Patterns and mechanisms of belowground carbon responses to changes in precipitation

Abstract

It is well known that aboveground productivity usually increases with precipitation. However, how belowground carbon (C) processes respond to changes in precipitation remains elusive, although belowground net primary productivity (BNPP) represents more than one-half of NPP and soil stores the largest terrestrial C in the biosphere. This paper reviews the patterns of belowground C processes (BNPP and soil C) in response to changes in precipitation from transect studies, manipulative experiments, modeling and data integration and synthesis. The results suggest the possible existence of nonlinear patterns of BNPP and soil C in response to changes in precipitation, which is largely different from linear response for aboveground productivity. C allocation, root turnover time and species composition may be three key processes underlying mechanisms of the nonlinear responses to changes in precipitation for belowground C processes. In addition, microbial community structure and long-term ecosystem processes (e.g. mineral assemblage, soil texture, aggregate stability) may also affect patterns of belowground C processes in response to changes in precipitation. At last, we discuss implications and future perspectives for potential nonlinear responses of belowground C processes to changes in precipitation.

摘要

地下碳过程对降水变化的响应模式及机制

众所周知，植物地上生产力通常随降水的增加而增加。地下净初级生产力(BNPP)占植物净初级生产力(NPP)总量的一半以上，并且土壤储存了生物圈中最大的陆地碳库，但地下碳(C)过程如何响应降水变化仍不明确。本文从样带研究、模拟实验、数据整合与集成等方面综述了地下碳过程(BNPP和土壤C)对降水变化的响应模式。结果表明，BNPP和土壤C含量对降水变化表现出非线性响应，这与地上生产力的线性响应存在较大差异。C分配、根系周转时间和植物物种组成可能是调控地下C过程对降水非线性响应的3个关键因素。此外，微生物群落结构和长期生态系统过程(如矿物组合、土壤质地、团聚体稳定性)也可能影响地下C过程对降水变化的响应模式。最后，本文讨论了地下C过程对降水变化非线性响应的潜在生态意义及未来工作展望。

INTRODUCTION

The Earth’s surface temperature has increased by 1.09°C since 1850 and is projected to become increasingly apparent due to anthropogenic buildup of greenhouse gases in the atmosphere (IPCC 2021). A warming climate will likely amplify global water cycle, which may largely alter precipitation regimes (Donat et al. 2016; Huntington 2006; Hwang et al. 2011). The anticipated increases in precipitation are very likely in high latitudes, while decreases are likely in most subtropical land regions in this century but with more frequent extreme rainfall events over most land areas (Donat et al. 2016; IPCC 2021). The altered precipitation regime would likely have greater impact on ecosystem carbon (C) dynamics, both above- and belowground processes, than the singular or combined effects of rising CO₂ and temperature (Beier et al. 2012; Weltzin et al. 2003).

It has been well known that aboveground net primary production (ANPP) usually increases with precipitation either along its natural gradients (Bai et al. 2008; Huxman et al. 2004; Knapp and Smith 2001; Rosenzwe 1968; Zhou et al. 2009), or across temporal changes in precipitation within a site (Jobbagy et al. 2002; Lauenroth and Sala 1992; Oesterheld et al. 2001), or across rainfall treatments of manipulative experiments (Yahdjian and Sala 2006). This relationship has been explicitly explored since 1970s, but its slopes largely varied between temporal and spatial series (Guo et al. 2006; Lauenroth and Sala 1992), even among the temporal series in different ecosystems (Huxman et al. 2004; Knapp and Smith 2001). Moreover, the correlation may converge across biomes during the dry years (Huxman et al. 2004). Even if we have perfect understanding of ANPP responses to precipitation, our knowledge on ecosystem C responses is incomplete because ANPP accounts only for less than one-half of productivity in terrestrial ecosystems, especially in grasslands (Scurlock et al. 2002; Zhang et al. 2022).

The belowground NPP (BNPP) represents more than one-half of NPP, especially in dry ecosystems (Lauenroth and Sala 1992; Luo et al. 2009), and soil stores the largest terrestrial C in the biosphere (Bradford et al. 2016), which is critical for ecosystem biogeochemical cycles (Sáez-Sandino et al. 2023). However, only a small percent (<3%) of articles published in ecological journals were focused on belowground processes (Wardle 2002), although the number of studies have considerably increased recently. However, we really do not know whether BNPP also linearly increase with precipitation as ANPP does. Does soil C linearly increase with precipitation? What is the interannual variability in those processes in response to temporal changes in precipitation? We almost have no knowledge on those processes due to major hindrances to field experiments as well as modeling of belowground root dynamics. The scarcity of knowledge about BNPP and soil C storage under varying precipitation limits model development to predict future ecosystem function in response to global change.

