Occurrence of Irreducible Water and Its Influences on Gas-Bearing Property of Gas Shales From Shallow Longmaxi Formation in the Xishui Area, Guizhou, Southern China

Systematic studies are quite rare on the gas-bearing property and its controlling factors of the shallow Longmaxi shale outside the Sichuan Basin. In a previous study, the gas in place contents of a suit of Longmaxi shale samples with a depth range of 362–394 m from the well XK2, which was drilled in the Xishui area, Guizhou, southern China, were reported. In the present study, the pore structure parameters and irreducible water occurrence characteristics of those samples, and their influences on the gas-bearing property were further investigated. The results show that, compared to the dry sample, the non-micropore specific surface areas and micropore volumes of the moist sample are significantly reduced by an average value of 61 and 30%, respectively, and that the water averagely occupies 82 and 41% of the inorganic and organic non-micropore specific surface areas, respectively, and 44 and 18% of the inorganic and organic micropore volumes, respectively. The shallow shale reservoir is dominated by adsorbed gas. It accounts for 66–93% of the total gas. The water significantly decreases the adsorption capacity of the inorganic matte (mainly clay minerals) pores, but has a limited influence on that of the organic matter pores. The adsorbed gas occurs mostly in the organic matter nanopores, and even if the shales were highly saturated with the water, they can still store a certain amount of the adsorbed gas. These results are to provide some guides for the evaluation and exploration of the shallow Longmaxi shale located in the strongly tectonic transformation areas of southern China.


INTRODUCTION
Shale gas, mainly stored as free and adsorbed states, is a typical unconventional natural gas (Jarvie et al., 2007;Bustin et al., 2008;Ross and Bustin, 2009). For different shale reservoirs, not only do their gas contents vary greatly, but the percentages of the adsorbed and free gasses can differ widely. For examples, the percentages of adsorbed gas in the Barnett and Lewis shales (United States) are approximately 20% and in a range of 60-85%, respectively (Curtis, 2002), and the free gas percentage of Longmaxi shale samples from the Jiaoshiba Block in the Sichuan Basin, varies from 60 to 75% (Dong et al., 2014). The gas-bearing property (GBP) and gas in place (GIP) content of shales are not only related to their geological conditions (temperature and pressure), but also restricted by their geochemical properties. For a specific gas shale, the free gas content is mainly related to its porosity, gas saturation, and geological temperature and pressure Gasparik et al., 2012;Pan and Connell, 2015;Ye et al., 2016;, but the controlling mechanism of the adsorbed gas is more complex. Except for the factors mentioned above, the adsorbed gas content is also correlated with its pore type, property and structure, which are comprehensively controlled by the organic matter content and type, mineral composition, and thermal maturity (Yang and Aplin, 1998;Dewhurst et al., 1999;Bustin, 2007, 2008; Ross and Bustin, 2007;Curtis et al., 2010Curtis et al., , 2011Song et al., 2013;Zhang et al., 2013;Shan et al., 2017;Chen et al., 2018;Hou et al., 2018;Li et al., 2018;Bhowrnik and Dutta, 2019;Klewiah et al., 2019;Wang and Guo, 2019;Gou et al., 2020;Zhang Y. F. et al., 2020). High-pressure adsorption experiments based on dry samples are generally carried out to evaluate the adsorption capacity of the targeted shales. The achieved results show that, for high and over mature shales, nanopores are extremely developed in their OM (organic matter), and the TOC (total organic carbon) content presents an obvious positive linear relation to the adsorption capacity Ross and Bustin, 2009;Gasparik et al., 2012Gasparik et al., , 2014Tan et al., 2014;Pan et al., 2016;Tian et al., 2016;Fu, 2017;Wang and Guo, 2019), with a R 2 (correlation coefficient) of 0.67-0.99 Fu, 2017). Moreover, clay minerals can also have some gas adsorption capacities to methane (Aringhieri, 2004;Sondergeld et al., 2010;Curtis et al., 2011;Ji et al., 2012;Gasparik et al., 2014;Chen et al., 2016). According to the methane adsorption experiments carried out by Ji et al. (2012), kaolinite, chlorite, and illite could adsorb methane with an amount up to 2.5, 2.3, and 1.8 m 3 /t, respectively.
