Baseline Subsoil CO 2 Gas Measurements and Micrometeorological Monitoring: Above Canopy Turbulence Effects on the Subsoil CO 2 Dynamics in Temperate Deciduous Forest

Featured Application: This work emphasizes the importance of the deep soil zone and above groundwater that can be applied to create a CO 2 baseline for a wide range of geological storage applications. Abstract: Accurate and continuous measurement of the subsoil CO 2 is critical to better understand the terrestrial and atmosphere gas transfer process. This work aims to develop and ﬁeld test a speciﬁc ﬂow system to continuously measure the soil gas concentration ( χ c) and understand its main physical drivers. Hourly data measured in situ were collected through two dedicated wells at 1 m and 6 m depth coupled with micrometeorological measurement. Our study shows that χ c at -1 m was at the lowest in winter and highest in summer. Meanwhile, the seasonal variation of χ c at -6 m is somewhat unclear. While it is inevitable that temperature plays a signiﬁcant role, this factor related to biological activity cannot fully explain the variation. The decrease in χ c at both depths in summer coincides with an increase of friction velocity, especially during dry periods with R 2 of 0.68, which shows strong empirical evidence that wind turbulence plays a signiﬁcant role in driving the deep soil CO 2 . A monitoring strategy for gas measurement combining borehole and micrometeorological measurement offers excellent long-term monitoring possibilities to derive the vertical distribution of CO 2 and better understand the main physical drivers of gas resulting in increased CO 2 storage in deep soil during low turbulence. This study also demonstrates that permeability signiﬁcantly reduced during wet periods diminishing the molecular diffusion and advection. All of these highlight the importance of


Introduction
The gas exchange between the atmosphere and terrestrial ecosystem represents a large proportion of CO 2 and CH 4 atmospheric budgets [1]. Soil CO 2 budget and efflux due to land-use change averaged globally from 2009-2018 were estimated at 5496-8793.6 PgCO 2 and 5.49 PgCO 2 yr −1 respectively [2], as for the comparison, the emission due to fossil fuel combustion and the cement industry is around 33.4 PgCO 2 yr −1 [3]. Adequate and trustworthy monitoring strategies of natural gas exchange will help understand the local variability of gas concentration in the soil, e.g., in anticipation of the CO 2 injection [4,5] and promote public acceptability [6], or for ecological studies on terrestrial greenhouse gases emissions [3,[7][8][9]. Generally, GHG (Greenhouse gasses) emissions from the soils Measurement of the gas concentration using the low-resolution FTIR (Fourier-transform infrared spectroscopy) requires a calibration procedure by characterizing all of the spectrophotometer's instrumental line shape [39]. The chemical functional groups that absorb fundamental bands located at 2350 cm −1 for CO 2 and a combination of 1400-1200 cm −1 and 3200-2800 cm −1 for CH 4 can be used to quantify their concentration. FTIR technique takes into account the fact that the majority of the molecules absorb IR radiation, and the degree of the absorption is proportional to the concentration of the molecule. There are some significant conveniences in using FTIR including, (i) few optical elements; (ii) no need slits to attenuate radiation; (iii) outstanding stability and sensitivity (iv) high-speed data collection. FTIR also offers various benefits, such as near real-time operation, low maintenance, low cost, and simple structural specification [40,41]. Nonetheless, in term of its operation, FTIR spectroscopy has some limitations as follows: (i) This spectrometry can only be used for heteronuclear gases; (ii) it is considered as low resolution (1 cm −1 ) that requires specific laboratory calibration for every single gas; (iii) it is superimposed by the presence of impurities [5,42], (iv) its implementation on long-time span in the field is difficult due to the required energy power.
The main objectives of this study are three-fold; (i) To develop and field test a specific flow system equipped with a membrane probe to continuously measure the soil gas concentration and auxiliaries data; (ii) to combine borehole with micrometeorological monitoring; (iii) to understand the main physical drivers of gas exchange in a deep soil layer (−6 m) and a shallow soil layer (−1 m) regardless of the water saturation conditions. In this paper, both the gas calibration procedure in the laboratory are described, and the observational setup of the measurement campaign, aspects of the data analysis, and analysis of the recorded time series of soil CO 2 and CH 4 concentration are shown.

