The greenhouse experiment was conducted at the campus of Kyung Hee University, South Korea (37°14′24″ N, 127°5′2″ E) (He et al., 2019b). The soils used in this experiment were a 1:1 (v/v) mixture of potting and mineral soils. Commercial potting soil was procured from the Korea Association of Seedbed Media in South Korea. Mineral soil was collected from the Environmental Impact Evaluation Test Facility in South Korea (36°57′44″ N, 127°28′3″ E). This facility is an artificial gassing site that was established to develop environmental management techniques for soil, groundwater, atmosphere, and ecosystems in CO₂ storage sites. The basic physicochemical properties are summarized in Table 1. Three-leaf stage kale (Brassica oleracea var. viridis), purchased from West Suwon Agricultural Products Inc. (Suwon, South Korea), was prepared for the experiment. Kale was selected as our testing plant for the following reasons: (1) it is known to be sensitive to environmental stress (Tang et al., 2014), and (2) the leaves are sufficiently large to be covered by the chambers of measurement devices.

A specially designed acrylic pot consisted of an upper soil chamber (15 × 30 × 30 cm, i.e., height × length × width, respectively) and bottom injection space (5 × 30 × 30 cm, i.e., height × length × width, respectively) (Fig. 1). A clapboard between the soil and gassing chambers had 16 holes drilled into it and was covered with mesh (pore diameter: 250 µm) for optimal gas diffusion. The soil chamber was filled with 8 kg of soil mixture to a depth of 15 cm. Seedlings were transplanted into the soil chamber with 12 plants per pot on September 13, 2018.

Three CO₂ treatments included 10%, 20%, and 40% soil CO₂ concentrations (labeled 10% CO₂, 20% CO₂, and 40% CO₂, respectively). Pure CO₂ gas was continuously injected into the gassing pots from October 10 to 20, 2018. The soil CO₂ concentrations were adjusted by controlling the injection flow rate using a flow meter. A control plot was prepared without any CO₂ injection. Soil water contents in the CO₂ treatments and control were maintained at 55–65% of soil water-holding capacity (WHC) by watering daily at 500–800 ml per pot during the experimental period. To differentiate the plant responses to CO₂ stress to those with water stress, we prepared a water stress treatment (WATER) and water stress with 40% CO₂ treatment (WATER + CO₂). Soil water contents in the WATER and WATER + CO₂ treatments were maintained at 25–35% of WHC by watering daily at 100–150 ml per pot. These soil water contents were considered to be a mild water stress condition for kale (Li et al., 2013; Xavier et al., 2017). Each treatment included three replicated pots. During the experimental period, the ambient temperature was 19 ± 3 °C during the day and 6 ± 3 °C at night. The greenhouse remained open and two fans were used to prevent ambient CO₂ accumulation which might affect the plant leaf level CO₂ concentration (He et al., 2019b).

The CO₂ and O₂ concentrations were measured daily using a portable GA5000 gas sensor (for CO₂%) ranging from 1 to 100% (volume with ±1% accuracy) (Geotechnical Instruments UK Limited, Coventry, UK) at 10 cm depth. Soil water content (% volumetric basis) was measured daily using a Decagon 5TM soil moisture sensor (Decagon Devices, Inc., Washington, DC, USA).

The overall visual changes in the plants in each treatment were recorded by taking photographs every 2 days. Measurements of reflectance signature, photosynthetic parameters, chlorophyl content, and chlorophyl fluorescence parameters were measured on the third to the fifth fully expanded leaves. These selected leaves were in a similar developmental stage, which would minimize variation due to different physiological ages of the leaves (Mendelssohn, McKee & Kong, 2001). Plant measurements were conducted between 10:00 am and 16:30 pm on October 10, 13, 15, and 18 (i.e., 0, 3, 5, and 8 days after experiment onset, respectively), 2018. The hyperspectral reflectance signatures from wavelengths 350 to 2500 nm (spectral resolution: 3 nm VNIR, 30 nm SWIR) were measured using ASD FieldSpec Pro (Malvern Panalytical Ltd. Cambridge, Malvern, UK). Leaf level measurements were conducted on 10 leaves in each pot. The spectral signatures were converted to readable reflectance values using the ViewSpecPro software (Vision 6.2; ASD, Inc., Falls Church, VA, USA). The vegetation indices, including the NDVI, photochemical reflectance index (PRI), and enhanced vegetation index (EVI), were calculated to monitor leaf pigment changes according to the equations in Table 2. The NDVI is a popular vegetation index, which has been shown to be strongly related to chlorophyl light interception (Tucker, 1979; Hatfield et al., 2008). On the other hand, the PRI is an indicator of changes in carotenoid pigments, which could also imply environmental stress (Gamon, Peñuelas & Field, 1992; Penuelas et al., 1997). The EVI is an alternative index to assess vegetation greenness and has also been used to detect CO₂ leakage (Bateson et al., 2008). Furthermore, the NDWI and modified normalized difference water index (mNDWI) were correlated with the water content of plants (Xu, 2006; Male et al., 2010; Gautam et al., 2015), and were also calculated following the equations in Table 2.

