Introduction

Crop diversification, including temporal diversification (e.g., rotations, multiple cropping, and cover crops after harvest of the main crop) and spatial crop diversification (e.g., intercropping, species mixtures, cultivar mixtures, cover crops and agroforestry), have repeatedly been shown to enhance ecosystem functioning, such as enhanced yield and stability, increased resource-use efficiency, enhanced soil fertility, reduced crop disease, and minimized environmental costs (Li et al. 2021b; Renard and Tilman 2019; Tamburini et al. 2020; Yang et al. 2021a). Competition, complementarity, and facilitation play pivotal roles in shaping community structure and driving productivity; the net outcome is determined by their relative strengths (Callaway 2007; Jiao et al. 2021a). A large number of case studies and reviews indicate that enhanced ecosystem functioning in diverse agroecosystems is derived from temporal and spatial niche complementarity (Brooker et al. 2015; Homulle et al. 2021; Li et al. 2020b; Yang et al. 2021a; Yu et al. 2015, 2016).

Aboveground complementarity promotes light interception and use efficiency by differences in crop height and light requirement; the greater temporal complementary use of resources also enhances yield benefits (Yu et al. 2015; Zhang et al. 2008). Belowground complementarity increases water and nutrient acquisition by niche differentiation (e.g., hydraulic lift by deep-rooting species) and resource partitioning (e.g., crop species use different forms of a given nutrient) (Homulle et al. 2021; Li et al. 2021a). Interspecific facilitation is also a main driver enhancing ecosystem functioning via resource enrichment, disease and pest protection, and toxin alleviation in multi-cropping systems (Brooker et al. 2021; Li et al. 2014, 2021a; Yang et al. 2021b).

Our evidence supports the complementarity theory for above- (Zhang et al. 2015b, 2017, 2021b) and belowground components (Li et al. 1999; Liu et al. 2015, 2020; Xia et al. 2013b). Numerous studies have emphasized the importance of belowground processes in positive biodiversity-ecosystem functioning (BEF) relationships in both natural ecosystems and agroecosystems (Brooker et al. 2015, 2021; Homulle et al. 2021; Li et al. 2014; Wright et al. 2017b; Yu et al. 2021). Therefore, this review mainly summarizes and synthesizes the advances in belowground processes that drive yield benefits and sustainability of intercropping systems.

We have spent 30 years exploring belowground mechanisms underlying enhanced ecosystem functioning in several intercropping combinations. Some experimental methods are widely used to uncover belowground interactions in intercropping systems. More evidence and belowground mechanisms have been uncovered around the world which broadened the view and promoted advances in crop-diversification research. Therefore, below, we first review some widely-accepted methods to study root-root interactions, and then summarize recent advances in belowground interspecific competition and facilitation that determine the outcome of diverse agroecosystems. We also focus on the effect of intercropping on mechanisms underlying maintaining or improving soil fertility. Then, we review the yield stability and acclimation to climate change via belowground processes in agroecosystems with intercropping. Finally, we provide some perspectives for future crop diversification research from the point of view of root-root interactions.

Methods to explore root-root interactions in intercropping

Root sampling and measurement are critical for understanding belowground interactions between plants in a real soil environment or under field condtions. Researchers have gradually overcome the difficulty in measuring and quantifying root systems (Freschet et al. 2021).

Excavation of roots is a method that can directly show the architectural, morphological and chemical traits of the root system (Böhm 2012; Freschet et al. 2021). Trenches are dug in vertical and horizontal planes to study root distribution in intercropping and monoculture plots in a field experiment (Li et al. 2006). The depth extends to 100 to 130 cm in different periods, and the width is 100 cm to facilitate excavation. After digging, the working face or wall of the profile is smoothed by carefully removing the soil. A cleaned profile is used to observe the root distribution of different species. Monolith excavation such as a “shovelomics approach” (Böhm 2012) usually takes soil monoliths from the soil surface and collects the root systems. Standard shovels are used to excavate a soil cylinder of 40 cm diameter and a depth of 25 cm with the plant in the center. Root crown architecture is evaluated after removing the soil adhering to the root crown (Trachsel et al. 2011).

Soil core samplings with a soil auger take a cylindrical volume of soil, and the roots are then separated from the soil by washing. In different experiments, the auger diameter should be adjusted to the mean root diameter and root length density, ranging from 2 to 10 cm. Soil cores (5.5 cm diameter) are collected at 10 cm intervals to a maximum depth of 100 cm to determine the vertical root distribution in intercropping and monocropping plots in different growing stages (Liu et al. 2015; Xia et al. 2013b; Wu and Guo 2014) developed a conventional soil core sampling method with wide corer (50 cm diameter, 55 cm deep), sampling machinery, and analysis software, providing a novel technique for quantifying the root architecture of field-grown crops.

Rhizoboxes are usually designed for glasshouse experiments to study root-root interaction with high precision. Rhizoboxes and pots with different barrier types are designed to determine the interactions between roots. A solid barrier eliminates root interactions and solute movement, making it similar to a monoculture; a mesh barrier prevents root intermingling but permits the exchange of root exudates. No barrier allows full root-root interactions, i.e. true intercropping (Garland et al. 2017; Li et al. 1999). Similar experiments have also been constructed in the field to study the effect of root interactions on crop growth; a trench (1 m deep, 0.1 m wide, and 8 m long) is excavated between the two crop strips and the root barrier material (nylon mesh or a flexible plastic sheet) are buried in each trench before sowing (Li et al. 2007; Wang et al. 2018). However, the method causes severe soil disturbance in field experiments in the first year. It probably gives more reliable results in the second year.

Minirhizotrons provided non-destructive, repeated monitoring of fine roots over time with tubes (typically 2–5 cm in diameter and up to 2 m in length) installed in the soil in which small video cameras or scanners are inserted at regular intervals (Hendrick and Pregitzer 1996; Iversen et al. 2012). Combined with results from minirhizotrons in the field and glasshouse experiments, researchers estimated the P facilitation by cluster roots to neighbor plants with non-cluster roots (Teste et al. 2020). In a rhizobox experiment, Zhang et al. (2020a) used a minirhizotron camera and showed that maize (Zea mays L.) roots grow along with faba bean (Vicia faba L.) roots in maize/faba bean intercropping. Compared with minirhizotrons, root windows are transparent viewing planes installed in soil with much larger viewing areas (Smit et al. 2000), which may be more suitable in root-root interaction studies.