Gas Analyzer
IN-SITU IR: Infrared spectra of gases were recorded in-situ using a portable infrared spectrometer ALPHA provided by BRUKER (Bruker Optik GmbH, Ettlingen, Germany) with high humidity optics and zinc selenide (ZnSe) beamsplitter. The spectra were acquired with 16 scans in the mid-infrared range (5000-600 cm −1 ) with a spectral resolution of 4 and 1 cm −1 and a zero-filling factor of 2. A stainless-steel gas cell with a short path length (5 cm) with ZnSe windows was placed in the sampling module. The low-power NDIR sensor (home-made engineered) without moving part allows the continuous monitoring of solely CO 2 was also employed. In situ samples were also analyzed with a gas chromatography analyzer (CP-4900 Micro-Gc, Varian BV, Netherlands) to determine the soil gas composition, mainly CO 2 and CH 4 . A more detailed explanation regarding the calibration technique of the primary gas analyzer (FTIR) is given below.
EX-SITU FTIR: Because of the low resolution used (1 cm −1 ) for IR measurements, calibration curves for CO 2 and CH 4 analyses must be determined in the laboratory [43]. Two reference gases (CO 2 : 99.7% ± 2% and 5000 ppm ± 5%) were used to create a CO 2 calibration curve, and one reference gas (CH 4 : 5000 ppm ± 5%) was used to develop a CH 4 calibration curve. Furthermore, the reference gases were mixed using AlyTech GasMix™, an on-site customized gas standard preparation system with a relative uncertainty of less than 2% for various mixtures such as CO 2 and Ar, CO 2 and N 2 , CH 4 and Ar, and CH 4 and N 2 . The gas mix instrument is based on mixing, diluting, and injecting up to 12 gas standards at different concentrations that can be used for instrument calibration purposes. The gas standard was controlled in the software interface allowing the user to define the gas concentration, flow rate, and pressure. A G2201-I gas analyzer (Picarro Inc., Santa Clara, CA, USA) based on Cavity Ring-down Spectroscopy (CRDS) was used to verify the concentration indicated by the gas mix coupled with Gas Chromatography.
During the calibration procedure, the gas pressure inside the cell was measured using a built-in piezoresistive-based pressure gauge manufactured by Keller, ranging from 1 mbar to 1 kbar with an accuracy of 0.01-1%. A Pfeiffer vacuum model MVP 015-2 was Appl. Sci. 2021, 11, 1753 4 of 19 used during the measurement of references. The calibration was conducted at a constant room temperature of 22 • C and pressure of 1 ± 0.002 bar considering the total pressure most likely encountered ranges from 0.950 to 1.050 bar. Besides, the gas's temperature in the borehole is relatively stable, especially at −6 m, which was around 10 • C. The leak testing was carried out where this configuration managed to maintain its pressure of 0.02 bar within 15 min of the timeframe. The subtraction of atmospheric CO 2 was done for each recorded infrared spectrum to eliminate the ambient air background spectrum. The different spectral areas of CO 2 and CH 4 were calculated, as shown in Figure 1. As for the CO 2 concentration, the analysis is performed using the fundamental band located at 2350 cm −1 (ν 3 ) [2400; 2220] and for CH 4 using the fundamental bands at 1400-1200 cm −1 (ν 4 ) and 3200-2800 cm −1 (ν 3 ). room temperature of 22 °C and pressure of 1 ± 0.002 bar considering the total pressure most likely encountered ranges from 0.950 to 1.050 bar. Besides, the gas's temperature in the borehole is relatively stable, especially at −6 m, which was around 10 °C. The leak testing was carried out where this configuration managed to maintain its pressure of 0.02 bar within 15 min of the timeframe. The subtraction of atmospheric CO2 was done for each recorded infrared spectrum to eliminate the ambient air background spectrum. The different spectral areas of CO2 and CH4 were calculated, as shown in Figure 1. As for the CO2 concentration, the analysis is performed using the fundamental band located at 2350 cm −1 (ν 3) [2400; 2220] and for CH4 using the fundamental bands at 1400-1200 cm −1 (ν4) and 3200-2800 cm −1 (ν3).

On-Field Implementation
The gas measurement was performed in the boreholes dedicated to investigating soil-gas concentration in the unsaturated zone with two depths of 1 m and 6 m. Those boreholes were drilled in June 2019 using a geotechnical rig capable of performing rotary Appl. Sci. 2021, 11, 1753 5 of 19 drilling. A borehole consists of a perforated pipe and a permanent well that was fitted with drilling tubes. Figure 2 shows the subsoil measurement system composed of three main parts, including completion (A), gas circulation module (B), and calibrated FTIR Bruker Alpha (C), with an additional GC analysis for cross-validating purpose. This well is dedicated to recording the gas phase's geochemical properties in the subsoil, which operates continuously and simultaneously equipped with a logs data station. The configuration of the subsoil measurement system was subject to previous research [5,37].

On-Field Implementation
The gas measurement was performed in the boreholes dedicated to investigating soilgas concentration in the unsaturated zone with two depths of 1 m and 6 m. Those boreholes were drilled in June 2019 using a geotechnical rig capable of performing rotary drilling. A borehole consists of a perforated pipe and a permanent well that was fitted with drilling tubes. Figure 2 shows the subsoil measurement system composed of three main parts, including completion (A), gas circulation module (B), and calibrated FTIR Bruker Alpha (C), with an additional GC analysis for cross-validating purpose. This well is dedicated to recording the gas phase's geochemical properties in the subsoil, which operates continuously and simultaneously equipped with a logs data station. The configuration of the subsoil measurement system was subject to previous research [5,37]. Monitoring concept for simultaneous and continuous gas measurement, including schematic diagram (i) and realistic view (ii) of monitoring infrastructure. The schematic diagram describes in more detail the process flow, which includes completion/boreholes (A), the circulation module (B), and calibrated FTIR Bruker Alpha with an additional GC sensor (C) for cross-validating. The circulation module consists of (1) temperature sensor, (2) pressure sensor, (3) water trap, (4) fluxmeter, (5) packer pressure, (6) borehole temperature indicator, (7) atmospheric temperature sensor, (8) atmospheric pressure sensor, (9) CO2 NDIR sensor, and (10) soil moisture sensor.
The soil gases monitoring equipment (A) consists of three significant subparts: chamber, packer, and circulation lines. The collection chamber is an essential component that allows continuous sampling of the gas in the soil using an electric pumping system and a manual sampling port. The length and the diameter of the collection chamber is 60 cm and 50 mm, respectively. The packer made of an annular inflatable device of 64 mm long prevents any interferences with the ambient atmospheric gases by fully isolating the collection chamber. The circulation module consists of two lines of stainless steel, which is 4 mm in diameter allowing the gas to be transferred to the sensor and reintroduced into the borehole. The other two lines are equipped with pressure and temperature sensor. The gas flow module permits the gas sample's transportation from the sensor's completion Monitoring concept for simultaneous and continuous gas measurement, including schematic diagram (i) and realistic view (ii) of monitoring infrastructure. The schematic diagram describes in more detail the process flow, which includes completion/boreholes (A), the circulation module (B), and calibrated FTIR Bruker Alpha with an additional GC sensor (C) for cross-validating. The circulation module consists of (1) temperature sensor, (2) pressure sensor, (3) water trap, (4) fluxmeter, (5) packer pressure, (6) borehole temperature indicator, (7) atmospheric temperature sensor, (8) atmospheric pressure sensor, (9) CO 2 NDIR sensor, and (10) soil moisture sensor.
The soil gases monitoring equipment (A) consists of three significant subparts: chamber, packer, and circulation lines. The collection chamber is an essential component that allows continuous sampling of the gas in the soil using an electric pumping system and a manual sampling port. The length and the diameter of the collection chamber is 60 cm and 50 mm, respectively. The packer made of an annular inflatable device of 64 mm long prevents any interferences with the ambient atmospheric gases by fully isolating the collection chamber. The circulation module consists of two lines of stainless steel, which is 4 mm in diameter allowing the gas to be transferred to the sensor and reintroduced into the borehole. The other two lines are equipped with pressure and temperature sensor. The gas flow module permits the gas sample's transportation from the sensor's completion and controls all parameters associated with the fluid flow (pressure, temperature, and flow). The gas circulation module ( Figure 2B) was installed equipped with (a) circulation pump with a flow rate of 20 mL/min allowing the extraction and reintroduction of the gas in the collection chamber; (b) a water trap ensuring the control of the potentially liquid water in the circuit by trapping the water and redirect its flow; (c) the flow meter and a pressure sensor measuring the flow rate and the pressure of the gas sample inside of the circulation module; (d) sample cells connected in parallel to the gas circuit allowing samples of gas for laboratory analysis without causing the ambient atmospheric gases to enter the system; (e) various types and sizes of valves; and (f) an inlet and an outlet to FTIR Bruker Alpha. Since the monitoring site had no access to grid power, three batteries with a total capacity of 300 Ah were used to supply the monitoring system's power.