Recent research from the transect and modeling studies indicates that response patterns of belowground C processes to variation in precipitation may be nonlinear (Knapp et al. 2017; Scheffer et al. 2001; Zhou et al. 2008, 2009), which is largely different from that of aboveground one (Reid 2005). Differential responses of above- and belowground C processes to precipitation may constrain biogeochemical transformations at the ecosystem scale, likely resulting in changes in ecosystem services to human society (Potts et al. 2006). The Millennium Ecosystem Assessment (Reid 2005) have identified nonlinearities in ecosystem processes as a high-priority research area, listed them as one of the most important uncertainties for decision-making, and raised significant research needs. If nonlinear patterns of belowground C processes to precipitation variation are not adequately addressed, ecosystem C responses to climate change cannot be fully understood and policies to mitigate climate change will fall short in meeting targets of the Kyoto protocol. Therefore, identifying response patterns of belowground C processes to changing precipitation is critical to develop a reliable capability toward predictive understanding of ecosystem responses to climate change in the future.

Belowground root biomass or BNPP in an ecosystem is determined by root growth and death. Root growth of a plant is regulated by C allocation to roots vs. shoots (i.e. root:shoot ratio) while root death is related to root turnover time, all of which are influenced by plant species compositions in an ecosystem. Moreover, root quality and quantity (e.g. C:N ratio), rhizodeposition and root decomposition directly and indirectly affect soil organic matter (SOM) stability (Poirier et al. 2018; Rasse et al. 2005). However, it is largely unclear whether these processes (e.g. root:shoot ratio, root turnover time and plant community structure) mainly contribute to the underlying mechanisms for responses of belowground C processes to changes in precipitation.

To better understand belowground responses to changes in precipitation, we review potential responses and mechanisms underlying belowground C processes. In this review, we mainly focus on BNPP and soil organic carbon (SOC). At first, we characterize possible response patterns of belowground C processes (e.g. BNPP, soil C) and compared them with those of aboveground processes to changes in precipitation from transect studies, manipulative experiments, modeling and data integration and synthesis. Then, the key processes underlying mechanisms responsible for response patterns of belowground C processes to changes in precipitation are probed toward predictive understanding of precipitation effects on ecosystem structure and functioning. This will guide both the experimentalist and modeling communities to better coordinate efforts to conceptualize and parameterize belowground dynamics to improve scientific understanding of the terrestrial C cycle and the accuracy of terrestrial responses to climate change. At last, implication and future perspectives are discussed for potential nonlinear responses of belowground C processes to changes in precipitation.

RESPONSE PATTERNS OF BELOWGROUND C PROCESSES TO CHANGES IN PRECIPITATION

Comparing with the studies on ANPP, the responses of BNPP or SOC to changes in precipitation are relatively lack in the context of global change. Here, we summarize the response patterns of belowground C processes to changes in precipitation from the evidence of transect study, manipulative experiments, data integration and synthesis as well as modeling studies. Based on these evidences, we propose a possible nonlinear response of belowground C responses to changes in precipitation.

Transect studies

Natural rainfall gradients are invaluable to understand mechanisms underlying precipitation control on ecosystem C processes when other environmental variables do not affect the interpretation of spatial patterns (Austin and Sala 2002; Jenny 1980). In these systems over the gradient, plants have adapted to the local precipitation regime, which may reflect the long-term effects compared to short-term manipulative or observation experiments (Meier and Leuschner 2008). To date, numerous transect studies have been carried out to examine precipitation effects on ecosystem processes, mainly focusing on aboveground ones (Austin and Sala 2002; Huxman et al. 2004; Lauenroth and Sala 1992; Santiago and Mulkey 2005; Sun and Du 2017). It is still scarce about how belowground C processes respond to precipitation along the gradients.

Some transect studies had examined belowground C processes and found that their responses to changes in precipitation along the gradients often did not vary at a narrow range of temperatures. For example, along a precipitation gradient from 430 to 1200 mm in tallgrass prairie when mean annual temperature (MAT) ranged from 13.0 to 16.5°C, total root biomass at 0–30 cm remained relatively constant, although aboveground biomass (AGB) linearly increased with mean annual precipitation (MAP, Fig. 1a and b), resulting in a decrease in root:shoot ratio (Fig. 1c, Zhou et al. 2009). Similarly, Zerihun et al. (2006) found that AGB declined with increased aridity whereas root biomass was relatively stable in northeast Australia when MAT was from 19.5 to 22.1°C, leading to a great increase in root:shoot ratio toward arid sites along the gradient. In the western Yucata’n peninsula, Mexico, root biomass also did not significantly change along the precipitation gradient from 530 to 1035 mm in seasonally dry tropical forests when MAT was almost same (25°C, Roa-Fuentes et al. 2012). These results were probably due to a decrease in the proportion of C allocation to roots and an increase in root turnover with increasing precipitation (Comeau and Kimmins 1989; Pietikainen et al. 1999). However, the relatively constant trend in root biomass may not last along the gradients if precipitation goes lower (e.g. deserts or MAP = 0 at the extreme condition), which may result in the nonlinear patter.

