| 23 Jan 2026
A Preliminary Study on a Synergistic Assimilation Scheme for Multi-band Satellite Soil Moisture Data
Abstract. Soil moisture retrievals from satellite sensors operating at different microwave frequencies provide diverse and complementary data sources for assimilation. However, fully exploiting the advantages of each frequency band while increasing the volume of assimilated observations remains a challenge. This study assimilates soil-moisture retrievals from three dominant-frequency instruments—SMAP (Soil Moisture Active–Passive), ASCAT (Advanced Scatterometer), and MWRI (Microwave Radiation Imager)—into the Common Land Model (CoLM) via the Simplified Extended Kalman Filter (SEKF). On the basis of a systematic assessment of the disparate impacts of each single-band product, we propose and test a synergistic multi-satellite assimilation framework that optimally combines the complementary information inherent in the multi-frequency observations. Results show that assimilating soil-moisture retrievals significantly improves the accuracy of the CoLM land-surface model; nevertheless, the effectiveness of each product exhibits a pronounced dependency on vegetation type. Analyses of simultaneous multi-source assimilation indicate that, when SMAP and ASCAT products are already ingested, the additional introduction of MWRI data over low-stature vegetation further enhances the joint assimilation performance. Validation against in-situ observations across China demonstrates that the largest improvements occur in the central and western parts of the country: the domain-mean correlation coefficient rises by about 0.25, while the error declines from 0.068 to 0.058 m3m-3. This indicates that improvements from multi-sensor assimilation stem not only from increased data volume but also from the complementary characteristics of the assimilated products. These findings provide valuable insights into the design of synergistic multi-sensor land data assimilation systems and contribute to improving land surface modeling, as well as weather and climate prediction accuracy.
This paper investigates how soil moisture data assimilation can be improved by synergistically combining multi-frequency satellite observations. The authors assimilate soil moisture retrievals from SMAP (L-band), ASCAT (C-band), and FY-3D MWRI (X-band) into the Common Land Model (CoLM) using a Simplified Extended Kalman Filter (SEKF). Overall, the paper provides a meaningful contribution by shifting multi-sensor assimilation from a data-volume-drivenapproach toward a complementarity- and vegetation-aware strategy. Well fitted and written for publication.
1) The vegetation types used in Figures 5–7 are central to the paper’s conclusions. Adding a brief description of how these vegetation classes are defined (e.g., source dataset and thresholds) would improve clarity for readers not familiar with the classification scheme.
2) Some figures (e.g., Figures 5, 6, and 9) contain many panels and colors, which makes interpretation difficult. Increasing font sizes and simplifying legends would enhance readability.
3) A small number of grammatical issues remain (e.g., “improving the shill of the CoLM” in Section 3.4). A final round of language proofreading would further improve the manuscript’s presentation quality.
4) The experiments are conducted over a relatively short period (June–August 2022). While the results are convincing for summer conditions, soil moisture dynamics and microwave retrieval performance can vary substantially across seasons (e.g., frozen soils, snow cover, phenological changes). This limitation raises questions about the generalizability of the proposed vegetation-adaptive scheme. Extending the analysis to additional seasons or providing a clearer discussion of this limitation would strengthen the conclusions.
5) I strongly suggest that the authors acknowledge related studies that "assimilation of Sentinel-derived leaf area index to improve the representation of surface–groundwater interactions in irrigation districts." Citing and briefly discussing such work would strengthen the linkage between the proposed framework and existing literature, and help position the study within the broader context of land data assimilation research.