Precipitation-induced soil properties and plant communities mediate root strategies in an alpine meadow

Abstract

Plant roots show flexible traits to changing precipitation, but the factors driving root trait covariation remain poorly understood. This study investigated six key root traits and explored the potential driving factors, including plant community characteristics and soil properties, in the Zoige alpine meadow across five precipitation gradients: natural precipitation (1.0P), a 50% increasing precipitation (1.5P), and 30%, 50% and 90% decreasing precipitation (0.7P, 0.5P and 0.1P, respectively). Our results demonstrated distinct root trait responses to changes in precipitation. Both increasing (1.5P) and decreasing precipitation (0.1P, 0.5P and 0.7P) inhibited root diameter (RD), specific root length (SRL) and specific root area compared with 1.0P. Conversely, root tissue density and root nitrogen content increased under decreasing precipitation but declined under 1.5P. With increasing precipitation, root foraging strategies shifted with thinner RD and larger SRL to that with a larger diameter. Shifts in root strategies were primarily influenced by soil properties, specifically soil water content and available nitrogen. Additionally, root strategies in surface soils (0–10 cm) were mainly related to the grass and sedge coverage, whereas in deeper soils (10–20 cm) root strategies were related to overall plant community coverage and biomass. Our findings indicate that root trait variations and strategies in alpine meadows are co-driven by soil properties and plant communities in response to changing precipitation.

摘要

降雨量引发的土壤性质和植物群落变化调节高寒草甸根系策略

植物根系对降雨变化表现出较强的可塑性，但驱动根系性状变化的因素尚不完全清楚。本研究以若尔盖高寒草甸植物群落根系为研究对象，沿5个降雨梯度(1.0P：自然降雨量、1.5P：增加50%降雨量、0.7P：减少30%降雨量、0.5P：减少50%降雨量和 0.1P：减少90%降雨量) 对6种关键根系性状及其潜在影响因素(植物群落特征与土壤性质)进行分析。研究结果表明，根系性状对降雨变化有显著的响应。与1.0P相比，无论是增加降雨(1.5P)还是减少降雨(0.1P、0.5P、0.7P)均抑制了根直径、比根长和比根面积。相反，在降雨减少的处理下，根组织密度和根氮含量增加，但在1.5P时下降。随着降雨量的增加，根系策略从较细的根直径和较大的比根长逐渐转变为较粗的根直径和较小的比根长。根系策略的变化主要受土壤性质的影响，尤其是土壤含水量和有效氮含量。此外，表层土壤(0–10 cm)中的根系策略主要与禾草和莎草的覆盖度有关，而在深层土壤(10–20 cm)中，根系策略则与整体植物群落的覆盖度和生物量相关。上述结果表明，高山草甸中植物根系性状的变化和根系策略在应对降雨量变化时受土壤性质和植物群落的共同驱动。

INTRODUCTION

Fine roots with a diameter of less than 2 mm are crucial for the water and nutrient absorption of plants. Various root traits-such as the architectural, morphological, chemical and biotic traits, along with their correlations, reflect how roots acquire resources and represent the root strategies (Bergmann et al. 2020; Kong et al. 2019). By adjusting these root traits and root strategies, plants can enhance their adaptability to future climate changes (McCormack et al. 2017), which in turn impacts plant growth, survival and distribution and drivers of ecosystem processes and functions (Kuzyakov and Blagodatskaya 2015; Wang and Guo 2008).

In line with the leaf economic spectrum (Wright et al. 2004), previous studies have proposed a dimension of root trait variation that ranges from a ‘Fast’ strategy characterized by a high root nitrogen content (RNC) and a fast root turnover rate to a ‘Slow’ strategy characterized by a high root tissue density (RTD) and a long lifespan (Long et al. 2013; McCormack et al. 2012; Reich 2004; Roumet et al. 2016; Zhou et al. 2019). The ‘Fast–Slow’ strategy is a trade-off between root investment and nutrient uptake (Kong et al. 2021), termed the ‘Conservation’ gradient (Bergmann et al. 2020). However, studies have demonstrated that another independent gradient exists (Erktan et al. 2018; Weemstra et al. 2016) called the ‘Collaboration’ gradient, which ranges from ‘Do-it-yourself’ soil exploration through high specific root lengths (SRLs) to outsourcing by investing carbon into the mycorrhizal partners for the return of limiting resources, thus requiring a thicker root diameter (RD) (Bergmann et al. 2020; Ding et al. 2023; Han and Zhu 2021). Thus, the ‘Collaboration’ and ‘Conservation’ gradients, which are independent of each other, form the root economic space (RES) primarily due to the nonlinearity of root traits (Zhang et al. 2023; Zhao et al. 2024). It is also the nonlinear relationships that support plants developing more strategies for climate change adaptation (Kramer-Walter et al. 2016).

