Plant height as an indicator for alpine carbon sequestration and ecosystem response to warming

Growing evidence indicates that plant community structure and traits have changed under climate warming, especially in cold or high-elevation regions. However, the impact of these warming-induced changes on ecosystem carbon sequestration remains unclear. Using a warming experiment on the high-elevation Qinghai-Tibetan Plateau, we found that warming not only increased plant species height but also altered species composition, collectively resulting in a taller plant community associated with increased net ecosystem productivity (NEP). Along a 1,500 km transect on the Plateau, taller plant community promoted NEP and soil carbon through associated chlorophyll content and other photosynthetic traits at the community level. Overall, plant community height as a dominant trait is associated with species composition and regulates ecosystem C sequestration in the high-elevation biome. This trait-based association provides new insights into predicting the direction, magnitude and sensitivity of ecosystem C fluxes in response to climate warming.


Disentangling height variability and community composition effects
Because variation of community-weighted trait can be caused by both trait variability and changes in the community composition, we disentangled community-weighted height (CWH) into both height variability and changes in community composition following Leps et al. (2011)  81 : where  is community weighted height defined in Methods section in the main text.  is 'fixed' community weighted height caused by changes in community composition alone.Specifically, it was calculated using 'fixed' mean height of each plant functional group across all communities and weighted by plot-specific functional composition data.As the difference between them,  is the height variability which represents changes in height alone.By disentangling their roles, we can evaluate the effects of height variability and community composition on net ecosystem productivity, which are shown in Supplementary Figure 1.

Soil C content measurement in warming experiment
Soil samples were randomly collected from each plot once a year from 2015 by using a soil auger (7.5 cm in diameter, 10 cm in depth).All samples were air-dried and sieved with 2 mm mesh to remove stones and plant roots.We used a LECO macro-CN analyzer (LECO, St. Joseph, MI, USA) to measure soil total C. Changes in soil total C content under three warming treatments are shown in Supplementary Figure 8.

Light intensity measurement in warming experiment
We measured light intensity by using a digital light meter (TES-1335, TES Electrical Electronic Corp., Taiwan, China) at the control plots in August 2021 when biomass peaked.The light intensities at five heights from ground surface to canopy (0, 10, 20, 30, 40 and 50 cm) of plant community were measured under cloud-free condition.The differences in light intensity among different heights are shown in Supplementary Figure 12.

Plant traits investigation
Plant traits data were obtained from a large-scale field survey when biomass peaked during the growing season from 2019 to 2021 on the Qinghai-Tibetan Plateau.In brief, 1564 sites were surveyed by using a rasterized sampling method.These sites were distributed with a 0.5 o grid size in latitudinal and longitudinal directions.In each site, plant samples for each species of were collected from three plant communities, in addition, plant samples of all visible species within 1 km area around the sample site were also collected.20 mature leaves of each plant species were collected for the measurements of plant traits.Details of the field survey and method are described in Li et al. (2022) 82 , Wang et al. (2021) 83 and Zhang et al. (2020) 84 .We used available plant traits that related to ecosystem C uptake or sequestration from this field trait survey.For example, chlorophyll content captures light energy and converts CO2 and light into organic C compounds 85,86 .Leaf C content represents plant's investment in photosynthetic tissues and refers to the amount of C stored in leaves which can contribute to soil C sequestration 87,88 .The leaf area index provides an estimate of the surface area available for photosynthesis 17,89 .Stomata regulate gas exchanges on the surface of leaves 90,91 .Larger stomatal size would promote CO2 uptake while smaller stomata exhibit faster opening and closing rates, enabling shorter plants to respond rapidly to light fluctuations beneath the canopy and efficiently maximize their photosynthesis [90][91][92][93][94][95] .All these traits are associated with ecosystem C dynamics.

Stomatal size measurement
In brief, 6-8 leaves of each species were cut into 0.5 cm pieces along the main vein and fixed in formalin -acetic acid -alcohol (FAA) solution.Scanning electron microscope (S-3400N, Hitachi, Japan) was used to observe stomatal morphological parameters.
Three replicates of pieces for each species were photographed.Stomatal guard cell length (L, μm) and guard cell width (W, μm) were measured by using MIPS software (Optical Instrument Co., Ltd., Chongqing, China).The measurement details are described in Liu et al. (2018) 96 .Stomatal size (S, μm 2 ) was calculated as:

Chlorophyll content measurement
For the measurement of chlorophyll, 0.1 g fresh leaves were extracted three times by 95% ethanol with four replicates for each plant species.The filtered extraction solutions were adjusted to 50 ml.Then, the chlorophyll content (chlorophyll a and chlorophyll b) of the solution was measured by using a spectrophotometer (Pharma Spec, UV-1700, Shimadzu, Japan).Based on the Lambert-Beer law, the relationships of the optical density with chlorophyll concentration are calculated as:  98 .

Leaf C content measurement
The dried leaf samples were finely ground by an agate mortar grinder (RM200, Retsch, Haan, Germany) with a ball mill (MM400 Ball Mill, Retsch).Then, we used an elemental analyzer (Vario Max CN Element Analyzer, Elementar, Hanau, Germany) to measure leaf C content (%) 84 .Plant functional traits at the community level were calculated as the same as community-weighted height.In the warming experiment, we used the trait values collected from the field survey under natural condition weighted by community compositional data in the warming communities, we estimated changes in plant functional traits (stomatal size, chlorophyll content and leaf C content) at the community level in the warming plots, assuming that changes in community weighted traits were mainly caused by species turnover rather than intraspecific trait variation among warming treatments 21,24 .The relationships of community weighted height with other plant functional traits at the community level are shown in Figures 2 and 4.

Obtaining leaf area index and net ecosystem productivity from remote-sensed data
The MODIS (Moderate Resolution Imaging Spectroradiometer) provides satellitebased observations of carbon fluxes and surface measurements from 1982 to present 99,100 .We extracted leaf area index (LAI) for each site of 1 km spatial resolution, and 8day temporal resolution from LPDAAC.The daily net ecosystem productivity (NEP) data was extracted from SMAP (soil moisture active passive) level-4 NEE product with a 9-km resolution 101 .To be consistent with the timing of our transect study, annual average LAI and NEP in 2019 for each site were calculated.

Factors controlling the variation in NEP
We used structural equation model (SEM) to estimate the pathways through which warming influenced the NEP in both the warming experiment and the transect study (Supplementary Fig. 16).In the warming experiment, the biotic and abiotic factors included community weighted height, community weighted chlorophyll content, community weighted stomatal size, community weighted leaf C content, mean soil temperature during the growing season and mean annual precipitation.In the transect study, the factors included community weighted height, community weighted height chlorophyll content, community weighted height stomatal size, community weighted height leaf C content, leaf area index, mean annual temperature, mean annual precipitation, soil available P and soil total N.All potential causal variables and relationships were considered in a priori model, and we simplified the model by removing redundant variables and non-significant pathways.Model performance was assessed through Chi-square (χ 2 ) test, root mean square error of approximation (RMSEA) and Akaike information criterion (AIC).The final optimal mode was selected based on the lowest AIC value.Moreover, we used ridge regression to control for collinearity among covariates and to evaluate the relative importance of each variable on NEP in both the warming experiment and the transect study (Supplementary Fig. 17).The ridge regression was performed using the "ridge (v3.3)" package in R statistical software v 3.4.3(The R Foundation for Statistical Computing, Vienna, Austria), and the SEM models were conducted using AMOS 21.0 (Amos Development Corporation, Chicago, IL, USA).
F and P values) on the effects of warming treatment, year, plant functional type, and their interactions on mean plant height (n = 5).
and (a) height variability and (b) changes in plant community composition in the controlled warming experiment.Linear regression with two-sided test was used for the statistical analysis.The coefficient of determination (R 2 ) and the exact P values for all the regressions were indicated.The error bands are 95% confidence intervals (± 1.96 s.e.m.) around the fitted regression lines, sample size n = 60.and plant community traits for (a) chlorophyll content, (b) stomatal size, and (c) leaf C content in the controlled warming experiment.Linear regression with two-sided test was used for the statistical analysis.The coefficient of determination (R 2 ) and the exact P values for all the regressions were indicated.The error bands are 95% confidence intervals (± 1.96 s.e.m.) around the fitted regression lines, sample size n = 60.Supplementary Figure 13 Light intensity of different height from ground surface (0 cm) to canopy (50 cm) of plant community in the control plots.Data are presented as mean ± s.e.m., sample size n = 4 for 0 cm to 40 cm and n = 3 for 50 cm.One-way ANOVA followed by two-sided LSD test was used for multiple comparison.Different letters close to the bars indicate significant differences among height levels at α = 0.05.
649 and  665 are the optical densities of the chlorophyll solution at wavelengths of 649 and 665 nm;   ,   and  represent the concentrations of chlorophyll a, chlorophyll b and total chlorophyll (g L -1 ) in the solution; the coefficients 24.54 and 44.24 are the specific absorption of chlorophyll a and chlorophyll b at the wavelength of 649 nm, 83.31 and 18.60 are at 665 nm.ℎ is the total chlorophyll 24.54   + 44.24   (3)  665 = 83.31  + 18.60   (4) content per gram of dry weight (mg g -1 ) and % is the percentage of species leaf water content.Details on the measurement are described in Li et al. (2018) 97 and Zhang et al. (2020) Information of investigation sites.MAT: mean annual temperature; MAP: mean annual precipitation.Effects of community weighted height, mean annual temperature, mean annual precipitation, soil available P and soil total N on net ecosystem productivity, and effects of these environmental factors on community weighted height in the transect study (n = 45).The effects were estimated by using the linear mixed-effects model with two-sided test.