Connection Between Root Zone Soil Moisture and Surface Energy Flux Partitioning Using Modeling, Observations, and Data Assimilation for a Temperate Grassland Site in Germany

Abstract

Land surface models (LSMs) with different degrees of complexity are in use as lower boundary conditions for atmospheric models with the simpler LSMs preferentially used in numerical weather forecasting. This study evaluates the second‐generation TERRA Multi‐Layer and the third‐generation Community Land Model (CLM) to better understand the connection between root zone soil moisture and surface energy fluxes, which is important for predictions. Both LSMs were compared in multiyear, observation‐driven simulations at the Falkenberg grassland site (Germany), and their results were compared to observations. With their default settings for the site, both LSMs tend to overestimate the Bowen ratio, while CLM additionally exhibited a wet bias and a too low soil moisture variance. With modified photosynthetic parameters in CLM, the Bowen ratio improved considerably, but the soil moisture bias and its too low variance remained. Joint data assimilation with soil parameter update significantly improved the soil moisture variance but degraded the Bowen ratio. We could identify the default shallow root fraction distribution to be responsible for the overestimated Bowen ratio, which could be largely reduced by increasing the root fractions in deeper layers. This study demonstrates how observations and data assimilation with joint state‐parameter updating can be used to improve the realism of third‐generation LSMs and thus our understanding of the connection between root zone soil moisture and surface energy flux partitioning.

Modeled Land Atmosphere Coupling Response to Soil Moisture Changes with Different Generations of Land Surface Models

“…For shallow drier regimes, only with additional modified deeper root fraction distribution, the TSMP response tends to converge with COSMO results. In CLM, the static root fraction distribution needs to be modified to bring larger portion of root fractions to available water in deeper layers [9]. Besides, the permanent wilting point in TERRA-ML depends on the soil type, and is slightly lower than in CLM with respect to the relative soil moisture content; hence the shift between the model responses in CSDR experiments.…”

“…In the following sections, we briefly describe the atmospheric model COSMO and the two LSMs. A more detailed discussion of the differences between the two LSMs is also available in [9].…”

“…The initialized relative soil moisture content (S w ) represents transitions of soil states from very dry ( S w ∼ 0.2) to wet regimes ( S w ∼ 0.6). Additional experiments are designed to illustrate the impact of relatively shallow dry regimes (S w,z≤30 cm < 0.5), typically observed during recession periods (e.g., [9]) on the model response, including the sensitivity to the vertical profile of the root distribution in CLM, as discussed above.…”

“…The root depth was set to the default 0.6 m for TERRA-ML. For CLM, the default and a modified [9] root distribution was used. The default vertical profile of root distribution contains 80% of the root fraction within the top 50 cm of the soil depth, while the modified root distribution contains 85% of the root fractions within the top 150 cm of the soil.…”

“…Additionally, the changes in the model response when using different LSMs coupled with the atmospheric model in real-world data settings, becomes less tractable due to the complexity in the spatial patterns of soil-vegetation-atmosphere states and the resulting two-way interactions. Recently, a second (TERRA-Multi Layer (TERRA-ML), [9][10][11][12]) and a third-generation land surface model (Community Land Model (CLM) version 3.5, [13]) were evaluated with multi-year observation driven simulations at the Falkenberg grassland site in Germany. The study showed that both models tend to overestimate the Bowen ratio, while CLM exhibited a wet bias.…”