Shale reservoirs always have some water in their pore system (Li et al., 2016a;Hu et al., 2018). Recently, quite a few papers have been related to the occurrence characteristics and mechanism of the pore water in gas shales and its influences on their adsorption capacities Ross and Bustin, 2009;Gasparik et al., 2014;Chen et al., 2016;Li et al., 2016b;Liu et al., 2018;Bai et al., 2020;Wang et al., 2020). The results imply that water can coexist in non-micropores and micropores in absorbed and condensed states, respectively (Newsham et al., 2003;Charriere and Philippe, 2010), and distributes in both OM and IM (inorganic matter) pores (Cheng et al., 2017(Cheng et al., , 2018, which significantly reduces the methane adsorption capacity (Ross and Bustin, 2009;Gasparik et al., 2014). Nevertheless, those findings are mainly obtained from the laboratory data, with determination results calculated from relevant theoretical models for samples with different water saturations derived from equilibrium water adsorption experiments (Chareonsuppanimit et al., 2012;Gasparik et al., 2012Gasparik et al., , 2014Ji et al., 2012;Zhang et al., 2012;Yaguzhinsky et al., 2013), lacking the constraint and validation of data from geological gas shales (Li et al., 2016b).
After the commercial development of the Longmaxi shale gas in the Sichuan Basin, southern China (Li and Zhang, 2015;Guo, 2016), the exploration of shallow marine shale gas outside the Sichuan Basin has made significant progresses. For instance, the GIP content of the Longmaxi shale in the Zhaotong Block, Yunnan Province can reach 3.81 m 3 /t, with an initial gas production of 0.39-1.12 × 10 4 m 3 /d although its burial depth is less than 1,000 m (Tian, 2018;Yang, 2018;Liang et al., 2020). The shallow marine shale gas in southern China is expected to become a new exploration field (Li, 2018;Liang et al., 2020).
The shallow Longmaxi shale in northern Guizhou develops widely, and is expected to have a huge amount of shale gas resources (Liang et al., 2008;Li et al., 2012;Hu et al., 2014). Sun et al. (2020) reported the data of geochemistry and GIP contents of a suit of Longmaxi shale samples taken from the well XK2 in this area, and revealed that the shallow Longmaxi shale really contains a high content of shale gas. On the basis of the previous work, the pore structures of those shale samples under the moist and dry conditions were analyzed, respectively, the pore water occurrence characteristics in nanopores were investigated, and the influences of them on the GBP were discussed.

Samples
There are totally sixteen Longmaxi shale samples investigated in this study, which were collected from the well XK2 in northern Guizhou, southern China (Figure 1). The collection and preparation methods of those samples, and their geochemical characteristics, pore water contents and total porosity were reported by Sun et al. (2020) and are briefly summarized here. The 16 shale core samples (with a depth interval of 362-394 m) were collected at an interval of about 2 m. Each sample was quickly divided into two parts. One part was immediately conducted on the GIP content test, and the other part was sealed in plastic bags and sent to the laboratory for the analysis of the pore water content, TOC content, mineral composition, total porosity . The pore water content of the studied shale samples was measured by the oven-drying and weight-loss method, each sample was drilled from its middle to obtain a small cylinder sample (15 mm in diameter and 20 mm in length). The shale TOC content was measured by a LECO CS-230 carbon-sulfur analyzer, the samples were crushed into powders of 80 mesh sizes. X-ray diffraction (XRD) analysis of the sample powders (200 mesh) was carried out using a Bruker D8 Advance X-ray diffractometer. The total porosity was determined by the differences between the shale skeletal density and the apparent density, the two densities were measured with a helium pycnometer (Ultrapyc 1200e) and a hydrometer (DAHO-120M), respectively.
