Partitioning of root, litter and microbial respiration by plant input manipulation in forests

Soil respiration (R s) is the largest carbon (C) flux from terrestrial ecosystems to the atmosphere and is of great significance to the global C budget. An increasing number of studies have assessed R s through in situ observations and model estimates over the last decades, but the sources and pathways of soil carbon dioxide (CO2) are not fully understood, and great uncertainty remains in R s partitioning of soil CO2 sources. Here, we compiled 236 paired observations that measured soil CO2 fluxes after concurrently removal of living roots (and rhizosphere), litter, and both roots and litter in plant input manipulation experiments conducted at 14 forest sites to partition root + rhizosphere (R r), litter (R l) and soil organic matter-derived microbial respiration (R m) in total soil CO2 flux. We found that R r, R l and R m accounted for 20.1%, 21.8% and 62.7% of the total R s, respectively. Mean annual precipitation (MAP) was the most important factor driving R r/R s, R l/R s and R m/R s, and MAP was positively correlated with R r/R s and R l/R s but negatively correlated with R m/R s, suggesting a significant climatic control over the proportions of R s components. Across all sites, the proportions of R r/R s and R l/R s increased but R m/R s decreased with the increase in soil CO2 flux, suggesting that the proportions of root- and litter-derived soil CO2 are generally higher in the tropics than in cold temperate and boreal forests. More accurate partitioning of R r, R l and R m to elucidate different sources and pathways of soil CO2 flux will provide important insights for the global R s assessment and terrestrial C budget.


Introduction
Soil organic matter (SOM) is the largest terrestrial organic carbon (C) pool, containing more C than global vegetation and the atmosphere combined (Friedlingstein et al 2022). SOM is primarily derived from plant input, which also stimulate microbial decomposition of SOM (Bastida et al 2019) and release part of the SOM into the atmosphere as carbon dioxide (CO 2 ). Soil respiration (R s ) is the largest CO 2 flux from the terrestrial C pool to the atmosphere (Davidson and Janssens 2006) and is of great significance in regulating the global C budget and atmospheric CO 2 concentration, with values estimated from 72.6 to 91 Pg C/year (Hashimoto et Lu et al 2021). In recent decades, a growing number of studies have evaluated R s from in situ measurements and model estimates across a broad range of terrestrial ecosystems, but the sources, pathways and contributions of soil CO 2 flux are not fully understood. It still remains challenging to partition the various components of R s , such as root (and root-associated rhizosphere) respiration (R r ), litter respiration (R l ) and SOM-derived microbial respiration (R m ).
Partitioning of the different sources of R s is a key control incorporated in Earth system models (Wieder et al 2015, Luo et al 2016. R s has historically been divided into heterotrophic respiration (R h ) by microbial decomposition of SOM and autotrophic respiration (R a ) from living roots and rhizosphere (Hanson et al 2000, Bond-Lamberty et al 2004. The trenching method has been widely used to separate R h and R a by cutting the rooting system in the plots (Savage et al 2013), and R h can be estimated as the soil CO 2 flux without living roots (i.e. plots with root removal, or girdled plots with trenching) or be disentangled by isotopic fractionation (Hanson et al 2000, Jian et al 2022. Recent syntheses have found that there was no significant difference between these methods in evaluating R a , which accounts for an average of approximately 41%-42% of R s at the global scale (Lu et al 2021, Jian et al 2022, but the turnover time varies greatly between R h and R a . For example, C turnover from progressive decomposition of stable SOM (R h ) was estimated to be hundreds or even thousands of years, particularly in cold biomes (Luo et al 2019), but recent photosynthate C delivered from phloem could be metabolized by living roots (R a ) only several hours (Prescott et al 2020). Therefore, it is critical to isolating R h and R a because their C sources represent dramatic residence times. In the last three decades, there has been concerted scientific effort into understanding how much C was lost from plant litter (Parton et al 2007, Soong et al 2015, but the contribution of litter decomposition to total soil CO 2 is usually not separated from R h (SOM-and litter-derived CO 2 fluxes) in current estimates. Therefore, a consistent method in partitioning of R s from different pathways (root-, litter-and SOM-derived soil CO 2 flux) remains highly uncertain.
Detritus input and removal treatment experiments provide an excellent opportunity to partition the components of R s from different sources (roots and litter) by manipulating plant inputs into soils. However, some previous studies synthesized only the proportion of root-(R r /R s ) or microbial-respiration (R m /R s ) from unpaired observations measured at individual sites without incorporating litter-derived respiration (R l /R s ) (Lu et al 2021, Jian et al 2022, resulting in uncertainty in the partitioning of global soil respiration. Here, we synthesized 236 paired observations to assess the proportions of R r /R s , R l /R s and R m /R s by concurrently removing living roots + rhizosphere, litter, and both roots and litter from current detritus input and removal treatment experiments conducted at 14 forest sites. Our objectives were (a) to distinguish the partitioning of forest soil respiration from different sources through plant input manipulations, and (b) to elucidate the underlying mechanisms of variations between forest types with different climates and edaphic factors and the underlying mechanisms of uncertainty in current estimates in forest soil CO 2 flux in a global context.

