Data assimilation for sea-ice models on adaptive non-conservative meshes
Abstract
The issue of estimating the state of a system based on a large, but limited, set of information is ubiquitous in science: the methods designed for this issue are the central theme of nonlinear filtering and optimal control theory. Fields ofapplication include applied physics, navigation, telecommunications, genetics, engineering, and every situation where the aim is to extract as much information as possible about a natural or laboratory-scale phenomenon from partial andscattered initial knowledge. These conditions are typically met in geosciences, where state estimation methods, referred to as data assimilation (DA), have experienced a flourishing trend of improvement. Nowadays, in addition tothe increase in numerical model resolution and of the desired prediction time horizon, we are also experiencing a dramatic growth and refinement in the supply of observations. The Earth is now observed over a wide range of spatialand temporal scales, thanks to an increasingly wider variety of sustained observing systems, which include satellites, but also new, non-stationary, oceanmeasurements such as floats, drifters and more recently gliders. The assimilation of data derived from instruments that are not stationary has come to be known as Lagrangian DA (LaDA). These methods have gained popularity,particularly in oceanography, as they offer a suitable way to incorporate measurements from floats, drifters and gliders.More recently, the Lagrangian dimension of the DA problem has also involved the model component,and not just the data, with the appearance of numerical models discretized on a spatio-temporal varying mesh, a feature that has shown to enhance the accuracy of the representation of a wide variety of phenomena, usually characterized by a higher degree of variability and of physical complexity. The use of an adaptive mesh presents a great methodological challenge to DA as no longer do just the values of the physical variables at the grid points need updating, but also their positions. Furthermore, some of the models using adaptive meshes, in particular those of sea ice movement, are particularly interesting in that they use a remeshing process to remove and insert mesh points at various points in their evolution. This helps to modulate the degree of the resolution where it is more necessary in the course of the simulation. Nevertheless, the non-conservative character of the mesh presents another key challenge in developing compatible DA schemes, as the dimension of the state space we wish to estimate can change over time due to the remeshing. Standard state-of-the-art DA methods cannot work in such a framework and a drastic re-thinking of the formulation of the DA problem is required. The objective of this project is to study and develop novel approaches to carry out DA with models based on an adaptive non-conservative mesh. The ultimate goal is to successfully apply the new methods for DA to the state-of-the-art Lagrangian sea-ice model neXtSIM developed in-house at NERSC so as to contribute to enhance prediction skill in such a strategic area as the Arctic sea. DASIM II comes as the natural step forward after the 1-year long, ONR-funded, project DASIM (see Section 5 in the project description) where the sensitivity analysis of neXtSIM to external forcing, its probabilistic forecasting, skill and capability have been developed for search and rescue operations on the sea-ice. It is an interdisciplinary project between geoscientists and mathematicians whose main objective is to contribute to the birth of novel DA and uncertainty quantification methods for Lagrangian sea-icemodels, but the impact will go beyond the sea-ice prediction enterprise to other areas where DA is applied in conjunction with models using adaptive non-conservative grids. The project is a partnership between NERSC (Norway)and RENCI-UNC (USA), and brings together leading scientists in mathematics, dynamical systems, DA, climate science and sea-ice modeling. NER
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 26, 2018
- Source ID
- N000141812493
Entities
People
- Alberto Carrassi
Organizations
- Nansen Environmental and Remote Sensing Center
- Office of Naval Research
- United States Navy