The samples have the TOC, clay mineral, quartz, and carbonate mineral contents of 1.50-8.04%, 20.70-57.00%, 25.00-60.00%, and 4.80-15.80%, respectively, with the porosity of 3.03-5.50% ( Table 1). The measured pore water content of those samples is between 5.01 and 12.38 mg/g ( Table 1). It was deemed to be the connate irreducible water content by Sun et al. (2020). This is because the samples were taken from a gas-bearing shale stratum with an ultra-low water saturation, and the gas in the shales prevented the inflow of outside water during the drilling and sampling processes. Therefore, it should be kept in mind that the pore water content of the studied samples can be basically representative for that in their reservoir conditions.

Low Pressure N 2 and CO 2 Adsorption Experiments
The non-micropore (mesopore + macropore) and micropore structures of the moist and dry shale samples were tested using low-pressure N 2 and CO 2 adsorption experiments, respectively, by a Micromeritics ASAP 2020 apparatus. The main experimental methods are summarized below. The samples were crushed into 20-40 meshes (380-830 µm), and each sample was prepared into two parts. One part retained the irreducible water (i.e., the moist sample), and the other part was dried in a vacuum oven for 12 h at a temperature of 110 • C to remove the water and/or volatile substances (i.e., the dry sample). For the dry sample, loaded approximately 1 g into sample tube for further degassing under the high vacuum conditions (<10 mmHg) at 110 • C for 12 h in degassing station of the apparatus, and then it was moved to the analysis station for the adsorption experiment. The moist sample was directly placed in the analysis station for the freezing, degassing and adsorption experiments. The samples were tested at a P/P 0 (the relative pressure) of 0.005-0.995 and a liquid nitrogen temperature of 77.35 K for the N 2 adsorption, and at a P/P 0 of 1 × 10 −5 -0.03 and a temperature of 273.15 K (icewater mixture environment) for the CO 2 adsorption. For the N 2 adsorption, the adsorption and desorption isotherms were tested to analyze the type of hysteresis. According to the N 2 adsorption isotherms, the modified BET (Brunauer-Emmett-Teller) equation was utilized to determine the non-micropore specific surface area (Brunauer et al., 1938;Gregg and Sing, 1982;Tian et al., 2015), and the BJH (Barrett-Joyner-Halenda) model was adopted to determine the non-micropore pore-size distribution (PSD) (Barrett et al., 1951;Tian et al., 2013;Pan et al., 2015) for the dry and moist samples. The micropore volume was determined by DA (Dubinin-Astakhov) equation (Dubinin, 1989;Clarkson and Bustin, 1999;Ross and Bustin, 2009;Cheng et al., 2017), and the DFT (Density Functional Theory) was utilized calculating the micropore PSD on the basis of the CO 2 adsorption data from the both sample sets (Mastalerz et al., 2013;Lin et al., 2014).

Calculation of Water Saturation
According to the pore water content, porosity and apparent density data of the shale samples from the well XK2 (Table 1), their water saturation (S W , %) can be determined by Eq. 1: where C IW is the pore water content (mg/g); is the porosity (%); ρ sample is the apparent density (g/cm 3 ); ρ w is the pore water density (g/cm 3 ), and it is supposed that its density is 1 g/cm 3 for all samples.

Calculation of the Shale Free and Adsorbed Gas Contents
According to the shale gas composition from the well XK2 reported by Sun et al. (2020), the shale gas from the XK2 well shale samples consists of CH 4 , C 2 H 6 , and CO 2 with an average content of 93.99, 1.21, and 4.80%, respectively, while other gasses, such as C 3 , N 2 and so on were not detected due to their negligible contents. Thus, the average molar mass (M) of the shale gas was calculated using Eq. 2. The free gas density (ρ b ) was obtained by a PVTsim software. Combined with the porosity, water saturation and apparent density data, the free gas content (Q free ) was determined using Eq. 3. The adsorbed gas content (Q ads ) was obtained by the difference between the total gas content (Q total ) (i.e., the GIP content) and the Q free (Eq. 4). These equations are as follows: where M is the average molar mass of the shale gas (g/mol); Q free , Q ads , and Q total are the free, adsorbed, and total gas contents, respectively (m 3 /t); M 1 , M 2 , and M 3 are the molar mass of CH 4 , C 2 H 6 and CO 2 , respectively (g/mol); a, b, and c are the volume percentages of CH 4 , C 2 H 6 , and CO 2 , respectively (%); ρ b is the free gas density (g/ml), and it can be determined using a PVTsim software based on the reservoir temperature, pressure, and gas composition data.

