Long-Term Human Disturbance Accelerates Soil Carbon Loss in Earth's Driest Ecosystems

Abstract

Drylands cover over 41% of Earth's terrestrial surface and support nearly 38% of the global population, yet they have long been overlooked in global carbon cycle assessments due to their low net primary productivity (Chen et al. 2024). Despite this, drylands store substantial soil organic carbon (SOC), often deeply buried and stabilized by vegetation and microbial communities adapted to arid conditions. In hyperarid deserts, deep-rooted plants such as Alhagi sparsifolia create vertical SOC stratification, with labile particulate organic carbon (POC) near the surface and persistent mineral-associated organic carbon (MAOC) at depth (Zhao et al. 2025). This deep carbon pool is a slow-cycling reservoir that can sequester carbon for centuries, making drylands potentially important carbon sinks despite low productivity. However, intensifying human disturbances, biomass harvesting, burning, and irrigation pose increasing risks to the stability of these carbon stocks (Ali and Xu 2025; Chen et al. 2024; Delcourt et al. 2025). The fate of the more stable microbial-derived carbon and its mineral associations under chronic disturbance remains poorly understood.

Recent work by Gao et al. (2025), published in Global Change Biology, provides valuable new insights through a rare 16-year field experiment along the southern margin of the Taklimakan Desert, one of the world's driest and most fragile ecosystems. Their study applies disturbances mimicking local human activities, seasonal biomass harvest, fire, and artificial irrigation, to reveal how long-term anthropogenic pressure drives SOC loss. The study site, a desert–oasis transition zone, is stabilized by perennial shrubs like Alhagi sparsifolia, which also provide forage for local herders during spring and autumn harvests. Vegetation burning and artificial floodwater channeling are common disturbances whose impacts on SOC were unclear before this work.

Starting in 2008, Gao et al. (2025) applied five treatments annually: control (no disturbance), spring harvest, autumn harvest, fire, and irrigation simulating flood events. Each 30 × 50 m plot was buffered to prevent cross-contamination. This uncommon long-term, consistent disturbance experiment allowed a detailed investigation of chronic impacts in a hyperarid environment. In 2024, the team sampled plant biomass, litter, fine roots, and soils to 150 cm depth at six intervals. SOC was fractionated into POC (> 53 μm) and MAOC (< 53 μm). Plant-derived carbon was traced using lignin phenol biomarkers, and microbial-derived carbon quantified via amino sugars from fungal and bacterial residues. Soil mineralogy, enzyme activities, microbial biomass, and community composition from metagenomic sequencing were also measured, enabling a comprehensive mechanistic view of SOC dynamics.

Results showed consistent SOC depletion across all disturbance types. Total SOC declined by 13.2% relative to controls, with POC decreasing by 16.3% and MAOC suffering a striking 41.1% loss. The POC/MAOC ratio rose 46.2%, indicating destabilization: a shift from stable, mineral-protected carbon to a more labile, decomposition-prone pool. The 0–15 cm surface soil, rich in plant residues and weakly protected carbon, was most vulnerable, showing steep declines. Although deeper soils (> 100 cm) were less affected, significant losses occurred in microbial-derived carbon, the main stabilizer in subsoils. In absolute terms, disturbances removed an average of ~5.6 Mg C ha⁻¹ of SOC across the 0–150 cm profile, with losses reaching ~9 Mg C ha⁻¹ under autumn harvest and irrigation treatments. If extrapolated across the hyperarid margins of the Taklimakan (~100,000 km² of shrubland), this could translate into regional SOC losses on the order of 0.9–1.2 Pg C, suggesting potentially important implications for the global carbon cycle.

Not all disturbances impacted SOC equally. Autumn harvest and irrigation caused the greatest losses (~20%–21%), far exceeding spring harvest and fire. Autumn harvest likely removes biomass during a critical phase, reducing surface litter and fine root inputs needed to replenish POC. Irrigation introduced large water pulses, increasing leaching of dissolved organic carbon and promoting microbial conditions favoring rapid turnover over stabilization. Fire surprisingly had smaller net effects on SOC than in many forest biomes. In this hyperarid system, autumn burning left root systems mostly intact and sometimes enhanced post-fire germination. Ash inputs might contribute nutrients and undecomposed fragments, though benefits depend on context; in windy deserts, ash can be lost, and fire may increase erosion risk.

