The rainfall thresholds and soil characteristics of large geological disasters in Zhejiang, China

Based on observed minute precipitation data during Lekima influencing Zhejiang (from 2019–08-07 00:00:00 to 2019–08-12 23:55:00), the rainfall thresholds for debris flow and landslide are investigated. The rainfall intensity and duration (I-D) thresholds are I = 1247.73D−403.44 + 19.09 and I = 142.63D−0.58–3.37 for debris flow and landslide, respectively. Hourly meteorological data and soil data reveal that (1) the soil volume water content (SVWC) at deep soil layers (50–100 cm) fluctuated continuously during the occurrence of debris flow and landslide. At the end of the landslide period, SVWC at deep soil layers returned stable. (2) At the same soil layer, there was little difference between the upper adjacent values of SVWC in the periods of debris flow and landslide, but the lower adjacent values in the period of landslide were smaller. However, different from the distribution of soil moisture, the lower adjacent values of the soil temperature at all layers were basically the same when debris flow and landslide occurred, while the upper adjacent values were higher when landslide occurred than when debris flow occurred. (3) Compared with the whole typhoon influencing period and debris flow occurrence period, ground surface temperature and surface air temperature, soil temperature, and surface air temperature showed better correlations in landslide occurrence period. (4) Both during debris flow and landslide occurrence time, the soil temperature at all layers did not change with the change of soil depth. Soil temperature and moisture at shallow soil layers owned better correlations than at deep soil layers.


Introduction
As the variety wide distribution and great harm, the safety of people's life and the social development are seriously affected by geological disasters (He et al., 2011;Zhang et al., 2015). Debris flow and landslide caused by rainfall are the most common large-scale geological disasters and are extremely destructive Li et al., 2010;Saez et al., 2011;Di et al., 2017;Shu et al., 2020;Sun et al., 2021). China is one of the countries with the most serious geological disasters in the world; meanwhile, Zhejiang Province is a province prone to geological disasters in China (Yue et al., 2010;Zhang et al., 2015). Especially in recent years by the frequent impact of typhoons, geological disasters are particularly serious and debris flows and landslides occur frequently (Li et al., 2010Yue et al., 2010;Ma et al., 2015;Zhu et al., 2019). Recent researches on debris flow and landslide in Zhejiang mainly focus on internal dynamic conditions of soil (Zhi et al., 2016;Tian et al., 2020a, b), and some tries to forecast disasters by artificial intelligence (Li et al., 2010;Tian et al., 2020a, b), hazard sensitivity mapping (Wu et al., 2014;, disaster monitoring (Zhi et al., 2014), and risk management (Jia et al., 2019). These studies mainly forecast and provide a decision-making basis for debris flow and landslide through phenomena, but pay less attention to the weather and soil conditions when induced.
The observation and statistical data show that there is a critical rainfall threshold for the occurrence of regional geological disasters (Campbell, 1975;Giannecchini et al., 2012;Li et al., 2012;Dowling and Santi, 2014;Ma et al. 2011). Rainfall threshold is one of the most important parameters for debris flow prediction (Gariano et al., 2015;Ma et al., 2015). As the main factor inducing debris flow and landslide, few studies focus on the relationship between precipitation and large-scale disasters. Some limited researches like Li et al. (2011) found that there is a good exponential relationship between the cumulative frequency of the landslide occurrence and rainfall, and about 80% of landslides in Zhejiang Province during 2004-2007 are induced when the cumulative 1-day rainfall is 200 mm. Dependent on the strong correlation between debris flow location and rainfall distribution, Xu et al. (2012) established two prediction models to estimate precipitation. Chen et al. (2005) analyzed the influence of regional rainfall pattern (antecedent rainfall, duration, intensity and accumulated rainfall) and geological conditions on triggering debris flow and found the combination of averaged rainfall intensity and duration is a more practical index for debris flow monitoring. Ma et al. (2015) counted the I-D thresholds during landslides in 62 different regions of Zhejiang Province in 1990-2013. Clearly, examining the spatiotemporal characteristics of debris flows and landslides constitutes the first step towards understanding the relationship between disasters and rainfall and some details of meteorology and soil situation are needed for the modeling of disaster occurrence. The objective of this study is twofold: (1) examining the meteorological and soil conditions before and after the occurrence of debris flows and landslides caused by typhoon, and (2) investigating the possible connections between rainfall and debris flows and landslides with the aim of explaining the relationship between rainfall pattern and debris flows and landslides.
