Elevated Tropospheric Ozone Concentration Alters Soil CO 2 Emission: A Meta-Analysis

: Elevated tropospheric ozone (O 3 ) concentration may substantially inﬂuence the belowground processes of terrestrial ecosystems. Nevertheless, a comprehensive and quantitative understanding of O 3 impacts on soil CO 2 emission remains elusive, making the future sources or sinks of soil C uncertain. In this study, 77 pairs of observations (i.e., elevated O 3 concentration treat-ment versus control) extracted from 16 peer-reviewed studies were synthesized using meta-analysis. The results depicted that soil CO 2 efﬂux was signiﬁcantly reduced under short-term O 3 exposure ( ≤ 1 year, p < 0.05), while it was increased under extended duration (>1 year, p < 0.05). Particularly, soil CO 2 emission was stimulated in nonagricultural ecosystems, in the free-air CO 2 enrichment (FACE) experiment, and in the soils of lower pH. The effect sizes of soil CO 2 efﬂux were signiﬁcantly positively correlated with experimental duration and were signiﬁcantly negatively correlated with soil pH, respectively. The ozone effect on soil CO 2 efﬂux would be enhanced at warm temperatures and high precipitation. The duration of O 3 exposure was the fundamental factor in analyzing O 3 impacts on soil CO 2 emission.


Introduction
Ozone (O 3 ) in the troposphere is not just a greenhouse gas but also an air pollutant that is prejudicial to human health and injures vegetation [1,2]. Tropospheric O 3 concentrations (hereinafter referred to as [O 3 ]) have significantly increased by 1-2% per year in recent decades due to accelerated industrial development and intensive combustion of fossil fuels [3].
As a highly oxidizing phytotoxic pollutant, O 3 could inhibit plant growth and cause considerable reduction in crop yields and quality [4]. Moreover, elevated [O 3 ] may impact the C allocation to roots [5], alter the rhizodeposition [6], change the quantity and composition of root exudate [7], modify the nutrient and energy supply to soil microorganisms [8], influence the microbial community diversity [9,10], and accordingly, influence soil CO 2 emission flux [11,12].
The rising levels of atmospheric O 3 make it difficult to assess the future soil C source/sink function. A few studies have reported increased CO 2 efflux under elevated [O 3 ] [13][14][15][16], but there are also results to the contrary [11,17]. These inconsistencies could be attributed to the variations in ecosystem type [18], O 3 exposure duration [19], fumigation facility [15], and the complexity of soil and climate conditions [20].
The purposes of this study were (1) to assess the responses of soil CO 2 emission flux to elevated [O 3 ] using meta-analytic techniques and (2) to reveal the possible sources of variation.

Data Sources
All peer-reviewed literature on soil CO 2 emissions with reference to elevated O 3 concentration were extensively searched through Scopus (https://www.scopus.com (accessed on 19 February 2021)) and ISI Web of Science (http://apps.webofknowledge.com (accessed on 19 February 2021)) with the following key terms: (O 3 OR ozone) AND (soil OR rhizosphere* OR terrestrial OR land) AND (carbon dioxide OR respiration* OR CO 2 ) AND (elevate* OR impact* OR affect* OR effect* OR alter* OR respond* OR decrease* OR increase*). The relevant Chinese literature were searched through the China National Knowledge Infrastructure Database (CNKI) (http://www.cnki.net/ (accessed on 19 February 2021)). The search produced a total of 293 publications in ISI, 244 publications in Scopus, and 279 publications in CNKI. After examining the relevance, eliminating the duplications in both ISI and Scopus, and cross-checking the reference lists to avoid missing relevant studies, 28 publications were identified for further screening.
Data in the published sources were screened according to the following criteria: (1) only field studies were included; (2) experiments of control (ambient [O 3 ]) and O 3 enrichment treatments were carried out at the same site, with similar microclimate, vegetation and initial soil conditions; (3) the means, standard deviation (SD), and replicate numbers (n) could be obtained; (4)  In the present study, the cumulative or average soil CO 2 efflux over a year or across a growing season was included as a single data point [21,22]. Thus, if the original study put forward the temporal dynamics of the variable, the annual or growing-seasonal averaged mean (M) and standard error (SE) were calculated as [23] where j is the observation times with each year or growing season (≥2), and M i , SE i , and n i represent mean, standard error, and sample size in the ith observation, respectively. Soil CO 2 efflux data were considered to be independent if they referred to multiple O 3 levels or to different additional treatments within a single study [24]. Values from context or tables were directly obtained, and data from the figures were digitized and extracted using GetData Graph Digitizer version 2.24 (http://getdata-graph-digitizer.com (accessed on 26 March 2008)). The unidentified error bar was assumed to be standard error [25]. Unidentified replicate numbers were assumed to be the replication of the plot [25]. The detailed dataset could be found in the Mendeley Data (doi:10.17632/jxfkh54hkh.1).

