Data assimilation refers to the combination of disparate observations to provide a comprehensive and internally consistent data set that describes how a system is changing over time. Improvements in data assimilation systems have led directly to substantial improvements in numerical weather prediction over the past several decades by improving the realism of the initial conditions used to run weather forecast models. Improved data assimilation techniques have also led to improved data sets for analyses of climate change.
Climate data records (see NRC, 2004a) are generated by a systematic and ongoing process of climate data integration and reprocessing. Often referred to as reanalysis, the fundamental idea behind such efforts (see, e.g., Kalnay et al., 1996) is to use data assimilation methods to capitalize on the wealth of disparate historical observations and integrate them with newer observations, such as space-based data. Data assimilation, analysis, and reanalysis are also becoming increasingly important for areas other than regional and global atmospheric models, such as ocean models, land models, marine ecosystems, cryosphere models, and atmospheric chemistry models.
Improvements have occurred in all components of data assimilation and reanalysis, including data assimilation models, the quality and quantity of the observations, and methods for statistical interpolation (see, e.g., Daley, 1991; Kalnay, 2002). However, additional advances are needed. For example, data for the ocean, atmosphere, and land are typically assimilated separately in different models and frameworks. Given that these systems are intrinsically coupled on climate time scales, for instance through exchanges of water and energy, coupled data assimilation methodologies are needed to take into account their interactions. Next-generation data assimilation and reanaly-
sis systems should aim to fully incorporate all aspects of the Earth system (and, eventually, human systems) to support integrated understanding and facilitate analyses of coupled human-environment systems.
Finally, and critically, all observing systems and data analysis activities depend on effective data management—including data archiving, stewardship, and access systems. Historically, support for data management has often lagged behind support for initial data collection (NRC, 2007d). As the demand for sustained climate observations is realized and actions are taken to improve, extend, and coordinate observations, there will be an increase in the demands on both technology and human capacity to ensure that the resulting data are securely archived, quality controlled, and made available to a wide range of users (Baker et al., 2007; NRC, 2004a, 2005e, 2007d). Likewise, as data volume and diversity expand new computational approaches as well as greater computing power will be needed to process and integrate the different data sets on a schedule useful for planning responses to climate change. Finally, because some data have the potential for violating personal privacy norms and legal guarantees, proper safeguards must be in place to protect confidentiality.
Was this article helpful?