The structure of this paper is as follows. The "Study domain, data, and methods" section describes the study domain, datasets, and methodology. The analysis results are presented in the "Results" section. The "Discussions" section contains the discussions and the "Conclusions" section lists the major conclusions.

Study domain
Zhejiang Province (Zhejiang, 118° 02′-123° 08′ E, 27° 03′-31° 11′ N; Fig. 1) is located in the southeast coast of China. The southwest of Zhejiang is mostly mountainous areas with an altitude of more than 1000 m, and the transition area in the middle is hilly and basin. The eastern and northern parts are plains with low altitudes. As located in the subtropical monsoon zone, typhoon influence, and geological conditions prone to geological disasters, geological disasters occur frequently in Zhejiang, especially debris flows and landslides triggered by extreme precipitation occur frequently in recent years.
Super Typhoon Lekima (No. 1909) was developed from a tropical depression in the western North Pacific on August 3, 2019. At about 1:45 on August 10, Lekima landed in the coastal area of Zhejiang with the maximum wind force near the center 16 (52 m/s) and then it crossed Zhejiang and Jiangsu provinces and moved into the Yellow Sea at 12:00 on August 11; at about 20:50 on August 11, Lekima landed again in Shandong Province, China, with maximum wind force near the center 9 (23 m/s), and then it moved into the Bohai Sea and continued to weaken (Xha et al., 2020). From 1:45 on August 10 to about 22:00 on August 10, Lekima landed in Zhejiang for more than 20 h. Severe storms and seawater flooding in Zhejiang had been caused during Lekima landing, waterlogging in some urban and rural areas, the collapse of some houses, damage to a large number of infrastructure, and secondary disasters such as debris flow and landslides in many places, which have had a very serious impact on Zhejiang (Zheng et al., 2021). Lekima caused an economic loss of at least RMB 51.53 billion yuans and 56 fatalities in mainland China (Wu et al., 2020). Figure 1 shows the Lekima landing track in Zhejiang and the locations of landslide and debris flow caused by Lekima. The location information of debris flow and landslide in Zhejiang when Lekima landed is shown in Table 1. During Lekima landing in Zhejiang, there were 139 landslides and 99 debris flows respectively, which were mainly located in the southeast and northwest Zhejiang (Fig. 1, Table 1).

Data
Observed minute precipitation data, hourly meteorological data, and soil data are obtained from the website of Climate Data Center of China Meteorological Administration (CMA; http:// data. cma. cn/). Hourly soil data include soil temperature data and soil moisture data. The data quality is controlled by removing any abnormal values such as negative and unreasonably large values and identifying missing values and filling in the gaps by interpolation.
Derived precipitation variables based on minute level precipitation during the time completely covering the period of typhoon Lekima affecting Zhejiang (2019-08-07 00:00:00 to 2019-08-12 23:55:00) include the following: (1) rainfall intensity, that is, the rainfall per unit period; (2) effective antecedent precipitation calculated in Eq. (2), that is, the accumulated precipitation over disaster sites up to the time of debris flow or landslide; (3) present hourly precipitation, that is, the hourly rainfall over disaster sites when debris flow or landslide occurs; (4) maximum accumulated rainfall in one hour (1 h) and three hours (3 h), that is, the maximum accumulated precipitation of 1 h and 3 h at the disaster sites during the typhoon Lekima influencing; (5) regional average accumulated precipitation (acc_PRE), that is, the accumulated value of hourly average precipitation in Zhejiang during the typhoon Lekima influencing.