Analysis
The meta-analyses were performed using MetaWin 2.1 software (Sinauer Associates, Inc., Sunderland, MA, USA [26]). For each pair of observation, the effect size and variance were calculated as where lnR is the effect size; X E and X C are annual or growing-seasonal averaged means in the experimental (elevated [O 3 ]) and control treatments, respectively; n E and n C are the replicate numbers; and SD E and SD C are the standard deviations. Before proceeding with the weighted analyses, normal quantile plots and the frequency distributions of lnR ( Figure 1) were plotted to check the data normality [27]. A weighted random-effects model was used to calculate the overall and grouped effect size [28]. The 95% bias-corrected confidence intervals (CIs) around the effect size were assessed using a bootstrapped resampling technique with 64,999 iterations. The response to elevated [O 3 ] was considered significant if the 95% bootstrapped CIs did not overlap zero. The O 3 -induced percentage change was calculated as (R−1) × 100% [24]. Positive percentage change indicates an increase in the soil CO 2 efflux in response to elevated [O 3 ], and a negative value indicates a decrease. The total heterogeneity of each variable (Q T ) was tested against a χ 2 -distribution with n−1 degrees of freedom (Table 1) [26]. It was also assessed using an I 2 index that quantifies the ratio of true variation caused by real differences between studies to total variation (true variation + sampling error) ( Table 1) [29]. A positive I 2 value indicates true heterogeneity [30]. Afterwards, the categorical analysis was proceeded across all data by dividing the total heterogeneity (Q T ) into the within-group heterogeneity (Q W ) and between-group heterogeneity (Q B ). The dataset was then subdivided into different levels if the Q B of categorical variables was significant, i.e., the randomized p-values < 0.05 (Table 1) [31]. Significant differences between two levels were reported if their 95% bootstrapped CIs did not overlap [32]. Table 1. The total heterogeneity (Q T ); percentage of heterogeneity due to true variation among effect sizes (I 2 ); and between-group heterogeneity (Q B ), estimated using resampling tests with 64,999 iterations to generate a randomized p-value for responses of soil CO 2 efflux to elevated O 3 relative to controls in a weighted random-effects meta-analysis with categorical structure. A continuous meta-analysis model was employed to examine the relationship between lnR and environmental or forcing factors, such as duration of ozone exposure, soil pH, MAT, and MAP. Weighted regressions were used to test whether the slopes differed from zero using a parametric mixed model approach [26]. The regression relationship was considered significant if p < 0.05.
The possibility of publication bias within each group was firstly evaluated statistically with Kendall's tau rank correlation and Spearman's rho rank correlation between the standardized effect size and replicate number of each study (Table 2) [26]. Then, a Rosenthal's fail-safe number at α = 0.05 was calculated ( Table 2) [33]. A publication bias was confirmed only if all above methods showed positive results simultaneously (namely, if rank correlation tests were significant (p < 0.05) and Rosenthal's fail-safe number less was than 5k + 10, where k is the observation number).

Results and Discussion
Although soil CO 2 efflux shows an increasing trend (4.6%) at elevated [O 3 ] compared with the control [O 3 ] across all studies (the grand mean change in Figure 2), it did not reach a significant level since its bias-corrected 95% bootstrapped Cis overlapped zero. However, the significant Q T (p < 0.05) and positive I 2 values suggested heterogeneity among studies (

Effect of O 3 Exposure Duration
As illustrated in Figure 2, there were significant differences in the responses of soil CO 2 efflux to elevated [O 3 ] among three O 3 exposure duration levels. Statistically significant increases in soil CO 2 efflux were found in both medium (15.1%) and long (11.2%) O 3 fumigation (p < 0.05), while the opposite was observed in the short-term exposure (−16.3%, p < 0.05). The weighted meta-regression analysis showed a significantly positive correlation between lnR and O 3 exposure duration (p < 0.01, Figure 3a), which also showed that prolonged O 3 exposure could stimulate soil CO 2 emission.
Elevated [O 3 ] is known to influence the transport of photosynthates from leaves to roots, reduce the quantity of root residue and exudates, and decrease belowground carbon allocation from root exudates to soil labile carbon pools [34]. Therefore, soil carbon contents generally decreased under elevated [O 3 ] [11]. This may explain the decreased soil CO 2 efflux under short O 3 fumigation (≤1 year) since soil organic matter is the main carbon source for soil respiration and the energy source of microbial activity [35]. However, the responses of belowground processes to elevated [O 3 ] is much less and slower than that of aboveground processes. Short-term O 3 exposure may give a false estimate of the O 3 effects, and longer-term studies would be more meaningful when the experimental results are interpreted by cumulative status [36]. Pregitzer et al. [13] reported an increase in fine-root biomass over a long-time O 3 exposure. The greater root biomass increased the autotrophic root component of soil respiration, and the root detritus increased the carbon source of soil microorganisms, hence improving the soil heterotrophic respiration. This may be the possible reason for the enhanced soil CO 2 efflux under prolonged O 3 exposure.

