Unexpected high carbon losses in a continental glacier foreland on the Tibetan Plateau

Glacier retreat due to global warming is exposing new foreland soils that are typically cold, dry, and nutrient-poor, with carbon being a key element for soil development. Many glacier forelands worldwide act as carbon sinks, with SOC generally accumulating rapidly in early stages. However, regional differences in climate, initial carbon inputs, and soil factors can alter this pattern. On the Tibetan Plateau, temperate glacier forelands (e.g., Hailuogou) show marked SOC accumulation, whereas sub-continental glacier forelands (e.g., Urumqi Glacier No. 1) show modest increases. Continental glaciers, extensive in China and highly sensitive to climate change, remain understudied regarding their carbon dynamics. Laohugou Glacier No. 12 has retreated substantially, offering a chronosequence to test whether continental forelands also accumulate SOC. Because early-stage carbon dynamics depend heavily on microbial communities—particularly the balance of microbial respiration and fixation and shifts in community composition and functional genes—the study hypothesized that, despite cold and dry conditions, SOC would accumulate along the continental glacier foreland chronosequence, linked to microbial community changes.

Prior studies generally report SOC increases during primary succession in glacier forelands across regions including the Alps, Norway, Iceland, the Antarctic Peninsula, and the High Arctic. On the Tibetan Plateau, a temperate glacier foreland (Hailuogou) developed into coniferous forest with high SOC accumulation rates, while a sub-continental foreland (Urumqi Glacier No. 1) saw only slight SOC increases over decades. Microbial processes are central to early-stage carbon accumulation before vegetation establishes, with bacterial pioneers mediating carbon fixation and respiration. Shifts in microbial community composition (e.g., autotrophs vs heterotrophs) and functional gene abundances are known to influence nutrient cycling and carbon storage. However, there has been a lack of studies focused on continental glacier forelands in China, despite their large area and sensitivity to climate change.

Study site and sampling: Laohugou Glacier No. 12 (39°26.4′N, 96°32.5′E; elevation up to 5483 m) on the northern slope of the western Qilian Mountains is a large continental glacier with mean annual temperature −11.8 °C, ~390 mm annual precipitation mainly in late summer–early autumn, and prevailing westerly winds. The terminus retreated ~403 m from 1960 to 2015 (~7.3 m yr−1). Five sites were positioned by distance from the ice tongue corresponding to exposure ages of 0, 10, 15, 31, and 50 years (S0, S10, S15, S31, S50). S0–S15 were unvegetated; S31 and S50 had sparse patches of Thylacospermum caespitosum and Ajania scharnhorstii. In November 2020, surface soils were collected in triplicate per site, sieved to 2 mm, split for physicochemical and molecular analyses, and transported at 4 °C. Physicochemical analyses: Soil pH (1:5 soil:water slurry) measured with electrode; gravimetric water content for fresh soil. SOC determined after carbonate removal by acidification with 1 M HCl (1:2.5 m/v). Total N by combustion (Vario Macro cube). Total P by acid solution Mo–Sb anti-spectrophotometry. DOC extracted in water (1:5 m/v) after 24 h shaking at 30 °C, filtered (0.45 μm), quantified by TOC analyzer (Vario TOC). NH4+ and NO3− extracted in 2 M KCl (soil:extractant 1:5 m/v), shaken 2 h, settled 30 min, filtered (0.45 μm), measured by continuous flow analyzer (AA3). DNA extraction and 16S rRNA sequencing: DNA extracted from 0.5 g soil (MOBIO PowerSoil). DNA quantity/quality by Qubit 4.0 and NanoDrop 2000. Bacterial communities amplified using primers U341F/U806R targeting 16S V3–V4, PCR conditions specified (94 °C denaturation, 52 °C annealing, 72 °C extension; 30 cycles). Libraries sequenced on Illumina Nova6000. Reads processed via Galaxy: primer removal, merging (FLASH), quality filtering; OTU clustering at 97% (UPARSE), chimeras removed; taxonomy assigned with Greengenes 13.8. Samples rarefied to minimum OTU count after removing low-frequency OTUs. Quantitative microbial element cycling (QMEC): High-throughput qPCR chip quantified 32 carbon-cycling functional genes using 16S rRNA gene as reference. Amplifications run on SmartChip system with triplicate processing, including non-template controls. Genes retained if efficiency 1.8–2.2 and NTC showed no amplification; Ct=31 as detection limit. Absolute gene abundances calculated from Ct and absolute 16S rRNA gene quantities. Statistics: Alpha diversity (observed richness, Chao1, Shannon, Pielou’s evenness) computed in R 4.0.5. One-way ANOVA with Fisher’s LSD tested differences across successional ages for diversity, environmental factors, and functional gene abundances. Pearson correlations examined relationships among SOC, dominant bacterial phyla, and carbon-related functional gene groups. Community differences assessed by PCoA (Bray–Curtis) and PERMANOVA. Heatmaps via vegan and pheatmap; plots via ggplot2. Structural equation modeling (SEM): Conducted in AMOS 24 to evaluate effects of retreat time, soil pH, bacterial abundance (16S rRNA copy number, as biomass proxy), and functional potential (FD ratio = carbon fixation to degradation gene abundance) on SOC. Model fit criteria: non-significant χ² (p>0.05), RMSEA<0.05, GFI and NFI>0.90.

