table 2 Experiences about odors heard from victims of the 2018 West Japan heavy rain disaster (Kaibori and Yanagisako,2019)
Full size table

According to the knowledge of soil biochemistry and microbiology, forest soil contains numerous microorganisms, and it is estimated that tens of thousands of species and tens of millions to hundreds of millions of bacteria live in 1 g of soil (Roesch et al. 2007; Shibata et al. 2018). In relation to soil odors, Geosmin, a metabolite produced by Streptomyces species among actinomycetes, which are microorganisms living in soil, has been reported to produce earthy odors (Gerber and Lechvalier 1965). In recent years, research using genetic information has revealed that soil contains many unknown microorganisms with metabolic functions. Considering the diversity of microorganisms in the ground, the odor in the ground is considered to be a complex of odorants produced by the metabolism of various microorganisms (Hug et al. 2016; Shibata et al. 2018). In the field of geotechnical engineering, the activity of microorganisms in soil has attracted considerable interest from the perspective of new ground improvement technologies. Biocementation, in which calcium carbonate cement soil particles are crystallized by the enzyme urease of ureolytic bacteria in the ground, has been extensively studied as a ground improvement technology that does not impose a burden on the environment (DeJong et al. 2010; van Paassen et al. 2010; Jiang et al.,2020). In this technique, non-native ureolytic bacteria are often introduced into the soil, but methods of stimulating indigenous ureolytic bacteria in the soil in situ to achieve improvement have also been investigated (Gowthaman et al. 2019) And more, some of the techniques will promote unconfined compression strength of sandy soil with low calcium carbonate precipitation rate, which can be combined with skim milk, casein etc.,. These techniques can reduce toxic byproduct sources named ammonium that will promote heavy odor in the soil (Almajed et al. 2019, Miyake et al., 2022).

As described above, soil has a peculiar smell owing to a metabolite produced by bacteria in the soil. However, the experiences of survivors suggest that the odor of the soil did not occur locally, but occurred at least in a range of several tens of square meters around their residence. To date, insufficient research has been conducted on odors that occur on such a scale before landslide disasters. To clarify the relationship between landslides and odors, the authors measured the intensity of odors on the ground of dangerous natural slopes with a history of landslide disasters. We examined the distribution of odors in the ground and investigated the relationship between the odor intensity in the ground and geotechnical constants.

2 Measurement and distribution of odors in the ground on a slope with a collapse history

2.1 Method of odor measurement

As an odor frequently exists as a complex in which various odor substances are mixed, the odor concentration and odor index in the olfaction measurement method specified by the Ministry of Environment (Ministry of Environment, 2003) do not have a unit, and a relative value, which is a dilution factor when the odor is no longer perceived, is used. In this study, an odor device XP-329IIIR (New Cosmos Electric Co. Ltd.) was used to measure the odor of the soil in the field. Figure 1 shows the appearance and specifications of XP-329IIIR, which quantifies the odor of the sucked air relative to the odor of the air deodorized with deodorizing charcoal as the zero point and immediately displays the odor intensity as a numerical value from 0 to 2000.

Fig. 1
figure 1

Odor measurement device XP-329IIIR

Full size image

The sensor used in XP-329IIIR has a structure in which a platinum wire coil is coated with a metal oxide semiconductor in a ball shape with a diameter of about 0.4 mm and subsequently sintered. When odor molecules are adsorbed on the surface of a metal oxide semiconductor, the electrical conductivity of the semiconductor improves, and the resistance decreases. The change in the resistance value is obtained as a deviation voltage using a bridge circuit and quantified. In this study, the value measured by this sensor was called the odor intensity. During the measurement, the sensor was washed with a standard gas (carbon dioxide) to set a zero point, and then the measurement was performed.

Table 3 shows the odor intensities of various objects measured using XP-329IIIR. As shown in the table, leather shoes and coffee with relatively low odors have intensities of 50–160, and fermented soybeans (natto), kimchi (Korean pickles), and soy sauce, which are typical foods with strong odors, have intensities of 450, 850, and 990, respectively. Thus, when this device indicates a relative numerical value of 100 or more, the existence of an odor is recognized with almost no individual differences.