Field Test
The subsoil gas was collected using a collection chamber via the pump at the rate of 20 mL/min. It operated for 10 min every hour to allow well-mixed gas in the system and conserve energy consumption. The gas passed through the temperature, pressure, and NDIR sensor before it was analyzed hourly through the IR gas cell and re-injected to the borehole's collection chamber. For every follow-up visit for replacing the battery, some gas samples were also collected to be cross-validated with gas chromatography. Subsoil gas spectra were recorded using FTIR for 16 scans over the range of 5000 to 600 cm −1 with a spectral resolution of 1 cm −1 . The Opus software allows integrating the various processing, including a continuous record of the spectra, eliminating the H 2 O fingerprints, and calculating CO 2 concentration. Even though the water condensation can occur inside the gas circuit, it only interferes strongly with the combination band of 3609 cm −1 ( Figure 1). Therefore, an additional correction is unnecessary since it does not interfere with the fundamental vibrational mode of CO 2 at 2400 cm −1 . In addition, the data collected and recorded are automatically uploaded and synchronized to the online platform of Mission OS once the Internet network is available. The obtained field data were further processed using open-source statistical computing and graphics software [44].

Site Description and Additional Recorded Data
The study was conducted at the beech forest described extensively in previous work [45]. This site was chosen due to some main criteria, including a high diversity of soil sequence, well-managed beech trees, homogeneous (species, age, and stem density), and having a minimum influence of the past forest management practices. The three studied soils' physicochemical properties have also been previously studied [45,46]. The dystric cambisol lies on the valanginian layer to around 200 cm of the depth with a pHw of 4.9 at the surface and 5.1 at a depth of 45-60 cm. The cationic exchange capacity (CEC) ranged from 0.9-6.7 cmolc kg −1 up to a depth of 60 cm of the vertical section. Sandy and clayey passages were found in the lower soil layer indicating the textural and structural heterogeneity caused by the complex sedimentary source material. A flux tower, 355 m apart from the boreholes (Figure 3), was used to record air temperature (T air ) using sensors installed at six different levels (HMP155, Vaisala, Vantaa, Finland), friction velocity (u*), soil water content (SWC) and soil surface temperature (T soil ), and precipitation (P). It was built at the end of 2012 in the framework of the ICOS (integrated carbon observation system) and as part of the Andra's long-term observatories (Ope). The tower reaches 45 m height above the ground. It allows the continuous measurement of the wind's speed, direction, and friction velocity (u*) at 35 m elevation (around 10m above forest canopy) using six 3D sonic anemometers (HS-50 Gill, Hampshire, UK) with a sampling rate of 10 Hz, and data were compiled per 30 min interval.
In this site, the wind direction is dominated by South and Southwest directions. The annual time series of environmental drivers and atmospheric χc (35 m) in 2018 showed that the monthly mean χc is highest in March with 431 ppm and the lowest in June at 401 ppm. The pattern across the soil temperature, air temperature, and net radiation unsurprisingly follows a seasonal pattern with the highest in July and the lowest in February with the range of the monthly average of 2.2 • C to 17.7 • C, −1.1 • C to 21.6 • C, 8.21 wm −2 to 665.8 wm −2 Respectively. Also, the precipitation is found to be intense in January, May, August, and December. Appl. Sci. 2021, 11, x FOR PEER REVIEW 7 of 19 In this site, the wind direction is dominated by South and Southwest directions. The annual time series of environmental drivers and atmospheric χc (35 m) in 2018 showed that the monthly mean χc is highest in March with 431 ppm and the lowest in June at 401 ppm. The pattern across the soil temperature, air temperature, and net radiation unsurprisingly follows a seasonal pattern with the highest in July and the lowest in February with the range of the monthly average of 2.2 °C to 17.7 °C, −1.1 °C to 21.6 °C, 8.21 wm −2 to 665.8 wm −2 Respectively. Also, the precipitation is found to be intense in January, May, August, and December.