In a natural ecosystem, plants usually exhibit various root strategies to acquire resources. For example, root trait data from 1810 plant taxa have confirmed that most variation is explained by the ‘Collaboration’ gradient (Bergmann et al. 2020). However, these root strategies would shift with environmental changes or biotic factors (Pierick et al. 2023; Spitzer et al. 2023). In a Picea conifer forest, with increasing leaf nutrient resorption efficiency, plants shift from depending on the ‘Outsourcing’ strategy to the ‘Do-it-yourself’ with thinner roots. This shift in RES has been correlated with temperature and precipitation (Ding et al. 2023). Additionally, soil physicochemical properties might have a profound effect on plant root traits and strategies (Bruelheide et al. 2018; Leuschner et al. 2004). Soil aggregate stability is positively correlated with RD and root mass density but negatively correlated with the SRL (Garcia et al. 2019). Therefore, the root strategies are not constant, and plants can adjust their root strategies to the changing environments. Previous studies have shown that precipitation changes would result in root trait variations (Bai et al. 2010; Li et al. 2013; Post and Knapp 2020; Zheng et al. 2021). For instance, increasing precipitation can increase the RNC (Freschet et al. 2017) while decreasing precipitation can inhibit the SRL of Artemisia frigida in temperate steppe (Zhou et al. 2019). The response of plant roots to precipitation gradients is not linear (Chen et al. 2024), especially the root biomass (Bond-Lamberty and Thomson 2010; Li et al. 2019a).

Furthermore, the effects of precipitation changes on root traits are interspecific; thus, the plant community structure is a key driver of root strategies (Li et al. 2019b) . In an alpine meadow, decreasing precipitation promotes the deeper growth of roots of Leymus secalinus and also increases the root biomass and diameter, whereas it had no effects on the roots of Stipa purpurea; thus, the community composition shifted in dominance from S. purpurea to L. secalinus (Zheng et al. 2021). In addition, precipitation changes indirectly affect the root traits and root strategies in different soil depths caused by changing soil properties (Ibrahim et al. 2020). Decreasing precipitation reduces the topsoil water content and forces plant roots to grow to deeper soil layers to acquire the water (Padilla et al. 2019), which would further increase the difference between root traits in different soil layers. In forest ecosystems, the roots in deeper soil layers exhibited the ‘Slow’ strategy with lower nitrogen contents and respiration rate (Wang et al. 2016), while the roots in the surface soil exhibited the ‘Fast’ strategy with stronger potential resource acquisition availability under higher nitrogen contents and greater root branching density (Wang et al. 2020, 2023). In conclusion, the changes in plant species and soil properties caused by precipitation changes might result in variations in root traits and thus root strategies. However, relatively few studies have explored the relative contribution of the plant communities and soil properties to the root traits in the RES framework. Such studies can help to recognize and predict the adaption and distribution of plants under future climate changes.

The Qinghai-Tibetan Plateau (QTP) is characterized by high altitude and perennial low temperatures, and the local plants are more sensitive to climate changes. Over the last 30 years, the QTP has been experiencing rapid climate change, and precipitation has increased by 22 mm per decade (Chen et al. 2013b). However, the precipitation patterns are significantly different in the different regions of the QTP. Increasing precipitation has been recorded in the eastern region, while decreasing precipitation was occurred in the western region (Gao and Niu 2018). Soil water availability is the main driver of root trait variation in the alpine meadow (Tang et al. 2022). Previous studies have explored the responses of the plant communities (Qiao et al. 2023) and soil microbial communities (Xiao et al. 2020; Xu et al. 2022) to precipitation changes in alpine meadow of the QTP, with less attention on belowground roots, especially in the context of precipitation changes. Exploring the variability in root traits, strategies and their underlying drivers across different precipitation gradients is crucial for enhancing our understanding of plant adaptation mechanisms.

To address this knowledge gap, we investigated changes in six key root traits in the 0–10 and 10–20 cm soil layers, and explored how the plant communities and soil properties affect the root strategies along the precipitation gradients. Based on the existing RES framework, we hypothesized that with increasing precipitation, plant roots would become thinner and exhibit a higher SRL due to the enhanced availability of resources, leading plants to rely more on the ‘Do-it-yourself’ strategy rather than the ‘Outsourcing’ strategy. Otherwise, the roots in the surface soils would exhibit higher RNC, indicative of the ‘Fast’ strategy, which would be enhanced under increasing precipitation treatments with more available nutrients and exhibit a trade-off under the decreasing precipitation treatments with drought stress. We also hypothesized that the shifts in RES in response to precipitation changes might result from both alterations in soil properties and plant communities, with the effects of soil properties potentially being more significant than those of plant communities.