Soil properties: Soils were alkaline, with pH slightly increasing to a maximum at S50 (8.93 ± 0.02). Water content was very low (0.38–0.91%). SOC varied with age: it peaked at S10 (37.80 ± 0.07 g kg−1) then declined to 10.77 g kg−1 at S50, which is 11.44 g kg−1 lower than S0 (22.21 g kg−1). TN increased from 0.22 to 0.45 g kg−1 over 50 years; TP remained relatively stable (0.51–0.66 g kg−1). C:N and C:P ratios were highest at S10 (211.24; 192.32) and declined to ~27.86 and 49.94 at S50. NH4+ and NO3− decreased markedly (S50 values about half and one-sixth of S0, respectively). DOC increased from 23.33 mg kg−1 (S0) to 37.85 mg kg−1 (S31), then dropped to 27.02 mg kg−1 (S50). Microbial functional potentials: Bacterial 16S rRNA gene copy number increased from 1.07×10^5 to 4.95×10^6 copies g−1 soil over 50 years, peaking at S31 (5.38×10^6). A total of 31 carbon-cycling genes (18 degradation, 13 fixation) were detected. Absolute abundance of carbon fixation genes increased ~7-fold (from 21.77×10^5 to 152.79×10^5 copies g−1 at S50), and carbon degradation genes increased 5.7-fold to 44.83×10^5 copies g−1 at S50. The FD ratio (fixation:degradation) peaked at S10 (5.48) then fell to 3.41 at S50, similar to S0 (2.77). Carbon fixation pathways: WL (reductive acetyl-CoA) genes dominated (>50% across sites), increasing in proportion from 50.01% (S0) to 57.81% (S50). 3HP cycle genes increased from 26.10% to 30.41%. CBB cycle genes slightly increased from 3.37% to 5.81%. rTCA cycle genes declined from 20.52% to 5.96% (S50), indicating reduced potential for high-CO2-assimilating pathways. Carbon degradation pathways: Hemicellulose degradation genes were most prevalent and increased in proportion from 46.03% (S0) to 64.70% (S50). Lignin and starch degradation gene proportions decreased overall (lignin from 28.95% to 13.12%; starch fluctuated, 21.09% to 19.34% at S50). At S50, absolute abundance of starch-degradation genes (867.02×10^3 copies g−1, 19.34%) exceeded lignin-degradation genes (588.09×10^3, 13.12%). Community composition and diversity: From 437,265 quality-filtered reads, 7693 OTUs across 35 phyla were identified. Proteobacteria decreased from 39.64% (S0) to 12.32% (S50); Bacteroidetes decreased by 12.64%. Cyanobacteria declined to 2.05% at S31 from 8.69% (S0), then rose to 4.83% at S50. Actinobacteria increased to 51.51% at S50; Acidobacteria rose from 0.84% to 13.41%. Alpha diversity indices increased notably at S31 and S50. Beta diversity differed significantly among sites (PCoA axes: 36.60% and 27.25% variance; PERMANOVA F=6.71, p<0.001). Correlations and SEM: Acidobacteria and Planctomycetes were negatively correlated with SOC; Acidobacteria positively correlated with hemicellulose, pectin, and 3HP genes and negatively with lignin genes. Proteobacteria positively correlated with lignin degradation and rTCA cycles; Bacteroidetes showed similar patterns. Actinobacteria and Gemmatimonadetes correlated positively with CBB and WL and negatively with rTCA. SEM explained 81% of SOC variance; FD ratio had a strong direct positive effect on SOC (path coefficient 0.60); bacterial abundance had a smaller direct positive effect (0.12). Retreat time had a negative direct effect on SOC (−0.56) with small positive indirect effects (total negative), and soil pH affected SOC indirectly (−0.35). Model fit was good (χ²=1.37, p=0.50, df=2; RMSEA<0.0001; GFI=0.96; NFI=0.96).

Contrary to the common pattern of increasing SOC in many glacier forelands, the Laohugou Glacier No. 12 continental foreland showed an overall SOC decline across 50 years of succession. The findings implicate microbial processes in this carbon loss. The FD ratio decreased after an early peak, indicating lower relative carbon fixation potential versus degradation over time. A marked reduction in rTCA-associated genes—capable of fixing more CO2 per cycle—combined with dominance of WL and 3HP pathways, implies lowered carbon fixation efficiency. Concurrently, autotrophic taxa (e.g., Proteobacteria subgroups, Cyanobacteria, Chloroflexi) decreased, while K-strategists (Actinobacteria, Acidobacteria) increased, suggesting reduced carbon availability and higher microbial competition. Rising bacterial biomass (copy number) points to increased respiratory carbon demand. Together, lower fixation efficiency, fewer autotrophs, strengthened microbial respiration, and a shift toward hemicellulose-centric degradation indicate insufficient replenishment of fixed carbon and elevated consumption, explaining SOC loss. These results challenge assumptions that glacier forelands universally serve as carbon sinks and highlight the need to account for continental forelands’ potential positive feedbacks to climate warming in models.

This study reveals unexpected SOC loss along a 50-year chronosequence in a continental glacier foreland on the Tibetan Plateau. Microbial community shifts from r- to K-strategists, reduced rTCA carbon fixation gene representation, decreased autotrophic taxa, and increased microbial biomass collectively point to declining fixation efficiency and increased carbon demand, resulting in net carbon loss. These insights underscore that continental glacier forelands may not be carbon sinks and that their dynamics should be incorporated into climate and carbon cycle models to avoid overestimation of carbon storage. Future work should expand to additional continental glacier forelands, incorporate direct measurements of carbon fluxes alongside functional potentials, and assess seasonal and interannual variability to generalize and refine predictions.

The study focuses on a single continental glacier foreland with five successional sites sampled once (November 2020), which limits generalization. The authors note that special cases may arise due to the limited number of studied subjects, implying broader uncertainty when extrapolating to all continental glacier forelands. Functional gene abundances indicate potential capacities rather than direct process rates, which may affect interpretation of carbon fixation versus degradation dynamics.