Pore Structures of the Moist and Dry Samples
Low Pressure N 2 Adsorption and Pore Distribution . Figure 2 indicates that the adsorption quantity of the moist sample is definitely less than that of its corresponding dry sample. Furthermore, the isotherm shapes of the dry and moist samples are significantly different. For the dry sample, the isotherm of N 2 adsorption is very close to the type II isotherm according to the IUPAC (International Union of Pure and Applied Chemistry) (Brunauer et al., 1940;Sing et al., 1985;Rouquerol et al., 1994). It shows a distinct hysteresis loop, without plateau at the p/p 0 approximately 1, and a notable adsorption quantity, besides, can be seen at the lowest p/p 0 (<0.01), implying that the mesopore, macropore, and micropore are well developed in the dry sample (Sing et al., 1985). The difference between the moist and dry samples indicates that water influences the N 2 adsorption capacity and pore structure. Besides, the N 2 adsorption quantities of different samples are likely related to their TOC contents, which will be discussed in see section "Controlling Factors of the Shale Pore Structure." Frontiers in Earth Science | www.frontiersin.org Note that water in sample JX-2 (B) occupies a broader pore size range compared with other samples shown in A,C-H, up to 110 nm, may be due to, in part, its greater pore water content (Table 2).
Meanwhile, the shape of hysteretic loop for the dry sample is typically Type B, indicating the fracture-type pore is dominant (Brunauer et al., 1940;De, 1958). However, the hysteresis loop of the moist sample is less developed than that of the dry sample, especially the samples with low TOC contents (such as samples JX-1, JX-2, and JX-4 concomitantly with a higher water content), their hysteresis loops are not notable. It can be seen from the above results that not only the irreducible water in the shales takes up some pore spaces, but it also changes the pore structure characteristics to some extent, especially for the low TOC samples. The BET surface area (S BET ), representative for nonmicropores, of the dry sample is 6.16-14.83 m 2 /g and averages 10.17 m 2 /g. The S BET of the moist sample is obviously lower than that of its corresponding dry sample, and has a range of 2.15-6.66 m 2 /g (the average value is about 4.00 m 2 /g) ( Table 1). Compared with the dry sample, the S BET of the moist sample is reduced averagely by 61%. Figure 3 presents the BJH pore-size distributions (PSD) for the moist and dry samples. For the dry sample, the PSD tends to decrease with increasing pore size, the nonmicropores are mainly distributed in the pores <10 nm, and meanwhile, an overall increase in PSD value was observed with increasing TOC. However, for the moist sample, the PSD shows a unimodal peak, and the main peak appears in the pore range of 4-7 nm (Figure 3). Compared with the dry sample, the PSD of the moist sample decreases notably in the pores <10 nm. The difference between the dry and moist samples is not significant for the pores >10 nm, indicating the water is mostly distributed in the pores <10 nm. It should be noted that, compared with samples JX1 and JX7, the adsorption quantity and PSD value of sample JX2 are slightly higher under the dry condition and lower under the moist condition, which is likely attributed to its greater clay mineral content (51.60%) ( Table 2).  Figure 4 shows the CO 2 adsorption isotherms of the dry and moist samples. The shape of the adsorption isotherms of the dry sample exhibits Type I, which likes that of the moist sample, suggesting both the dry and moist samples present abundant micropores. Furthermore, the CO 2 adsorption quantities of the moist sample are slightly less than those of the corresponding dry sample. Combined with Figures 2, 3, it can be considered that the influence from the irreducible water on micropores is relatively less than that on non-micropores. Frontiers in Earth Science | www.frontiersin.org The micropore volume (V mic ) of the dry sample is 0.0060-0.0131 cm 3 /g (an average of 0.0099 cm 3 /g), while the moist sample has a V mic of 0.0039-0.0099 cm 3 /g (an average of 0.0069 cm 3 /g) ( Table 1). Compared to the dry sample, the V mic of the moist sample cut down about 30% on average, much less than the reduction extent of the S BET (61% on average).