Methods
Through the statistical analysis of the historical rainfall data that has caused or may cause geological disasters, the rainfall threshold that causes geological disasters is determined, which is used as the reference value of realtime rainfall observation to predict the possible occurrence time of geological disasters. That is referred to as rainfall threshold or rainfall lower limit method (Cui et al., 2003;Wei et al., 2005;Guzzetti et al., 2007). The core idea of the method is to establish the critical combination discriminant of geological hazards by using rainfall intensity, duration, and average rainfall.
The I-D threshold (rainfall intensity-duration threshold) (Campbell, 1975;Starkel, 1979;Caine, 1980) is widely used to determine the critical rainfall in the fields of hydrology and geography (Hadadin, 2005;Vyver 2015; radar stations over the research area and its surroundings. The lightyellow background represents the weather radar control range. Purple and gray points represent the national and regional meteorological stations, respectively   Zhang et al., 2015;Karki et al., 2018;Wang and Bi, 2020). The I-D threshold is expressed as follows: where I is the average rainfall intensity (unit: mm /h) from the beginning of rainfall to the occurrence of geological disasters; D is the duration (unit: h), which refers to the duration of a rainfall event or rainfall period; a, b, and c are constant parameters calculated by the regression method.
Besides the I-D threshold, the E-D (accumulated rainfall-duration) threshold is also proposed to classify the rainfall thresholds (Cannon and Ellen, 1985;Joyce et al., 2004;He et al., 2020). D is the duration (unit: h) and E is the accumulated rainfall (unit: mm) during this duration. The E-D threshold is equivalent to the I-D threshold, while the E-D threshold can avoid unnecessary conversions. Effective antecedent precipitation is an important index to estimate the rainfall threshold of debris flow and landslide, which is defined as: where R is the effective antecedent precipitation (unit: mm); R n is the rainfall on the nth time step before the occurrence of debris flow; a is the attenuation coefficient, during the Meiyu period in Zhejiang, a = 0.8 for daily attenuation (Du et al., 2006). In this study, hour is taken as a time step for the study of effective antecedent precipitation, so a = 1 is used in this paper; that means, there is basically no precipitation attenuation among time steps.

Meteorological and soil conditions at the time of disaster occurrence
The debris flow occurrences concentrated in 00:00 August 10 to 00:00 on August 11, 2019, and the landslide occurrences concentrated in 21:00 August 9 to 09:30 August 11, 2019, in Zhejiang (Fig. 2). It means that before Lekima landed, the heavy rainfall had already caused debris flows and landslides. After landing, it caused more geological disasters. During the whole process of Lekima influencing Zhejiang (2019-08-08 00:00 to 2019-08-11 00:00), the regional average accumulated precipitation (acc_PRE) of Zhejiang ranged from 0.0 to 164.5 mm and the regional average precipitation in the past 1 h (PRE_1h) ranged from 0.0 to 6.9 mm (Fig. 2a). The obvious rising stage of acc_ PRE was from 2019-08-09 07:00 to 2019-08-10 14:00, and then it changes slowly. The PRE_1h process showed a single peak curve. And the maximum PRE_1h occurred at 2019-08-09 19:00, which means the maximum precipitation appeared before the Lekima landing. In the whole process of Lekima influencing Zhejiang, PRS and VAP ranged from 965.2 to 987.3 hPa and 28.9 to 33.0 hPa, respectively (Fig. 2b). The change of PRS directly related to typhoon; PRS decreased obviously before the occurrence of debris flow and landslide, and then increased obviously during the occurrence. PRS went up from 965.1 to 979.0 hPa in the period of debris flow and landslide occurrences. The reason for the above PRS change was that when typhoon approached, the PRS decreased; then typhoon landing, decreasing wind speed affected by increasing ground friction and reduced heat source and water vapor supply made the convergence weaken and the central air of typhoon increase, which led to pressure rising. Different from PRS, the VAP increased slightly and then decreased (2) R = aR 1 + aR 2 + ... + aR n continuously during the occurrences of debris flow and landslide, which was related to the positive correlation between water vapor and typhoon.