Data compilation
We searched peer-reviewed journal articles that reported the soil CO 2 fluxes after concurrent removal of living roots, litter and both roots and litter in forests using the Web of Science and China National Knowledge Infrastructure databases in December 2021. The search terms were as follows: (a) 'soil respiration OR soil CO 2 flux' , AND (b) 'litter/root removal OR litter/ root manipulation OR litter/root alteration OR litter/ root input' , AND (c) 'forest' . All observations of the soil CO 2 flux included in our dataset met the following criteria: (a) All data were retrieved from detritus input and removal treatment experiments and included paired observations with concurrent removal of living roots + rhizosphere, litter, and both roots and litter to ensure that the partitioning of R r /R s , R l /R s and R m /R s were comparable at independent forest sites and during the same experimental period. The R r /R s was measured by the root trenching method, and several previous studies did not find systematic bias in the trenching method compared with other partitioning methods (Bond-Lamberty et al 2011, Comeau et al 2018, Jian et al 2022. (b) The soil-atmosphere CO 2 fluxes were measured in filed plots with soil chambers by using gas chromatography, in situ infrared gas analysis, or by soda lime methods, and those from laboratory incubations or model estimates were excluded. Given that most living roots, litter and microorganisms are distributed in topsoil and most of the soil CO 2 is produced in surface soils (Wordell-Dietrich et al 2020), all CO 2 fluxes were measured at the soil-atmosphere interface. (c) The proportions of R r /R s , R l /R s and R m /R s were evaluated under natural conditions, and there were no other global change manipulations, such as elevated temperature, CO 2 enrichment, nitrogen (N) deposition, and increased or decreased precipitation.
The coordinates, elevation, climatic factors (mean annual temperature (MAT) and precipitation (MAP)) and edaphic factors (bulk density, clay, pH, and soil C and soil N concentrations) were obtained from the articles directly or retrieved by using the WorldClim and SoilGrids databases. Experimental duration did not significantly change the proportions of R r /R s , R l /R s and R m /R s (all P > 0.05, figure S1), so the soil CO 2 flux data collected continuously over a time series at a certain forest site were included in our dataset. Moreover, the proportions of R r /R s , R l /R s and R m /R s did not vary significantly among seasons (spring, summer, autumn and winter) across the studied sites (all P > 0.05, figure S2), so we pooled the data measured at different seasons in this study.
A total of 236 paired observations (root + rhizosphere removal, litter removal, and both root and litter removal for R r /R s , R l /R s and R m /R s , respectively) were retrieved from 17 articles (appendix S1) with detritus input and removal treatment experiments conducted at 14 forest sites (table S1). The latitude ranged from 9.06 • N to 47.90 • N, elevation from 186 m to 2476 m, MAT from 3.8 • C to 27.0 • C, and MAP from 548 mm to 3500 mm.

Data visualization
R r and R l were estimated as the reductions in soil CO 2 fluxes (g CO 2 /m 2 /d) derived from living root + rhizosphere removal and litter removal relative to the control plots (without removing roots or litter), respectively. R m was estimated as the soil CO 2 flux after removing both roots and litter.
The percent of R r /R s was calculated as follows: The percent of R l /R s was calculated as follows: The percent of R m /R s was calculated as follows: where C c is the soil CO 2 flux in control plots (without removing roots or litter), and C r , C l and C r+l are the soil CO 2 fluxes in roots + rhizosphere removal, litter removal, and both roots and litter removal plots, respectively.