Chlorophyll content was measured for 10 leaves in every pot using a chlorophyl meter (SPAD 502plus; Konica Minolta, Tokyo, Japan). This is a well-known method for estimating leaf chlorophyl concentration (Süß et al., 2015). The measurement head of the SPAD emits light with peak wavelengths at 650 nm and 940 nm when it clips a leaf. Part of the light was absorbed by chlorophyl, while the transmittance light was measured by a photodiode detector to calculate the relative chlorophyl content automatically (Süß et al., 2015).

The conventional photosynthetic parameters (i.e., photosynthetic rate (Pn), stomatal conductance (Gs), and transpiration rate (Tr)) were measured in three to five leaves per pot using an infrared gas analyzer by portable devices, LI-6400 and LI-6800 (LI-COR, Lincoln, NE, USA). The measurements were made with a red-light source at 1,000 µmol m^–2 s^–1 photosynthetic photon flux density (PPFD), which was near the plant photosynthetic saturation point (Kalaji et al., 2018; Casanova-katny, Barták & Gutierrez, 2019).

Chlorophyll fluorescence was measured using the WinControl-3 controlled MONI-PAM Fluorometer system (Heinz-Walz, Eifeltrich, Germany). The MONI-PAM system comprised three emitter-detector heads (MONI-head/485). These units can measure three leaves simultaneously because each head is an independent fluorometer. The head can provide an actinic light pulse up to 1500 μmol m⁻² s⁻¹ PPFD (Porcar-Castell et al., 2008; Janka et al., 2015). The MONI-PAM device measured the maximal fluorescence yield (F_m′), steady-state fluorescence yield (F_s′), and photosynthetically active radiation (PAR) on six leaves in each pot. The fluorescence parameters of Y(II) was calculated to estimate the plant stress responses by the formula: Y(II) = (F_m′ – F_s′)/F_m′ (Maxwell & Johnson, 2000; Murchie & Lawson, 2013; Janka et al., 2015). Y(II) was analyzed using a modified regression method by plotting the light response curve of Y(II) to PAR for each leaf (Sven, Mates & John, 2014; Casanova-katny, Barták & Gutierrez, 2019). The leaf level PAR values were adjusted by controlling the emitter-detector head to emit actinic light pulse incrementally (i.e., 100, 200, 300, 450, 650, 800, 1,200, and 1,500 µmol m⁻² s⁻¹). The Y(II) values was obtained from the curve at PAR = 1,000 μmol m⁻² s⁻¹.

After the injection stopped, plants were harvested and oven-dried at 70 °C for 3 days to measure biomass.

The indicator evaluation score (IES) was developed to compare the efficiencies of the parameters for CO₂ leakage monitoring. Based on extensive reviews, we set up four criteria for IES, including early detection, CO₂ specificity, field applicability, and cost (Male et al., 2010; Patil, 2012; Al-Traboulsi et al., 2012; Noble et al., 2012; He et al., 2019a). The scores in each criterion, ranging from 1 to 5, were allocated for each parameter based on the evaluation standards in Table 3.