There was an insignificant negative correlation between DPT and TEM (p < 0.05), during Lekima influencing Zhejiang (Fig. 2b). DPT changed little in the whole process and ranged from 22.7 to 25.1 ℃. TEM, ranging from 25.0 to 33.9 ℃, showed a larger change. During the occurrence of debris flow and landslide, TEM was lower than that in the adjacent period, which may be related to the rainy day caused by typhoon. After the typhoon leaving Zhejiang, the TEM increased obviously, followed by another precipitation at about 2019-08-11 10:00, TEM dropped again, and it rose again on August 12 (Fig. 2a, b).
Except for 320-cm soil temperature (GST_320cm), the temporal characteristics of ground surface temperature and soil temperature were similar to that of TEM (Fig. 2c). GST, GST_5cm, GST_40cm, GST_80cm, and GST_160cm had significant positive correlations (p < 0.05) with TEM in the landslide occurrence time (2019-08-08 00:00 to 2019-08-11 00:00) (Fig. 3). Their correlation coefficients ranged from 0.70 to 0.89. Meanwhile, soil temperature at 40 cm, 80 cm, and 160 cm also significantly correlated with TEM during debris flow occurrence. However, GST_10cm, GST_15cm, GST_20cm, and GST_320cm showed worse correlations, especially GST_15cm and GST_20cm owning discrete points. GST_320cm was not sensitive to TEM, which may be related to less energy exchange between air and soil surface as its deeper depth (Fig. 2i). Because of the direct energy exchange with the air, the variation of the ground surface temperature with the TEM was particularly obvious (Figs. 2a). And there was a significant correlation between GST_80cm and TEM (Fig. 2g). In debris flow occurrence time, GST, GST_40cm, GST_80cm, and GST_160cm correlated significantly with TEM (Fig. 2a, f-h). Compared with the whole period and debris flow occurrence period (2019-08-10 00:00 to 2019-08-11 00:00), better fitting relationships between ground surface temperature and soil temperature and TEM were shown in the landslide occurrence period (2019-08-09 21:00 to 2019-08-11 09:30) in general (Fig. 3).
In terms of air temperature and humidity, with the increase of TEM, RHU decreases, and vice versa. However, soil volume water content (SVWC) did not change with TEM (Fig. 2d). From 10-to 50-cm soil layer, soil moisture increased with depth, then from 60 to 100 cm, there was no large difference among layers and no obvious SVWC change with depth. Under the influence of typhoon, SVWC at all layers increased and reached the maximum within the debris flow occurrence time. Then, with the end of the typhoon process, the soil moisture decreased. It is worth noting that SVWC at deeper soil layers (50-100 cm) fluctuated continuously for about 24 h (about from 2019-08-10 10:00 to 2019-08-11 10:00) during the occurrence of debris flow and landslide. At the end of the landslide period, SVWC at deeper soil layers returned stable.
The correlation between soil temperature and humidity should be analyzed to further observe the soil conditions of debris flow and landslide occurrence. Affected by the precipitation caused by typhoon, the soil moisture of all and debris flow (2019-08-10 00:00 to 2019-08-11 00:00) respectively, and they coincide in the occurrence period of debris flow. a The precipitation in the past hour (PRE_1h) and the accumulated precipitation (acc_PRE) in the whole process (2019-08-08 00:00 to 2019-08-11 00:00). b Air pressure (PRS), vapor pressure (VAP) in the whole process. c Air temperature (TEM), ground surface temperature (GST), ground temperature in depths of 5 cm, 10 cm, 15 cm, 20 cm, 40 cm, 80 cm, 160 cm, and 320 cm (GST_5cm,GST_10cm,GST_15cm,GST_20cm,GST_40cm,GST_80cm,GST_160cm,and GST_320cm) in the whole process. d Air relative humidity (RHU)and soil volume water content (SVWC) in different soil depths (10 cm, 20 cm, 30 cm, 40 cm, 50 cm, 60 cm, 80 cm, 100 cm) in the whole process layers at the time of debris flow and disaster were all higher than the annual average soil moisture