The micropore PSDs of the moist and dry samples were presented in Figure 5. It is clear that the dry sample displays a similar PSD to that of the moist sample, and both of them show bimodal characteristics. The main peaks of the PSD values of the dry sample occur at approximately 0.52-0.58 nm and 0.84-0.89 nm, respectively, and those of the moist sample occur at 0.54-0.59 nm and 0.79-0.87 nm, respectively. Moreover, the PSD value of the moist sample differs little from that of the corresponding dry sample, further illustrating that the influence of the irreducible water on micropores is less than that on non-micropores. It should be pointed out that the maximum equilibrium pressure (P/P 0 ) of the CO 2 adsorption experiment in the present study was about 0.03 (1 bar), which is significant lower than the CO 2 saturation pressure at 273 K (34.5 bar), hence the effective micropore range characterized by our CO 2 adsorption experiment is only 0.36-0.93 nm, while the larger micropores (0.93-2.00 nm) were not detected. Figure 6 displays the relationships of TOC and clay mineral contents with the S BET (Figures 6A,B) and V mic (Figures 6C,D).

Controlling Factors of the Shale Pore Structure
To avoid the impact of clay minerals on the pore structure parameters being masked by TOC, the S BET , and V mic were normalized to per unit TOC (1%), so they can directly reflect the contribution of clay minerals to the S BET and V mic . For the dry sample, the S BET is positively linearly correlated with TOC, having a correlation coefficient(R 2 )of 0.92 (Figure 6A), and a linear relationship also exists between the S BET and the clay mineral (R 2 = 0.47) (Figure 6B). These results imply the S BET of the dry sample is controlled by both TOC and clay minerals, and the former is the main controlling factor, which is because the clay minerals in the studied samples consist of chlorite and illite  with low specific surface areas (Gasparik et al., 2012;Ji et al., 2012;Liu et al., 2018). For the moist sample, the S BET still has an obvious positive linear correlation with TOC (R 2 = 0.93) (Figure 6A), however, its correlation with clay minerals decreases (the R 2 reduces from 0.47 to 0.36) ( Figure 6B). Therefore, a conclusion can be made that the irreducible water shows a more obvious influence on the non-micropores of clay minerals compared with that of organic matter.
The V mic of the dry sample shows a notable positive linear correlation with TOC (R 2 = 0.96) (Figure 6C), and a weak positive linear correlation with clay minerals (R 2 = 0.32) ( Figure 6D). For the moist sample, the V mic is still apparently correlated to TOC (R 2 = 0.95) (Figure 6C), while the weak linear relationship seems exist between the V mic and clay minerals, but having a lower R 2 value (0.29) compared to FIGURE 7 | Percentages occupied by irreducible water in the non-micropores (A) and micropores (B) for the selected 8 samples. It should be pointed out that pores smaller than 0.40 nm were not counted since the diameter of H 2 O molecules is approximately 0.40 nm, and the percentages occupied by water in the pores of 0.93-2.00 nm were not observed since the effective micropore range is only 0.36-0.93 nm in the present study. the dry sample ( Figure 6D). Combined with Figure 6B, it is clear that the effect of the irreducible water on micropores is less significant than that on non-micropores. This is because clay minerals have a greater contribution to mesopores and macropores than that to the micropores Bakshi et al., 2017Bakshi et al., , 2018.

