Li et al. (2025) A new approach for joint assimilation of cosmic-ray neutron soil moisture and groundwater level data into an integrated terrestrial model

Identification

• Journal: Hydrology and earth system sciences
• Year: 2025
• Date: 2025-11-18
• Authors: Fang Li, Heye Bogena, Johannes Keller, Bagher Bayat, Rahul Raj, Harrie‐Jan Hendricks Franssen
• DOI: 10.5194/hess-29-6419-2025

Research Groups

• Agrosphere Institute, IBG-3, Forschungszentrum Jülich GmbH, Jülich, Germany
• Centre for High-Performance Scientific Computing in Terrestrial Systems: HPSC TerrSys, Geoverbund ABC/J, Jülich, Germany
• College of Geographical Science, Inner Mongolia Normal University, Hohhot, China
• Center for Remote Sensing and GIS Research, Faculty of Earth Sciences, Shahid Beheshti University, Tehran, Iran

Short Summary

This study developed a novel weakly coupled multivariate data assimilation framework to jointly assimilate cosmic-ray neutron soil moisture and groundwater level data into an integrated terrestrial model (TSMP). The new approach significantly improved the accuracy of both soil moisture and groundwater level predictions, outperforming single-variable assimilation and conventional fully coupled methods.

Objective

• To evaluate the effectiveness of simultaneously assimilating Cosmic-Ray Neutron Sensor (CRNS)-based soil moisture (SM) and groundwater level (GWL) observations using a new multivariate data assimilation (DA) method.
• To compare assimilation performance across different multivariate DA strategies.
• To demonstrate the advantages of the proposed approach over conventional single-variable assimilation in improving SM, GWL, and evapotranspiration (ET) predictions.

Study Configuration

• Spatial Scale: Rur catchment, western Germany, approximately 2354 square kilometers. Model discretized at 500 meter x 500 meter horizontal resolution and 100 meter vertical extent with 25 layers.
• Temporal Scale: Data assimilation experiments were conducted over a three-year period, from 1 January 2016 to 31 December 2018. Soil moisture was updated daily, and groundwater level was updated weekly.

Methodology and Data

• Models used:
  • Terrestrial System Modeling Platform (TSMP), a fully coupled land-energy-hydrology model.
  • Community Land Model (CLM, version 3.5) for terrestrial surface dynamics.
  • ParFlow for three-dimensional saturated-unsaturated subsurface flow.
  • Coupling via Ocean Atmosphere Sea Ice Soil Model Coupling Toolkit (OASIS-MCT).
  • Data assimilation implemented using the Parallel Data Assimilation Framework (PDAF) with the Localized Ensemble Kalman Filter (LEnKF).
• Data sources:
  • Atmospheric Forcing: COSMO-REA6 reanalysis dataset (approximately 6 kilometer spatial resolution, hourly temporal frequency).
  • Terrestrial and Subsurface Data: Shuttle Radar Topography Mission (SRTM) version 4 (90 meter resolution) for digital terrain; Sentinel-2 imagery for land cover; Sentinel-2 Level 2 Prototype Processor (SL2P) for monthly Leaf Area Index (LAI); BK50 soil map (1:50,000) for soil texture; European Soil Database for bulk density; HK100 subsurface geology map (1:100,000) for aquifer hydraulic conductivity.
  • Field Measurements:
    • Soil Moisture (SM): 12 Cosmic-Ray Neutron Sensor (CRNS) sites (TERENO framework, COSMOS-Europe project).
    • Groundwater Level (GWL): 78 observation wells for assimilation and 465 for independent validation (Geoportal NRW platform).
    • Evapotranspiration (ET): Flux measurements from three eddy covariance monitoring sites (Selhausen, Rollesbroich, Wüstebach) (TERENO infrastructure).

Main Results

• Univariate assimilation of SM reduced the unbiased root mean square error (ubRMSE) for SM by over 45% at monitored sites, with limited impact on ET (less than 3% reduction) and a minor negative effect on GWL (average 3.87% increase in ubRMSE).
• Univariate assimilation of GWL reduced GWL ubRMSE by approximately 60% at assimilation sites, with improvements extending up to 5 kilometers from assimilation points (ubRMSE reductions between 2% and 50%). However, it had a negative impact on SM (average 0% to 25% increase in ubRMSE) and ET.
• Joint state and parameter estimation (e.g., hydraulic conductivity, log Ks) consistently outperformed state-only assimilation, further reducing ubRMSE by up to 17% for GWL.
• The novel weakly coupled multivariate DA framework (WCDAr_PAR), which uses separate update modules and distinct localization radii (5 km for GWL, 100 km for SM) and asynchronous updates, achieved the best overall performance. It reduced GWL ubRMSE by nearly 75% at assimilation sites (from 7.23 meters to 2.05 meters) and SM ubRMSE by 50%.
• Conventional fully coupled multivariate DA approaches were less efficient and did not provide additional benefits beyond single-variable assimilation.
• Improvements in ET estimation were observed only when SM was included in the assimilation process (univariate or multivariate), with the positive impact remaining comparable across these scenarios.
• Independent validations using the revised hydraulic conductivity (Ks) from parameter-updating experiments confirmed enhanced predictions of both GWL and SM, demonstrating the value of correcting systematic model biases.

Contributions

• This study represents the first attempt to simultaneously assimilate in-situ Cosmic-Ray Neutron Sensor (CRNS) soil moisture and observed groundwater level (GWL) data within the integrated Terrestrial System Modeling Platform (TSMP) at the catchment scale.
• A novel weakly coupled multivariate data assimilation (DA) framework was developed, allowing for independent assimilation of SM and GWL through separate modules with variable-specific localization radii and asynchronous updates, enhancing update stability and robustness.
• The new approach successfully integrates the strengths of individual univariate assimilation models, leading to more accurate and balanced predictions of SM, GWL, and evapotranspiration (ET) compared to both single-variable and conventional fully coupled DA methods.
• The research highlights the importance of including parameter estimation (specifically hydraulic conductivity, Ks) alongside state updating in multivariate assimilation frameworks to achieve more reliable hydrologic predictions and correct systematic model biases.

Funding

• China Scholarship Council (CSC) (grant number 201904910448)
• "Light of West China" Program of the Chinese Academy of Sciences (CAS) (Project No. xbzglzb2022020)
• Interinstitute Youth Joint Fund project of the Lanzhou Branch (CAS) (Project No. E4400404)
• DETECT project (SFB 1502/1-2022, grant number 450058266)
• CosmicSense project (FOR 2694, grant number 357874777)
• Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)
• Forschungszentrum Jülich (covered article processing charges)
• Helmholtz Association (TERENO program)

Citation

@article{Li2025new,
  author = {Li, Fang and Bogena, Heye and Keller, Johannes and Bayat, Bagher and Raj, Rahul and Franssen, Harrie‐Jan Hendricks},
  title = {A new approach for joint assimilation of cosmic-ray neutron soil moisture and groundwater level data into an integrated terrestrial model},
  journal = {Hydrology and earth system sciences},
  year = {2025},
  doi = {10.5194/hess-29-6419-2025},
  url = {https://doi.org/10.5194/hess-29-6419-2025}
}