MATERIALS AND METHODS

Study site

This study was conducted in an alpine meadow on the QTP, China (32.83° N, 102.59° E). The study site has an average altitude of 3500 m above sea level, with a mean annual precipitation of 680 mm and a mean annual air temperature of 1.1 °C. Approximately 80% of the precipitation falls from May to August, which is the growing season of this alpine meadow (Hu et al. 2017). The vegetation cover is higher than 80%, and the plant height is approximately 45–60 cm for the tallest grasses. The dominant species at the study site are the grass Elymus nutans and sedge Kobresia humilis.

Experimental design

We selected a fenced alpine meadow with relatively flat terrain and a slope of less than 5° to conduct the simulated precipitation experiments in 2015. Five precipitation gradients with decreases of 90% (0.1P), 50% (0.5P), 30% (0.7P), natural precipitation (1.0P) and an increase of 50% (1.5P) were set up based on the regional variation in precipitation over the last 41 years (1970–2010) (Yang et al. 2014). Each precipitation treatment comprised six plots in a randomized block design. Each plot was 2 m × 2 m and included at least 2 m between adjacent plots. There were 14, 7 and 5 pieces of the ‘V’ shaped rainwater interceptor with a length of 2.4 m and width of 14 cm that were constructed from highly transparent organic glass with 95% transmittance covering 90%, 50% and 30% of the area above the plot to achieve precipitation of 0.1P, 0.5P and 0.7P, respectively. The remaining rainwater from the 0.5P plots was collected from six buckets, and it was then supplied to the 1.5P plots using a spray bottle after a rain event in 2015. Each plot was surrounded by an aluminum sheet of 45 cm in height and buried to a depth of 40 cm to prevent lateral water movement in the topsoil (Supplementary Fig. S1). We measured the soil water content (SWC, %) and soil compaction (kPa) in 0–10 and 10–20 cm using the TDR 300 and SC 900 (Spectrum Field Scout, USA) once every 7 days since 2018 to quantity the effect of different simulating precipitation treatments.

The experiment achieved the expected effect on soil moisture represented by the gravimetric SWC under different precipitation gradients, and the gravimetric SWC in 0–10 and 10–20 cm soil depth increased with increasing precipitation from 2019 to 2021 (Supplementary Fig. S2 and Table S1), The precipitation treatment effects were detailed in Tang et al. (2022) and Qiao et al. (2023), who conducted experiments at the same site. However, the natural precipitation in 2022 was extremely low; thus, the simulated precipitation effects of different precipitation treatments were not significant expect for 0.1P in 2022. Nonetheless, the root traits and strategies were mainly affected by the consistent and significant precipitation gradient treatments from 2019 to 2021.

Plants, fine roots and soil sampling

In 2022, when the simulated precipitation gradients had lasted for 7 years, all six plots with the same precipitation gradient treatment were selected for harvesting the aboveground plants. First, the plant height and coverage of each plant species were recorded to characterize the plant communities. Then, all aboveground plants were mowed by species level and dried at 65 °C for 48 h to obtain the biomass. The important value (IV) of individual plant species was calculated by the average value of the relative height, relative coverage and relative biomass (Hu et al. 2017). Subsequently, the IV was used to calculate the Shannon–Wiener index, Pielou J index and Simpson index. Meanwhile, we divided the aboveground plants into four functional groups, namely, grasses, sedges, legumes and forbs, and calculated their coverage and biomass to characterize the plant communities.

After removing the aboveground plants, three of the six plots were used to sample fine roots and soils using a soil auger with an inner diameter of 5 cm following the ‘X’ shape at two soil depths (0–10 and 10–20 cm) in each plot. These soil samples and fine root samples were sieved using 2 mm mesh and then separated. The fine root samples (diameter ≤2 mm) were selected to measure the fine root traits and the soil samples were used to measure the soil physicochemical properties.

The fine root samples were promptly washed off the soils for several times using deionized water and then used to distinguish the live and dead roots. The brown and white roots were considered live, while black roots were considered dead (Bai et al. 2010; Majdi and Öhrvik 2004). The live roots were separated in the root disc (40 cm length, 30 cm width and 2 cm height) with the deionized water. Thereafter, the root disc was placed in the root scanner (400 DPI, Epson Expression 10000XL, Seiko Epson Corp., Tokyo, Japan) to analyse the average RD, root length, root volume and root surface using WinRHIZO Pro software (Regent Instrument, Inc., 2012, Quebec, Canada). These fine root samples were oven-dried at 65 °C for 48 h and then weighed.

Measurements of soil and fine root samples

Soil pH was measured using a pH meter (FE20-FiveEasyTM pH, Mettler Toledo, Germany) in a 1:2.5 ratio of soil to deionized water. Soil total carbon (TC) and soil total nitrogen (TN) were measured using an elemental analyser (Elementar vario EL cube, Elementar, Germany). The ammonium nitrogen (NH₄-N) and nitrate nitrogen (NO₃-N) contents were assayed by a continuous-flow autoanalyser (Auto Analyzer III, Bran+Luebbe GmbH, Hamburg, Germany). Soil total phosphorus (TP) was measured by the molybdenum blue method with an ultraviolet–visible spectrophotometer (UV-2550, Shimadzu, Kyoto, Japan). Soil available phosphorus (AP) was extracted using 0.5 M NaHCO₃ and measured using the molybdenum blue method.

The SRL (m g⁻¹) was calculated using the root length and root dry weight. The specific root surface area (SRA, cm² g⁻¹) was calculated according to the root surface area and root dry weight. The RTD (g cm⁻³) represented the ratio of the root dry weight to the root volume. The root carbon content (RCC, g kg⁻¹) and RNC (g kg⁻¹) were measured using an elemental analyser (Elementar vario EL cube, Elementar, Germany).

Statistical analyses

The two-way analysis of variance (ANOVA) was conducted to explore the precipitation gradients, soil depths and their interaction effects on the soil properties and root traits, and the ANOVA was conducted to explore the difference of plant communities to precipitation gradients using the IBM SPSS statistics 25.0. Principal component analysis (PCA) was applied to explore the dimension patterns of the key root traits (RD, SRL, SRA, RTD, RNC and RCC) and to determine the shift in root strategies across precipitation gradients using the stats package in R 3.5.3. Redundancy analysis (RDA) was performed to explore the effects of soil physicochemical properties and plant community characteristics on root traits across precipitation gradients using the vegan package in R 3.5.3. The indices of soil and plant community that contributed more than 10% to root trait variation were selected using the interactive-forward-selection method. Then, these indices were used to perform variation partition analysis to qualitatively determine the relative explanations of plant community and soil properties on root trait variation using CANOCO 5.0 software.

RESULTS

Plant community characteristics and soil properties across precipitation gradients

The precipitation treatments changed the plant community characteristics, with significant effects on the plant community biomass and legume coverage (Supplementary Fig. S3 and Table S2). On the whole, the plant biomass was significantly decreased in 0.1P (Supplementary Fig. S3a), in which the grass, sedges and forbs showed a decreasing tread (Supplementary Fig. S3g, h, j). Although the plant community coverage was not significantly different across the precipitation gradients, the legume coverage was significantly increased in 1.5P (Supplementary Fig. S3m). Meanwhile, plant richness was also affected by precipitation changes, and the plant richness, Shannon–Wiener index and Pielou J index showed increasing trends in 1.5P, while the Simpson index showed a decreasing trend in 1.5P (Supplementary Fig. S3c–f).

The precipitation treatments also significantly affected the SWC, soil NH₄-N, NO₃-N and AP. Meanwhile, soil depth had significant effects on soil bulk density (SBD), soil pH, soil TC and soil TN, while the interaction between precipitation treatments and soil depth had a significant effect on the content of NH₄-N (Supplementary Table S2). Specifically, the SWC in 0.1P was significantly lower than that in other precipitation gradients, but showed an increasing trend with precipitation gradients from 0.5P to 1.5P (Supplementary Fig. S4a and Table S2). The NO₃-N and AP among the decreasing precipitation treatments (0.1P, 0.5P and 0.7P) were not significantly different but significantly decreased compared with that in the 1.0P and 1.5P treatments (Supplementary Fig. S4h, i and Table S2). SBD, pH, TC, TN and TP did not show significant responses to the precipitation gradients. NH₄-N in both soil layers was significantly decreased in the decreasing precipitation treatments (0.1P, 0.5P and 0.7P) compared with that in the 1.0P. However, the NH₄-N in 10–20 cm from the 1.5P was significantly increased and that in 0–10 cm was significantly decreased (Supplementary Fig. S4g).

In general, the soil depth effects on soil properties were consistent. SWC, TC, TN, NO₃-N and AP were higher in the 0–10 cm soil layer than the 10–20 cm, and SBD, soil pH and TP were higher in the 10–20 cm soil layer, although not significantly higher (Supplementary Fig. S4). Additionally, the interacting effects of precipitation treatments and soil depth narrowed the difference of NH₄-N between the two soil layers (Supplementary Fig. S4g).