Statistical analysis
The average proportions of R r /R s , R l /R s and R m /R s were estimated by a linear mixed effect model by assigning 'site' as an independent variable and by nesting the 'observation' within each site in the lme4 package in R 4.2.2. These values were reported as the percent contributions to total soil CO 2 fluxes, and 95% bootstrap confidence intervals (CIs) were calculated to denote the ranges of variations. Given that the number of paired observations is limited, a resampling test with 999 iterations was used to assess the 95% bootstrap CIs (Ni et al 2022). The frequencies of these data were assumed to follow normal distributions and to fit Gaussian functions (figure S3). A linear mixed effect model was used to determine the relationships of R r /R s , R l /R s and R m /R s with the total soil CO 2 flux, which was standardized as g CO 2 /m 2 /d across all of the studied forest sites. A Kruskal−Wallis test was used to assess differences among forest types. If the Kruskal−Wallis test results were significant (P < 0.05), the differences between two independent samples were then tested by the Wilcoxon rank-sum test with Bonferroni corrections in MATLAB R2012a (MathWorks Inc., Natick, MA, USA). A partial least square (PLS) model was used to distinguish the relative importance of climatic (MAT, MAP) and edaphic (clay content, pH, and soil C and N concentrations) factors with respect to the proportions soil respiration components (R r /R s , R l /R s and R m /R s ) at site-level. Standardized coefficient and relative importance of those variables estimated based on Akaike information criterion were used to assess the effects of climatic and edaphic variables on these proportions.

Results
Root + rhizosphere respiration and litter respiration accounted for 20.1% (95% CI of 13.0% to 27.2%) and 21.8% (95% CI of 14.6% to 29.2%), respectively, of the total soil respiration, and SOM-derived microbial respiration accounted for 62.7% (95% CI of 55.9% to 69.5%) of the total soil CO 2 flux across the studied forest sites (figure 1). The fluxes of R r (r = 0.52), R l (r = 0.59) and R m (r = 0.82) were positively related to the total soil CO 2 flux (all P < 0.001, figure S4). The proportions of R r /R s (P = 0.044, figure 2(a)) and R l /R s (P = 0.022, figure 2(b)) increased significantly with the increase in soil CO 2 flux, while the proportion of R m /R s (P < 0.001, figure 2(c)) decreased with the increasing soil respiration. The proportion of R r /R s was consistent across forest types (P = 0.830, figure 3(a)), but R l /R s was significantly lower in mixed (14.6%, with 95% CI of −3.6% to 33.6%) and deciduous forests (16.9%, with 95% CI of 7.8% to 26.1%) than in evergreen (30.6%, with 95% CI of 20.8% to 40.3%) forests (P = 0.001, figure 3(b)). The proportion of R m /R s differed slightly between forest types (P = 0.025), with higher value observed in deciduous than in evergreen forests ( figure 3(c)). PLS analysis showed that MAP was a dominant factor affecting R r /R s , R l /R s and R m /R s (figure 4). R r /R s was strongly controlled by MAP and soil N concentrations ( figure 4(a)), which were positively  figure 4(c)). Neither R l /R s nor R m /R s were significantly controlled by soil

Discussion
Partitioning of the sources and pathways of soil CO 2 flux is of significance for soil C sequestration and greenhouse gas mitigation. In this study, we evaluated the soil respiration derived from root-(and rhizosphere-), litter-and SOM-derived microbial pathways from paired observations in current detritus input and removal treatment experiments and found that R r , R l and R m accounted for 20.1%, 21.8% and 62.7%, respectively, of the total soil CO 2 flux in forests.

