Paleoclimate data assimilation with CLIMBER-X: An ensemble Kalman filter for the last deglaciation.

Using the climate model CLIMBER-X, we present an efficient method for assimilating the temporal evolution of surface temperatures for the last deglaciation covering the period 22000 to 6500 years before the present. The data assimilation methodology combines the data and the underlying dynamical principles governing the climate system to provide a state estimate of the system, which is better than that which could be obtained using just the data or the model alone. In applying an ensemble Kalman filter approach, we make use of the advances in the parallel data assimilation framework (PDAF), which provides parallel data assimilation functionality with a relatively small increase in computation time. We find that the data assimilation solution depends strongly on the background evolution of the decaying ice sheets rather than the assimilated temperatures. Two different ice sheet reconstructions result in a different deglacial meltwater history, affecting the large-scale ocean circulation and, consequently, the surface temperature. We find that the influence of data assimilation is more pronounced on regional scales than on the global mean. In particular, data assimilation has a stronger effect during millennial warming and cooling phases, such as the Bølling-Allerød and Younger Dryas, especially at high latitudes with heterogeneous temperature patterns. Our approach is a step toward a comprehensive paleo-reanalysis on multi-millennial time scales, including incorporating available paleoclimate data and accounting for their uncertainties in representing regional climates.

Particle filters for high-dimensional geoscience applications: A review.

Particle filters contain the promise of fully nonlinear data assimilation. They have been applied in numerous science areas, including the geosciences, but their application to high-dimensional geoscience systems has been limited due to their inefficiency in high-dimensional systems in standard settings. However, huge progress has been made, and this limitation is disappearing fast due to recent developments in proposal densities, the use of ideas from (optimal) transportation, the use of localization and intelligent adaptive resampling strategies. Furthermore, powerful hybrids between particle filters and ensemble Kalman filters and variational methods have been developed. We present a state-of-the-art discussion of present efforts of developing particle filters for high-dimensional nonlinear geoscience state-estimation problems, with an emphasis on atmospheric and oceanic applications, including many new ideas, derivations and unifications, highlighting hidden connections, including pseudo-code, and generating a valuable tool and guide for the community. Initial experiments show that particle filters can be competitive with present-day methods for numerical weather prediction, suggesting that they will become mainstream soon.

Towards reliable Arctic sea ice prediction using multivariate data assimilation.

Rapid declines in Arctic sea ice have captured attention and pose significant challenges to a variety of stakeholders. There is a rising demand for Arctic sea ice prediction at daily to seasonal time scales, which is partly a sea ice initial condition problem. Thus, a multivariate data assimilation that integrates sea ice observations to generate realistic and skillful model initialization is needed to improve predictive skill of Arctic sea ice. Sea ice data assimilation is a relatively new research area. In this review paper, we focus on two challenges for implementing multivariate data assimilation systems for sea ice forecast. First, to address the challenge of limited spatiotemporal coverage and large uncertainties of observations, we discuss sea ice parameters derived from satellite remote sensing that (1) have been utilized for improved model initialization, including concentration, thickness and drift, and (2) are currently under development with the potential for enhancing the predictability of Arctic sea ice, including melt ponds and sea ice leads. Second, to strive to generate the "best" estimate of sea ice initial conditions by combining model simulations/forecasts and observations, we review capabilities and limitations of different data assimilation techniques that have been developed and used to assimilate observed sea ice parameters in dynamical models.