Root traits across precipitation gradients

The precipitation treatments, soil depths and their interactions significantly affected the root traits, among which the SRL, SRA, RTD and RNC changed significantly in response to the precipitation treatments. Meanwhile, the RD, SRL, SRA and RCC were also significantly affected by soil depth (Supplementary Table S2).

Although not statistically significant, the RD in the 0–10 and 10–20 cm soil layers exhibited the opposite responses to the changing precipitation treatments (0.1P, 0.5P, 0.7P and 1.5P, Fig. 1a). Compared with the 1.0P treatment, the RD in the 0–10 cm soil layer under the changing precipitation treatments showed a certain decrease, while the RD in the 10–20 cm soil layer showed a certain increase. Additionally, the RD in the 0–10 and 10–20 cm soil layers did not show significant changes under changes in precipitation treatments compared with that under natural precipitation (1.0P, Fig. 1a). Although the SRL, SRA and RTD in the 0–10 cm soil layer were not significantly changed by the precipitation changes, they were significantly altered in the 10–20 cm soil layer. For example, SRL and SRA at 10–20 cm were significantly decreased (Fig. 1b, c) while RTD was significantly increased (Fig. 1d). The RNC in the two soil depths increased in the 0.1P, 0.5P and 0.7P treatments but decreased in the 1.5P treatment (Fig. 1e). RCC was relatively consistent across the precipitation gradients from 0.1P to 1.5P (Fig. 1f).

Root strategies across precipitation gradients

PCA showed a multidimensional root economics space across the precipitation treatments, with over 90% of the total variance being explained by the first two axes (PC1 and PC2, Supplementary Table S3). In the 0–10 and 10–20 cm soil layers, the RD indicative of the ‘Outsourcing’ strategy occupied the positive end of the first axis (PC1), where increasing precipitation (1.5P) was also located (Fig. 2a, b). However, on the opposite end of PC1, which corresponded to the SRL and SRA indicative of the ‘Do-it-yourself’ and the RNC indicative of the ‘Fast’ strategy, the decreasing precipitation treatments (0.1P, 0.5P and 0.7P) were located in the 0–10 cm soil layer (Fig. 2a). In the 10–20 cm soil layer, the SRL and SRA indicative of the ‘Do-it-yourself’ strategy occupied the opposite end of PC1, which corresponded to natural precipitation (1.0P, Fig. 2b). For the second axis (PC2), the RTD indicative of the ‘Slow’ strategy was located on the positive end in the 0–10 cm soil layer, which corresponded to extreme drought (0.1P, Fig. 2a). However, in the 10–20 cm soil layer, the RNC indicative of the ‘Fast’ strategy was located on the negative end of PC2, which corresponded to extreme drought (0.1P, Fig. 2b). Expansively, the roots in the 0–10 and 10–20 cm showed different root strategies, which resulted from the precipitation gradient treatments, especially in the 0.1P.

Links between plant communities, soil properties and root traits

The RDA showed that the plant community characteristics explained the 80.53% variation in root traits of 0–10 cm, with RDA 1 and 2 accounting for 59.94% and 20.59% of the total variance, respectively (Fig. 3a). RDA1 was mainly driven by variations in plant community coverage and monocotyledon (grass and sedge) coverage and richness, which significantly affected root traits, especially SRL and SRA (Supplementary Table S4 and Fig. S5a). In addition, the plant community coverage was negatively correlated with RD and RTD but positively correlated with RNC and RCC (Supplementary Fig. S5a). The RDA also showed 85.60% of the variation in root traits in the 10–20 cm soil layer, with RDA 1 and 2 accounting for 51.42% and 34.18% of the total variance, respectively (Fig. 3b). RDA1 was mainly driven by variations in legume biomass, and the RDA2 was mainly driven by variations in the plant community biomass and coverage (Supplementary Table S4 and Fig. S5b).

Soil properties also explained 83.78% of the variation of root traits at 0–10 cm (Fig. 3c) and 79.15% at 10–20 cm (Fig. 3d). In the 0–10 cm soil layer, RDA1 was mainly driven by variations in SWC, TP and AP, which were significantly correlated with root traits (Supplementary Fig. S5c). Along the RDA2 axis, soil NH₄-N showed the greatest contribution to root trait variance (Supplementary Table S4 and Fig. 3c), especially RNC (Supplementary Fig. S5c). In the 10–20 cm soil layer, RDA1 was mainly driven by variations in SWC, which was significantly correlated with SRL, RTD, SRA and RCC (Fig. 3d and Supplementary Fig. S5d). RDA2 was mainly driven by several soil indexes, such as TP and NO₃-N, which showed significant correlations with SRL and RNC, respectively (Fig. 3d and Supplementary Fig. S5d).