Tracer techniques are usually applied by adding tracers to the soil dissolved in water either by spraying an area around target plants or by injecting them into the soil, then harvesting the plant after a certain period and measuring the tracer concentrations (Xiao et al. 2004; Hoekstra et al. 2014). Tracers are usually included in two classes: radioisotopes and stable isotopes. Radioisotopes were popular a few decades ago, they were used to determine root growth, activity, and distribution by injecting tracers into the soil at different depths and lateral distances from the target plants. For example, Hauggaard-Nielsen et al. (2001) used radioactive phosphorus (32P) to assess the root depth and lateral root distribution in barley (Hordeum vulgare L.)/pea (Pisum sativum L.) intercropping. Stable isotopes are usually used in more recent studies, which focus on nutrient and water uptake. For example, fertilizer labeled with enriched 15N was used to study the nitrogen (N) uptake of different plants at different times or depths or of different chemical forms which can determine niche differentiation between neighboring plants (von Felten et al. 2009).

The 15N natural abundance method is often used in studies of N cycling in ecosystems. Based on differences in 15N abundance between legumes, non-legumes, and atmospheric N2, this method can assess the reliance upon N2 fixation for legumes and the transfer of N from the legume to neighboring non-legumes (Peoples et al. 2015). For example, Chapagain and Riseman (2014, 2015) used this method to quantify N2 fixation by faba bean and pea, and subsequent transfer to non-legume neighbors. Similarly, 13C natural abundance can be used to separate roots from C3 and C4 plants (Hobbie and Werner 2004).

The collection and measurement of root exudates can help uncover nutrient-mediated and signal-mediated root-root interactions in intercropping systems. Oburger and Jones (2018) compared the methods to sample root exudates in different substrates. Measuring root-released or rhizosphere soil phosphatases, protons and carboxylates, shows the differences in P mobilization among species and reveals interspecific P facilitation between efficient P-mobilizing species and inefficient P-mobilizing species in intercropping (e.g., Li et al. 2007). The measurement of hormones and secondary metabolites may further help detect potential signaling components that trigger root-root competition (e.g., Kong et al. 2018) and facilitation (e.g., Li et al. 2016a).

Adopting these wide-accepted methods improves the exploration of root-root interaction in intercropping, which helps uncover belowground competition and facilitation between crop species in both field and glasshouse experiments.

Competition

Competition for resources is one of the fundamental mechanisms determining the structure and development of plant communities. In diverse agroecosystems, intra- and interspecific competition among crop individuals is vital for the growth and yield of components in multi-cropping systems (Vandermeer 1992). Compared with shoot competition for light, root competition is generally greater in nutrient-poor environments. Interspecific root competition is generally more important when the neighbor is a strong competitor, like a Poaceae species (Kiaer et al. 2013; Ravenek et al. 2016). Due to the simultaneous use of two or more crop species/cultivars in the same field, competition, especially between root systems, is considered a key factor that impacts growth and yield in diverse agroecosystems compared with monocultures (Caballero et al. 1995; Nassab et al. 2011). For example, when maize is grown next to wheat (Triticum aestivum L.), shoot growth and horizontal root distribution of maize are inhibited by the presence of wheat (Fig. 1).

Fig. 1
figure 1

Field observations of competition-based and facilitation-based intercropping systems. (a) Above- and belowground performance of maize intercropped with wheat (at the left) and faba bean (at the right). Maize grown with wheat experiences strong competition. In contrast, the leaves of maize are greener when intercropped with faba bean. Photo credit: Long Li. (b) Compared with a monoculture, the lateral distribution of maize roots is suppressed in intercropping, while the roots of maize grow toward faba bean. Photo credit: Long Li. (c) Maize root distribution in monoculture, maize/wheat intercropping, and maize/faba bean intercropping. Maize roots cannot expand beneath the wheat area, while they grow in the faba bean area (Li et al. 2006)

Full size image

Intra- and interspecific competition can be driven through resource depletion, i.e. scramble competition, and by root secondary metabolites, allelochemicals that inhibit the access of neighbors to resources, i.e. contest competition (Fig. 2, Schenk 2006). In scramble competition, root-root interactions are thought to mainly occur via water and nutrient competition whereby these resources acquired by an individual plant would no longer be available to a neighboring plant (Bechmann et al. 2014; Chen et al. 2012; Fransen et al. 2001; Guderle et al. 2018; Mommer et al. 2011). Competitive ability, earlier emergence, and the competition-recovery processes together determine the outcome of competition-based intercropping. The response of root plasticity in morphological and physiological to belowground resource use may be driven by competition for nutrients and water (Bechmann et al. 2014; Guderle et al. 2018).

Cereals show greater competitive ability than neighbors in belowground competition in barley/white lupin (Lupinus albus L.), wheat/white lupin, barley/vetch (Vicia sativa L.), wheat/vetch, wheat/maize intercropping systems, thus promoting shoot growth of cereals (Li et al. 2001a, b; Mariotti et al. 2009; Yin et al. 2020). The roots of intercropped wheat expand in the soil area under maize plants, exhibiting much greater root length density than in monoculture (Fig. 1c, Li et al. 2006; Ma et al. 2019a). The greater phenotypic plasticity of wheat roots attributes to the greater N uptake in intercropping (Liu et al. 2015). A field study further shows that rainfed condition promotes the horizontal distribution of intercropped wheat roots than supplementary irrigation treatment, strengthening the advantages of wheat (Ma et al. 2019a). In agroforestry systems, tree species exhibit a greater root competitive ability than crop species. Belowground competition in wheat/jujube (Zizyphus jujuba Mill.) and wheat/walnut (Juglans regia L.) agroecosystems results in variation of roots traits of both intercropped species (Wang et al. 2014a; Zhang et al. 2013, 2015a). The strength of interspecific competition is linked to the distance between different species. In addition, higher irrigation levels in apple (Malus pumila Mill.)/soybean (Glycine max (L.) Merr.) intercropping divert the fine roots of apple trees to the topsoil layer and intensify competition. The strong plasticity and the growth and distribution of roots increase water-use efficiency compared with monocultures (Fig. 2, Zheng et al. 2021).