When MAT has the wide range along the precipitation gradient, both shoot and root biomass increased due to effects of confounding factors (e.g. temperature). For example, Wang et al. (2005) found that root biomass linearly increased with MAP when MAT ranged from −3 to 7°C. In addition, in 14 mature forest stands of European beech with a range of MAP from 520 to 970 mm despite similar MAT (7.1–8.3°C), fine root biomass significantly increased along the precipitation gradient likely due to nutrient difference, while leaf biomass remained constant, resulting in a significant increase in fine root/leaf biomass ratio toward wetter sites, probably due to increased fine root mortality at low precipitation (Meier and Leuschner 2008, 2010).

Compared to root biomass, SOC content showed the similar responses to increases in precipitation along the gradients. In tallgrass prairie, SOC did not significantly change as root biomass along the precipitation gradient from 430 to 1200 mm (Fig. 1d; Zhou et al. 2009). This likely results from that plant allocation to belowground, which may leave a more distinct imprint on SOC content than aboveground production and litter mass (Jobbagy and Jackson 2000) and/or that increases in litter inputs were roughly balanced by litter decomposition. Nonetheless, SOC increased with increasing annual rainfall when temperature or nutrients varied largely (Amundson et al. 1989; Burke et al. 1991; Quilchano et al. 1995; Schuur et al. 2001). Although transect studies provide an important approach to explore ecosystem responses to precipitation change, the confounding factors (i.e. temperature, nutrient condition) may bias the relationship between precipitation and response variables along the gradient.

Manipulative experiments

Many manipulative experiments have been done to examine ecosystems’ response to precipitation change. Belowground C processes are sensitive to precipitation change, but most often in a different way from the responses of aboveground processes. For example, reduced amounts of precipitation or drought usually restricted aboveground plant production (Fay et al. 2003; Griffin-Nolan et al. 2018; Penuelas et al. 2007). However, root growth and belowground biomass showed contradictory results, with increased root production and biomass (Qaderi et al. 2006), or declined root production (Bai et al. 2010; Peek et al. 2006; Xu et al. 2012), or no change in dry conditions (Gaul et al. 2008), which probably depend on antecedent soil moisture. Increased precipitation often stimulated both above- and belowground production but with less change in belowground biomass or productivity (Harpole et al. 2007; Wall et al. 2006). The different responses of above- and belowground production to precipitation largely result from allometric changes, in which BNPP:NPP ratio was negatively correlated with MAP in global grasslands (Hui and Jackson 2006).

Most belowground studies are conducted in grasslands due to much easier access. Summer drought drastically reduced ANPP but did not substantially affect BNPP in a Mongolian steppe. ANPP recovered quickly after the drought likely because BNPP was not severely damaged (Shinoda et al. 2010). In Jasper Ridge multiple global change experiments (MAP = 655 mm), supplemental precipitation (+50%) increased shoot production but decreased root production, leading to little changes in NPP (Dukes et al. 2005). In the semiarid grassland of Inner Mongolia, China, water irrigation increased ANPP much more than BNPP (Gao et al. 2011). Across most of precipitation manipulative experiments, a meta-analysis study showed that increased precipitation stimulated BNPP much less than ANPP (Wu et al. 2011). However, some studies found substantial changes in BNPP in response to precipitation change. For example, in different grassland ecosystems, experimental drought significantly decreased root production in nearly all studied grasslands (lowland, highland, mountain regions), with a 17% decrease in mountain grassland and a 43% decrease in lowland grassland (Fiala et al. 2009). Our synthesis of all precipitation manipulative experiments showed that both root biomass and BNPP nonlinearly respond to relative changes in precipitation, which increased at the beginning and then reached a plateau (Fig. 2a and b).