Table 3 Odor intensities of various objects measured using XP-329IIIR
Full size table

Odor measurements in the ground were performed using a penetration hole made using a lightweight dynamic cone penetration test (LWDCPT) device. The LWDCPT is a dynamic cone penetrometer with variable energy and was designed and developed in France during the 1990s (Langton 1999). A schematic of the LWDCPT is shown in Fig. 2. The device weighs 20 kg, and its major feature is that it can be operated and tested by one person, even on steep slopes and other places with poor footing. The LWDCPT primarily consists of an anvil with a strain gauge bridge, a central acquisition unit, and a dialogue terminal. The hammer is a rebound-type hammer and weighs 1.73 kg. The stainless-steel rods are 14 mm in diameter and 0.5 m in length, and the area of the cone is 2 cm2. In the LWDCPT, when the head is struck with a hammer, the accelerometer installed on the head measures the acceleration of the collision, and this acceleration and the penetration amount obtained from the retractable tape are measured. The penetration resistance qd is calculated by a central processing unit using the data and is displayed and recorded on the data logger together with the penetration length. Most geotechnical survey methods are not applicable to steep, heavily wooded, natural slopes. The advantage of the LWDCPT is that it is lightweight; therefore, it can be carried around in mountains and can be operated by a single person, even on natural slopes. (Athapaththu et al. 2007, 2014). Tsuchida et al. (2011) proposed a method of estimating the shear strength parameters of a weathered granite layer on a natural slope from the penetration resistance of the LWDCPT. The LWDCPT has also been used to measure the layer thickness and shear strength parameters of layers that have collapsed and flowed down, even after landslide disasters caused by heavy rains (Tsuchida et al. 2014, 2015).

Fig. 2
figure 2

Lightweight dynamic cone penetration test (LWDCPT)

Full size image

The Ministry of the Environment stipulated the method of investigating gases in the ground as the “collection and measurement method for soil gas investigation” based on the Soil Contamination Countermeasures Law (Ministry of Environment, 2003). In this study, referring to this method, the odor in the ground was measured using the penetration hole of the LWDCPT, as shown in Fig. 3. The measurement procedure was as follows: First, after the penetration test, the rod was pulled out and an aluminum pipe was inserted into the hole. Next, a Teflon tube with a diameter of 5 mm, connected to an XP-329IIIR odor intensity tester, was inserted straight through the pipe to the tip. The odor intensity was measured using carbon dioxide placed in the flex sampler as the zero point. After the odor intensity at a certain depth was measured, the odor sensor XP-329IIIR was idled until the odor substances stabilized. Figure 4 shows the typical odor intensity-elapsed time relationship. The measured odor intensity reached the peak value at 30–90 s, began to decrease, and converged to a steady value at approximately 120 s after the peak time. The peak and steady values of odor intensity were measured at each point and depth. Figure 5 shows the situation in which the LWDCPT was performed on a natural slope, and Fig. 6 shows the situation in which the odor intensity was measured with time using penetration holes after the LWDCPT was performed.

Fig. 3
figure 3

Measurement method of odor intensity in ground using penetration hole of LWDCPT

Full size image
Fig. 4
figure 4

Measured odor intensity with time

Full size image
Fig. 5
figure 5

LWDCPT in the mountain stream

Full size image
Fig. 6
figure 6

Measured odor intensity with time

Full size image

2.2 Odor intensity measured at the Mt. Gagara research field at Hiroshima University

The intensity of the odor in the ground and its horizontal and vertical distributions were investigated on a natural slope. The observation site was the Mt. Gagara research field on the premises of Hiroshima University. Figure 7 shows a photograph of the survey site. The mountain stream on the right side of the study site collapsed, and debris flow occurred because of the 1999 heavy rainfall disaster. From 2002 to 2009, continuous measurements using tensiometers, soil moisture meters, and rain gauges were conducted on the slopes (Thi et al. 2003, 2008). The relationship between depth and penetration resistance using an LWDCPT over a wide area and the changes in penetration resistance before and after rainfall have been reported for the test site (Athapaththu et al. 2007, 2014; Tsuchida et al. 2011). Furthermore, in 2018, debris flow occurred again at two nearby locations during the West Japan torrential rain disaster, as shown in Fig. 7. In this study, the distribution of the odor in the ground in the depth and plane directions was measured using an odor sensor.