Species that are sown earlier may explore the soil volume freely and occupy the dominant niche, while the later-sown species likely experience interspecific competition and inhibition during the co-growth stage (Homulle et al. 2021; Li et al. 2001b). After the early-maturing crop harvest, the late-maturing crops recover via root plasticity and use of extra nutrients under the area where the early-maturing crop was growing (Liu et al. 2015, 2020; Ma et al. 2020). Interspecific interactions lead to an increase for dominant species (e.g., wheat in wheat/maize intercropping), but a decrease for subordinate species (e.g., maize in wheat/maize intercropping) in terms of yield and nutrient acquisition during the co-growth stage (Li et al. 2001a). The yield benefit of wheat comes from the competitive advantages in the co-growth stage, and the yield and nutrient acquisition advantages of the subordinate species (e.g., maize) is derived from the recovery in the late growth stage after the dominant species has been harvested (Li et al. 2001b). A recovery effect for the late-maturing crop that follows the early-maturing crop has also been reported in wheat/soybean (Li et al. 2001a, b), maize/pigeon pea (Cajanus cajan L.) (Dalal 1974), maize/chickpea (Cicer arietinum L.) (Xia et al. 2013b), and maize/pea (Tan et al. 2020; Zhao et al. 2016) intercropping systems. Studies also showed that the growth trajectories in wheat/maize and barley /maize exhibit a temporal niche differentiation, which compensates the growth of intercropped maize after the earlier-maturing crop harvest (Zhang et al. 2015b, 2017). Intercropped crops with different growth periods lead to shifts from competition to complementarity via improved spatial-temporal niche differentiation.

In contest competition, allelopathy via root exudates affects belowground interactions in a species- and dose-specific manner and is common in both natural and agricultural ecosystems (Fig. 2, Bertin et al. 2003; Wang et al. 2021c). Previous studies have reviewed the main allelochemical compounds (Macias et al. 2019) and signaling molecules (Wang et al. 2021c) in root-root interactions. The main allelochemicals include phenolic compounds (e.g., simple phenolics, flavonoids, coumarins, quinones), terpenoids (e.g., monoterpenes, sesquiterpenes, diterpenes, triterpenes and steroids), and N-containing allelochemicals (e.g., alkaloids, benzoxazinoids) (Macias et al. 2019). The identification of signaling molecules is much more difficult than that of allelochemicals. Recent studies show that jasmonic acid, (-)-loliolide, ethylene, allantoin, canonical strigolactones and some noncanonical strigolactones are root-secreted signaling chemicals in root-root interactions (Lopez-Raez et al. 2017; Wang et al. 2021c). Allelochemicals have a negative effect on the receiving plants. For example, Kong et al. (2018) showed that when wheat is grown with 100 other plant species, wheat can detect neighbors and respond by increasing allelochemical release in a density-dependent manner. They further identified that (-)-loliolide and jasmonic acid are signaling molecules triggering allelochemical production in cereals (Kong et al. 2018; Li et al. 2020c).

A shift of the soil microbiome may mediate the effects of allelochemicals on target species (Fig. 2). A previous review indicates that microbiomes may enhance or alleviate allelopathic effects on the receiving plant species (Cipollini et al. 2012). For example, microbes may enhance the distribution of allelochemicals via common mycorrhizal networks which intensifies the negative effects; microorganisms may also degrade allelochemicals and enhance plant tolerance to allelochemicals in diverse plant communities (Cipollini et al. 2012). Another case study shows that cover crops may suppress weeds through allelochemicals, mainly at the early stage. The effects only last a limited time because soil microbes exhibit an antagonistic interaction with allelochemicals. Instead, microbiomes may suppress weeds in the later phase in agroecosystems (Luo et al. 2016a).

Plant species detect and recognize neighboring species and respond intraspecifically between kin and non-kin individuals via root exudates (Biedrzycki and Bais 2010; Biedrzycki et al. 2010; Dudley and File 2007; Murphy et al. 2017a). The first example of kin recognition was found in Cakile edentula, which exhibited different performances when grown with the same population than when growing with unrelated plants. When grown with a non-relative neighbor, C. edentula produced more root biomass and directed its root towards the stranger. However, the species tends to avoid belowground competition in the presence of a kin (Dudley and File 2007). Similar kin recognition patterns are also found in rice (Oryza sativa L.) (Yang et al. 2018) and soybean (Murphy et al. 2017b).

Yang et al. (2018) showed that an N-rich allantoin compound secreted by roots is important in kin recognition in rice. Xu et al. (2021) further grew allelopathic rice cultivars with kin pairs and non-kin pairs; they found that roots of target rice individuals grow more towards neighboring weeds than towards their relatives. The altered root replacement enhances interspecific competition with weeds and decreases the production of allelochemicals to inhibit weeds. This indicates that a desirable combination of relatedness allows plant cultivars to inhibit weeds more economically without compromising yield. Another study found that Pseudotsuga menziesii shares more carbon (C) to neighbor individuals through ectomycorrhizal networks in kin pairs than in non-kin pairs which may be mediated by signaling molecules (Pickles et al. 2017).

Therefore, when selecting crop combinations with a strong competitor, a sufficient recovery stage and great root trait plasticity are needed for the less competitive crop to overcompensate yield after the competitor is harvested (i.e. temporal niche differentiation). We also need to consider secondary metabolites-mediated contest competition when establishing intercropping systems. Notably, allelochemicals released by the target crop species is a double-edged sword, which may suppress the growth of neighboring crop species, while the desirable cultivar mixtures also inhibit the growth of weeds without comprising yield.

Fig. 2
figure 2

Mechanisms that drive interspecific competition and facilitation in intercropping. Interspecific competition may be driven by resource depletion of competition-dominated species via root traits (scramble competition). The phenotypic plasticity of root traits also plays an important role for competitive-dominated species to get advantages at co-growth stages and for subordinate species to recover after the harvest of dominant species. It is also mediated by the effects of secondary metabolites (e.g., allelochemicals) on the soil microbiome. Belowground facilitation might result from resource enrichment (e.g., N, P, micronutrients, and water). The root traits (e.g., root morphological traits and root exudate signaling molecules) that match facilitating and facilitated species may play important roles. The enrichment of beneficial microbiome, which attenuates negative plant-soil feedback, also contributes to interspecific facilitation involving disease, pathogen or pest protection

Full size image

Facilitation

It is well-documented that enhanced performance in diverse plant communities is mainly derived from complementarity, which is usually considered as resource complementarity, niche partitioning and facilitation (Barry et al. 2019; Loreau and Hector 2001). A large body of case studies (e.g., Cardinale et al. 2007; Ceulemans et al. 2017; Giles et al. 2017; Tang et al. 2019; Zhang et al. 2014), theoretical research (e.g., Wang et al. 2021d), and reviews (e.g., Barry et al. 2019; Homulle et al. 2021; Li et al. 2020a, b) have illustrated the importance of resource complementarity and niche partitioning in the positive biodiversity-ecosystem functioning (BEF) relationship in natural systems and agroecosystems. For example, in a maize/faba bean intercropping system, the maize grows better than of maize grown next to wheat (Fig. 1a). We showed that the shallow roots of faba bean have a limited lateral root distribution and allow intercropped maize roots to extend under faba bean plants and explore a larger soil volume for nutrients (i.e. spatial complementarity, Fig. 1b-c, Li et al. 2006).

Facilitation occurs when the presence of one or more species increases the performance of neighboring species (Callaway 2007). In a facilitation-dominated crop combination, the first-harvested crop species (e.g., faba bean) obtains a yield advantage, and the neighboring crop species (e.g., maize) may have the same yield as the monocrop, or overyield in the co-growth stage, i.e. a win-win situation. Facilitative mechanisms that explain species-specific overyielding in diverse plant communities include direct abiotic facilitation (e.g., aboveground microclimate amelioration and belowground resource enrichment), and indirect biotic facilitation (e.g., microbe-mediated nutrient mobilization, pathogen and pest protection, and biotic feedback) (Fig. 2, Brooker et al. 2021; Li et al. 2014; Wright et al. 2017b; Yu et al. 2021).

Interspecific N facilitation through N2 fixation by legume is a widespread mechanism in diverse natural ecosystems and agroecosystems (Fig. 2). Legume/cereal intercropping has been widely practiced as a sustainable agricultural system, because legumes can increase ecosystem N input by symbiotic N2 fixation. A recent study showed that yield advantages in maize/soybean and maize/peanut (Arachis hypogaea L.) are related to a positive complementarity effect in the absence of N fertilization. This suggests facilitation of N acquisition by intercropped legumes (Zhang et al. 2021b). Nitrogen can be transferred from legumes to neighbors through root exudates, common mycorrhizal networks, or rhizodeposits (Li et al. 2009a; Thilakarathna et al. 2016; van der Heijden and Horton 2009; Wahbi et al. 2016; Zhang et al. 2020b). Notably, N facilitation acts in a reciprocal manner in intercropping systems, i.e. legumes not only enhance N uptake of neighboring cereals, but specific cereals also promote N2 fixation of legumes by root exudates (Li et al. 2016a).

Intercropping legumes with cereal crops enhances the biological N2 fixation of legumes (Fan et al. 2006; Xiao et al. 2004). This is because of the greater competitive ability of maize for soil mineral N which reduces nitrate accumulation (Li et al. 2005, 2011), and alleviates the inhibitory effect of N on legume nodulation and N2 fixation, enhancing N2 fixation in intercropped legumes (Li et al. 2009b). Wheat and barley exhibit a greater competitive ability for soil N than maize (Zhang et al. 2015b, 2017); however, a stimulatory effect on faba bean is not observed in wheat/faba bean and barley/faba bean intercropping systems (Fan et al. 2006). Thus, we surmise this relationship is species-specific. Combining field studies with glasshouse experiments, we found that root exudates of maize promote faba bean nodulation, whereas exudates of wheat and barley do not. Faba bean treated with root exudates of maize elicits a two-fold increase in exudation of flavonoids, and an increased expression of genes mediating nodulation and auxin (Li et al. 2016a). Genistein, a flavonoid released from maize roots, plays an important role as a signaling molecule in this facilitative root-root interaction.

Intercropping may also increase the abundance of some N2-fixing microbes, e.g., Rhizobium hainanense, Rhizobium leguminosarum, and Frankia (Chen et al. 2018). Wheat exudates probably enhance nodulation of neighboring faba bean by specific flavonoids such as flavonol, isoflavone, chalcone and hesperetin (Liu et al. 2017). Using 13 C-DNA-based stable isotope probing, Hu et al. (2021) further identified nine key bacterial genera (i.e. Agromyces, Arthrobacter, Bacillus, Gemmatimonas, Heliobacillus, Lysobacter, Natronocella, Paenibacillus and Sorangium) in the rhizosphere of the faba bean that are highly related to N2 fixation. These key bacteria may be affected by root exudates and contribute more than 50% of N2 fixation in intercropped faba bean (Hu et al. 2021). Therefore, there are three pathways for efficient use of N in cereal/legume intercropping via facilitation. One is facilitative effects on non-legume species via N transfer by legumes; the second is enhanced symbiotic N2 fixation of legumes which results from soil mineral N dilution due to competition for soil mineral N by cereals; the third is enhanced nodulation of legumes, driven by signaling molecules in root exudates from neighboring species and a shift of the soil microbiome.

Facilitation of P and micronutrient acquisition via root-root interactions and microbial processes are widely-accepted mechanisms driving overyielding in intercropping (Fig. 2, Dai et al. 2019; Dissanayaka et al. 2015; Khashi u Rahman et al. 2021; Li et al. 2014, 2016b; Tang et al. 2014, 2016; Xue et al. 2016; Zhang et al. 2016, 2020a). Efficient P-mobilizing legumes may mobilize either inorganic P, e.g., faba bean, or organic P, e.g., chickpea, by releasing large amounts of protons and carboxylates, or acid phosphatases, respectively (Li et al. 2003a, 2004, 2007, 2014; Zhou et al. 2009). The increased soil P availability facilitates the growth of neighboring species (Li et al. 2014).

Our field studies showed that grain yield and P uptake of intercropped faba bean and maize are greater than those of the corresponding crop species in monoculture (Li et al. 2003b; Zhang and Li 2003). In a meta-analysis, Tang et al. (2021b) showed that 90% of intercropping combinations exhibit P-uptake advantages. Interspecific P facilitation and P-use efficiency are greater when fertilizer input is reduced and may be associated with more N2 fixation of legumes (Tian et al. 2020; Xia et al. 2013a). We showed that the enhanced soil P availability by root exudates from legumes facilitates neighboring inefficient P-mobilizing species with root interactions in intercropping systems under supplies of both organic P and sparingly-soluble inorganic P (Li et al. 2003a, 2004, 2007, 2014). This facilitative interaction may not only save P fertilizer (Li et al. 2007), but also enhance apparent recovery of fertilizer P in intercropping in both infertile (Mei et al. 2012) as well as fertile soils (Xia et al. 2013a).

The strength of P facilitation also requires greater plasticity of root traits that allow the facilitated species to interact with the facilitator (Fig. 2, Yu et al. 2020a). Using different barrier types, we showed that species exhibiting a better match of root traits to a facilitator can obtain more benefits in a community (Yu et al. 2020a). In response to the facilitator, root trait plasticity is also affected by soil nutrient availability and neighbor identity (Abakumova et al. 2016; Schneider and Lynch 2020; Zhang et al. 2016, 2020a). In a tomato (Solanum lycopersicum L.) / onion (Allium cepa L.) intercropping system, root exudates of intercropped tomato shifted bacterial composition under different P-application sources. The increased abundance of P-solubilizing microbes is positively associated with plant-available soil P concentration, suggesting an interaction between root exudates and soil microbial composition to increase soil P availability in intercropping systems (Khashi u Rahman et al. 2021). Using a 41 K-labeling method, asymmetric interspecific facilitation via K exchange through common mycorrhizal networks in tomato/onion intercropping was shown (Gao et al. 2021). Based on our previous studies and recent advances in belowground root-root interactions, we proposed that belowground facilitation requires trait matching in natural ecosystems as well as agroecosystems (Yu et al. 2021).

A large body of research has indicated that maize enhanced iron (Fe) and zinc (Zn) uptake of legumes in intercropping (Zuo et al. 2000, 2003, 2004; Zuo and Zhang 2008, 2009). Field and glasshouse experiments showed that the phytosiderophores secreted by maize facilitate Fe acquisition of peanut in maize/peanut intercropping (Jiao et al. 2021b; Xiong et al. 2013). The dynamics in rhizosphere effects and molecular mechanisms in maize/peanut intercropping was summarized in a previous review (Dai et al. 2019). The shifts of microbial community composition and functioning may also directly or indirectly contribute to micronutrient facilitation (Dai et al. 2019), while belowground processes need further work to help better use intercropping to secure micronutrient supply in agroecosystems.

Beyond nutrient-uptake facilitation, diverse agroecosystems also promote water acquisition, protect plants against hazardous substances (e.g., metals and allelochemicals), diseases and pests via root-root interactions (direct abiotic facilitation) and plant-soil biota interactions (indirect biotic facilitation) (Fig. 2). Intercropping enhances water-use efficiency and saves water with mulching in arid regions (Mao et al. 2012). Hydraulic redistribution by roots and common mycorrhizal networks increases ecosystem functioning in agroforestry and intercropping systems (Dawson 1993; Pang et al. 2013; Singh et al. 2019), as has been covered in recent reviews (Bayala and Prieto 2019; Homulle et al. 2021). Specific species mixtures might also mitigate autotoxic effects and enhance nutrient availability through chemical facilitation and a shift of microbial composition (Xia et al. 2016, 2019). Intercropping upland rice with Solanum nigrum, a cadmium (Cd)-hyperaccumulating species, may mitigate Cd accumulation in rice. This is associated with shifts in rhizosphere processes in rice and S. nigrum, together with interactions between root exudates and AMF that block Cd uptake of rice and enhance Cd acquisition of S. nigrum (Yang et al. 2021b).

Cultivar mixtures and intercropping also reduce diseases and pests, increasing productivity via asymmetric facilitation of target species or cultivars and reducing the need for pesticides. The role of root exudates and microbiome in disease and pest inhibition in diverse plant communities includes negative density dependence, direct inhibition by root exudates, and indirect biotic facilitation by soil microbiome (Brooker et al. 2021; Mwakilili et al. 2021; Stomph et al. 2020; Zhang et al. 2019a, b; Zhu et al. 2000; Homulle et al. 2021; van Ruijven et al. 2020). Beyond the above mechanisms, root exudates of wheat may also induce positive microbial legacy effects, which suppress pathogens and increase microbes that are beneficial to intercropped faba bean in the long run (Wang et al. 2021a). Interactions among trophic levels may also contributes to plant health. For example, co-inoculation of AMF and earthworms may interactively enhance yield via increased nutrient availability (Zarea et al. 2009). Earthworms also inhibit infection by nematodes and increase plant resistance to herbivores by modulating plant chemistry and soil properties (Xiao et al. 2019). However, in a tomato/leek (Allium porrum L.) intercrop, the authors found that inoculation of AMF increases root-knot nematode infection, and earthworm and AMF co-inoculation does not facilitate plant growth (Detrey et al. 2022). Another study also showed that facilitative pairwise microbial interaction promotes pathogen invasion, while antagonistic interaction suppresses pathogen invasion in the tomato rhizosphere (Li et al. 2019). Therefore, specific microbiome and interkingdom combinations are needed to protect target species from pathogens and herbivore attacks in diverse agroecosystems.

This section reviews the mechanisms underlying belowground direct abiotic facilitation (e.g., resource enrichment) and indirect biotic facilitation (e.g., disease inhibition) in intercropping systems. Belowground facilitation may function in a universal manner through the amelioration of soil environments or in a species-specific fashion via the trait matching between facilitator and facilitated species (e.g., trait plasticity and root exudate signaling, Yu et al. 2021). Therefore, specific belowground processes between crop species should be considered to strengthen interspecific facilitation in future intercropping design.

Belowground processes and soil fertility in agroecosystems with intercropping

As reviewed above, intercropping enhances grain yield and nutrient uptake via belowground processes, while it also means multi-cropping removes more nutrients from the soil than monocultures. Therefore, it is crucial to understand whether soil fertility would increase, stay the same or decrease in the long run. Therefore, we established four long-term field experiments with substantial soil fertility gradients in Gansu province and Ningxia Autonomous Region, Northwest China, to examine the effects of intercropping on soil fertility and sustainability. Two experiments, conducted in fertile soils in Baiyun village and Wuwei City, Gansu province, were started in 2003 and 2009, respectively; the third experiment, on moderately fertile soils in Jingtan village, Baiyin City, Gansu province, was started in 2009; and the fourth experiment, in newly-reclaimed desert soils in Xingsheng village, Hongsibu County, Wuzhong City, Ningxia Autonomous Region, was also started in 2009 (Fig. 3). Here, we review the advances of the effects of diverse cropping systems on soil physical, chemical, and biological properties.