Laboratory Calibration Results
The calibration curve of CO2 and CH4 was established in the laboratory after calculating the areas of the selected IR bands of CO2 and CH4, as presented in Figure 1. The calibration procedure consists of 11 measurement points for CO2 ( Figure 4a) and 13 measurement points for CH4 (Figure 4b), on which the measurements were repeated three times with new references IR absorption spectrum for each point. The reference gas concentrations have also been cross-validated using the Picarro CRDS G2201-i analyzer and gas chromatography. The calibration curve of CO2 (a) shows a polynomial fitting of the CO2 calibration curve on FTIR ranging from 101 ppm to 59,806 ppm. As for the CH4 (b), the highest concentration that was applied was 5000 ± 99 ppm, with an average absorbance area (Abs) of 3.97 ± 0.012. The lowest concentration tested was 50 ppm, with an average absorbance area of 0.046 ± 0.024 (resolution: 1 cm −1 ).
The lowest concentration of the standard gas tested is 101 ppm with an absorbance area (Abs) of 0.22. In comparison, the highest concentration tested is around 59,806 ppm, with an average absorbance area (Abs) of 94, As shown in Figure 1. In Table 1, the CO2 concentration and the unit area have a large goodness fit of linear correlation with R 2 of 1 and a Y-intercept of −21.7 ppm. It also has a significant deviation from zero and a p-value of less than 0.0001. The CH4 concentrations (50 to 5000) and the associated absorbance area have a large goodness fit of linear correlation with adjusted R 2 of 0.9999 and residual standard

Laboratory Calibration Results
The calibration curve of CO 2 and CH 4 was established in the laboratory after calculating the areas of the selected IR bands of CO 2 and CH 4, as presented in Figure 1. The calibration procedure consists of 11 measurement points for CO 2 ( Figure 4a) and 13 measurement points for CH 4 (Figure 4b), on which the measurements were repeated three times with new references IR absorption spectrum for each point. The reference gas concentrations have also been cross-validated using the Picarro CRDS G2201-i analyzer and gas chromatography. In this site, the wind direction is dominated by South and Southwest directions. The annual time series of environmental drivers and atmospheric χc (35 m) in 2018 showed that the monthly mean χc is highest in March with 431 ppm and the lowest in June at 401 ppm. The pattern across the soil temperature, air temperature, and net radiation unsurprisingly follows a seasonal pattern with the highest in July and the lowest in February with the range of the monthly average of 2.2°C to 17.7 °C, -1.1°C to 21.6°C, 8.21 wm -2 to 665.8 wm -2 Respectively. Also, the precipitation is found to be intense in January, May, August, and December.

Laboratory Calibration Results
The calibration curve of CO2 and CH4 was established in the laboratory after calculating the areas of the selected IR bands of CO2 and CH4, as presented in Figure 1. The calibration procedure consists of 11 measurement points for CO2 (Figure 4.a) and 13 measurement points for CH4 (Figure 4.b), on which the measurements were repeated three times with new references IR absorption spectrum for each point. The reference gas concentrations have also been cross-validated using the Picarro CRDS G2201-i analyzer and gas chromatography. . The calibration curve of CO 2 (a) shows a polynomial fitting of the CO 2 calibration curve on FTIR ranging from 101 ppm to 59,806 ppm. As for the CH 4 (b), the highest concentration that was applied was 5000 ± 99 ppm, with an average absorbance area (Abs) of 3.97 ± 0.012. The lowest concentration tested was 50 ppm, with an average absorbance area of 0.046 ± 0.024 (resolution: 1 cm −1 ).
The lowest concentration of the standard gas tested is 101 ppm with an absorbance area (Abs) of 0.22. In comparison, the highest concentration tested is around 59,806 ppm, with an average absorbance area (Abs) of 94, As shown in Figure 1. In Table 1, the CO 2 concentration and the unit area have a large goodness fit of linear correlation with R 2 of 1 and a Y-intercept of −21.7 ppm. It also has a significant deviation from zero and a p-value of less than 0.0001. The CH 4 concentrations (50 to 5000) and the associated absorbance area have a large goodness fit of linear correlation with adjusted R 2 of 0.9999 and residual standard error of 32.14 ppm. It also has a significant deviation from zero and a p-value of less than 0.0001.