Occurrence of Pore Water in Different Size Nanopores
Occurrence of the irreducible water in different size nanopores can be reflected by their volumes occupied by the water in different pore sizes relative to the total pore volume occupied by the water. For a specific shale sample, these volumes occupied by the water can be largely calculated by the difference between its dry and moist conditions. As seen from the results of the representative samples in Figure 7, in the range of nonmicropores, the percentages occupied by the water decrease with increasing pore size. The water occurs mainly in the pores of 2-10 nm, and the occupation percentage varies from 85.50 to 97.34%. This is followed by the pores of 10-50 nm, with an occupation percentage range of 1.44-10.88%, while the percentage occupied by the water in macropores (>50 nm) is the lowest, which is 0.42-3.62% (Figure 7A). In the micropore range, the water occurs mostly in the pores of 0.40-0.60 nm and 0.80-0.93 nm, with average percentages of 57.54 and 30.07%, respectively. The occupied volume in pores between 0.60 and 0.80 nm is the least, with an average percentage of about 12.39% ( Figure 7B). According to the above results, the irreducible water distributes mainly in mesopores of <10 nm and micropores of 0.40-0.60 nm and 0.80-0.93 nm.

Occurrence of Irreducible Water in OM and IM Nanopores
It is conventionally believed that inorganic matter (such as clay minerals) pores are hydrophilic and the organic pores are hydrophobic. However, a few recent papers document that water occurs widely in organic pores especially in high and over mature shales (Cheng et al., 2017(Cheng et al., , 2018. There  Frontiers in Earth Science | www.frontiersin.org are mainly two understandings on the mechanism of the hydrophilicity of organic pore. One is that some organic pore surfaces contain hydrophilic adsorption sites, such as oxygencontaining groups (Bahadur et al., 2017), water molecules can be adsorbed to the organic pore surfaces by forming hydrogen bonds with oxygen-containing groups. Moreover, even if the organic pore contains no containing-oxygen groups, it is inherently hydrophilic (Kozbial et al., 2014;Wei and Jia, 2015). Therefore, water can occupy both in shale inorganic and organic pores.
It is difficult to determine the OM and IM pores of shale samples precisely (Löhr et al., 2015). The means based on the linear positive relationship between TOC and total porosity for a set of shale samples is commonly adopted at present. According to this method, the OM porosity of a shale sample is the product of its TOC and the regressed line slope, and the difference between its total porosity and OM porosity is considered to be the IM porosity Tian et al., 2013Tian et al., , 2015Hu et al., 2017;Sun et al., 2020). In the present study, the S BET and V mic of both the moist and dry samples exhibit strong positive linear correlations with TOC (R 2 values of 0.91-0.95) (Figures 6A,C). Therefore, the above linear regression method can be also applied to estimate the OM and IM pore structure parameters (S BET and V mic ) of the studied samples, as reported by Cheng et al. (2018). The OM S BET of a shale sample is the product of its TOC and the regressed line slope (Figure 6A), and the difference between its total S BET and OM S BET is considered to be the IM S BET ; the estimation of OM, IM V mic is based on the same method.
The determined results indicate that the S BET−OM (BET surface area of OM) and S BET−IM (BET surface area of IM) values of the dry sample are 1.71-9.16 and 3.26-5.93 m 2 /g, respectively ( Table 1), and that the S BET−OM is greater than the S BET−IM for the samples with TOC contents >4.25% (Figure 8A). However, for the moist sample, the S BET−OM and S BET−IM values are 1.01-5.41 and 0.26-1.30 m 2 /g, respectively (Table 1), and the S BET−OM is greater than the S BET−IM for all studied samples ( Figure 8B).
The V mic−OM (OM micropore volume) and V mic−IM (IM micropore volume) values of the dry sample are 0.0017-0.0088 and 0.0039-0.0055 cm 3 /g, respectively, and those of the moist sample are 0.0014-0.0072 and 0.0019-0.0034 cm 3 /g, respectively (Table 1). Moreover, the V mic−OM is greater than the V mic−IM for the dry samples with TOC contents >4.34% ( Figure 8C) and for the moist samples with TOC contents >3.09% (Figure 8D).