Fig. 3
figure 3

Aerial photos of four long-term field intercropping experiments. These long-term field experiments are located at Baiyun village (upper left and right), at Jintan village (left bottom), and at Hongsibu (bottom right). Photo credit: Long Li

Full size image

Soil physical properties are of fundamental importance in ensuring soil functions and ecosystem services (Rabot et al. 2018). As a key indicator of soil physical properties, soil aggregates significantly determine soil functions, such as protecting soil organic matter from microbial decomposition, affecting soil water and oxygen movement, as well as nutrient cycling and conditions for crop growth (Six et al. 2004; Tisdall and Oades 1982). Soil aggregation is affected by many factors, such as root traits, soil microorganisms and fauna, inorganic binding agents, and environmental variables (Six et al. 2004). Our long-term field experiments show that intercropping systems exhibit a greater macro-aggregate (> 2 mm) content than monocultures (Li et al. 2021b; Tian et al. 2019), which is in line with other experiments (Garland et al. 2017; Nyawade et al. 2019). This change could be associated with the changes of root traits, microorganisms, and soil environments in crop diversification (Fig. 4). First, root traits play a fundamental role in soil aggregation, including root biomass, root architecture, and root physiology. More root production in intercropping provides more binding agents for aggregate formation and stabilization (Chen et al. 2019b; Nyawade et al. 2019). More root exudates associated with root biomass act as binding agents of soil particles that contribute to soil aggregation (Peres et al. 2013). Species with finer roots enhance soil aggregation through entangling particles by roots and root hairs, e.g., Lolium perenne (grass), while species with thicker roots have a negative effect on soil aggregation, e.g., Trifolium repens (legume) (Gould et al. 2016). Second, changed soil microbiome and fauna in diverse systems affect aggregate formation. Tian et al. (2019) showed that intercropping enhances soil macro-aggregates by decreasing the relative abundance of soil Nitrospirae in poor-fertility soil, by increasing AMF biomass in medium-fertility soil, or by increasing the relative abundance of soil Sordariales in high-fertility soil. Fungal mycelium acts as a “sticky string bag” that entangles soil particles and cements particles together by producing extracellular polysaccharides. Mucilage exuded by bacteria and fungi also improves soil aggregation (Six et al. 2004). In contrast, increased earthworm biomass in diverse systems may destabilize soil aggregates due to increased mechanical breakdown (Peres et al. 2013). Finally, soil environmental variables induced by diversification, such as increased soil organic matter content and a decreased soil C: N ratio, indirectly affect aggregate formation. Soil organic matter is a crucial binding agent of soil aggregates, and a lower C: N ratio increases the formation of soil aggregates through enhanced microbial activity (Gamboa et al. 2020; Peres et al. 2013).

Fig. 4
figure 4

Belowground processes and mechanisms that may change soil fertility in intercropping. The greater root biomass likely drives belowground processes by increasing carbon (C) input in multi-species and multi-cultivar systems. The diverse root exudates may, directly and indirectly, affect soil C sequestration by specific soil microbiomes, macro-aggregate formation and physical protection, quality of root residual and decomposition rates. The biological N2 fixation by legumes also directly enhances nitrogen (N) sequestration or affects soil N storage through a shift of the soil microbiome. The changes of C and N sequestration via belowground processes in intercropping then may affect greenhouse gas emissions. However, the effects of root exudates on the soil microbiome, enzyme activity, deep soil C and N processes, or microbial effect on soil C and N sequestration, and the linkage between soil C and N storage and greenhouse gas emission needs further research (“?” and dashed lines)

Full size image

Soil chemical properties, especially soil organic carbon (SOC) and total soil N, are essential for soil fertility and sustainability (Fig. 4, Lal 2004). Species diversity enhancing SOC and total soil N have been shown in grasslands (Lange et al. 2015; Yang et al. 2019), forests (Diaz et al. 2009), and agroecosystems (Cong et al. 2015b). For example, in a long-term intercropping experiment, Cong et al. (2015b) found that SOC content in the topsoil was 4% greater in intercropping systems than in monocultures. In general, SOC storage represents the net outcome of C inputs and losses (Houghton 2007). In agroecosystems, root inputs are the major source of SOC, thus increased root biomass in intercropping systems compared with monocultures provides a possible mechanism for increased SOC content (Cong et al. 2015b). However, some studies also show that intercropping does not affect SOC (Wang et al. 2014b, 2015, 2020a). This may contribute to greater microbial activities, increased litter decomposition of mixed roots in intercropping and negatively affect C storage (Cong et al. 2015a). A recently published study also suggests that changes in SOC by intercropping depend on soil characteristics (Li et al. 2021b).

In addition to topsoil, plant diversity also affects SOC stocks in deeper soil which probably contributes to a deeper root distribution in diversified systems compared with monocultures (Chen et al. 2020). Soil organic carbon in deeper soils is more stable and difficult to access for microbes. Thus, root inputs in deeper soil are the major source of SOC. In addition, by using advanced equipment, the root-root interaction research is going deeper (Thorup-Kristensen et al. 2020). However, the effects of root- and microbe-driven changes on deep soil C and N processes remain largely unknown (Fig. 4, Lal 2018). Sequestration of SOC is a slow process; more long-term studies are needed to accurately estimate the relationships between crop diversity and SOC.

Greater species diversity increases soil N content (Cong et al. 2014; Fornara and Tilman 2008). In a long-term intercropping experiment, Cong et al. (2015b) observed that soil organic N content in the topsoil was 11% greater in intercropping systems than in monocultures, and lower soil δ15N values in intercropping systems with legumes indicated that increased biological N2 fixation probably contributed to the increased soil N content (Fig. 4). However, no studies have quantified the relative contribution of biological N2 fixation in changes in soil N sequestration in intercropping. In the intercropping system without legumes, mechanisms of N retention, e.g., complementary N-acquisition strategies of intercropped crops and reduced N leaching, might explain the increased total soil N content (Cong et al. 2015b). Specific species and cultivar may recruit distinct microbiomes via root exudates in multi-species and multi-cultivar combinations, shifting soil microbiome composition and function (Sasse et al. 2018). However, the effects of root exudates on soil microbiome and their linkage with soil C and N sequestration remain largely unknown in crop diversification (Fig. 4).

Soil C and N cycling are also tightly linked with greenhouse gas emissions. Less C emissions and a lower C footprint may result in more C sequestration in intercropping systems than in monocultures (Fig. 4, Chai et al. 2021; Hu et al. 2015, 2016; Wang et al. 2020a). Thus, we surmise intercropping has the potential to sequester more SOC and reduce CO2 emission to the atmosphere. In a four-year investigation, sugarcane (Saccharum officinarum L.)/soybean intercropping had an insignificant impact on N2O emissions, while reduced N fertilizer exhibited a greater effect on mitigating N2O emissions than intercropping (Luo et al. 2016b). Intercropping could save N fertilizer, and the reduced N fertilizer input by intercropping might enhance N-use efficiency and N storage, thus decreasing N2O emission and mitigating climate change. Thus, further research is necessary to determine the linkage between soil C and N sequestration and greenhouse gas emission in crop diversification systems (Fig. 4).