Partitioning of soil respiration in forests
Root trenching has historically been used to partition R r from R s (Hanson et al 2000, Sulzman et al 2005, Subke et al 2006. Our synthesis compiled data from paired observations that removed living roots and associated mycorrhizal fungi and priming effects by root exudates, showing a slightly lower R r /R s than several previous assessments across a broad range of terrestrial ecosystems. For example, an earlier estimate showed that the average contribution of autotrophic sources to total soil CO 2 flux was 48%, ranging from 40% to 50% (Hanson et al 2000). A recent study found that a global autotrophic CO 2 flux of 43.8 ± 0.4 Pg C/year based on the Global Soil Respiration Database (version 4.0) (Tang et al 2019), and this flux ranged from 43.6% to 57.2% of the total soil respiration (Konings et al 2019, Warner et al 2019, Tang et al 2020, Lu et al 2021. The R r /R s has recently been estimated to be 0.42 based on a random forest model (Jian et al 2022) but varies greatly among ecosystems, with a range from 0.18 to 0.48 (Lu et al 2021, Jian et al 2022. On the one hand, forests have greater amounts of roots relative to other terrestrial ecosystems, and the decomposition of dead roots after trenching leads to an apparent increase in heterotrophic respiration, which in turn results in an underestimation in the proportion of R r /R s . On the other hand, the removed mycorrhizal fungi at the trenched plots has been found to contribute an average of 15% of total soil respiration (Han et al 2021). Therefore, even a small difference in R r /R s could have a strong influence on the global C budget, and a more accurate evaluation of the partitioning of soil CO 2 flux will greatly improve the predictions of Earth system models.

Litter respiration contributes to soil CO 2 flux
The CO 2 flux released from decomposing litter on the soil surface has not been separated from heterotrophic respiration in most assessments of soil respiration, and R l has been considered predominantly to be the CO 2 efflux induced by faunal and microbial respiration during litter decomposition. Although soil fauna are considered 'soil ecosystem engineers' and contribute to the decomposition of plant litter, few studies have evaluated the proportion of faunal respiration in the total soil CO 2 flux. A recent model simulation found that soil fauna contribute to an average of 3% of soil respiration (Fatichi et al 2019). Another study using 13 C isotope labeling found that plant litter contributed an average of 33% to total soil respiration in a warm Mediterranean forest (Albanito et al 2012). These results are slightly higher than our value of 21.8% for R l /R s estimated from the paired observations from detritus input and removal experiments across 14 forest sites. Assuming a global soil CO 2 flux of 72.6-91 Pg C/year (Hashimoto et al 2015, Jian et al 2018, Warner et al 2019, Huang et al 2020, Lu et al 2021, R l is estimated to range from 15.8 to 19.8 Pg C/year. This estimate suggests that CO 2 flux during litter decomposition is a tremendous pathway of terrestrial C emission, particularly in warm tropical/subtropical forests with higher litter production and faster decomposition rate than in cold temperate and boreal forests (Parton et al 2007). More accurate evaluations across a broad range of terrestrial ecosystems should be addressed on the partitioning of CO 2 flux released from decomposing litter.

Climatic control over soil respiration in forests
Microbial decomposition of SOM has been estimated to be 39-57 Pg C/year (Warner et al 2019, Tang et al 2020, Lu et al 2021, Yao et al 2021 and is largely dependent on temperature, as most soil C degradation relies on extracellular enzyme activity (Davidson and Janssens 2006). The Arrhenius function states that reactants with higher activation energies (i.e. less reactivity and more refractory) should have higher temperature sensitivity (Tjoelker et al 2001), which increases with increasing substrate molecular complexity (Lehmann et al 2020). Michaelis−Menten kinetics show that the maximum decomposition rate increases as temperature increases, but the enzymes begin to denature, and the decomposition rate decreases rapidly when the temperature exceeds the optimal threshold (Moorhead and Weintraub 2018). This could explain why R m flux gradually decreased with increasing soil CO 2 flux, which is greater in tropical/subtropical forests with higher temperatures than in temperate and boreal forests (figure S5; Jian et al 2018, Huang et al 2020. By contrast, R m /R s was highly independent of substrate quality, such as soil C and N concentrations (figure S6), suggesting a climatic (not substrate) control over microbial decomposition of SOM in forests.
The proportion of R m /R s was lower at sites with higher R s levels (figure 2), suggesting that soil CO 2 flux in tropical/subtropical forests may be dominantly controlled by root and litter respiration due to greater microbial metabolism and faster C turnover. By contrast, microbial decomposition of SOM could be limited in cold temperate forests because the extracellular enzyme activity is constrained by low temperature (Huang et al 2020). By contrast, both R r /R s and R l /R s increased but R m /R s decreased with increasing soil respiration (figure 2). Globally, soil respiration is generally higher in the tropics than in cold temperate and boreal forests (Huang et al 2020), suggesting that the proportions of living roots and litter decomposition contributing to total soil CO 2 flux were greater in warmer and wetter tropical/subtropical forests with higher microbial metabolism that contributes to SOM turnover.
It is noteworthy that the proportions of root-(and rhizosphere-), litter-and SOM-derived microbial respiration was strongly correlated with precipitation (figure 4), which has been considered a proxy for soil moisture (Hursh et al 2017, Yao et al 2018. The nexus between soil CO 2 flux and MAP is more complicated than that between MAT (enzymatic kinetics) and soil CO 2 flux, as root and microbial activity vary greatly with inherent physiological tolerance, initial soil moisture, soil texture and the magnitude and duration of precipitation events (Ni et al 2022). A previous study found that both root metabolism and microbial catabolism are physiologically dependent of water and nutrient availability, particularly in dry upland soils with a low water potential and in coarse-textured soils, as pore connectivity strongly controls hydraulic conductivity and solute diffusivity (Manzoni and Katul 2014). Therefore, soil CO 2 flux has been found to be highly sensitive to precipitation change in upland forests (Hursh et al 2017, Yao et al 2018, even for short-term precipitation pulses (Ni et al 2019). Moreover, plant C input has been found to be positively correlated with soil respiration across biomes (Hursh et al 2017), but litter production is generally greater in tropical/subtropical forests (Jia et al 2016), showing different spatial and temporal patterns of litter-derived CO 2 flux in various climatic zones. In addition, litter substrate quality has widely been demonstrated to be the primary factor controlling C release from decomposing litter (Adair et al 2008, Wickings et al 2012 and further controls R l /R s , yet this proportion of litter-derived CO 2 flux remains uncertain.