Seasonal, Daily, and Hourly Variations
The micrometeorological measurement was performed from summer to winter (19 July 2019 to 23 July 2020). The subsoil measurement system of CO 2 using FTIR started on 23 July 2019, sustaining approximately two days of continuous measurement on each visit due to power supply limitation. Meanwhile, the auxiliaries data both from the subsoil monitoring system and lattice tower which include soil temperature at various depth (Figure 5a  Overall, the atmospheric temperature ranged from −2 • C to 38 • C (Figure 5a). The atmospheric temperature at 35 m shows a clear seasonal pattern where it was highest in summer 2019 at 38 • C, and it started to decline significantly during the Autumn. The Temperature variations at −1 m and −6 m were found at 14.98-19.80 • C and 9.73-11.13 • C, respectively. The diurnal atmospheric temperature variations are also considered more significant than soil temperature (0 m), subsoil temperature at −1 m, and −6 m.
Even though there is a small difference in magnitude between the atmospheric pressure (35 m) and soil-air pressure (−6 m) around 150 mbar, the general pattern seasonally over the monitoring periods is similar. Although small differences in magnitude were identified, the daily average of atmospheric pressure (35 m and 0 m) and soil-air pressure (−6 m) shows a visually similar general pattern (Figure 5b). The atmospheric pressure at 35 m high indicates the lowest value, followed by soil air pressure at 6 m depth and soil interface (0 m) consecutively. The atmospheric pressure ranged from 0.933 to 0.998 at 35 m, 0.971 to 1.026 bar at 0 m, and 0.964 to 1.021 at −6 m. It is clear that the highest atmospheric pressure (35 m) was found in Summer and the lowest in winter. The daily average of soil water content (Figure 5c) demonstrates a clear correlation with precipitation and anticorrelation with temperature and very pronounced seasonal variation. The average lowest SWC occurred in summer at around 10%, with some spikes due to the rainfall episodes. However, this increase in SWC of approximately 20% lasted only around five days after the rainfall. During the winter, the average SWC remained high at around 40% due to more frequent rainfall and low temperatures. A slight reduction of SWC by approximately 5% was identified from January to February due to less frequent rain events. Figure 5d, the daily χc (ppm) at −1 and −6 m and friction velocity (u*) above the canopy are provided. The daily average of friction velocity (u*) was relatively higher during the winter (0.7 m/s) than in summer (0.4 m/s). The χc generally increases as the depth, from the near-atmospheric layer (−1 m) up to a few meters to the subsoil layer (−6 m). This study demonstrates that subsoil CO 2 at 1 m depth is constantly lower than subsoil CO 2 at 6 m depth ranging from 1000-7000 ppm at −1 and 10,000-60,000 ppm at −6 m. The χc at −1 m was at the lowest in winter and highest in summer in June 2020. Meanwhile, the seasonal variation of χc at −6 m is somewhat weak. However, it can be seen that the rise of χc at −6 m depth occurs together with the fall of u* and the decrease in SWC due to the absence of rainfall. The seasonal variation of u* is also weak, where it is found to be low during summer and higher during winter, with an exception in early December 2019 to the end of February 2020. To better understand the general relationship between variables statistically, the matrix correlation based on Pearson's is provided in Figure 6.
The heatmap based on hourly data ( Figure 6) between observed variables from 1 June 2019 to 23 July 2020 shows that the influence of temperature m on subsoil CO 2 concentration at −6 m was very pronounced with an R of 0.94 (T soil at −1 m) and 0.83 (T soil at 0 m). The CO 2 concentration at −6 m was also weakly inversely correlated with atmospheric pressure at 0 m (R: −0.59), soil-air pressure at −6 m (R: −0.48), SWC (R: −0.50), and soil temperature at −6 m (R: −0.59). However, a strong correlation between temperature and soil CO 2 does not extend to the shallower depth of −1 m. Pearson's CO 2 concentration at −1 m with soil temperature at −1 m and 0 m were −0.36 to 0.14 consecutively. This relatively low correlation is most likely due to time window measurement inconsistency. While all temperature data were measured hourly during the monitoring campaign, χc at −1 m was measured only once every visit during midday. Furthermore, the heat map demonstrates a pronounced pattern between soil temperature at various depths and SWC with R more significant than −0.7, except for soil temperature at −6 m. To better understand the driving mechanism of CO 2 concentration at −6 m during dry and wet periods and weather correlation between subsoil CO 2 concentrations extends to seasonal timescales, Principle Component Analysis (PCA) and time series of hourly data corresponding to observed variables are provided in Figure 7. The heatmap based on hourly data ( Figure 6) between observed variables from 1 June 2019 to 23 July 2020 shows that the influence of temperature m on subsoil CO2 concentration at -6 m was very pronounced with an R of 0.94 (Tsoil at -1m) and 0,83 (Tsoil at 0m). The CO2 concentration at -6 m was also weakly inversely correlated with atmospheric pressure at 0 m (R: -0.59), soil-air pressure at -6 m (R: -0.48), SWC (R: -0.50), and soil temperature at -6m (R: -0.59). However, a strong correlation between temperature and soil CO2 does not extend to the shallower depth of -1 m. Pearson's CO2 concentration at -1 m with soil temperature at -1 m and 0 m were -0.36 to 0.14 consecutively. This relatively low correlation is most likely due to time window measurement inconsistency. While all temperature data were measured hourly during the monitoring campaign, χc at -1 m was measured only once every visit during midday. Furthermore, the heat map demonstrates a pronounced pattern between soil temperature at various depths and SWC with R more significant than -0.7, except for soil temperature at -6 m. To better understand the driving mechanism of CO2 concentration at -6 m during dry and wet periods and weather correlation between subsoil CO2 concentrations extends to seasonal timescales, Principle Component Analysis (PCA) and time series of hourly data corresponding to observed variables are provided in Figure 7. Figure 7 focuses on a principal component analysis (PCA) and hourly variations of χc at 6 m depth auxiliaries data. The color affected the different parameters refers to the significance of the variables (cos 2 ) expressed between 0 (bad defined) and 1 (well defined). It can be seen from PCA that CO2 at 6 m depth during low SWC (Figure 7.b) is inversely correlated with u* with cos 2 above 0.75. A graph showing intraday variation firmly confirms the inverse relationship with identical rises and falls on the right side. The χc at 6 m depth decreased by more than 10000 ppm, following an increase of u*, from 0.2 m/s to 0.8 m/s. However, this pattern does not apply in observation during the wet period ( Figure  7.a), where PCA fails to define the relationship between CO2 at -6m with other variables. The intraday graph on the right side does not indicate a pronounced negative correlation pattern.

Field Performances and Processing Infrastructure
Even though this work aims to develop a continuous flow system to continuously measure the soil gas concentration and the variability of physical parameters, it can also be used as preliminary guidelines to better understand the CO2 variations in the forest ecosystem. Figure 8 shows a typical infrared profile of gas coming from the borehole at 6  The color affected the different parameters refers to the significance of the variables (cos 2 ) expressed between 0 (bad defined) and 1 (well defined). It can be seen from PCA that CO 2 at 6 m depth during low SWC (Figure 7b) is inversely correlated with u* with cos 2 above 0.75. A graph showing intraday variation firmly confirms the inverse relationship with identical rises and falls on the right side. The χc at 6 m depth decreased by more than 10,000 ppm, following an increase of u*, from 0.2 m/s to 0.8 m/s. However, this pattern does not apply in observation during the wet period (Figure 7a), where PCA fails to define the relationship between CO 2 at −6 m with other variables. The intraday graph on the right side does not indicate a pronounced negative correlation pattern.