In forests, fine roots are more sensitive to precipitation changes than coarse ones. Manipulative experiments with young plants in both pots and field studies provide some evidence. For example, experimental drought with temperate tree saplings or seedlings grown in pots or containers substantially reduced fine root biomass, but root:shoot ratio typically increased (Aspelmeier and Leuschner 2006). A meta-analysis of fine root biomass data from 16 Central European beech forests showed a reduction in fine root biomass when exposed to summer drought (Leuschner and Hertel 2003). The reduction in fine root biomass may result from a reduction in fine root lifespan and relative fine root growth rate, and hence an increase in fine root turnover and the root:shoot ratio under drought (Leuschner and Hertel 2003).

The change in SOC content induced by altering precipitation may be slow. We compiled all data from available precipitation manipulative experiments up to 35 years in the literature to examine response patterns. We found that all drought experiments did not significantly affect SOC content. Increased precipitation experiments also showed no change in SOC for most of studies but it decreased or increased SOC in some studies (Zhou et al. 2016). When we pooled all the data together and found that SOC content nonlinearly responded to relative precipitation change (Fig. 2c). Although the various responses of BNPP and SOC to precipitation occurred in different ecosystems partly due to antecedent precipitation and manipulative magnitudes as well as the vegetation types, nonlinear response patterns appeared in BNPP and SOC with increasing precipitation (Fig. 2). However, more extensive studies are needed to reveal the underlying mechanism on belowground C responses.

Data integration and synthesis

Some synthesized studies were conducted to explore regional and global response patterns of belowground C process to precipitation change. Two global syntheses showed that root biomass have scattered distribution with precipitation (Cairns et al. 1997; Mokany et al. 2006). The scattered responses of root biomass to precipitation change may result from the confounding effects from temperature, soil texture and others. However, when MAT was set within a range of 10–15°C, global BNPP increased at the low range of precipitation, reached a plateau between 1000 and 2500 mm, and then decreased (Fig. 3a, Bond-Lamberty and Thomson 2010), while BNPP in global grasslands showed a sigmoidal response to precipitation when MAT was at 5–15°C (Fig. 3b, Hui and Jackson 2006).

SOC content generally increases with increasing precipitation at a global scale (Post et al. 1982), while temperature may confound the response of global soil C distribution to precipitation. When temperature was restricted at the low range, the response pattern of SOC may be nonlinear. For example, in Tibetan grasslands, SOC content showed a nonlinear response to precipitation with the relatively constant values when MAP <300 mm, steeply increases when MAP >300, then reach a plateau (Fig. 3c, Yang et al. 2008). In these regional/global synthesized studies, biotic factors (e.g. vegetation change) may also considerably confound the belowground response to precipitation change. For example, root depths in different plant functional types may affect BNPP responses to precipitation (Schenk and Jackson 2002). The vegetation regulation makes the belowground responses to precipitation change even much more complex for prediction.

Modeling

Although considerable efforts have been done to model belowground processes, large improvement is hindered by methodological difficulties for estimating belowground C cycling, the complexity of belowground interactions, and an incomplete understanding of belowground response to precipitation change (Jackson et al. 2000; Litton and Giardina 2008; Wullschleger et al. 2001). Currently, due to the limited data of belowground C dynamics (Scurlock et al. 2002), most terrestrial ecosystem models still assume that: (i) C allocation patterns are fixed across large geographic, climatic and taxonomic scales (Litton et al. 2007; Pendall et al. 2004); (ii) BNPP is approximately 50% of total belowground C flux (Law et al. 1999; Vitousek 2004; Zhou et al. 2008) and (iii) belowground C processes respond to climate change in a similar way with aboveground ones (Jackson et al. 2000). In global community land model, new fine root growth tied directly to new leaf growth while new coarse root growth tied to new stem growth (Lawrence et al. 2011). Allocation patterns are constant in time and space within plant functional type, while the horizontal distribution of roots is assumed to match canopies for a mixture of plant functional types (PFTs) (Lawrence et al. 2011). These assumptions are not consistent with the results from transect studies, manipulative experiments and data integration and syntheses, which are different for the response patterns of above- and belowground processes to precipitation. This challenges modelers how to incorporate these knowledge of distinct belowground response patterns to implement new root capability in global models.