It was supposed that the difference of pore structure parameters between the moist and dry samples could present the space taken by the irreducible water. A further calculation based on the data in Table 1 reports that the water occupies 82% of the S BET−IM and 41% of the S BET−OM , and 44% of the V mic−IM and 18% of the V mic−OM for the studied shale samples. Therefore, it is believed the irreducible water occurs preferentially in the IM pores, occupying a greater part of their non-micropore surface areas, and near a half of their micropore volumes, but its influence on the OM pores is limited, and most of the OM pores can still provide spaces for the storage of shale gas.

Control of TOC on Gas-Bearing Property
The water saturation (S W ) of the studied shale samples varies greatly, ranging from 27.71 to 84.28%, with an average of 45.83% (Table 3). It is close to that of the Longmaxi shales from the wells Shao101 and YQ1 (>60%) outside the Sichuan Basin (Liu and Wang, 2013;Fang et al., 2014), and greater than that of the Longmaxi shales in the Sichuan Basin (10-45%) (Liu and Wang, 2013;Fang et al., 2014;Wei and Wei, 2014) and the gas shales in North America (10-35%) (Boyer et al., 2006;Bowker, 2007; Cipolla et al., 2010;Wang, 2012). As illustrated in Figure 9, the S W is positively related to the total clay minerals (R 2 = 0.76, Figure 9A), but presents a negative linear relationship to some extent with TOC (R 2 = 0.47, Figure 9B). This implies the irreducible water not only mostly occurs, but also has a greater saturation in the pores of clay minerals. The calculated free gas (Q free ) and adsorbed gas (Q ads ) of the studied samples are presented in Table 3. The Q free is 0.08-0.63 m 3 /t and averages 0.37 m 3 /t, while the Q ads is 0.75-2.56 m 3 /t and averages 1.30 m 3 /t, accounting for 66.06-93.45% of the total gas (77.62% on average). Pan et al. (2016) reported a GIP model based on a Longmaxi shale (TOC = 3.98%) from the Pengshui Block, South China. According to their GIP model, at the depth of 400 m, the Q ads is about 88% of the GIP content without taking into account the influence of the pore water. For the well XK2 in the present study, the adsorbed gas percentage of the shale samples with TOC≈4% is in the range of 70-80%, indicating the irreducible water should reduce the Q ads to some extent. Figure 10 presents the positive linear relationships of TOC with the Q free and Q ads for the studied samples, with R 2 values of 0.67 and 0.80, respectively. The correlation coefficient between the TOC and Q ads is significantly lower than that between the TOC and methane adsorption capacity reported by Pan et al. (2016) on the basis of dry Longmaxi samples (R 2 : 0.96-0.97), and similar to that proposed by Chalmers and Bustin (2008) based on the moisture-equilibrated shale samples (R 2 = 0.67). Therefore, compared with the dry shale samples, not only the actual adsorption capacity of the moist sample is reduced, but its correlation with TOC is also decreased, and the highpressure CH 4 adsorption experiment results conducted on dry shale samples can not be directly applied to the geological adsorption gas evaluation.
Based on the linear regression equations presented in Figure 10, the Q ads and Q free in OM and IM pores for the studied samples were also estimated by a similar method discussed in section "Occurrence of Irreducible Water in OM and IM FIGURE 10 | Correlations of TOC with free, adsorbed, and total gasses for the studied shale samples. Their positive correlations indicate that TOC directly controls the gas-bearing property, so the adsorbed and free gas contents in the OM and IM pores could be determined from their linear regression equations.
Nanopores". The calculated Q ads and Q free in OM pores are 1.16 and 0.35 m 3 /t on averages, respectively, accounting for 89% of the total absorbed gas and 95% of the total free gas, and those in IM pores are 0.14 and 0.02 m 3 /t on averages, respectively, accounting for 11% of the total absorbed gas and 5% of the total free gas. From those results, a conclusion could be draw that the shale gas stored mostly in OM pores, and IM pores contain a little gas due to the effect of irreducible water.