Field Performances and Processing Infrastructure
Even though this work aims to develop a continuous flow system to continuously measure the soil gas concentration and the variability of physical parameters, it can also be used as preliminary guidelines to better understand the CO 2 variations in the forest ecosystem. Figure 8 shows a typical infrared profile of gas coming from the borehole at 6 m depth, where only the absorption bands of CO 2 and H 2 O are clearly visible. Even though CH 4 concentration was always below our IR sensor's sensibility limit, it was still detectable with an ex-situ GC analysis at a range of 0.9-2.7 ppm with minor depthdependence. Moreover, these monitoring stations have shown that soil CO 2 concentration monitoring can be maintained long term without grid power. After deployment, there were follow-up visits to change the batteries every month before December 2019 and every week after December 2019, but cumulatively these equated to only a few fieldwork days. During the deployments, the average power consumption for two days of continuous and simultaneous measurements was roughly 300 A. This includes the power demand from a laptop, FTIR, auxiliary sensors, and circulation module (pumps). Nonetheless, the advantages outweighed the drawback of its power consumption, where the other actively IR gas interfinger can be identified individually. Interferences weak bands of H2Ovapour are also visible on the spectrum at the low and high-frequency range.
The FTIR calibration was deemed very simple considering CO2 vibrational mode at the wavelength of 2400 cm −1 (v3). The high-frequency data hold many exciting features that are not captured in the synoptic data sets and will be useful for tuning our understanding of soil carbon. With this study, this monitoring infrastructure was considerably successful in terms of survivability and management. In the case of continuously accessible power, the system could record and store high-frequency measurements (<1 min). There is also an opportunity to reduce power consumption by eliminating an FTIR Bruker The FTIR calibration was deemed very simple considering CO 2 vibrational mode at the wavelength of 2400 cm −1 (v 3 ). The high-frequency data hold many exciting features that are not captured in the synoptic data sets and will be useful for tuning our understanding of soil carbon. With this study, this monitoring infrastructure was considerably successful in terms of survivability and management. In the case of continuously accessible power, the system could record and store high-frequency measurements (<1 min). There is also an opportunity to reduce power consumption by eliminating an FTIR Bruker alpha and a laptop while relying only on low power consumption, an IR sensor, and a data logger. Since the FTIR measurement with a resolution of 1 cm −1 and 4 cm −1 is visually identical with a polynomial model, the linear model fitting was applied for both resolutions provided in Table 2. Therefore, it is important to understand how FTIR's lower resolution at 4 cm −1 performs in conjunction with the Beer-Lambert law. Table 2. The linear fitting comparison between FTIR resolution of 1 cm −1 and 4 cm −1 .

FTIR Resolution
Equation The resolution of 1 cm −1 has a slightly higher correlation index using a linear model than 4 cm −1 at 0.987 and 0.976, respectively. It has been reported that the equation between the partial pressure of CO 2 and CH 4 does not follow the beer-lambert law, especially if it is measured using the spectral resolution of fewer than 1 cm −1 [43,47]. Unlike solid or liquid, which does not possess a rotational mode, the gas state involves vibrational and rotational transitions. In the case of gases calibration, the rotation of the bands is often thin (FWHM [Full width at half maximum] < 0.2 cm −1 ), and all the spectral lines that make them are only detectable if a very high resolution (of the order of 0.1 cm −1 ) is used. The area derived from lower resolution FTIR does not precisely match the sum of all the theoretical spectral lines of vibration-rotation. Hence, it is almost certain the calibration curves fit with a three-degree polynomial equation rather than a linear equation result from its low-resolution measurement. However, considering the longer scanning time resulting in lower temporal resolution, the 3-degree polynomial is deemed acceptable for producing a calibration curve.

CO 2 Molar Fraction Dynamics
It is almost certain that CO 2 concentrations vary depending on physicochemical properties, soil cover types, and climatic conditions. However, the CO 2 change between forest land and grassland is similar to approximately 100 to 1300 mg CO 2 m 2 h −1 than barren land and cropland [3]. The ranges of CO 2 concentration measured at the grassland in Kyushu Island, Japan was around 1300-8700 ppm (0.1 m depth), 2500-10,800 ppm (0.2 m depth), 4200-13,200 ppm (0.3 m depth), and 5800-16,500 ppm (1.0 m depth) [48]. It has also been reported that the soil CO 2 concentration measured at a depth of 0.3 m in the South-Moravian region was around 5000 to 8000 ppm [49]. The soil CO 2 concentrations in our forest site are more comparable with a study conducted previously in a temperate deciduous forest using NDIR that ranged between 774 ppm and 6521 ppm at 1 m depth [50]. Higher CO 2 concentration at a depth less than 1 m has also been reported in forest site Meurthe et Moselle, France, with around 20,000 ppm at 60 cm depth [51] and a study conducted in a temperate forest located at Shale Hills Critical Zone Observatory, Pennsylvania, US, reported that the CO 2 concentration reached up to 50,000 ppm at a depth of 1.5 m, attributed to the greater root respiration depth partly due to higher macroporosity of sandstone lithologies [52]. The CO 2 concentration above 10% is generally found in the active geothermal area with a significant pressure gradient transport instead of the solely diffusive process [24,25]. Some studies suggest a strong positive correlation between χc and temperature [3,49,53], where temporal CO 2 concentration is affected by soil temperature on a biweekly basis [50]. Our work demonstrates that CO 2 and temperature are positively correlated at 1 and 6 m depth. In a temperate forest, the temperature is considered the primary driver of heterotrophic soil respiration, releasing CO 2 as organisms decompose soil organic carbon (SOC) [54][55][56]. However, the effects of soil temperature toward heterotrophic respiration are rather complex [57]. It might also indirectly affect the microbial, substrate quantity, and substrate availability on a daily to annual basis [53,54]. On minutes to an hourly basis, it might also influence the enzyme activity produced by bacteria and fungi [58].
Furthermore, most of the studies carried on in the forest soil mainly focused at a depth of 0 to 1 m, where the soil CO 2 production is often a correlation to biological activity is more instantaneous. Very little data are available on the evolution of the CO 2 concentration in the deepest layers of the soil. In the present study, the abiotic drivers' relationship, including SWC and u*, is further explored and provided in Figure 9. m, attributed to the greater root respiration depth partly due to higher macroporosity of sandstone lithologies [52]. The CO2 concentration above 10% is generally found in the active geothermal area with a significant pressure gradient transport instead of the solely diffusive process [24,25]. Some studies suggest a strong positive correlation between χc and temperature [3,49,53], where temporal CO2 concentration is affected by soil temperature on a biweekly basis [50]. Our work demonstrates that CO2 and temperature are positively correlated at 1 and 6 m depth. In a temperate forest, the temperature is considered the primary driver of heterotrophic soil respiration, releasing CO2 as organisms decompose soil organic carbon (SOC) [54][55][56]. However, the effects of soil temperature toward heterotrophic respiration are rather complex [57]. It might also indirectly affect the microbial, substrate quantity, and substrate availability on a daily to annual basis [53,54]. On minutes to an hourly basis, it might also influence the enzyme activity produced by bacteria and fungi [58].
Furthermore, most of the studies carried on in the forest soil mainly focused at a depth of 0 to 1 m, where the soil CO2 production is often a correlation to biological activity is more instantaneous. Very little data are available on the evolution of the CO2 concentration in the deepest layers of the soil. In the present study, the abiotic drivers' relationship, including SWC and u*, is further explored and provided in Figure 9.