Possible nonlinear patterns of belowground responses to precipitation

Several lines of evidence from manipulative experiments, transect study and data synthesis show that responses of belowground C processes to changes in precipitation may be nonlinear, which is largely different from linear response for aboveground one. For example, several transect studies found that SOC content and root biomass did not change along precipitation gradients (Zerihun et al. 2006; Zhou et al. 2009). If these transect studies go further to regions with lower precipitation (e.g. deserts), the constant patterns of SOC content and root biomass might not hold. Conceptually, there should be no roots and soil C as precipitation approaches to zero in barren areas. Indeed, when data points from deserts (i.e. Majove, Chihuahua, Sonoran and Colorado Plateau in USA, Gurbantonggut and Kelamayi in China) are plotted into the results from our transect data in southern Great Plains (Bell et al. 2009; Fernandez et al. 2006; Li et al. 2010; Scheffer and Carpenter 2003; Wheeler et al. 2007; Zhang et al. 2009; Zhou et al. 2009), SOC content shows a nonlinear curve with precipitation (Fig. 3d). Similarly, the nonlinear curve of SOC occurred in the Tibetan grasslands (Fig. 3c, Yang et al. 2008). In addition, BNPP also clearly showed the nonlinear curve in global grasslands when MAP was set at a range of 5.0–15.0°C (Fig. 3b, Hui and Jackson 2006). Similarly, the results from precipitation manipulative experiments showed nonlinear responses of root biomass, BNPP and SOC to relative change in precipitation (Fig. 2).

MECHANISMS UNDERLYING RESPONSES OF BELOWGROUND PROCESSES

Belowground C processes mainly include roots, SOC and their interaction. Root production is the major supply of SOC (Rasse et al. 2005). Root biomass and turnover are largely determined by root growth and death. Root growth is regulated by C allocation of photosynthates to roots (represented by root:shoot ratio). These processes are largely influenced by plant species compositions in an ecosystem. Therefore, root:shoot ratio, root turnover time and species composition would considerably affect ecosystem belowground C dynamics (Aerts et al. 1992; Andrews et al. 1999; Matamala et al. 2003; Vose 1962), which may be three possible key processes underlying mechanisms for the responses of belowground processes to changing precipitation (Fig. 4, Andrews et al. 1999; Gao et al. 2008; Gill et al. 2002; Vose 1962). Their changes may potentially lead to nonlinear response of belowground C processes to changes in precipitation. In addition, microbial community structure and long-term ecosystem processes (e.g. mineral assemblage, soil texture, aggregate stability) may also affect response pattern of belowground C processes to changing precipitation (Fig. 4).

Root:shoot ratio

The root:shoot ratio represents the distribution of foliage photosynthetic products to belowground plant parts (Hui and Jackson 2006; Jackson et al. 1996) as a key parameter for models of terrestrial C cycling (Piao et al. 2004; Warnant et al. 1994; Zhou et al. 2022). However, the resource balance/optimality theory and experimental results suggests that plants will adjust C allocation strategy (above- vs. belowground) to adapt to precipitation changes (Bhattachan et al. 2012; Bloom et al. 1985; Friedlingstein et al. 1999; Luo et al. 1995). Root:shoot ratio thus often increases from mesic to xeric environments in transect studies (Fan et al. 2009; Friedlingstein et al. 1999; Rutter and Whitehead 1963; Zerihun et al. 2006; Zhou et al. 2009) as well as with the increased drought in manipulative experiments (e.g. Niu et al. 2008; Pace et al. 1999). Our synthesis of manipulative experiments showed that drought increased root:shoot ratio by 25.0 ± 1.7% (n = 38) but irrigation decreased it by 13.3 ± 2.6% (n = 7) (Zhou et al. 2018). In addition, a global synthesis study also found that root:shoot ratio was negatively and nonlinearly related to MAP from 786 observations with a wide range of vegetation types (Fig. 5a, Mokany et al. 2006). The nonlinearly increased root:shoot ratio with decreasing precipitation has been applied to biogeochemical models in grasslands to simulate SOC dynamics (Parton et al. 1987).

The decreasing trend in root:shoot ratio likely resulted from changes in the relative importance of limiting resources (i.e. water, light and nutrients) with changing precipitation (Tilman 1988; Vinton and Burke 1997). In the xeric environment, plants increase root:shoot ratio for water to optimize growth. Conversely, plant production is limited more by light and nutrients than water in the mesic environment, resulting in low root:shoot ratio (Knapp and Seastedt 1986; Lane et al. 2000; Schimel et al. 1991). A high root:shoot ratio reflects high water-absorbing capacity (i.e. high water use efficiency), which allows deep rooting plants to have a better chance to survive under drought conditions (Blum 2011). The different responses of above- and belowground biomass to precipitation change may result from the different morphologies, anatomies, physiologies and functions of shoots and roots to some extent (Gregory 2006; Zhou et al. 2018). The responses of root:shoot ratio to changes in precipitation may partly cause the nonlinear responses of BNPP or root biomass to precipitation change, and then that of SOC content (Fig. 3).

Root turnover time

Belowground root growth and death are closely related to root turnover time, which plays an important role in C allocation and nutrient cycling in terrestrial ecosystems. Root turnover time did not show a significant relationship with MAP at a global scale (including grassland, forest, shrublands and wetland) probably due to the confounding effects of other environmental factors (e.g. temperature, Gill and Jackson 2000). However, with increasing precipitation, root turnover time at a specific temperature range (12–20°C) showed a parabolic curve with a turning point at about 450 mm of precipitation (Fig. 5b, Gill and Jackson 2000). The responses may partially contribute the nonlinear responses of BNPP or root biomass to precipitation change (Fig. 3).

In a gradient study in southern Great Plains (Bloom et al. 1985; Friedlingstein et al. 1999; Luo et al. 1995), constant root biomass indicates that root turnover time may increase with precipitation along the gradient based on a prediction of the resource balance/optimality theory (Friedlingstein et al. 1999; Zhou et al. 2009). Since the quantity of belowground C inputs to soil was significantly and positively correlated with soil C storage (Lu et al. 2011; Rasse et al. 2005; Russell et al. 2009; Zhang et al. 2017) and root biomass has a longer residence time in soil than shoot one, root-derived materials may dominantly contribute to stabilize SOM currently stored in soil compared to aboveground litter. To better understand the control of root turnover time on belowground C processes, the well-designed greenhouse and manipulative experiments are required to be conducted in different ecosystems.

Species composition

Root growth and death are considerably influenced by plant species compositions (Espeleta et al. 2009; Gregory 2006). The changes in precipitation are expected to affect composition and ecosystem function of plant communities (Otsu et al. 2023; Suttle et al. 2007; Tylianakis et al. 2008), which may, in turn, affect root biomass or BNPP and then soil C content. The extreme drought of the 1930s resulted in large and rapid decrease in species composition and total plant cover in central grassland region of USA. The resistance to drought was closely correlated with root extent and depth in most prairie grasses (Weaver and Albertson 1936). Along the precipitation gradient from 430 to 1200 mm in southern Great Prairie, species richness linearly increased as those in other studies (Fig. 5c, Adler and Levine 2007). Relative cover of C₄ plants increased firstly, reached a plateau, and then decreased along the precipitation gradient, while relative cover of C₃ plants exhibited the opposite curve (Fig. 5d). Shannon diversity index also displayed a parabolic curve (data not shown), which is similar to results from Naveh and Whittaker (1980). As we discuss above, changes in plant C allocation to belowground, root turnover time and root profiles may cause shifts in species composition, which likely contribute to the relatively constant root biomass and soil C content along the gradient (Zhou et al. 2009).

However, in precipitation manipulative experiments, how changes in precipitation affect species composition and diversity is lack of consensus among studies. The relationships between increased precipitation and species richness/diversity are positive (Sternberg et al. 1999; Stevens et al. 2006; Yang et al. 2011; Zavaleta et al. 2003), negative (Engel et al. 2009; Suttle et al. 2007) or no change (Grime et al. 2008). The effects of the antecedent conditions may be partly attributed to the diverse response of species composition to increased precipitation in short-term manipulative experiments. The opportunistic (e.g. an extreme drought year) and transect studies may be more suitable to examine the long-term responses of plant species compositions to precipitation change, which may contribute to the nonlinear responses of belowground C processes to changing precipitation. Currently, the modeling study did not incorporate species shift with precipitation changes to predict the responses of ecosystem processes, especially global land surface models.

Adjustment in species composition along the gradient is probably one of the key variables determining the response pattern of BNPP to changes in precipitation. In addition, species composition influences SOM dynamics through belowground processes that control transformation of biomass into SOM. The decrease in precipitation results in increased C allocation to belowground root growth by individual plants and/or a shift of community structure to plant species with higher root:shoot ratio. The shift to plant species with higher root:shoot ratio may be more important than increases in plant allocation to belowground growth for maintaining relatively constant root biomass in the high precipitation range (Zerihun et al. 2006; Zhou et al. 2009). As precipitation decreases further to the point that can’t support enough biomass production for root growth regardless of shifts in species composition, a threshold occurs (Fig. 3). The adjustments in root:shoot ratio and root turnover time in the dominant species and shifts in species composition likely contribute to the relatively constant root biomass along the gradient.

Microbial communities

Precipitation-driven differences in belowground plant quality and quantity have direct (C inputs) and indirect (i.e. aggregation) effects on soil C storage. Rhizodeposition (e.g. exudates, border cells, mucilage), litter and root decomposition and sloughing off of root cells, which all are related to soil microbial characteristics, are primary sources of SOM (Rasse et al. 2005). Microbial contribution to soil C storage is mainly affected by the interactions among microbial biomass, community structure, byproducts and soil properties (i.e. texture, clay mineralogy, Six et al. 2006). The composition, abundance and activity of microbes within the soil may thus regulate many belowground C processes, and in turn, influence SOC formation and stability (Castro et al. 2010; Das et al. 2023; Rice et al. 1998). For example, global microbial biomass was linearly correlated with soil C content (Das et al. 2023; Fierer et al. 2009). Soil moisture is one of the primary variables, which exert a large influence on microbial community structure and enzyme activities (Brockett et al. 2012; Guenet et al. 2012; Kardol et al. 2010), and then affect belowground C cycling processes.

The predicted changes in precipitation amount may considerably impact soil microbial communities (Castro et al. 2010; Cregger et al. 2012) and their function is important for predicting C feedbacks to global climate change. For example, in a constructed old-field ecosystem, increased precipitation decreased the relative abundance of Acidobacteria but increased that of Proteobacteria as well as altered fungal community composition (Castro et al. 2010), which may partly result in increased fine root detrital input to soil (Wan et al. 2007). In a piñon–juniper woodland (Pinus edulis–Juniperus monosperma), seasonal variability in rainfall, precipitation treatments and plant inputs of photosynthates from different aboveground community all altered microbial community composition and abundance, probably causing changes in root production (Cregger et al. 2012). Along a precipitation gradient from 100 to 500 mm in the Mongolian steppe, soil microbial biomass C decreased with increasing aridity, which was similar as changes in SOC (Li and Chen 2004; Wichern and Joergensen 2009).

Mycorrhizal fungi (MF) are nearly ubiquitous found in 90% of land plant species (Smith and Read 2008), which receive C from their host plants for nutrient exchange transferring to the roots and benefit plant growth (Shi et al. 2012; Vargas et al. 2010; Zhu and Miller 2003). At a global scale, fungal community structure of arbuscular mycorrhizal (AM) is significantly related to precipitation with higher abundance and activity under drought than wet conditions (Kivlin et al. 2011). In manipulative experiments, researchers found that decreased precipitation increased fungal root colonization with more vesicles and arbuscules (Martinez-Garcia et al. 2012) and had more MF abundant and diverse communities due to altered osmotic and elastic properties in roots compared to control and irrigation treatments (Auge 2004; Hawkes et al. 2011). These symbiotic fungi play crucial roles in regulating belowground C dynamics (e.g. root growth and soil C storage). For example, the ability of MF to form rhizomorphs and simulate root growth correlated with increased plant performance (Boyle and Hellenbrand 1991). Averill et al. (2014) also found that uptake of organic N by MF slowed the rate of SOM decomposition and then increased soil C storage with larger magnitudes for ectomycorrhizal and ericoid mycorrhizal than AM fungi.

Long-term ecosystem processes underlying precipitation responses (i.e. mineral assemblage, soil texture, aggregate stability)

Responses of belowground C processes to altered precipitation regime are complex and potentially idiosyncratic (Giardina et al. 2005), which are also regulated by long-term, slow ecosystem processes on decadal to centurial time scales. Those processes include mineral assemblage, soil texture and soil aggregate stability. In a California conifer forest, e.g., Rasmussen et al. (2005) found that soil mineral assemblage and aggregates were very important on controlling soil C dynamics. Along a precipitation from 160 to 3000 mm on the island of Hawaii, the amount of metastable noncrystalline (e.g. allophane, imogolite and ferrihydrite) increased with increasing precipitation due to high capacity of noncrystalline minerals to stabilize SOC for thousands of years in the wetter sites (Torn et al. 1997). Soil minerals thus play a stabilizing role on soil C storage.

Soil texture is a product of soil long-term development due to interactive effects of several state factors such as climate, parent material and biota, which considerably influences soil moisture retention and thus ecosystem productivity and decomposition (Jenny 1980; Torn et al. 2009). Research found that clay content was positively correlated with soil C storage (Percival et al. 2000), which is proxy for mineral control in soil C models (Parton et al. 1987). The binding and clumping of individual soil particles form aggregates. The interaction of SOM and clay particles forms stable complexes, while plant roots can improve aggregation by producing polysaccharides (Mamedov et al. 2007). Soil aggregation mediates many biological and chemical soil processes and hence decomposition and formation of soil C, but it is considerably affected by climate, soil texture, mineralogy and other factors (Six et al. 2002). Thus, soil mineral assemblage, development, texture and aggregates affect SOM stabilization (Feller and Beare 1997; Rasmussen et al. 2005; Six et al. 2002; Torn et al. 1997). However, their relative importance is difficult to be separated/quantified. Fully exploring those long-term processes need coordinated approaches and require unique projects, which are essential to test and constrain models in order to realistically project ecosystem dynamics at decadal to century time scales (Luo et al. 2011).

IMPLICATIONS AND FUTURE PERSPECTIVES

Accurately projecting future states of climate and ecosystems is our ultimate goal of global change research (Luo 2007). Although substantial research has been conducted on belowground processes from transect studies, manipulative experiments, modeling and data syntheses over the past decades, we still do not have an adequate understanding of the mechanisms underlying how belowground C pools and turnover rates respond to changes in precipitation. Here, we discuss a few future researches that are not mutually exclusive but may simultaneously contribute our predictive understanding of terrestrial climate–C cycle feedback, which are likely to be conducted at large scales and to explore complex ecosystem responses to precipitation change.

The previous discussion highlights some key processes (e.g. root:shoot ratio, root turnover time, species composition) from transect studies, manipulative experiments and data syntheses, which may indirectly contribute to the nonlinear responses of belowground C processes to changes in precipitation to some degree. However, a deep understanding of mechanisms underlying the threshold responses needs further work so we can realistically use the full extent of this knowledge for land management (e.g. ecosystem sustainability, biodiversity enhancement, etc.). Therefore, exploring possible mechanisms need fully designed and coordinated experiments from laboratory incubation, greenhouse and manipulative experiments, isotope tracers, species replacement, data syntheses and modeling to be conducted in the long term as well as remote sensing data. Simultaneous application of multiple approaches also has the potential to provide critical data sets for examining a variety of terrestrial feedback mechanisms to changing precipitation.

Meanwhile, although a large number of transect studies have been conducted, the data are still lack for belowground C processes, especially BNPP and its turnover time. Published root data are not always congruent with the nonlinear response pattern along the precipitation gradients probably due to confounding effects of climate, edaphic conditions and sampling time. More transect studies are needed to examine belowground C responses to precipitation change and understand the possible mechanisms within and across biomes when other environmental variables keep relatively constant (Austin and Sala 2002; Jenny 1980). In these systems, toward predictive understanding of precipitation effects on ecosystem structure and functioning, a comprehensive understanding of above- and belowground processes is required to investigate their responses to changes in precipitation and identify the mechanisms underlying the nonlinear responses.

Manipulative experiments on precipitation change may be more relatively realistic to simulate future climate conditions. However, the majority of experiments were currently manipulated in two treatment levels of precipitation. Although those results provided single-factor pulse response under altered precipitation, there was no information on nonlinearity of belowground C processes along the gradients of precipitation. In addition, most manipulative experiments have been conducted in low-stature grasslands (Luo et al. 2011; Song et al. 2019) but less in high-stature forests (e.g. boreal and tropical forests). Moreover, it is difficult to capture long-term adaptive responses of root system due to relatively short experimental duration. Therefore, long-term manipulative experiments are needed in different biomes with a common metric (Vicca et al. 2012) to examine how ecosystem processes underlying mechanisms interactively determine patterns of belowground responses to changes in precipitation. With several treatment levels, comparative studies along precipitation gradients and/or among ecosystems can produce valuable information on the long-term adaptive response of belowground processes to changes in precipitation.

The models generally incorporate current understanding of ecosystem processes and use those data for parameterization and validation. Terrestrial biogeochemical models offer the potential to integrate experimental results, test hypotheses on nonlinear responses, and advance mechanistic understanding on ecosystem C processes toward prognostic forecast of future ecosystem states (Cramer et al. 2001; Lawrence et al. 2011; Luo and Reynolds 1999; Newbold et al. 2020; Parton et al. 1987). The key processes underlying mechanisms of nonlinear responses of belowground C processes to precipitation change (e.g. root:shoot ratio, root turnover time and species composition) can be incorporated into ecosystem and land surface models to evaluate relative importance of various processes. In addition, vegetation dynamics (e.g. mortality of individual species, seedling establishment and inter- and intra-species competition), microbial community and long-term ecosystem processes (e.g. soil mineral assemblage, soil texture and soil aggregate stability) are also important to explore implication of various patterns and mechanisms on ecosystem functions and responses to climate change. The modeling studies may also enable us to quantify the relative importance of these processes and examine terrestrial feedback mechanisms and interactions among ecosystems, land use and the climate system.

Funding

This work was supported by the National Key Research and Development Program of China (2023YFF0806900), the National Natural Science Foundation of China (31930072, 32241032, 42203076) and the Natural Science Foundation of Heilongjiang Province of China (ZD2021C002).

Conflict of interest statement. The authors declare that they have no conflict of interest.

REFERENCES