Latent Space Data Assimilation of GLDAS Soil Moisture and Soil Temperature into E3SM Coupled Land-Atmosphere Simulations

Soil moisture (SM) and soil temperature (ST) have strong memory and influence surface-energy, water, and carbon fluxes. However, assimilating them into Earth system models remains challenging because they are high-dimensional, climatologically inconsistent across products and models, and partially observed. Here we develop and evaluate a latent space data assimilation (DA) framework for assimilating GLDAS-informed multilayer, monthly SM and ST anomalies into Energy Exascale Earth System Model (E3SM) land model (ELM) simulations. The framework uses an autoencoder to represent ELM states in latent space, an observation-to-latent network to map noisy, masked GLDAS-derived constraints into the same space, and an element-wise latent Kalman update to combine background and GLDAS-derived constraints before decoding the analysis back to physical states. We evaluate the method using synthetic validation, offline DA experiments, and Atmospheric Model Intercomparison Project simulations with online DA coupling. GLDAS comparisons assess assimilation consistency, while independent runoff, snow water equivalent, precipitation and near-surface air temperature products diagnose downstream impacts on unassimilated variables. Results show that latent space DA efficiently adjusts ELM SM and ST anomalies toward GLDAS-informed wet/dry and warm/cold departures while allowing subsequent evolution to remain governed by ELM physics. Independent responses are regionally and seasonally heterogeneous, indicating that land-state updates can influence unassimilated hydrological processes, precipitation and near-surface temperature, but do not guarantee uniform improvement across downstream variables. These results highlight the need to evaluate both land state consistency and downstream coupled impacts. The framework provides a scalable and computationally efficient pathway for high-dimensional land state updates in Earth system models.

Authors

Institutions

Publication Details

Published
2026-11-18
DOI
https://doi.org/10.5194/egusphere-2026-5703
Primary Topic
Soil Moisture and Remote Sensing
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Latent Space Data Assimilation of GLDAS Soil Moisture and Soil Temperature into E3SM Coupled Land-Atmosphere Simulations

Yilin Fang, Lingcheng Li, Gautam Bisht, Shixuan Zhang et al.
Soil Moisture and Remote Sensing
preprint

Latent Space Data Assimilation of GLDAS Soil Moisture and Soil Temperature into E3SM Coupled Land-Atmosphere Simulations

Yilin Fang, Lingcheng Li, Gautam Bisht, Shixuan Zhang, Hoang Tran, Lai-Yung Ruby Leung
preprint en

Abstract

Soil moisture (SM) and soil temperature (ST) have strong memory and influence surface-energy, water, and carbon fluxes. However, assimilating them into Earth system models remains challenging because they are high-dimensional, climatologically inconsistent across products and models, and partially observed. Here we develop and evaluate a latent space data assimilation (DA) framework for assimilating GLDAS-informed multilayer, monthly SM and ST anomalies into Energy Exascale Earth System Model (E3SM) land model (ELM) simulations. The framework uses an autoencoder to represent ELM states in latent space, an observation-to-latent network to map noisy, masked GLDAS-derived constraints into the same space, and an element-wise latent Kalman update to combine background and GLDAS-derived constraints before decoding the analysis back to physical states. We evaluate the method using synthetic validation, offline DA experiments, and Atmospheric Model Intercomparison Project simulations with online DA coupling. GLDAS comparisons assess assimilation consistency, while independent runoff, snow water equivalent, precipitation and near-surface air temperature products diagnose downstream impacts on unassimilated variables. Results show that latent space DA efficiently adjusts ELM SM and ST anomalies toward GLDAS-informed wet/dry and warm/cold departures while allowing subsequent evolution to remain governed by ELM physics. Independent responses are regionally and seasonally heterogeneous, indicating that land-state updates can influence unassimilated hydrological processes, precipitation and near-surface temperature, but do not guarantee uniform improvement across downstream variables. These results highlight the need to evaluate both land state consistency and downstream coupled impacts. The framework provides a scalable and computationally efficient pathway for high-dimensional land state updates in Earth system models.

Pacific Northwest National Laboratory (US)
U.S. Department of Energy, Battelle, Office of Science, Biological and Environmental Research, Pacific Northwest National Laboratory
Climate action
Soil Moisture and Remote Sensing
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon