Multivariate and multiscale data assimilation in terrestrial systems: a review.

Area of Science:

• Environmental Science
• Earth System Science
• Geosciences

Background:

• Terrestrial observational networks are expanding, providing high-resolution time-series data crucial for understanding climate, hydrological, and land-use changes.
• Increasing accuracy and availability of spaceborne sensor data, alongside automated terrestrial observations, enable advanced data assimilation (DA) for improved Earth system modeling and forecasting.

Purpose of the Study:

• To review the state-of-the-art in data assimilation (DA) for terrestrial systems, with a focus on integrating observational data from different spatial scales and types.
• To introduce and categorize major DA approaches, including univariate single-scale (UVSS), univariate multiscale (UVMS), multivariate single-scale (MVSS), and combined multivariate multiscale (MVMS).

Main Methods:

• The review introduces various DA techniques, such as the Ensemble Kalman Filter (EnKF), Particle Filter (PF), and variational methods (3/4D-VAR).
• It categorizes DA approaches based on data scale (single vs. multiscale) and data type (univariate vs. multivariate).

Main Results:

• The majority of current DA applications utilize univariate single-scale (UVSS) methods.
• The paper distinguishes four key DA approaches: UVSS, UVMS, MVSS, and MVMS, highlighting the increasing complexity and potential of integrating diverse data.

Conclusions:

• Existing methods can update multiple model states and parameters, but advancements are needed in understanding measurement errors and biases for different data types.
• Improved multiscale assimilation for nonlinearly scaling data, enhanced cross-validation, and ground truth verification are crucial for effective multiscale multivariate data assimilation.