Fig. 7
figure 7

Mt. Gagara research field in Hiroshima University

Full size image

The ground of Mt. Gagara is composed of two layers; the top layer has silty sand with some amounts of gravel, up to a depth of approximately 2 m, and weathered granite thereafter (Thi Ha 2005). Geologically, the area is underlined by highly fractured granitic rocks, overlain by weathered remnants. Owing to the abundant precipitation and adverse settings of granite rock structures in the area, weathering has occurred over the years, resulting in the formation of Masado soils, which is a residual sandy soil of heavily weathered granite that is widely spread over the southwestern part of Japan. The gradation curves of Masado soils in the research field are shown in Fig. 8. Masado primarily consists of sand with gravel particles interconnected with 0–10% fine particles. Figure 9 shows the soil horizons of the Mt. Gagara research field, observed by digging a pit (Thi Ha 2005). In soil science, soil horizons are classified into the O horizon (organic surface layer), A horizon (topsoil), B horizon (subsoil), C horizon (substratum), and R horizon (bedrock). As shown in Fig. 9, a depth of 2 m or more is substratum (C) of weathered granite, on which 10–20 cm thick topsoil (A) and 180–190 cm thick subsoil (B) are deposited. In the subsoil, the roots and stems of the plants are observed.

Fig. 8
figure 8

Grading curves of Masado soils in Mt. Gagara

Full size image
Fig. 9
figure 9

the soil horizons of Mt. Gagara research field observed by digging a pit (Thi Ha 2005)

Full size image

An LWDCPT and odor intensity measurements were performed. Figure 10 shows the measured odor intensity with time at sites B, B1, and B2. As mentioned earlier, the odor intensity reached a peak value OI1 in 30–90 s from the beginning of the measurement then decreased and reached a steady-state value OI2. Figure 11 shows the relationship between OI1 and OI2 at 94 measurement points. Variations in the relationship between the two were observed; however, overall, the steady-state value OI2 tended to become smaller than the peak value OI1 as the odor intensity increased. In the remainder of this paper, we use the maximum value of odor intensity.

Fig. 10
figure 10

Measured odor intensity with time

Full size image
Fig. 11
figure 11

Relationship between peak value of odor intensity and value after stabilization

Full size image

Figure 12 shows the relationship between the depth, penetration resistance, and measured odor intensity at seven sites. At site A, the cone penetration resistance increased with the depth above 2 MPa. This suggested that the ground was the residual soil of the weathered granite layer. The odor intensity at site A continued to be similarly high, from 250 at the surface to 350 at a depth of 1.8 m. At site B, a loose Masado soil layer with a cone penetration resistance of less than 1 MPa was deposited from the surface to 1.2 m. This loose layer on Mt. Gagara is presumed to be a talus layer formed by sediment that collapsed from the upper part of the slope in the past. From 1.2 m below, the penetration resistance value increased; therefore, it was assumed to be the residual soil of the weathered granite layer. The odor increased to 800 or more at a depth of 1.8 m. At both sites C and D, a talus layer was deposited with a thickness of 1.5 m from the surface and the residual soil of the weathered granite layer was below it, reaching the substratum at depths of 2.1 and 2.6 m, respectively. At site D, the odor intensity of the talus layer was less than 100 but increased to approximately 350 in the weathered granite layer. This was almost the same as that at site B. However, at site C, the intensity of each odor was 60–85 at all depths, and no significant change was observed in the upper talus layer and lower weathered granite layer.

Fig. 12
figure 12

Measured odor intensity and penetration resistance with depth at 7 sites in Mt. Gagara in Hiroshima University

Full size image

At site E, a talus layer appeared to be directly deposited onto the substratum. The measured odor intensity of the talus layer was small (60–80 on average) and almost constant. At sites F and G, the surface talus layer was thin (about 0.6–0.8 m), and there was a relatively thick weathered layer underneath. The odor intensity at site F was less than 100 in the talus layer and increased with depth in the weathered layer, exhibiting the same tendency as that at site B. In contrast, at site G, the talus layer had an intensive odor similar to that of the lower weathered layer. Because the odor intensity of the weathered granite layer at site G was as high as 1000 or more, the strong odor may have propagated to the pore air of the upper talus layer, which originally had a low odor intensity.