of Zhejiang in August (Fig. 4a). The annual average SVWC during 2000-2010 in August were 23.7 10 −2 g/m 3 at 10 cm, 27.9 10 −2 g/m 3 at 20 cm, 32.3 10 −2 g/m 3 at 40 cm, and 35.5 10 −2 g/m 3 at 80 cm (not shown in Fig. 4a). With the increase of soil depth, the SVWC increased during the occurrence of debris flow and landslide. Compared with the deeper soil layers (40 cm and 80 cm), the distributions of SVWC in the shallow soil (10 cm and 20 cm) were more concentrated when the debris flow occurred, but there was no such phenomenon when the landslide occurred. In the same soil layer, the median value Fig. 3 The correlation between air temperature (coordinate-axis X) and surface and soil temperature (coordinate-axis Y) (units: ℃). a-i represent relationship between TEM and GST,TEM and GST_5cm,TEM and GST_10cm,TEM and GST_15cm,TEM and GST_20cm,TEM and GST_40cm,TEM and GST_80cm,TEM and GST_160cm,and TEM and GST_320cm,respectively. Black, blue, and red points represent the whole process (2019-08-08 00:00 to 2019-08-11 00:00), occurrence time of landslide (2019-08-09 21:00 to 2019-08-11 09:30), and debris flow (2019-08-10 00:00 to 2019-08-11 00:00), respectively. Black, blue, and red lines are fitting lines of corresponding periods. Stars represent statistically significant correlations (p < 0.05) of SVWC during debris flow was larger than that during landslides. It was also confirmed that high water content was an important soil factor causing debris flow. Also, in the same soil layer, there was little difference between the upper adjacent values of SVWC in the periods of debris flow and landslide, but the lower adjacent values in the period of the landslide were smaller. No matter the stage of debris flow or landslide occurrence, soil temperature at each layer did not change with the change of soil depth (Fig. 4a). In the debris flow occurrence period, 10-cm soil temperature was distributed more discretely than that in other layers, while in the periods of landslide, except for the 20-cm soil layer, the soil temperature of the other three soil layers was relatively discrete. At every soil layer, the soil temperature median value of landslide was higher than that of debris flow. Different from the distribution of soil moisture, the lower adjacent values of the soil temperature at all layer were basically the same when debris flow and landslide occurring, while the upper adjacent values were higher when landslide occurring than in debris flow occurs. This shows that soil temperatures at all layers were lower during debris flow occurring than during landslides.
Compared with soil depth at 10 cm, 40 cm, and 80 cm, although soil temperature and SVWC at 20-cm layer during debris flow occurrence had a larger correlation coefficient (R 2 = 0.58), it did not meet the significance test (p < 0.05) (Fig. 4b). Except for no significant negative correlation between soil temperature and SVWC at 80-cm soil layer, soil Soil temperature and moisture during landslides and debris flows and their correlation. a represents the distribution of soil temperature and SVWC at different soil depths during debris flows and landslides. b, c represents the correlation between soil temperature and SVWC at different soil depths during debris flow and landslide, respectively temperature and SVWC in other soil layers showed insignificant positive correlation during debris flow occurrence. Different from debris flow scenario, soil temperature and SVWC under the landslide scenario showed insignificant negative correlation at 10 cm and 20 cm depths and showed insignificant positive correlation at 40 cm and 80 cm depths (Fig. 4c). Although there was no significant correlation, a higher correlation coefficient (R 2 = 0.67) at 10 cm depth appeared than at other layers during landslide occurrence. No matter during debris flow or landslide occurring, soil temperature and moisture at shallow soil layers showed better correlations than at deep soil layers (Fig. 4b, c).