Effect of Soil pH
As shown in Figure 2, the pre-existing soil pH significantly affected the magnitude and even direction of soil CO 2 responses to elevated [O 3 ]: the O 3 effect was positive in soils of pH ≤ 6.0 (13.4%), whereas it was significantly negative in soils of pH > 6 (−13.8%). There was also a significantly negative correlation between lnR and soil pH (p < 0.001, Figure 3b). O 3 exposure duration was longer in the group of soil pH ≤ 6.0 (5.2 years) than that in the group of soil pH > 6.0 (1.7 years), indicating that exposure duration may influence the results of soil pH categorization. In other words, the regression relationship between soil CO 2 efflux and soil pH may be essentially related to the regression relationships between soil CO 2 efflux and O 3 exposure duration.

Difference between Agricultural and Nonagricultural Ecosystems
In terms of ecosystem type, soil CO 2 efflux in response to elevated [O 3 ] was increased by 10.9% in nonagricultural soil and tended to decrease in agricultural soil ( Figure 2). The O 3 exposure duration for nonagricultural ecosystems (5.3 years) was generally longer than that for agricultural land (1.9 years). Extended O 3 exposure duration may be beneficial to soil CO 2 emissions. Moreover, crops on agricultural land are mostly annual plants. The reduced crop residue consequently decreased soil organic matter under O 3 exposure, which may lead to a decrease in soil CO 2 efflux (Figure 2) [14]. However, plants in nonagricultural ecosystems, such as forest and grassland, are mostly perennials and include both tolerant and sensitive species or genotypes. Over a relatively long time of O 3 exposure, the tolerant species become more able to live and occupy the living space of sensitive species, resulting in a relatively increased total root biomass [13]. Greater root biomass would be conducive to soil autotrophic and heterotrophic soil respiration [13]. Consequently, the nonagricultural soil CO 2 efflux increased and responded more significantly to elevated [O 3 ] than the agricultural soil.

Effect of O 3 Fumigation Method
There was a significant difference on O 3 -induced percentage change of soil CO 2 efflux between fumigation methods. Elevated [O 3 ] strongly increased soil CO 2 efflux by 11.5% in FACE experiments but did not induce significant changes in OTC experiments. The duration of O 3 exposure in FACE experiments (6.2 years) was generally longer than that in OTC experiments (1.5 years), and longer O 3 fumigation was intended to exacerbate the O 3 effect.

Effect of Climate Conditions
Weighted meta-regression analyses in Figure 3c,d showed that that the effect size of soil CO 2 efflux was significantly negatively correlated with MAT (p < 0.05) and MAP (p < 0.01), respectively. This means the O 3 effect on soil CO 2 efflux was more detrimental (i.e., more negative effect sizes) in higher temperature and more precipitation areas.
Previous studies have reported contrasting effects of warming and O 3 on plant physiology, rhizosphere chemical environment, and microbial communities [37,38]. However, the results of weighted meta-regression analyses in the present study revealed that the negative responses of soil CO 2 efflux to elevated [O 3 ] were generally enhanced in warmer areas. In addition, some studies showed that moderate drought might alleviate the O 3 effects on photosynthesis, physiology, stomata characteristics, fine-root dynamics, and soil respiration [39]. In other words, the detrimental response of soil CO 2 efflux to elevated [O 3 ] would be enhanced at higher precipitation conditions, which is consistent with the negative meta-regression results in the present study.

Conclusions
Understanding the effects of elevated [O 3 ] on soil CO 2 emission is of great importance for assessing future soil C source/sink status since soil plays an important role in the global C budget. The present meta-analysis revealed that O 3 exposure duration was an important factor that controlled the O 3 effect on soil CO 2 emission. Soil CO 2 efflux was significantly reduced in short-term (≤1 year) O 3 exposure but increased in prolonged (>1 year) duration. The response magnitudes of soil CO 2 efflux to elevated [O 3 ] were exacerbated in warm areas with high precipitation. The significantly stimulated soil CO 2 emission in nonagricultural ecosystems, in the FACE experiment, and in the soils of pH ≤ 6.0 was associated with a longer experimental duration. Such responses need to be validated in further studies since the confounding effects among classes reduces the effectiveness of meta-analysis in generalizing across studies.