The RDA showed that both the plant community characteristics and soil properties significantly affected root traits in the two soil layers (Fig. 3). In the 0–10 cm soil layer, 63% of the total root variation was explained by the plant community characteristics and soil properties, which accounted for 28% and 51% of the total variance, respectively (Fig. 4a). However, variations in root traits explained by the plant alone only accounted for 12%, which was much less than the 35% explained by soil properties alone (Fig. 4a). In the 10–20 cm soil layer, 47% of total root variation was explained by plant community characteristics and soil properties, which accounted for 16% and 34% of the total variance, respectively (Fig. 4b). In both soil layers, soil properties could better explain the root variations than the plant community characteristics.

DISCUSSION

Root traits across precipitation gradients

Our study explored the responses of root traits to precipitation changes and revealed the specific strategies roots employ in response to various water availability. We found that the RD and RCC exhibited relatively little variation across precipitation gradients, which is consistent with previous findings that RD presented limited changes in response to precipitation changes (Zhou et al. 2018). However, the RTD decreases as the precipitation changed, especially in the 10–20 cm soil layer. This might imply that the proportion of stele with higher tissue density in fine roots decreased, and the proportion of cortex and other nitrogen-rich tissues increased. Accordingly, the metabolic activity of the roots may tend to be enhanced. In addition, both RD and RCC were significantly correlated with plant community coverage, which was also not significantly changed with the precipitation gradient. In addition, changes in RD and RCC were closely linked to the plant community composition and carbon allocation between the aboveground and belowground parts of plants. The dominant species in our study, Saussurea nigrescens and K. humilis, accounted for over 50% of the plant community biomass across the five precipitation gradients (Qiao et al. 2023), which suggests a stable investment in root carbon despite varying water conditions. Thereafter, the stable dominant species biomass variation demonstrated that the plant might not increase the carbon investment into the belowground roots, which resulted in nonsignificant RCC. Although studies have demonstrated that the RCC was highly coordinated with root foraging and was driven by local temperature and precipitation, the root carbon referred to here was the nonstructural carbohydrates (Zhang et al. 2024), not the TC discussed in this article. Notably, we found that RNC was more sensitive to precipitation changes and increased slightly in the decreasing precipitation treatments but decreased significantly in the increasing precipitation treatment. This pattern aligns with the notion that higher precipitation can lead to potential nitrogen losses, which affect RNC (Freschet et al. 2017), because high precipitation usually results in high potential N losses via denitrification, leaching and runoff (Chapin et al. 2011). In contrast, decreasing precipitation usually led to slow organic matter cycling and nitrogen mineralization, which allowed plants to take up more nitrogen from soils. Roots with high N content indicating high root metabolism could result in a rapid and more efficient strategy for the acquisition of N and water (Liu et al. 2010).

In most studies, extreme drought conditions favoured the development of thinner roots with a higher SRL (Chen et al. 2013a) and higher RNC (Metcalfe et al. 2008), which corresponded to the root drought-avoidance strategy to coexist with perennial roots (Salguero-Gomez and Casper 2011). Higher SRL is generally correlated with more rapid root turnover (Eissenstat et al. 2015), suggesting a more rapid return on investment strategy (Eissenstat et al. 2000), which is especially beneficial in the cold or alpine regions where strong seasonality and soil freezing could result in intermittent nutrient availability (Bardgett et al. 2005). However, our finding showed that under a long-term drought scenario, the SRL was lower compared with that of natural precipitation (1.0P), especially in the 10–20 cm soil layer (Fig. 1), indicating a potential phenotypic adaptation to prolonged drought through a conservative ‘Slow’ strategy and an ‘Outsourcing’ strategy characterized by thicker roots (Zhou et al. 2019). Specifically, with the occasional drought event, roots with higher SRL were useful for the rapid acquisition of water and nutrients (Freschet et al. 2017). However, when simulating extreme drought over 7 years, the roots tended to exhibit the ‘Slow’ conservative strategy with higher RTD as well as the ‘Outsourcing’ strategy with higher RD to adapt to the persistent drought. Interestingly, we also found that the increased precipitation (1.5P) and the extreme drought (0.1P) had the same effect on SRL and SRA, both of which significantly reduced both traits in the 10–20 cm soil layer. This is because increased precipitation would improve soil water retention, initially enhancing water availability in the surface soils. As a result, the higher soil water content reduced the need for roots to extend extensively in search of water in the deeper soils, resulting in lower SRL and SRA. In conclusion, our research underscores the different responses of root traits to precipitation and indicates that while some root traits display limited plasticity, significant adjustments occur only beyond a certain precipitation threshold. This highlights the complex interplay between root trait adaptation and environmental conditions, thus offering valuable insights into plant survival strategies under varying climatic scenarios. Changes in root traits that indicated shifts in root strategies to precipitation changes were observed, and RD was strongly negatively correlated with SRL, thus supporting the ‘Collaboration’ gradient in the existing RES frameworks (Bergmann et al. 2020). However, only a partly ‘Conservation’ gradient was observed, which was represented by RNC, independent of the ‘Collaboration’ gradient as well as the RTD, especially in the 10–20 cm soil layer (Fig. 2). This finding diverges from existing RES frameworks, where the ‘Conservation’ gradient was represented by the negative correlation between RNC and RTD (Bergmann et al. 2020). In the ideal RES frameworks from Bergmann et al. (2020), RNC and RTD and RD and SRL were trade-offs and negatively correlated. However, according to the fine root construction theory, strong negative correlations have been found between SRL and RD (Chen et al. 2013; Kramer-Walter et al. 2016; Roumet et al. 2016; Zhou et al. 2020), while the relationships between SRL and RTD (Valverde-Barrantes and Blackwood 2016; Valverde-Barrantes et al. 2016) and between RD and RNC (Kong et al. 2019) have been less clear. In our study, RTD was negatively correlated with SRL but positively correlated with RD (Fig. 2 and Supplementary Fig. S5). Furthermore, the relationships between RNC, which represents the root metabolic rate to assure fast resource acquisition, and SRL, which reflects the rate of return per unit of investment, were decoupled (Chen et al. 2013; Zhang et al. 2023).

