LEMONTREE May Science Meeting : Soil Moisture

Every few months, the LEMONTREE team gathers virtually to discuss a specific research theme. In February, we heard from Prof. Pierre Gentine (Columbia University), who presented on how machine learning can merge soil moisture datasets more effectively. His approach aligns statistical distributions across different sensors, eliminating the need for additional bias correction. The result: more consistent and reliable long-term soil moisture data for trend and seasonal analysis. (Catch up on that talk here: Soil Moisture: Challenges, Trends and AI). Our May meeting built on this with four talks focused on cutting-edge research in soil moisture dynamics. Here’s a recap:

From Soil Moisture to Subsurface Hydrology: Abandoning the ‘Flat Earth’ Assumption

Dr. Beni Stocker (University of Bern)

Soil moisture acts as a key constraint on transpiration and photosynthesis, influencing surface energy balance and near-surface climate. Beni explored how soil moisture variability, both vertically and laterally, plays a critical role in land–atmosphere interactions, particularly affecting photosynthesis, transpiration, and evapotranspiration (ET). Key takeaways:

Vertical variation in soil moisture

Vertical variability (soil depth) is typically captured in land surface models. During dry periods, evaporation drops quickly as topsoil dries, but deep-rooted plants continue transpiring by accessing lower soil layers. This uptake depth varies significantly by biome and is purely a physical mechanism.

Lateral variation in soil moisture

The variations across large climatic gradients are accounted for by precipitation and radiation in land surface models but the variation across smaller-scale topographic gradients in neglected in global modelling. These variations are caused by differences in subsurface water flow, topography, and subsurface structure (soil depth and bedrock lithology) and groundwater table depth. He showcased hill-slope hydrological models that simulate topography-driven lateral water movement, revealing shallow water tables in valley bottoms, which is a crucial zone for vegetation.

Using fractional ET reduction data from flux tower sites, Beni’s team have found extreme variability in ET response to cumulative water deficits: some sites showed near-total shutdown, while others were barely affected, challenging the use of fixed 1.5m root-zone depth in models for simulating ET (and therefore, by implication Gross Primary Production, GPP). Remote sensing vs. modelled soil water storage showed little correlation, further questioning fixed-depth assumptions. While GPP can be reasonably simulated using current models, ET modelling accuracy demands a more spatially nuanced understanding of root-zone water availability.

In conclusion. It’s time to move beyond “Flat Earth” assumptions and to find solutions for predicting plant rooting depth, dependent on the hydroclimate. Whether through high-resolution models or by incorporating spatial variability into root-zone water storage estimates (as seen in Francesco’s work in the next summary), it’s clear a more realistic approach is needed to account for the large variability of plant rooting depth in relation to the substrate hydrology. Promising avenues are opened, e.g., by a high-resolution and Hillslope-Resolving LSM (Noah-MP) or subgrid-representation of hillslope hydrology.

Improving ET estimation using PM equation and data-inferred root zone water storage capacity 

Francesco Grossi (University of Bern)

Francesco presented his recent work on improving evapotranspiration (ET) simulations in the P-model by incorporating more realistic representations of water availability in the root zone. Building on recent advances, he explored how different model setups affect ET estimation, particularly in water-limited environments.

The baseline approach—using precipitation as a proxy for ET—tended to overestimate ET in drier regions. Switching to the Penman-Monteith (PM) equation, a physically-based method, offered improved ET estimates, especially under drought conditions. However, the most substantial improvement came when Francesco integrated updated data on root-zone water storage capacity, drawn from Stocker et al 2023. This addition helped reduce ET overestimation by better reflecting the actual water accessible to plants during dry periods.

At the catchment scale (across the US), the refined model captured interannual ET variability reasonably well across the CAMELS dataset, though some biases persisted and overestimation, likely due to uncertainties in observed runoff data. But it does seem to capture the interannual variability.

At the global scale, the model aligned well with other benchmarks in low- and mid-latitudes, but overestimated ET in high-latitude regions. The results also suggested that using gridded meteorological inputs may reduce model performance, pointing to a need for good quality data from flux sites.

Despite some overestimation of GPP during droughts—possibly due to the lack of photosynthetic downregulation—the results highlight how a more data-informed representation of root-zone water storage can significantly enhance ET simulation, especially under climate stress.

Soil Moisture Stress and Stomatal Conductance in Optimality Theory

Dr. David Sandoval (Imperial College London)

David examined how soil moisture stress affects stomatal behaviour and water use efficiency, focussing on the strengths and limitations of the P-model and related formulations rooted in optimality theory. A key issue he raised is that while most models include some representation of soil moisture stress, such as the beta β (θ) factor, they often implement it inconsistently or in ways that fail to influence key outcomes—such as water use efficiency.

In the current P-model structure, for instance, the beta β (θ) factor designed to modulate photosynthesis and transpiration under stress end up cancelling each other out, rendering water use efficiency insensitive to soil moisture. This runs counter to observational evidence.

To address this, David introduced a semi-empirical enhancement, a compromise between empirical and mechanistic representation: modifying the parameter ‘α’, the carbon cost of water transport, in response to soil moisture. By drawing from empirical data such as isotopic partitioning, evaporative fraction, and flux tower observations, he derived semi-empirical estimates of how ‘α’ varies under different conditions.

The results showed marked improvements in simulating both daily and seasonal transpiration dynamics. Sensitivity analyses confirmed that incorporating soil moisture directly into the water transport cost improved the model’s responsiveness to both short-term (e.g. viscosity changes) and long-term environmental stress. It was also found that there is a strong thermal acclimation of the leaf-specific sapwood respiration and a weak but significant increasing of ‘α’ with aridity.

Ultimately, David’s approach provides a promising path toward more biologically grounded and empirically consistent models. By integrating soil moisture constraints through a mechanistic lens, we get closer to realistically simulating how plants regulate water loss under stress.

How Soil Moisture Influences Sub-Daily GPP in Dry Climates

Mengdi Gao (Imperial College) 

1. that plants reduce stomatal conductance at midday to conserve water, leading to a temporary dip in internal CO₂ concentration and photosynthesis;
2. that water stress reduces photosynthetic capacity (φ0) more generally, lowering GPP throughout the day.