Uncertainty analysis
Although we quantified the proportions of R r /R s , R l /R s and R m /R s from 236 paired observations from detritus input and removal treatment experiments in forests, limitations and uncertainties remained in a few aspects. First, R r was evaluated by a trenching method, which had a potential effect on other R s components, increasing the uncertainty in separating the sources and pathways of total soil CO 2 flux. For example, the removal of living roots led to the recently dead roots remaining in the plots after trenching, increasing the C substrate for microbial decomposition of plant residues and ultimately resulting in an overestimate of R m (Savage et al 2013(Savage et al , 2018. Moreover, water uptake by living roots was reduced in the trenched plots, resulting in a change in soil moisture, which is a key control over R s in upland ecosystems (Subke et al 2006). In addition, the elimination of living roots could reduce R m by a lack of priming effect due to the reduced root exudates (Cheng 2009).
Second, although the proportions of R r /R s , R l /R s and R m /R s were consistent with experimental duration (all P < 0.05, figure S1), we cannot ignore the potential legacy effect of trenching because the dead roots in the trenched plots decompose faster at early stages but slower at later stages (Sun et al 2022). Third, several previous estimates showed that global autotrophic and heterotrophic respiration increased gradually (Tang et al 2019, Yao et al 2021, but the proportion of R r /R s decreased consistently over the past half-century (Jian et al 2022). However, the limited sample sizes (n = 236) collected from only paired observations with concurrently removal of roots (and rhizosphere), litter, and both roots and litter in forests may lead to a systemic bias in data calculation. Therefore, we did not calibrate the interannual variations in the proportions of R r /R s , R l /R s and R m /R s in this study due to the limited data measured in the detritus input and removal treatment experiments. Finally, the data included in this study were not completely independent at site level, although the study site was calibrated as a random effect in the linear mixed effect model. More in situ studies should be conceived at various climate zones with different forest and litter types to disentangle the proportions of soil respiration components.

Conclusion
Although uncertainty remains in this study, to our knowledge, this is the first attempt to concurrently assess the proportions of root (and rhizosphere), litter and microbial pathways to soil CO 2 efflux in forests through paired field observations. R r , R l and R m accounted for 20.1%, 21.8% and 62.7%, respectively, of total soil respiration across the studied forest sites. The proportions of R r /R s and R l /R s increased but R m /R s decreased with the total soil CO 2 flux, but all R s components were strongly associated with local temperature and precipitation across forest types. It is noteworthy that our estimates facilitate the partitioning of litter-derived CO 2 flux from heterotrophic respiration more accurately and could be incorporated in Earth system models to calibrate the sources, pathways and contributions of forest soil CO 2 flux (Shao et al 2013). The isolated proportion of litter derived CO 2 flux shed new light on our theoretical understanding of the efflux of soil organic C that is closely interacted with litter input in forests.

Data availability statement
The data that supports the findings of this study are available in the supplementary material of this article.