RNC representing the ‘Fast’ strategy could effectively differentiate variations in root strategies among the decreasing precipitation treatments (0.1P, 0.5P and 0.7P) (Fig. 2). The roots in 0.7P mainly depended on the ‘Fast’ strategy for higher RNC, especially in the 10–20 cm soil layer (Fig. 1). However, under extremely low precipitation (0.1P), roots depended on root self-modifications, such as higher RTD (Fig. 1), indicating the occurrence of the ‘Slow’ strategy. These results suggest that root strategies shifted from ‘Slow’ to ‘Fast’ from extreme to slight drought.

In addition, soil depth effects on root traits and associated strategies were also different, especially when precipitation changes were considered. Our study demonstrated that decreasing precipitation would enhance the uptake capacity through modified root traits, such as SRL and SRA. However, the phenomenon was only found in the roots from the 0–10 cm soil layer because of the higher water loss due to evaporation of surface soil compared with that in the deeper soil. Under decreasing precipitation, roots depend more on their own regulation strategies in the 0–10 cm than in the 10–20 cm soil layer. The difference in root strategies in both soil layers may have resulted from the nonsignificant effect of simulating precipitation in the 10–20 cm soil layer and the key factors, including ammonium, nitrogen and TP (Supplementary Fig. S4).

Soil properties and plant communities co-drive root strategies

Climate (i.e. temperature and precipitation) and soil properties (Ding et al. 2023; Freschet et al. 2017; Yu et al. 2024) strongly affect root trait variations and strategies. Among them, soil properties are particularly crucial to the root strategies (Zhang et al. 2021), even more so than the plant community (Li et al. 2019a; Valverde-Barrantes et al. 2017), which was also demonstrated in our study (Fig. 4). Among the soil properties, we found that SWC driven by precipitation gradients was a key determinant driving root strategy (Fig. 3), which was consistent with previous findings (De la Riva et al. 2016; Ding et al. 2020; Li et al. 2019a). Meanwhile, we also demonstrated that root strategies were affected by soil N, including NO₃-N and NH₄-N, and soil P, because alpine plant growth is co-limited by nitrogen and phosphorus (Xu et al. 2015). Moreover, soil nitrogen and phosphorus would affect root production and turnover (Hu et al. 2021), which indicates the changing root strategies. In our study, NH₄-N showed significantly negative correlations with RNC, which indicated that roots in soils with low NH₄-N content might exhibit the ‘Fast’ strategy. Several studies have also demonstrated that roots in infertile soil tend to show increased cell wall fractions, thus favouring a ‘Do-it-yourself’ strategy (Roumet et al. 2016; Wahl and Ryser 2000).

