Hidden Soil Layers Rule Water and Heat Fluxes in Qinghai–Tibetan Plateau Wetlands

High on the Qinghai–Tibetan Plateau, where the headwaters of the Yellow, Yangtze, and Lancang Rivers wind through some of the highest wetlands on Earth, the ground beneath the sedge meadows is quietly running a hydraulic experiment of its own. A new study published in Hydrology and Earth System Sciences by Rong Liu of the Chinese Academy of Sciences and colleagues provides the first depth-resolved, process-based evidence that these wetland soils are not uniform sponges at all. Instead, they are vertically stratified hydraulic systems in which surface layers drain almost instantly, middle layers hold water tenaciously, and deep layers act as long-term storage reservoirs. Ignoring that stratification, the researchers show, systematically distorts how land surface models simulate the exchange of water and heat between the plateau and the atmosphere.

The team worked at the Sanjiangyuan Land Surface Process Observation Station in Maqu County, at 3436 meters elevation, where mean annual precipitation is roughly 604 millimeters and the mean air temperature is just 1.8 degrees Celsius. The region, often called the Asian Water Tower, hosts extensive alpine wetlands that regulate regional climate through evapotranspiration and influence downstream river hydrology at the basin scale. The wetland soils there form under persistently cold, near-saturated conditions, and partially decomposed Sphagnum and woody residues build a loose, porous, three-dimensional network with enormous surface area. That architecture is nothing like the tightly packed mineral soils for which most standard soil parameterizations were developed, and the discrepancy turns out to matter far more than modelers assumed.

To capture the hidden structure, the researchers collected eleven undisturbed soil cores spanning the profile from the surface down to 160 centimeters, one core per depth interval. They measured saturated hydraulic conductivity with an automated falling-head KSAT instrument, determined saturated water content gravimetrically, and quantified bulk density, soil organic carbon by potassium dichromate oxidation, fiber content by the von Post method, and particle size distribution by hydrometer. X-ray computed tomography of a surface sample visualized the pore architecture in three dimensions: a reconstruction built from roughly 3000 slices revealed a dense micropore-dominated cap in the upper 4 centimeters giving way below to increasingly frequent and partly connected macropores, with root channels and biopores threading the matrix at multiple scales.

The centerpiece of the laboratory work was a transient evaporation experiment using the HYPROP-2 system, in which two tensiometers tracked matric potential at two depths inside each intact core while an electronic balance logged mass loss during free evaporation under controlled temperature and wind speed. Applying the simplified evaporation method, which assumes linear profiles of potential, water content, and flux over short time steps, the team inverted unsaturated hydraulic conductivity via the Darcy–Buckingham law. Because HYPROP-2 covers only the near-saturated suction range, dry-end data from minus 0.1 to minus 300 megapascals came from a WP4C dew point potentiometer, and the merged datasets were fitted to the Clapp–Hornberger formulation using the shuffled complex evolution global optimizer, with parameter uncertainty assessed by Markov chain Monte Carlo.

The derived Clapp–Hornberger parameters reveal coherent hydraulic contrasts along the profile. Surface layers, from 0 to 8 centimeters, showed saturated water contents of 0.64 to 0.70 cubic meters per cubic meter, far above the typical mineral soil value of less than 0.50, together with saturated conductivities of 36.6 to 86 centimeters per day, roughly an order of magnitude higher than most mineral soils. Bulk density was a mere 0.63 to 0.81 grams per cubic centimeter and soil organic carbon reached 45 to 99 grams per kilogram, confirming a loose, macropore-rich fabric that promotes rapid drainage and evaporation. The power law exponent b, which controls the slope of the water retention curve, measured about 3.2 near the surface, peaked at 5.08 in the 10 to 20 centimeter subsurface, and declined toward 1.05 at 130 to 150 centimeters, while air entry suction followed a partly opposite pattern from minus 0.378 meters near the surface to minus 0.518 meters at depth.

Translated into physical behavior, these numbers describe a vertical regime of rapid near-surface drainage, delayed mid-layer release, and long-lasting deep moisture storage. Pore size distributions derived from the water retention curves via the Young–Laplace relation sharpened the picture: surface peaks concentrated at 10 to 100 micrometers, indicating macropore connectivity that sustains fast infiltration and preferential flow; middle-layer pores narrowed to 1 to 10 micrometers, where mesopores enhance retention in finer pores; and deep-layer peaks shifted below 1 micrometer, marking a micropore-dominated fabric with strong capillary suction but poor permeability. Two diagnostic indices reinforced the state dependence of the system. The effective exponent b_eff, computed as the local logarithmic slope of the retention curve, jumped from 2 to 5 at intermediate saturation to 15 to 20 near saturation in surface and middle layers, while the near-saturated connectivity index, the logarithm of the ratio of conductivity at 90 percent to that at 50 percent saturation, reached 2.4 to 3.9 in shallow and middle layers, confirming that macropore flow dominates transport under wet conditions.