The first criterion of early detection was identified by the timing of when the parameter differentiated control and elevated soil CO₂ treatment (Table 3). The time intervals for scoring were defined based on the observations of the timing with which plants respond to elevated soil CO₂ concentrations in previous gassing studies (Patil, 2012; West et al., 2015; Lake et al., 2016b; He et al., 2019b). As 10% soil CO₂ was reported as a threshold level to negatively affect plants (Al-Traboulsi et al., 2012; Jones et al., 2015; West et al., 2015), the timing for the significant difference between CO₂ (≥10%) and control was used for scoring. Therefore, the parameter with a score of 5 was defined as capability to detect CO₂ leakage within 1 day. Subsequently, the parameters with scores of 4, 3, 2, and 1 can detect leakage within 2–4 days, 5–7 days, 8–10 days, and >10 days, respectively (Patil, 2012; West et al., 2015; Lake et al., 2016b; He et al., 2019b).

The second criterion of CO₂ specificity was identified by the ability to detect the difference between water and soil CO₂ stresses, or between water and water + CO₂ stresses. The identification of CO₂ specificity was based on previous studies that reported similar responses of plants to water and CO₂ stresses (≥40%) in the early stage of treatments (Lake et al., 2016b; Kim et al., 2017). Hence, the parameter, that differentiates water and CO₂ stresses, could be expected to be CO₂-specific. The parameter with a score of 5, in this criterion (Table 3), was defined by its capability to discriminate water stress from both 40% CO₂ and water + CO₂ stress (Bellante et al., 2014). On the other hand, the parameter with a score of 4 can differentiate water and 40% CO₂ stress. The one with a score of 3 can differentiate water stress from mild CO₂ stress (10% CO₂). The parameter with a score of 2 can detect CO₂ (≥10%) and water stresses but cannot differentiate both, while that with a score of 1 can only detect CO₂ stress.

The third criterion of field applicability was identified by examining whether the parameter meets the four sub-standards, which are nondestructive, remotely detectable, automatic, and continuous measurements (Table 3). The selection of sub-standards of field applicability for each parameter was based on the literature and product catalogs (He et al., 2019a, 2019b; Li et al., 2013; Male et al., 2010; https://www.walz.com; https://www.malvernpanalytical.com). The nondestructive sampling of plants allows sustainable monitoring of CO₂ leakage from plants (Noble et al., 2012). The remotely detectable parameters allow coverage of a large area (Male et al., 2010). Automatic and continuous measurements are helpful for ease of data collection and long-term monitoring in CCS sites (Male et al., 2010). The parameter with a score of 5 can meet four sub-standards, while the parameters with scores of 4, 3, 2, and 1 meet three, two, one, and none of the sub-standards, respectively.

The fourth criterion, cost, was identified by the average cost of sensor devices and time spent taking the measurement and data process (Table 3). The cost of devices was based on literature and sensor catalogs (Kim et al., 2019; Noble et al., 2012; https://www.walz.com; https://www.malvernpanalytical.com; https://licor.co.za) and the time cost for measurements was also referenced from above catalogs. In order to compare cost, the price of the device was categorized into three levels: low (<$5000), medium ($5000–$30,000), and high (>$30,000), following the extensive review by Noble et al. (2012), and the catalogs by the leading producers of sensors (https://www.walz.com; https://www.malvernpanalytical.com). In a similar manner, time for sampling and data processing was categorized into three levels, that is, low (<3 min), medium (4–20 min), and high (>20 min) (Noble et al., 2012). The parameter with a score of 5 (marked as low) requires that both the cost of device and time are classified as low level (Table 3). The parameter with a score of 4 (marked as low + medium) should be at least one low cost in device or time and one in medium cost in device or time. This included two conditions: (i) low price and medium time; (ii) low time and medium price (Table 3). The parameter with a score of 3 (marked as medium) should be the average cost of the device price and time at the medium level. This included three conditions: (i) medium cost in both device and time; (ii) low device price and high cost in time; and (iii) high device price and low cost in time. The parameter with a score of 2 (marked as medium + high) should be two conditions: (i) medium price and high time; and (ii) medium time and high price. The parameter with a score of 1 (marked as high) is high cost for both device and time (Table 3).

Finally, the IES value is calculated by adding the scores in the four criteria, following the equation:

(1)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$${\rm IES} = \mathop \sum \limits_{{\rm i} = 1}^{\rm n} {\rm V}i$$\end{document}IES=∑i=1n⁡Viwhere Vi is the score in the four criteria (n = 4) of early detection, CO₂ specificity, field applicability, and cost.