Rainfall threshold characteristics at the time of disaster occurrence
The I-D threshold (rainfall intensity-duration thresholds) corresponding to the occurrence of debris flow and landslide in Zhejiang caused by Lekima can be found in Fig. 5a. During the occurrence of debris flow and landslide, the rainfall intensity and duration were in power function, which was I = 1247.73D −403.44 + 19.09 and I = 142.63D −0.58 − 3.37 for debris flow and landslide, respectively. During the occurrence of debris flow caused by Lekima, the rain intensity increased with the increase of duration, while during the occurrence of landslide, the rain intensity decreased with the increase of duration in power function. The rainfall intensity and duration during landslides were more obvious than during debris flow occurring, which may be related to a large number of samples during landslides. The rainfall intensity of landslide and debris flow in 24 h were 18.1 and 19.1 mm, respectively. And rainfall intensity of landslide and debris flow in 48 h were 11.9 and 19.1 mm, respectively. There was no good correlation between accumulated rainfall and duration, no matter during landslide or debris flow (Fig. 5b). Meanwhile, the linear fitting thresholds between the accumulated rainfall and duration (E-D thresholds) under two scenarios are largely coincident. Compared with debris flow, the relationship between accumulated rainfall and duration of landslide was more discrete, which also showed that the threshold of landslide occurrence was lower and the occurrence scenarios were more abundant.
From the correlation between the effective antecedent precipitation and present hourly precipitation, no matter for debris flow or landslide points, the distribution was relatively concentrated, and the relationship satisfied y > 18.5x (Fig. 6a). In the case of small present hourly precipitation, the effective accumulated rainfall threshold of landslide was lower than debris flow. Most of the present hourly precipitation of debris flow disaster was between 0 and 10 mm, and the corresponding effective antecedent precipitation ranged in 400-800 mm. Compared with the debris flow points, the distribution of landslide point appeared more discrete, present hourly precipitation concentrated in 0-20 mm, and the corresponding effective antecedent precipitation also had a wider distribution (80-800 mm). By analyzing the relationship between the maximum 1-h accumulated rainfall and the maximum 3-h accumulated rainfall, the influence of extreme precipitation on the occurrence of debris flow and landslide can become more clear. The maximum 1-h accumulated rainfall had good linear correlations with the maximum 3-h accumulated rainfall during landslide and debris flow occurrence (Fig. 6b). With the increase of the maximum 1-h accumulated rainfall, the maximum 3-h accumulated rainfall also increased; the change rate was landslide scenario > debris flow scenario.

Different I-D thresholds in the southeast and northwest Zhejiang
The northwest and southeast of Zhejiang are prone to sudden geological disasters in case of rainstorm or typhoon due to large mountainous areas and poor mountain stability (Zhou et al., 2014). Debris flow and landslide also occurred mainly in the northwest and southeast of Zhejiang under Lekima influencing (Fig. 1). Caused by Lekima, 26 and 74 debris flow points located in the southeast and northwest Zhejiang respectively, and 100 and 38 landslide points located in the southeast and northwest Zhejiang respectively. In order to further explore the difference of I-D thresholds in different regions of Zhejiang, I-D thresholds of debris flow and landslide in the southeast and northwest are shown in Fig. 7. The northwestern and southeastern disasters are located in Linan city and Wenzhou city of Zhejiang Province respectively. The I-D thresholds given by different scholars can be found in Table 2.
Compared with studies of Ma et al. (2011Ma et al. ( , 2015 and Bao et al. (2016), I-D thresholds of debris flow and landslide caused by Lekima did not satisfy the power exponent correlation (Fig. 7). Shorter rainfall duration ranges (24 h) and longer time series (1990-2003 or 1990-2013) in studies of Ma et al. (2011Ma et al. ( , 2015 and Bao et al. (2016) may be the main reason for this, because a longer time series is more conducive to statistics, and a more concentrated duration can reduce the dispersion of samples. The power exponential E-D thresholds can be found for landslides in the southeast Zhejiang (E = 868.04D 0.16 -899.22) and for debris flow in both regions (E = 41.98D 0.47 − 372.02 in the southeast Zhejiang and E = 26.59D 0.71 − 214.56 in the northwest Zhejiang) (Fig. 7).