Relationship Between Irreducible Water and Gas-Bearing Property Figure 11 presents the relationships of the GBP (presented by Q total , Q ads , and Q free ) with the water content (Figures 11A,B) and water saturation (S W ) (Figures 11C,D). The Q ads shows a rapid drop and then a slow decrease ( Figure 11A) with increasing the water content. Since the shale gas of the studied samples is mostly in the adsorbed state, the trend of the Q total follows that of the Q ads (Figure 11B). In Figure 11A, the data of Lower Cretaceous shales from British Columbia, Canada reported by Ross and Bustin (2009) were included, and the two sets of data seem keeping the same trend with increasing water content.
The relationships of the S W with the Q total , Q ads , and Q free follow similar trends as described above. As the S W is less than 40%, the Q ads and Q tota decrease rapidly with increasing S W (Figures 11C,D), but these trends tend to slow with further increasing S W (Figures 11C,D). This result is in agreement with that of the Longmaxi shale in the Weiyuan-Changning Block, Sichuan Basin reported by Hu et al. (2018). They revealed that the Q ads declined significantly in the S W range of <30%, while it decreased slowly when the S W increased from 30 to 50%.
The influences of the irreducible water on the GBP can be explained by its occurrence characteristics in different nanopores. According to our results, the irreducible water occurs mostly in non-micropores, especially the non-micropores of clay minerals, with a great reduction of their non-micropore surface areas. Although OM pores also contain a small amount of irreducible water, the reductions of their non-micropore surface areas and micropore volumes are limited due to their hydrophobic nature. Therefore, the irreducible water mainly affects the adsorption capacity of IM pores, and OM pores contribute mostly to the shale adsorption capacity. In conditions of low S W , the water preferentially occupies the pores in clay minerals, significantly decreases their adsorption capacities, causing the adsorbed gas decrease rapidly with increasing the water content or S W . With the further increase of S W , the reduction of adsorption capacity is still mainly from clay minerals even when the S W reaches a high level, while the OM pores still have a certain adsorption capacity. This is why the absorbed gas content decreases slowly FIGURE 11 | Plots showing the relationships of irreducible water contents with the Q ads (A) and Q total (B), and the correlations of the S W with the Q ads (C) and Q total (D). The D-M shale data were measured at 30 • C (Ross and Bustin, 2009). It should be noted that, since the sample JX-16 is away from the trend of the regression line, the regression line have not included the data of this sample. with further increasing water content or S W as it exceeds a certain amount.
Figures 12A-D displays the correlations between the pore structure parameters (S BET and V mic ) and GBP (Q total, Q ads , and Q free ) for both the moist and dry samples. It is clear that their correlations are better for the moist sample than those for the dry sample. Figures 12E-H further represents the relationships of the Q total , Q ads , and Q free with either OM or IM pore structure parameters for the moist sample. They display obvious positive linear correlations with the OM pores, but reveal unclear correlations with the IM pores. Chalmers and Bustin (2008) made an explanation for a high CH 4 adsorption capacity of the moisture-equilibrated shale samples. They claimed gas and water took different sorption sites in shales. The present study results further demonstrate that the OM pores provide the most sorption sites for shale gas and control the Q ads , affecting the GBP.

CONCLUSION
The pore structure, irreducible water occurrence characteristics, gas-bearing property, and their relationships of the shallow Longmaxi shales from the well XK2 were investigated, and the main conclusions are as follows: • The micropore, mesopore, and macropore of the dry and moist samples are all well developed. Compared with the dry sample, the non-micropore BET surface area and micropore volume of the moist sample are reduced by 61 and 30% on averages, respectively. • The irreducible water occurs mainly in IM pores, especially smaller mesopores (2-10 nm). The water occupies averagely 82 and 41% of the inorganic and organic non-micropore BET surface areas, respectively, and 44 and 18% of the inorganic and organic micropore volumes, respectively.
• The irreducible water significantly reduces the adsorbed gas content in IM pores, but its influence on OM pores is limited, and the OM pores store approximately 89% of the adsorbed gas and 95% of the free gas.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.