The light response curve of Y(II) to PAR for each leaf was processed in MATLAB R2014a (The MathWorks Inc., Natick, MA, USA). Analysis of variance (ANOVA) of Y(II), NDVI, PRI, EVI, NDWI, mNDWI, Pn, Gs, Tr, chlorophyl content, and biomass among the control, CO₂, WATER, and WATER + CO₂ treatments were analyzed using SAS 9.1 (SAS Institute Inc., Cary, NC, USA). The least-square means were used to test for significant differences among treatments at the 5% probability level. All results are reported as mean ± standard error.

In the 10%, 20%, and 40% CO₂ treatments, soil gas concentrations were maintained at target levels during the experimental period (Table 4). Soil O₂ concentrations decreased in all the CO₂ treatments. This inference is consistent with the findings by Patil, Colls & Steven (2010), who found that the injected CO₂ can replace soil O₂. The photographs showed that the appearance of leaf chlorosis occurred in all the CO₂ treatments on day 5, and yellow leaves significantly increased in the higher CO₂ soil concentration treatments (Fig. 2). These results indicated that our experimental system was adequate for examining the effects of soil with different CO₂ concentrations on plants.

In the WATER and WATER + CO₂ treatments, soil water content was maintained at the target levels during the experimental period (Table 4). The leaves in both treatments slightly wilted and drooped (Fig. 2). This indicated that those plants were subjected to water stress (Naser et al., 2010; Xavier et al., 2017). The photographs showed that water stress also induced leaf chlorosis (Fig. 2). The number of yellow leaves in the WATER + CO₂ treatment was greater than those in the WATER treatment (Fig. 2), which implies that the effects of the combination of water and CO₂ stresses on plants were worse than those of water stress.

The timing for the first differentiation between the control and CO₂ treatments varied for different parameters. The hyperspectral reflectance parameters of NDVI, EVI, and PRI (within the pigment absorption bands) changed on day 8 in all CO₂ treatments compared to the control (Fig. 3), while chlorophyl content first changed on day 5 in the 40% CO₂ treatments (Fig. 4). These results imply that NDVI, EVI, PRI, and chlorophyl content can be used to monitor ecosystem changes by CO₂ leakage, but their responses were not quick enough to detect CO₂ leakage. Although the reflectance and chlorophyl parameters have been widely used to monitor CO₂ leakage points in artificial gassing sites and natural CO₂ spring areas (Bateson et al., 2008; Male et al., 2010; Patil, 2012; Feitz et al., 2014), the primary purpose of these parameters was not to detect CO₂ leakage but to monitor overall changes in plants due to leakage.

As hypothesized, the responses of photosynthetic parameters (Pn, Gs, and Tr) to the elevated soil CO₂ concentration were earlier than the changes in reflectance parameters and chlorophyl content. The changes in all the CO₂ treatments were first observed on day 3 compared to those in the control, and the significant difference continued until the end of injection (Fig. 5; Table 5). The recovery of Pn, Gs, and Tr in 10% CO₂ treatment from day 5 might be related to the plant compensation effects to the stress (Lake et al., 2016b; He et al., 2019a, 2019b). However, the overall lower values in the CO₂ treatments than control indicated that CO₂ leakage had adverse effects on the overall photosynthetic processes, which was consistent with the observations of Zhang et al. (2016). They reported that Pn, Gs, and Tr decreased after 4 days of CO₂ leakage. According to Lake et al. (2016b), elevated soil CO₂ concentration inhibits root water uptake and triggers the excretion of abscisic acid to close stomata, which instantly affects leaf photosynthesis and transpiration. Similar effects of elevated soil CO₂ on roots and leaves were also reported by He et al. (2019b).

Consistent with the photosynthetic process parameters, fluorescence parameters also showed early changes during the CO₂ treatments (Figs. 5 and 6). On day 3, Y(II) in the 20% and 40% CO₂ treatments were significantly lower than those in the control (Fig. 6). We consider Y(II) to be more sensitive than Pn, Gs, and Tr in detecting different levels of CO₂ leakage because Y(II) differentiated 10% and 20% CO₂ treatments on day 3, whereas Pn, Gs, and Tr did not.

As expected, biomass did not change between the CO₂ treatments and the control during this short-term incubation period. This indicates that the biomass is not helpful for early leakage detection and monitoring (Fig. 7). Consistent with our results, Kim et al. (2017) also reported no change in biomass after 10 days of 70% soil CO₂ exposure.