Biological properties of soil fertility in this review include soil enzyme activity, microbial biomass, composition, and function. Soil microbes increase soil fertility by enhancing nutrient availability, aggregation, and formation of SOC (Fig. 4, Bardgett and van der Putten 2014). Intercropping can shift soil microbial communities (Dai et al. 2013; Song et al. 2007a, b; Zheng et al. 2018). For example, sugarcane/peanut and sugarcane/soybean intercropping increases soil C-, N-, and P-related enzyme activities, soil bacterial and fungal biomass and unique microbe numbers compared with sole cropping, increasing nutrient uptake and soil fertility (Li et al. 2013; Tang et al. 2021a). A meta-analysis that focused on the effects of plant diversity on microbial biomass also showed that species mixtures enhance soil microbial biomass, fungal biomass, bacterial biomass, and the fungi: bacteria ratio (Chen et al. 2019a). This is because intercropping changes the type and amount of root exudates that can target control over soil microbial communities (Hartmann et al. 2009). Extracellular enzymes produced by soil microbes can decompose organic residues to release nutrients for plant production and microbial growth (Curtright and Tiemann 2021). A recent meta-analysis shows that soil enzyme activities related to C, N, and P cycling are significantly increased, and the intercropping effects are determined by plant types and environmental factors (Curtright and Tiemann 2021). Increased soil enzyme activities accounted for microbial biomass and microbial activity mediated by higher quality and quantity of plant residue inputs in intercropping (Curtright and Tiemann 2021). Although soil microbes produce extracellular enzymes, the links between the abundance of functional genes and extracellular enzyme production are often difficult to establish (Fig. 4, Burns et al. 2013). Future studies should focus on the effects of intercropping on microbial community composition and function, especially on the relationship of soil extracellular enzymes and enzyme-coding genes.

Belowground processes and sustainability in diverse agroecosystems under climate change contexts

Climate change threatens food security globally (Hasegawa et al. 2018; Ray et al. 2015; Rosenzweig and Parry 1994). Species diversity can stabilize productivity in both natural and agroecosystems via complementarity effects, while the selection effect may impair the stability (Gross et al. 2014; Isbell et al. 2009; Raseduzzaman and Jensen 2017; Renard and Tilman 2019; Río et al. 2017; Tilman and Downing 1994; Wang et al. 2021d). Case studies and synthesis studies show that temporal stability is regulated by complementarity effects (e.g., species asynchrony), compensatory effects (e.g., insurance hypothesis), the traits and stability of dominant species, and phylogenetic diversity, rather than diversity itself (Fig. 5, Bai et al. 2004; Craven et al. 2018; Ma et al. 2017; Sasaki and Lauenroth 2011).

Compared with the generally positive effects of species diversity on temporal stability, diversity-spatial stability relationships depend on functional trait diversity and small-scale heterogeneity (Weigelt et al. 2008). In this research, rooting depth and clonal growth form are important to determine spatial stability (Weigelt et al. 2008). However, another study indicates that diversity increases spatial stability across ecosystems at the global scale, with the greatest positive effects in the most productive communities (Wang et al. 2019).

The temporal and spatial stability in diverse agroecosystems has received little attention compared with that of natural ecosystems. A meta-analysis showed that intercropping enhances yield temporal stability compared with the corresponding monocultures, and cereal/legume intercropping exhibits greater yield stability than a combination without legumes (Raseduzzaman and Jensen 2017). Another meta-analysis showed that cultivar mixtures generally exhibit greater temporal stability than monocultures (Reiss and Drinkwater 2018). Our results also support the notion that diverse intercropping systems increase yield stability over a decade. This is probably related to the increased yield advantage and soil fertility via belowground interaction (Li et al. 2021b).

Compared with the positive BEF relationship in ambient environments, the relationship is usually greater under future climate conditions, indicating communities with higher diversity are more resistant and resilience to climate change, probably mediated by complementarity and facilitation (Hong et al. 2021; Isbell et al. 2015; Jourdan et al. 2019; Wright et al. 2017a, 2021). In the Jena Experiment, plant species are less negatively affected by flooding in more diverse plant communities. The positive BEF relationship is related to the enhanced soil aeration by species with more root aerenchyma, which facilitates the performance of flood-sensitive plant species (Wright et al. 2017a). A similar result was also found in positive BEF relationships in plant communities that experience drought stress (Wright et al. 2021).

Fig. 5
figure 5

Linking belowground processes and sustainability in intercropping. Climatic factors drive the shifts in soil properties and intra- and interspecific interactions affecting yield stability. The shifts of soil properties induced by climate change may also affect root traits of a given species, and the specific species combination then affects root-root interactions. The shift of root-root interaction may also help the plant community acclimate to soil environments (e.g., under flooding or drought), affecting sustainability in agroecosystems. Root-root interactions under climate change also affect yield stability through competition, complementarity or facilitation. “?” and dashed lines indicate that the processes are still unclear and in need of further study

Full size image

Acclimation in response to climate change in diverse plant communities is associated with the plasticity of root traits and belowground processes in response to climate change (Fig. 5). For example, in a maize/soybean intercropping with plastic film mulching, intercropped maize modified root growth with a confined strategy in drought environments and inhibited roots with excessive growth in wet environments. The greater intercropping benefits under varied soil water availability are related to the plasticity of root traits which may stabilize yield advantages in response to drought (Zhang et al. 2021a).

A warming treatment increased fine root biomass and N concentration and decreased root C: N ratio and nonstructural carbohydrate concentrations (Wang et al. 2021b). The effect of warming on plant growth also depends on specific traits of plant species. For example, tolerant C4 grasses tend to be less limited under warming than C3 grasses (Ward et al. 1999), and legumes may increase N2 fixation at higher temperatures (Gundale et al. 2012; Whittington et al. 2012). Elevated CO2 promotes root morphological development and root system expansion (Nie et al. 2013), while the impact of competitive species on co-occurring less-aggressive species becomes more severe under elevated CO2, because of increased interspecific root competition (Arnone and Kestenholz 1997). Therefore, we surmise elevated CO2 may intensify belowground interspecific competition of maize in intercropping systems (Fig. 5). Nitrogen deposition may significantly increase the proportion of coarse roots, root turnover and root respiration, while decreasing fine root C: N ratio and AMF colonization (Li et al. 2015). However, how these climate change factors separately and interactively affect root-root interaction, and how root-root interaction mitigate future climatic conditions via acclimation and amelioration of soil properties in diverse agroecosystems remains largely unknown (Fig. 5, Calleja-Cabrera et al. 2020).

Perspectives and conclusions

The ancient agronomic approach of intercropping in China is receiving increasing interest from plant scientists, ecologists, and agronomists worldwide. Although scientists have made considerable advances in diverse agroecosystems, we raise several perspectives in this thriving field.