The decisive test came from the Noah-MP land surface model, version 5.0, run in single-point mode at the station for June through August 2023 with half-hourly meteorological forcing from the eddy covariance and automatic weather station systems. The team designed a control experiment with the model’s default loamy sand parameters and six one-at-a-time sensitivity runs in which each measured Clapp–Hornberger parameter replaced its default counterpart in turn. Frozen-soil processes used the Niu and Yang supercooled liquid water parameterizations, the Jarvis stomatal conductance scheme was retained to minimize vegetation-related disturbances, and outputs were evaluated against observed soil moisture at 5 and 20 centimeters, latent heat flux, and sensible heat flux using root mean square error, bias, and coefficient of determination.

The sensitivity ranking was unambiguous: b, then air entry suction, then saturated water content, then saturated conductivity. Replacing the default exponent of 4.26 with the measured surface value of 3.215 cut the summer latent heat flux root mean square error from 9.45 to 3.87 watts per square meter during the early-morning-to-noon window, while the subsurface value of 5.076 worsened it to 11.06 watts per square meter. Sensible heat flux responded in a compensating pattern, suppressed during the day but overestimated overall by up to 24 watts per square meter, underscoring how strongly the exponent governs surface energy partitioning. Saturated conductivity, by contrast, proved nearly irrelevant under the wet summer background, shifting latent heat flux errors by less than a watt per square meter, because evapotranspiration was controlled by capillary water rather than saturated pathways. Saturated water content adjustments improved 20-centimeter soil moisture, reducing the root mean square error from 0.063 to 0.047 cubic meters per cubic meter, reflecting the role of the storage upper bound in retention.

Event-scale analysis for a wet episode on 15 to 16 July and a dry episode on 12 to 13 August 2023 exposed the structural limits of the single power law. During the wet period, when macropores were fully filled and gravity-driven drainage ruled, the low exponent scheme matched the observed latent heat flux peak far better, cutting the root mean square error from about 42 to about 28 watts per square meter, though it raised sensible heat bias to 28 watts per square meter and drained shallow soil too fast. During the dry period, when capillary and micropore retention governed, the high exponent scheme improved storage and flux accuracy but blunted the short-term evapotranspiration response after rainfall. No single exponent could simultaneously capture rapid macropore drainage and slow micropore release, producing opposite biases under wet and dry conditions. A companion test showed that swapping the Ball–Berry for the Jarvis stomatal scheme changed flux errors far less than the soil parameters did, confirming that in these sparsely vegetated wetlands, where leaf area index is below one, soil water availability rather than stomatal control dominates land–atmosphere exchange.

The authors conclude that pore-scale structure demonstrably regulates profile-scale hydrothermal responses in Qinghai–Tibetan Plateau wetlands, and that vertical hydraulic heterogeneity rather than any single-layer parameterization governs the timing and magnitude of evaporation and heat fluxes. They acknowledge limitations: the Clapp–Hornberger formulation cannot resolve macropore drainage near saturation, the inversion shifts inferred pore distributions toward medium and small pores, and the single-core-per-depth design limits quantification of local spatial variability. As a way forward, they recommend evaluating the more physical van Genuchten formulation and its Peters–Durner–Iden extension, which combines capillary and film flow pathways and achieved coefficients of determination above 0.95 in prior validations, to represent the bimodal pore networks of organic-rich wetland soils. For a plateau that supplies water to more than a billion people downstream, the message is clear: the hidden layering of its wetland soils is not a modeling detail but a first-order control on how the Asian Water Tower breathes.

Subject of Research: soil hydraulic heterogeneity

Article Title: Evidence of vertical soil hydraulic heterogeneity regulating hydrothermal simulations in Qinghai–Tibetan Plateau wetlands

Article References: Liu, R., Wang, X., Wang, Z., Pan, Y., & Song, Y. (2026). Evidence of vertical soil hydraulic heterogeneity regulating hydrothermal simulations in Qinghai–Tibetan Plateau wetlands. Hydrology and Earth System Sciences, 30(20), 6289-6313. https://doi.org/10.5194/hess-30-6289-2026

Image Credits: AI Generated

DOI: 10.5194/hess-30-6289-2026

Keywords: Qinghai–Tibetan Plateau, alpine wetlands, soil hydraulics, Clapp–Hornberger, Noah-MP, land surface modeling, evaporation experiments, water retention curve, pore size distribution, energy fluxes, Sanjiangyuan, Hydrology and Earth System Sciences