Synthesizing the timing of the changes in all the parameters by CO₂ treatments, the earliest responding parameters were Pn, Gs, Tr, and Y(II) (Figs. 5 and 6). Compared to reflectance parameters (NDVI, EVI, and PRI) and chlorophyl content, which initially differentiated control and 40% CO₂ treatment on days 5 and 8 (Figs. 3 and 4), respectively, the Pn, Gs, Tr, and Y(II) were 2–5 days earlier. Our results are consistent with those of Lake et al. (2016b) and Zhang et al. (2016), who reported that the effects of elevated soil CO₂ concentration on Pn, Gs, and Tr appeared within 4 days, which was earlier than that in chlorophyl content, observed within 7–15 days (Lakkaraju et al., 2010; Smith et al., 2017; He et al., 2019a, 2019b). As our experiment was the first to test the possibility of using fluorescence parameters to detect CO₂ leakage, there is no reference with which to compare our results. However, Rossini et al. (2015) reported that the effect of herbicide on chlorophyl fluorescence parameters was earlier than that on NDVI. In summary, we suggest that Pn, Gs, Tr, and Y(II) are early indicators of CO₂ leakage monitoring.

In the WATER treatment, the parameters of chlorophyl content, Pn, Gs, Tr, and Y(II) were significantly lower than those in the control from day 3 to day 8 (Figs. 4–6; Table 5), indicating that these parameters can be used as indicators for detecting water stress. These results of chlorophyl content, Pn, Gs, and Tr (Figs. 4 and 5) are consistent with previous observations that water stress would lead to leaf stomata closure and restrict CO₂ diffusion into the chloroplast to affect photosynthesis (Kozlowsk, 1972). The Y(II) were also reported that can differentiate plants in moderate or excessive drought stresses from those in non-stressed conditions (Li et al., 2013; Guo & Tan, 2015; Xavier et al., 2017). Unexpectedly, leaf-level measured NDWI and mNDWI (within the water-absorption bands) failed to detect water stress (Fig. 3). This result is inconsistent with the findings of Male et al. (2010) that canopy-level measured NDWI observed leaf water content reduction in a pasture field after the long, hot summer. Compared to Male et al. (2010), we treated broadleaf plants with mild water stress over a short period (10 days). Ling et al. (2019) reported that the leaf-level spectral measurement may be affected by the stress status and plant species due to different leaf structures, age, and front or back sides of leaves. Hence, we argue that NDWI and mNDWI may not detect short-term mild water stress in plants.

The negative effects of water stress on plant chlorophyl content, Pn, Gs, Tr, and Y(II) could be similar to those of CO₂ stress, at least in the initial stage, because the mechanism how underground CO₂ influence plant is related to plants’ root water absorption. He et al. (2019b) reported that high soil CO₂ could reduce root water absorption activity inducing leaf chlorophyl reduction. Lake et al. (2016b) explained the mechanism of how high soil CO₂ influence plant photosynthesis. The similarity between water and CO₂ stresses in early stage of exposure makes CO₂ leakage monitoring using plant more difficult in the field. The Y(II), differentiated WATER from 40% CO₂, and WATER from WATER+ CO₂ (Fig. 6). This verifies that chlorophyl fluorescence would be very helpful in distinguishing plants living in the CO₂ leakage area, which also suffered from water stress. Although chlorophyl content did not differentiate WATER from 40% CO₂ and WATER from WATER + CO₂, it differentiated 40% CO₂ from WATER treatments. This still implies that this would be useful in distinguishing water stress from extreme CO₂ stress in the field. On the other hand, the parameters of Pn, Gs, Tr had significant differences between any CO₂ treatment (≥10%) and WATER treatment (Fig. 5), which indicates that they could differentiate the effects of mild CO₂ leakage from water stress.

The field applicability of each parameter varied (Table 6). The fluorescence parameters of Y(II) and reflectance parameters of NDVI, EVI, PRI NDWI, and mNDWI had high field applicability because they can automatically and continuously monitor plant conditions without destruction (Delegido et al., 2011; Gautam et al., 2015; Janka et al., 2015) (Table 6). In particular, the reflectance parameters were remotely measurable in the artificial gassing sites by installing sensors on unmanned aerial vehicles (Male et al., 2010; Feitz et al., 2014). The photosynthetic parameters of Pn, Gs, and Tr and transmittance parameters of chlorophyll content had limited field applicability because these parameters were only measured at the leaf level, although the measurements would not destroy plants (Wu et al., 2014; Kim et al., 2017). Finally, biomass measurement destroyed plants and could not be remotely, automatically, and continuously detected, which had the lowest field applicability (He et al., 2016) (Table 6).

