Regional and Global Land Data Assimilation Systems: Innovations, Challenges, and Prospects

TLDR

Since 2004, the North American and Global Land Data Assimilation Systems have advanced markedly, and LDASs are now widely developed worldwide. This paper reviews the development, current status, applications, challenges, and future prospects of regional and global LDASs. The authors describe the evolution and innovations of various LDASs, evaluate and validate their performance for applications from numerical prediction to water‑resource management, and detail challenges related to data quality, model physics, assimilation, and scale mismatch. They highlight prospects such as land‑information‑system software, a unified global LDAS with nesting and hyper‑resolution, and uncertainty estimates for model structure, parameters, and forcing.

Abstract

Since the North American and Global Land Data Assimilation Systems (NLDAS and GLDAS) were established in 2004, significant progress has been made in development of regional and global LDASs. National, regional, project-based, and global LDASs are widely developed across the world. This paper summarizes and overviews the development, current status, applications, challenges, and future prospects of these LDASs. We first introduce various regional and global LDASs including their development history and innovations, and then discuss the evaluation, validation, and applications (from numerical model prediction to water resources management) of these LDASs. More importantly, we document in detail some specific challenges that the LDASs are facing: quality of the in-situ observations, satellite retrievals, reanalysis data, surface meteorological forcing data, and soil and vegetation databases; land surface model physical process treatment and parameter calibration; land data assimilation difficulties; and spatial scale incompatibility problems. Finally, some prospects such as the use of land information system software, the unified global LDAS system with nesting concept and hyper-resolution, and uncertainty estimates for model structure, parameters, and forcing are discussed.