Meteorological and soil conditions during land-in-Zhejiang typhoons
During 2000-2019, 18 tropical cyclones above tropical depression level (TD) have landed in Zhejiang (Table 3) (Ying et al., 2014;Lu et al., 2021). As the minute level soil moisture data started from 2015, soil volume water content (SVWC) in 2015-2019 of typhoons influencing Zhejiang were compared horizontally (Fig. 8). SVWC increased with the increase of soil depth. Caused by the precipitation before typhoons landing, SVWC increased before typhoons landing. In the "Meteorological and soil conditions at the time of disaster occurrence" section, we found that SVWC at each layer fluctuated during the typhoon Lekima influencing, but this phenomenon did not exist when the other three typhoons' influencing. SVWC fluctuating may be caused by typhoon Lekima's longer influence time, larger rainfall and wind speed, and wider influence range than the other three typhoons. The center of positive vorticity coincides with the center of strong divergence in the lower troposphere, and the height of convective development is high, which is conducive to sustained heavy rainfall (Zheng et al., 2021). The super large precipitation made the soil moisture at all layers affected by the typhoon, and the long influence time and large influence range make the SVWC of the whole Zhejiang continuously affected by Lekima. In addition, Zhang et al. (2021) pointed out that the rainfall of typhoons can reduce the depth of shallow groundwater in the affected area. The decrease of shallow groundwater depth further led to the change of groundwater recharge and soil flow, which may also cause the fluctuation of SVWC during Lekima landing.
With the increase of soil depth, the variation range of soil temperature during the typhoon influence period gradually decreased, which indicated that the influence of typhoon on deep soil temperature was less than on surface soil temperature (Fig. 9). Except for the typhoons in August, few typhoons in other months caused the change of soil temperature at 320 cm. No matter for the typhoon influencing period or the average soil temperature during 2000-2020, soil temperature at 5-80 cm depth in July and August were higher than that in September and October. The reason for this was that higher surface air temperature appeared in July and August and surface air temperature strongly influenced the shallow soil freeze-thaw status (Li et al., 2021). Due to the slow energy exchange in deep soil, the temperature of deep soil (160 cm and 320 cm) did not rise when the air temperature and shallow soil temperature increased in July. Also caused by the slow energy exchange, the temperature rise of  ( The regional average of soil temperature at different soil layers (5 cm, 10 cm, 15 cm, 20 cm, 40 cm, 80 cm, 160 cm, and 320 cm) during typhoons influencing Zhejiang deep soil from winter to summer was later than at shallow soil, which lead to lower temperature in July at 160 cm and 320 cm, and soil temperature gradually increases from July to October at 320 cm depth.

Conclusions
In this study, observed minute precipitation data, hourly meteorological data and soil data from 2019-08-07 00:00:00 to 2019-08-12 23:55:00 are employed to investigate rainfall thresholds of debris flow and landslide during Lekima influencing Zhejiang. The meteorological and soil conditions under debris flow and landslide scenarios are also examined. Correlations between soil temperature and moisture at different soil depths, between effective antecedent precipitation and present hourly precipitation, between maximum one hour accumulated rainfall and the maximum three hour accumulated rainfall during occurrence time of debris flow and landslide are further investigated. The major findings are summarized as follows.
1. The soil volume water content (SVWC) at deep soil layers (50-100 cm) fluctuated continuously during the occurrence of debris flow and landslide. At the end of the landslide period, SVWC at deep soil layers returned stable. 2. At the same soil layer, there was little difference between the upper adjacent values of SVWC in the periods of debris flow and landslide, but the lower adjacent values in the period of landslide were smaller. However, different from the distribution of soil moisture, the lower adjacent values of the soil temperature at all layers were basically the same when debris flow and landslide occur, while the upper adjacent values were higher when landslide occurs than they in debris flow occurs. 3. Compared with the whole typhoon period and debris flow occurrence period, ground surface temperature and surface air temperature, soil temperature, and surface air temperature showed better correlations in landslide occurrence period. 4. Both during debris flow and landslide occurrence period, the soil temperature at all layers did not change with the change of soil depth. Soil temperature and moisture at shallow soil layers owned better correlations than at deep soil layers. 5. During the occurrence of debris flow and landslide, the rainfall intensity and duration (I-D) threshold were in power function, which is I = 142.63D −0.58 − 3.37 and I = 1247.73D −403.44 + 19.09 for landslide and debris flow, respectively. The accumulated rainfall and duration (E-D) threshold did not show power as exponential as the I-D threshold showed, and the correlation was not significant.