Although the range of the device prices is large and there could be cheap sensors that are not commercially manufactured, we took the median price for simplicity (the median value between the lowest and highest prices) to evaluate and compare the cost of sensors. Based on our evaluation standards (Table 3), the parameters with low device cost were chlorophyl content and biomass because their median device prices were <$5,000 (Table 7) (https://www.specmeters.com; https://www.konicaminolta.com). The parameters of NDVI, EVI, PRI, NDWI, mNDWI, and Y(II) had medium device cost because the median price of the spectrometer and fluorometer was approximately $26,000 (i.e., ranged from $1,000 to $50,000) and $23,000 (i.e., ranged from $6,000 to $40,000), respectively (Table 7) (https://www.walz.com; https://fluorometers.psi.cz; https://www.hansatech-instruments.com; https://handheld.psi.cz; https://www.oceaninsight.com; http://en.li-ca.com). Finally, the parameters with high device cost were Pn, Gs, and Tr because the price LI-6400 or LI-6800 generally ranged between $40,000 and $60,000 (Table 7) (https://licor.co.za).

The parameter with low time cost was chlorophyl content because the total time for sampling and data processing ranged from a few seconds to 1 min (Table 7). The parameters with medium time cost were NDVI, EVI, PRI, NDWI, mNDWI, and Y(II). Their sampling times ranged from 10 s to 5 min (depending on sensor setup and user skills), and those of data processing time ranged from 3 min to 10 min (including MATLAB code writing) (Table 7). Finally, the parameters of Pn, Gs, and Tr and biomass had a high time cost due to the long sampling time (>20 min) (Table 7). After averaging the cost of device and time for each parameter, we can suggest chlorophyl content as a low-cost parameter for leakage monitoring (Table 7).

As our four criteria have equal importance, we did not give different weightings on them and we treated four criteria equal in mathematical way. So, the highest score means the best parameter. However, we did not only focus on the best score because we realize that the second or third best choices could also be useful in certain circumstances. Hence, we keep the scores of four criteria in the Table 8 and let users decide which would be the best choice of theirs depending on the conditions.

The fluorescence parameters of Y(II) had the highest IES values (Table 8), which suggests that it is the most efficient parameters. They had high scores in the criteria of early detection, CO₂ specificity, and field applicability, although the cost of measurement is considered to be medium (Table 8). Murchie & Lawson (2013) also suggested Y (II) as a useful indicator for field monitoring because it shows a rapid response to environmental stresses. Furthermore, the measurement of Y(II) does not require waiting for dark adaptation of plants, and thus has a fast detection. The second-highest IES values were the reflectance parameters of NDVI, EVI, and PRI (Table 8), which were still useful for CO₂ leakage monitoring. NDVI, EVI, and PRI had the highest score in field applicability and have been successfully applied in field detection (Male et al., 2010; Wimmer et al., 2011; Feitz et al., 2014). The third-highest IES was the SPAD-measured chlorophyl content, which has the major advantage of low cost; however, it has limited applicability to the laboratory or greenhouse (Table 8). The photosynthetic parameters of Pn, Gs, and Tr and reflectance parameters of NDWI and mNDWI had relatively low IES values (Table 8); therefore, they were less efficient for CO₂ leakage monitoring. Although Pn, Gs, and Tr were able to detect CO₂ leakage early, they did not differentiate CO₂ stress from other stresses. Moreover, the measurement can only be done at the leaf level, and the time and labor of measurement are not cost-effective. Compared to the photosynthetic parameters, NDWI and mNDWI can be applied to the field scale because they can cover large area of canopy by loading the sensor on the unmanned vehicles or drone. However, their ability to detect early leakage and accurately differentiate CO₂ stress was minimal. Finally, biomass had the lowest IES value, which indicates that it is not suitable for CO₂ leakage monitoring because it had the lowest scores in all the criteria except cost (Table 8).