Analysis of Wind Turbulence and Soil Water Content
Several periods with high and low friction velocity were detected during the monitoring period (Figure 9) where the relationship between friction velocity and soil CO2 concentration at −6 m is very pronounced, especially during dry periods with R 2 of 0.6 where 80,000 60,000 40,000 20,000 Figure 9. The relationship between hourly CO 2 molar fraction (χc) at 6m depth and friction velocity (u*) above the canopy for different levels of SWC.

Analysis of Wind Turbulence and Soil Water Content
Several periods with high and low friction velocity were detected during the monitoring period (Figure 9) where the relationship between friction velocity and soil CO 2 concentration at −6 m is very pronounced, especially during dry periods with R 2 of 0.6 where SWC is less than 0.2 m 3 /m −3 (CO 2 = −14,882u* + 55,066). However, this relationship does not extend during wet periods where SWC is above 0.2 m 3 /m 3 . This finding acts as empirical evidence of subsoil ventilation and the importance of advection in gas transport processes where the decrease of χc coincides with high u* that fits with the linear model (coefficients: b[0] 55,066, b [1] −14,882, R 2 : 0.68) (Figure 9). Some previous works focusing on soil CO 2 dynamics have reported the influence between χc and u* [26,31,59]. The lower soil of our site consists of sandy and clayey passages caused by the complex sedimentary source material. The calcareous bedrock mainly consisted of calcium carbonate and a few amounts of clay minerals surrounded by the weathered layer rich in clays due to decarbonization. It is a widely held view that limestone's solubility in weak acid and limestone, resulting in karst forming processes. It has been previously studied that the karst system stores a large amount of CO 2 underground, which later was emitted during windy days and filled quickly during low wind speed through cracks and fissures [59]. Even though some sinkholes in our site are visible, this hypothesis needs further investigation to understand the rate and magnitude of calcite dissolution. Also, monitoring data suggest that the subsoil's permeability is most likely reduced during wet periods diminishing molecular diffusion and advection. The lower the SWC value, the more enormous the ventilation's potential to remove CO 2 from the soil, as indicated by previous works in a karstic environment [59,60].
It is widely known that the increase of temperature by 10 • C leads to a higher respiration rate factor, known as temperature sensitivity of soil respiration, that is often expressed as the Q10 [61,62]. However, this factor related to biological activity cannot fully explain the variation. Instead, the CO 2 fluctuation is likely due to pressure pumping effects (PPE) promoted by u* fluctuation where the soil acts as a semi-permeable membrane storing and releasing CO 2 [33,63,64]. The continuous wavelet-based analysis ( Figure 10) was employed of the dataset from 17 June to 23 July 2020, which was chosen due to the dataset's continuity. The wavelet analysis was done using R software and Waveletcomp Package [44,65] to test our hypothesis regarding pressure pumping effects generated in a deep soil layer. cracks, and fissures, which play a significant role in regulating the vertical transport and gas storage suggested by previous works [33,67,68]. Nonetheless, propagation of atmospheric pressure fluctuations into the subsurface has been known to promote the air's migration in moderately permeable materials up to 10 m depth and even greater depth in a fractured rock system [69][70][71]. However, most works on assessing the PPE and wind turbulence CO2 focus on the top layer up to one-meter depth; our work provides empirical evidence of a strong partial relationship on a deeper soil layer (6 m). shows the multivariate statistical significance of (b1) temperature at 35 m over u* and (b2) ∆p over u*. In part, a reddish area and black lines represent high wavelet power levels. In part b, the red area with white contour lines demonstrates a high correlation with more than 5% significance levels. Arrows represent a multivariate phase-locked angle relationship where horizontal arrows pointing left (phase angle: 180°) indicates that two wavelet series are out of phase and horizontal arrows pointing right (phase angle: 0°) suggests that two wavelet series are in phase. A white shaded area indicates that the time series are influenced by edge effects.