Although the contribution of plant community characteristics to root trait variance was less than that of soil properties, plant community characteristics significantly affected root traits. Notably, root trait variations in the 0–10 cm soil layer were mainly determined by the coverage of grasses and sedges, while those in the 10–20 cm soil layer were mainly determined by the plant community coverage and biomass (Fig. 3 and Supplementary Table S4). In the studied alpine meadow, the dominant species were the grass E. nutans and the sedge K. humilis. Although their root traits were not differentiated and measured based on the cumbersome root system in our study, both the roots of grasses and sedges are well known to have shallow but extensive fibrous root systems, which are beneficial for nutrient uptake in the surface soil layers. As the root system size increased with the aboveground size, the root system sizes of forbs are usually less than that of grasses (Schenk and Jackson 2003). Therefore, most roots in the surface soils consisted of grasses and sedges in alpine meadows, and the root traits were mainly driven by the coverage of grasses and sedges. Specifically, monocotyledonous plants showed unique root traits, such as a higher SRL for quick nutrient acquisition (Birouste et al. 2012; Jiang 2022). This was supported by the SRL in the 0–10 cm soil layer, which was significantly positively correlated with the coverage of grasses and sedges (Supplementary Fig. S5). The RD and RNC in the 0–10 cm soil layer were affected not only by the grass coverage but also by the grass biomass. The RD of grass plants was strongly and negatively related to traits related to nutrient acquisition (Roumet et al. 2016), which also demonstrated the ‘Fast’ root strategy for the grasses. Additionally, in alpine meadows, the 0–10 cm soil layer is the primary zone of nutrient availability for most plant species; thus, the coverage of dominant plants with competitive root traits would be affected. In contrast, the roots in the deeper soil layer (10–20 cm) consisted of dominant monocotyledonous plant roots but also included dicotyledon plants, especially the legume plants in our study with tap root systems, which tended to penetrate deeper, thereby facilitating deeper nutrient and water uptake. Therefore, the roots in the deeper soil layer reflected the combined effects of most plant species rather than the dominant species alone. This differential influence of plant community characteristics on root traits across soil depths underscores the complexity of plant–soil interactions in alpine grassland ecosystems. Understanding these relationships is essential for elucidating the mechanisms driving ecosystem structure and function in high-altitude environments.

CONCLUSIONS

Our research revealed the responses of root trait variations and root strategies to both increasing and decreasing precipitation in an alpine meadow. Certain root traits, such as RD and RCC, were not significantly influenced by precipitation changes. However, the SRL, RNC and RTD demonstrated sensitivity towards precipitation changes. These variations in root traits could result in shifts in root strategies across precipitation gradients. For example, under conditions with increased precipitation, plant roots tend to rely more heavily on the ‘Outsourcing’ strategy. Conversely, under conditions of decreasing precipitation, roots favour the ‘Do-it-yourself’ and ‘Fast’ strategies. However, long-term extreme drought conditions could compel roots to depend more on the ‘Slow’ strategy. The distinct responses of root traits and strategies were co-driven by soil properties, such as the SWC and available nutrients, and the plant communities. In addition, our study demonstrated that root traits and root strategies also depend on the soil depth, with roots in the surface soils mainly driven by monocotyledon plants with their fibrous root system and those in deeper soils mainly driven by the plant communities. Future research should concentrate on differences in the sensitivity of root traits and rooting depth from different plant functional groups to climate changes because such data could lead to the development of more effective methods of predicting how plant communities respond to climate changes, particularly from the perspective of belowground roots.

Supplementary Material

Supplementary material is available at Journal of Plant Ecology online.

Figure S1: Sample site information.

Figure S2: The volumetric soil water content of different precipitation treatments from September 2019 to December 2021.

Figure S3: Response of plant community characteristics in different precipitation gradients.

Figure S4: Response of soil properties in different precipitation gradients.

Figure S5: The correlation among the root traits and plant community characteristics, soil properties.

Table S1: Gravimetric soil water content (%) in the 0–10 and 10–20 cm layers under different precipitation treatments (mean ± SE).

Table S2: The linear mixed effects model showing the effects of precipitation gradients, soil depth and their interaction on soil properties and root traits.

Table S3: The loading scores of root traits on the first axes (PC1) and second axes (PC2) of principal component analysis (PCA) in 0–10 and 10–20 cm soil layer.

Table S4: The analysis results from interactive-forward-selection showing the effects of plant communities and soil properties on the root traits in 0–10 and 10–20 cm soil layer.

Funding

This work was supported by the National Key R&D Program of China (2023YFF1304304), the National Natural Science Foundation of China (U20A2008), the Project of Grassland Multifunctionality Evaluation in Three-River-Source National Park (QHQXD-2023-28) and the Fundamental Research Funds for the Central Universities, Southwest Minzu University (ZYN2023072).

Authors’ contributions

Lei Hu, Chang-Ting Wang and Lerdau Manuel conceived the ideas and methodology; Lei Hu, Chang-Ting Wang and Min Liu designed the experiment; Yi-Heng Li and Xin-Di Zhang collected the data; Yi-Heng Li, Xin-Di Zhang and Lei Hu analyzed the data; Yi-Heng Li, Lei Hu and Min Liu wrote the manuscript.

Acknowledgement

We would like to thank Editage (www.editage.cn) for English language editing.

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

REFERENCES