To investigate the horizontal distribution of odor in the ground, we measured the odor intensity at a depth of 1 m at 19 points at 50 cm intervals in a 2 m×2 m area near site B. Figure 13 shows the planar distribution of the peak values of odor intensity at 19 points. As shown in the figure, the odor intensity was randomly distributed in the range of 109 to 962 within a narrow area of 2 m× 2 m, with an average of 340, a standard deviation of 190, and a coefficient of variation of 0.56.

Fig. 13
figure 13

Measurement points to investigate horizontal variation of odor intensity in Mt. Gagara site

Full size image

As a result of measuring the odor in the ground at Mt. Gagara at Hiroshima University, which has a history of collapse due to heavy rain, we verified that there was an odor with a maximum intensity of 1047 at a 0–2.5 m depth in the ground. The relationship between depth and odor intensity varied significantly depending on the site. In the range investigated in this study, the talus layer, which was considered to have been deposited by the collapse of the upper weathered layer in the past, often had a low odor intensity of 100 or less. The results of intensive measurements over a 2 m × 2 m area indicated that the odor intensity varied significantly, even within a narrow range in the ground.

3 Relationship between ground conditions and odor intensity

As a result of the odor survey on Mt. Gagara, we observed a strong odor in the ground that can be detected by the human sense of smell. To investigate the relationship between odor intensity and geotechnical constants that may be related to odor, we measured odor intensity in the ground at depths of 30 and 50 cm on eight natural slopes in Hiroshima Prefecture, where landslides have occurred over the last 30 years, and collected soil samples. We selected eight survey points in Higashi-Hiroshima City, Hiroshima City, Kure City, and Akitakata City. Figure 14 shows the locations of the survey points. As with the ground on Mt. Gagara, the soil was deposited with weathered granite.

Fig. 14
figure 14

Locations of the survey points in Hiroshima Prefecture

Full size image

3.1 Measured odor intensity

Figure 15 shows the measurement results for each point. Odor intensity was measured three times at each point. As shown in the figure, the largest odor intensity of 657 was observed at Miyakegawa, and no odor was observed at Mt. Yasumi, Kure City, and at Dontagawa, Etajima City. Although only two points were measured in the depth direction, the overall odor intensity remained constant or increased in the depth direction. Figure 16 shows a histogram of the odor intensity in the ground measured in this time. As the odor intensity was 0 at two of the eight locations, the average measured maximum odor intensity was 132, which was smaller than the measured value at Mt. Gagara.

Fig. 15
figure 15

Maximum odor intensity measured with depth

Full size image
Fig. 16
figure 16

Histogram of odors in the ground

Full size image

3.2 Relationship between odor intensity and geotechnical parameter of soil

As the odor in the ground is generated by the biochemical activity of microorganisms, a relationship should be determined between the activity of the microorganisms in the soil and the odor. In this study, water content, fine particle content, pH, and ignition loss were measured as items related to the activities of microorganisms in the soil, and the relationship between these and odor intensity was examined. The water content, particle content, and ignition loss were measured at a depth of 30 cm using 400–600 g samples. For the pH measurement, 4 L of distilled water was sprinkled near the odor measurement site and left for 30 min. The ground was then excavated, and the pH was measured by inserting a soil moisture meter and a soil acidity meter at depths of 30 and 50 cm, respectively. The reason for sprinkling distilled water is that accurate measurements cannot be obtained if the soil to be measured is dry.

Figure 17 shows the relationship between the LWDCPT penetration resistance qd and odor intensity at the Mt. Gagara research field and eight survey points. A weak but positive correlation was observed between the logarithm of qd of the ground at Mt. Gagara and the odor intensity, and the following approximation was obtained:

Fig. 17
figure 17

Relationship between odor intensity and water content of soil

Full size image
$${I_{o}{\rm{ }} = {\rm{ }}122{\rm{ }}ln\left( {{q_d}} \right){\rm{ }} + {\rm{ }}249}$$
(1)

Data from eight locations were included in the data range for Mt. Gagara, except for Dandagawa and Mt. Yasumi, where no odors were measured. The ground with a qd value of 0.2 to 0.5 was likely to be a talus layer with collapsed sediment deposited, and the fact that the odor intensity in the talus layer was low was considered to be the background of the above equation.