Conclusions
We describe the calibration technique, observational setup of the measurement campaign, aspects of the data analysis, and instrument field analysis of the recorded time series of soil gases and auxiliaries data within the study area. This soil gas measurement infra- Figure 10. Half-hourly wavelet-based analysis from 17 June to 23 July 2020 which include (a) wavelet power spectrum of (a1) temperature at 35 m, (a2) friction velocity (u*), (a3) air pressure at 35 m, and (a4) air pressure difference (∆p) between 35 m and −6 m. Moreover, wavelet coherence analysis (b) shows the multivariate statistical significance of (b1) temperature at 35 m over u* and (b2) ∆p over u*. In part, a reddish area and black lines represent high wavelet power levels. In part b, the red area with white contour lines demonstrates a high correlation with more than 5% significance levels. Arrows represent a multivariate phase-locked angle relationship where horizontal arrows pointing left (phase angle: 180 • ) indicates that two wavelet series are out of phase and horizontal arrows pointing right (phase angle: 0 • ) suggests that two wavelet series are in phase. A white shaded area indicates that the time series are influenced by edge effects.
Friction velocity (u*) and temperature (35 m) recorded from the 17 June to 23 July 2020 ranged from 0 m/s-1 m/s and 10-28 • C, respectively. Wavelet power analysis of temperature and u* (Figure 10a(1,2)) demonstrates substantial diurnal scale as expected shown by maximum power wavelet at periods of 24 (24 h). Wavelet power levels of u* in periods of 24 were slightly higher with 3.33 compared to temperature with 2.87, indicating more pronounce daily variation. As for the air pressure, even though the pressure above the canopy (35 m), soil-air interface pressure (0 m), soil-air pressure (−6 m) shows different values of magnitude, the general patterns are similar; therefore, a wavelet power analysis was employed only to the air pressure at 35 m. In contrast with the temperature (35 m) and u*, the air pressure at 35 m shows little or no daily variations. Instead, the wavelet power levels of air pressure at 35 m (Figure 10a(3)) were significant at the region of 128 (5 days) with only 1.74, which was also found in a wavelet power analysis of temperature (35 m). What is striking in the wavelet power analysis is that the pressure difference (∆p, 35 m and −6 m) provided in Figure 10a(4) shows strong daily fluctuation shown by dominant periods of 24 similar to temperature (35 m) and u*. However, these data should be interpreted carefully since the pressure sensor's resolution was only 0.002 bar and could not detect smaller pressure fluctuation. The pressure fluctuation in the frequency ranging from 0.01 Hz-0.1 Hz is highly dependent on the above canopy wind speed [66]. The pressure dynamics that might be generated from pressure pumping effects due to wind turbulence above the canopy must be further investigated. According to wavelet coherence analysis, the temporal correlation between the temperature at 35 m over u* (Figure 10b(1)) and ∆p (35 m and −6 m) over u* (Figure 10b(2)) demonstrated substantial variation, especially at a daily scale with time lags of temperature (35 m) and ∆p (35 m and −6 m) over u* ranged from 2 to 4 h. Besides the temperature, ∆p between soil air interface and soil-air is highly likely to promote the contraction and expansion of the air stored in soil pores, cracks, and fissures, which play a significant role in regulating the vertical transport and gas storage suggested by previous works [33,67,68]. Nonetheless, propagation of atmospheric pressure fluctuations into the subsurface has been known to promote the air's migration in moderately permeable materials up to 10 m depth and even greater depth in a fractured rock system [69][70][71]. However, most works on assessing the PPE and wind turbulence CO 2 focus on the top layer up to one-meter depth; our work provides empirical evidence of a strong partial relationship on a deeper soil layer (6 m).

Conclusions
We describe the calibration technique, observational setup of the measurement campaign, aspects of the data analysis, and instrument field analysis of the recorded time series of soil gases and auxiliaries data within the study area. This soil gas measurement infrastructure combining borehole measurement with micrometeorological measurement offers excellent possibilities for long-term in-situ and continuous gas monitoring to derive the vertical distribution of CO 2 . Thus, this infrastructure allowed the observation of the temporal dynamics in soil-gas CO 2 research. Relatively short-term events could be monitored, notably those associated with pressure and temperature at a deep soil layer, which is difficult to measure using a conventional method. Besides, the long-term data set of soil gas baseline will contribute to understanding better the influence of short-term events, including (i) spikes of short-term event, (ii) magnitudes of change caused by the suspected driving variables, (iii) the duration of the changing circumstances, and (iv) indirect variables influencing these short-term events. Such a metrology approach will contribute to the definition of good practice for the state of studies related to subsoil CO 2 molar fraction in conjunction with the soil-air pressure and temperature. Even though this approach has been applied to the forest ecosystem, it can be extended to any other natural or anthropic systems. Even though CH 4 concentration was always below our IR sensor's sensibility limit, this infrastructure might still be applied in CH 4 -rich site.
Furthermore, this study shows substantial empirical pieces of evidence that the increase of temperature leads to higher respiration rates, which further produced CO 2 stored in deep soil layers. However, this factor related to biological activity cannot fully explain the variation. We hypothesize pressure pumping effects also play a role, especially during dry summer. The decreases of CO 2 molar fraction in the subsoil coincided with high turbulence, resulting in increased CO 2 storage in deep soil during low turbulence. This study also demonstrates that permeability significantly reduced during wet periods diminishing the molecular diffusion and advection. All of these highlight the importance of spatial and temporal dynamics of deep soil CO 2 dynamics that reach a value of up to 100 times larger than atmospheric CO 2 . This empirical evidence also challenges the current assumption that the influence of PPE is only significant on a topsoil layer. However, it is still uncertain whether this strong relationship is site-specific or not since it is highly likely that the soil porosity, soil texture, stand type, and canopy height play roles. As flux tower establishment expands, there is a strong need to apply borehole measurement underneath to understand better the relationships between friction velocity and subsoil CO 2 dynamics on various sites.