First, new methods and proxies are needed to better understand root-root interactions in intercropping and cultivar mixtures. Multi-species genotyping by sequencing and Fourier-transform infrared spectroscopy are promising tools to quantify the relative proportion of root biomass in multi-species and multi-genotype communities (Streit et al. 2019; Wagemaker et al. 2021). Coupling zymography, planar optodes and different “layers” can visually and quantitatively unravel belowground processes for nutrient acquisition in sole cropping (Ma et al. 2019b, 2021). The plasticity of root traits has also not been linked with ecosystem functioning in diverse plant communities due to the difficulty of distinguishing roots among species. Therefore, machine learning and convolutional neural network are urgently needed to separate roots of different plant species (Shen et al. 2020). Another promising alternative approach to study the highly variable root exudates is leaf [Mn], which has repeatedly been shown to be an easily-measurable proxy for P-acquisition strategies involving the release of carboxylates or other P-mobilizing compounds. Leaf [Mn] can be used to assess carboxylate release in both native plants and crop species (Lambers 2022; Lambers et al. 2015, 2021; Pang et al. 2018; Yu et al. 2020a, b).

Second, diverse plant communities maintain multiple ecosystem functions and services simultaneously, i.e. multifunctionality, such as food production, C and N storage, and pollination. Recent studies have shown that most crop diversification systems enhance multiple ecosystem services and functions, while the components may involve trade-offs and offset positive effects via diversity (Beillouin et al. 2021; Tamburini et al. 2020; Wittwer et al. 2021). In addition, the evaluation of multifunctionality usually includes averaging approaches and multiple threshold approaches, while agroecosystems usually include limited species, and the weight of different components may not be equal in agroecosystems in different areas (Manning et al. 2018). Therefore, the approach to weaken trade-offs in multifunctionality and the evaluation methods are vital to understand the net effects of intercropping on agroecosystem functions.

Third, we suggest combining cultivar diversity and species diversity is a promising approach to enhance intraspecific complementarity and interspecific facilitation and reduce intraspecific competition (Cong et al. 2020; Yu et al. 2021). When integrating cultivar diversity and species diversity, multiple facets of diversity may affect multifunctionality via functional diversity, trait identity, and trait metrics (e.g., community-weighted trait, variance, skewness and kurtosis) (dos Santos et al. 2021; Mahaut et al. 2019; Zhang et al. 2019a, b). A better understanding of above- and belowground trait-related processes may help us link complementarity and enhanced ecosystem functions via efficient use of resources. As we reviewed, selecting cultivar mixtures with the aim of exudate-mediated kin recognition may efficiently and economically inhibit weeds. However, combining cultivars that vary in nutrient use and root depth also enhances nutrient uptake and grain yield, but this may require genotypes to be distantly related (Cong et al. 2020). In addition, a mixture of glutinous rice and rice hybrids that are resistant to rice blast (which are distantly related) may reduce disease by 94% compared with monocultures (Zhu et al. 2000). Therefore, enhancing weed or disease inhibition (i.e. reducing the need for herbicides and pesticides which enhances environmental profits) without compromising yield needs further work in cultivar mixtures.

In the longer term, we need to explore the effects of soil microbial legacies on ecosystem functioning (e.g., productivity and disease control) to manipulate species assembly order in rotational intercropping systems or mixtures (Delory et al. 2019, 2021; Wang et al. 2021a). This soil microbiome-mediated yield advantage in intercropping through inhibition of the negative plant-soil feedbacks of pathogens requires more attention, especially in communities with high chemical inputs (Brooker et al. 2021; Wang et al. 2021a). Therefore, long-term studies are needed to accurately estimate the effects of plant-soil feedback on positive BEF relationships.

The effects of structured soil on root growth and root-root interaction are often overlooked. Structured soil offers a mechanical resistance or physical anchorage for root growth and induces temporal and spatial heterogeneity in water and nutrient distribution and availability (Jin et al. 2017; Wang et al. 2020b). Heterogeneity in structured soils could alter root architectural, morphological (e.g., root hairs) and physiological traits (e.g., root exudate composition); it may also shift root-associated microbial communities and functions, which may further shift the effects of root-soil and root-root interactions (Wang et al. 2020b). Root growth and root-root interaction response to structured soils (e.g., soil mechanical impedance) may also depend on nutrient availability (Wang et al. 2021e). Future belowground process research in crop diversification should aim to understand root-root interactions in an integrated manner in a real soil environment.

Finally, it is urgent to explore the effects of root mixtures on soil C and N sequestration and environmental costs to mitigate climate change (Fig. 4, Lal 2004). Recent reviews indicate that the root effect on SOC is a double-edged sword, which depends on the outcome of rhizodeposition, root turnover and nutrient-acquisition strategies (Dijkstra et al. 2021; Ding et al. 2021). For example, root inputs increase new C production and facilitate organo-mineral complexation and aggregate formation, enhancing SOC stabilization. In addition, root exudates also have opposite effects on the above processes which causes SOC destabilization (Dijkstra et al. 2021). The effects also depend on P-acquisition and -use strategies and soil heterogeneity (Ding et al. 2021; Wang et al. 2020b). For example, carboxylate release may act as a most important factor affecting soil N mineralization via SOC decomposition, which may enhance loss of C to the atmosphere. Arbuscular mycorrhizal fungi, however, may stabilize organic-mineral associations and aggregates and facilitate C sequestration (Ding et al. 2021). When growing crop species with different nutrient-use strategies, soil C and N sequestration may increase, not change, or even decrease. Therefore, we need to consider the effects of root-root interactions on nutrient mobilization, soil fertility and mitigation of climate change when selecting crop combinations.

In this review, we summarized widely-accepted methods to study belowground interactions in intercropping systems which may be further developed combined with new techniques to uncover the hidden half of root-root interactions in both agroecosystems and natural ecosystems. Then, we summarized the belowground competition-recovery theory, root-root interactions driven by allelochemicals, direct facilitation by resource enrichment, and indirect facilitation by disease suppression via the microbiome, which shed light on the promising approach to establish desirable crop combinations in future intercropping systems. We strongly recommend the elegance of and need for integrating multiple complementary methods and experimental designs to understand processes across scales, including long-term experiments and spatial scales, to illuminate root-root interactions. To better understand the potential advantages of intercropping in response to future climate change, we also reviewed the advances in soil fertility changes with crop diversification, especially in the long run. How intercropping mitigates climate changes via root-root interactions and shifts in soil microbiomes may be a hotspot under a changing climate. We also need to pay more attention to the potential publication bias of positive intercropping effects on ecosystem functions. Gray literature that reports neutral or negative effects needs to be analyzed in future meta-analyses, likely with the help of new tools (e.g., CorTexT Platform).