Figure 18 shows the relationship between the water content and odor intensity. The water content of the soil surveyed in this study ranged from 15 to 26%, and the odor intensity tended to increase as the water content decreased. However, based on the relationship between the content of fine particles in the soil and the odor intensity shown in Fig. 19, no correlation was observed between the two.

Fig. 18
figure 18

Relationship between odor intensity and water content of soil

Full size image
Fig. 19
figure 19

Relationship between odor intensity and fine content of soil

Full size image

Figure 20 shows the relationship between odor intensity and pH value. Most actinomycetes in the ground prefer a neutral region and grow at pH 5.0–9.0, and a suitable pH for growth is approximately 7.0 (Shibata et al. 2018). As shown in the figure, the pH of the ground examined in this study was within a narrow range of 6.5 to 7.4, and except for Dondagawa and Mt. Yasumi, the odor intensity tended to increase as the pH decreased. Soil acidification is related to the activities of microorganisms and organisms, and a low pH is considered to be related to the magnitude of the action of microorganisms and the resulting odor generation.

Fig. 20
figure 20

Odor intensity and pH of pore water in soil

Full size image

Figure 21 shows the relationship between the odor intensity and ignition loss li. As shown in the figure, a relatively good correlation was observed between the two, and the following equation was obtained.

Fig. 21
figure 21

Relationship between odor intensity and ignition loss

Full size image
$$I_{o}{\rm{ }} = {\rm{ }}79{\rm{ }}\left( {{l_i} - 2.1} \right)$$
(2)

A loss on ignition is strongly correlated with the amount of organic matter in the soil (JGS, 2020). Assuming that odorants are produced during the decomposition of organic matter by microorganisms in the ground, we reasonably consider that the organic matter content in the soil is related to odor intensity.

In summary, although the odor intensity in the ground varies significantly depending on the location, there is a correlation between the pH and ignition loss, which appears to be linked to the activity of bacteria in the ground. However, although it is difficult to relate it to microbial activity, a tendency was observed in that a lower water content of the soil results in a higher odor intensity, and the fine particle content rate is not related to the odor intensity.

4 Conclusions

The smell of soil is a premonitory phenomenon of debris flows. Although microorganisms in soil produce compounds with unique odors, odors that occur before disasters are not localized but spread over a wide area according to the experiences of disaster survivors. In this study, to clarify the relationship between landslide disasters and odors, we measured odor intensity in a ground that has a history of debris flow disasters, and the relationship between odors and geotechnical characteristics of the soil was investigated. The conclusions obtained in this study can be summarized as follows:

  1. 1)

    The odor intensity in the ground was measured at depths of 0–2.5 m on the slope of Mt. Gagara, Hiroshima University, which has a history of collapse owing to heavy rain. The method involves inserting an aluminum pipe and Teflon tube into a hole with a diameter of 10 mm formed by the penetration of lightweight dynamic cone penetration test (LWDCPT) device and measuring the odor intensity of the sucked interstitial air using an odor measuring instrument. More than 100 odor intensities were measured at many points, which were considered recognizable without individual differences.

  2. 2)

    The odor intensity tended to be low in the surface talus layer and high in the weathered residual soil layer. However, the talus layer had a strong odor intensity in some cases and the weathered layer had a weak odor intensity in other cases, and large differences were observed depending on the location.

  3. 3)

    To investigate the odor distribution in the ground, we measured odor intensity at 19 locations within an area of 2 m × 2 m. The average odor intensity was 340 and the coefficient of variation was 0.60. This result showed that the odor distribution in the ground varied significantly.

  4. 4)

    The odor in the ground was measured at depths of 30 and 50 cm on natural Masado slopes at eight locations in Hiroshima Prefecture. An average odor intensity of 100 or more was observed at four locations, 100 or less at two locations, and no odor was observed at two locations. We observed that odor was not always present in the ground. A weak but positive correlation was observed between the logarithm of the LWDCPT penetration resistance and odor intensity.

  5. 5)

    The relationship between the odor intensity and ground constant was investigated at these eight locations. Although the odor intensity in the ground varies significantly depending on the location, a correlation between the pH and ignition loss was observed, which appears to be linked to the activity of bacteria in the ground. However, although it is difficult to relate it to microbial activity, a lower water content of the soil results in a higher odor intensity, and the fine particle content rate is not related to the odor intensity.