A key finding was the contrasting role of plant- and microbial-derived carbon by depth. Surface SOC variation correlated mainly with POC and plant carbon, while in the deepest layer (100–150 cm), microbial-derived carbon dominated. Disturbances reduced microbial carbon by 16.2%, linked to decreases in exchangeable calcium and noncrystalline iron/aluminum oxides, minerals critical for organic matter protection, and shifts in microbial communities. Fast-growing r-strategy bacteria (e.g., Actinobacteria, Proteobacteria) increased, while residue-rich fungi declined. Since fungal residues were almost four times more abundant than bacterial residues, this shift threatens long-term carbon stabilization.

The depletion of MAOC is concerning, as MAOC represents a slow-cycling, centuries-persistent carbon fraction; its loss could reduce soil's long-term carbon storage capacity. In low-productivity, hyperarid ecosystems, such losses may be difficult to reverse within feasible management timescales. The shift from MAOC to POC dominance suggests soils become more vulnerable to future losses, potentially triggering a self-reinforcing degradation cycle. Loss of vegetation exposes soils to wind erosion, reduces seed banks, and hampers recovery, accelerating SOC depletion. Methodologically, the study's 16-year duration, deep soil sampling, and integration of biochemical, mineralogical, and microbial data provide unprecedented mechanistic insights. Structural equation modeling revealed distinct disturbance pathways affecting SOC at different depths: direct litter removal and POC loss at the surface, and indirect impacts via mineral protection and microbial residues below. Such depth-specific understanding is critical for improving Earth system models' representation of SOC dynamics under disturbance.

However, some limitations remain. The experiment held disturbance intensity constant, preventing identification of threshold effects where SOC loss accelerates. Plots cover a small fraction of the desert–oasis interface, limiting broad extrapolation. The irrigation treatment simulated flood events but may not capture variability in natural floods. Combined or sequential disturbances were not tested, though these often co-occur in real landscapes and may interact in complex ways. These limitations mean that extrapolating results to larger regions must be done with caution: real-world disturbances vary in timing, frequency, and intensity, and interactions among them may amplify or dampen carbon loss. Thus, while the study provides invaluable mechanistic insight, the absolute magnitudes of loss at broader scales should be interpreted as indicative rather than predictive.

The implications of this study for land management are clear. First, autumn harvests should be minimized in fragile desert margins to protect topsoil carbon stocks. Second, artificial flooding should be reconsidered due to the potential for SOC leaching; if irrigation is necessary, smaller and more frequent applications may reduce harm. Although fire is less damaging than expected, it still poses erosion risks and should be tightly controlled. More broadly, subsurface SOC dynamics, especially microbial-derived carbon, must be incorporated into desertification control and global carbon accounting frameworks.

Perhaps most importantly, Gao et al. (2025) highlight that chronic, low-intensity disturbances can erode centuries-old carbon stocks in some of Earth's harshest environments. In hyperarid deserts, where organic matter accumulation is extremely limited, the tolerance for further loss is very low. This challenges the assumption that low-productivity drylands are inherently resilient to carbon depletion. As climate change intensifies and human pressures on drylands grow, these long-lived carbon reservoirs risk irreversible drawdown, fueling positive feedbacks that exacerbate global warming. Protecting them requires minimizing disturbances and embedding an understanding of depth-resolved SOC mechanisms in climate and land management policies. Gao et al. (2025) make it clear that in the driest places on Earth, every disturbance can contribute to carbon loss, and recovery of these carbon stocks may take centuries.

Hua Zhang: conceptualization, writing – original draft, writing – review and editing. Ganghua Li: conceptualization, funding acquisition, project administration, supervision, writing – review and editing.

The authors declare no conflicts of interest.

This article is a Commentary on Gao et al. (2025) https://doi.org/10.1111/gcb.70423.