Description of task
This task focuses on quantifying predictability and prediction capabilities with the help of ensemble runs and data assimilation and the associated development and improvement.
Improved prognostic models from the first activity of this task and specific data assimilation systems tailored for production of analyzed 4D datasets under the third activity of this theme (NAOSIMDAS, TOPAZ) are used here as input. Output is given to impact activities in the form of quantified predictability and prediction capability, as well as to activity 3 in the form of an improved NWP model (HIRLAM/HIROMB) to be utilized for a reanalysis.
- The improved predictive skill of coupled arctic climate models (as a result of activity 1 and based on the current integrated EU project GLIMPSE) will be explored in decadal scale hindcast runs based on ERA forcing at the lateral boundaries, including the DAMOCLES period. These runs will be carried out in an ensemble setup with two model systems, RCAO (SMHI/RC) and ORCM (met.no), and slightly varying boundary forcing. Predictability of the regional arctic system will be estimated in terms of inter-ensemble variability analysis. Such statistics represent an approach to the limit of natural inherent arctic predictability on seasonal to climate timescales. The quantification of arctic coupled predictability will allow conclusions to be drawn about the relative importance of regional processes compared to large-scale processes. Furthermore, these results support the identification of important arctic processes to be built into global models. In addition, causal relationships between teleconnected quantities with multi-annual time delays will be explored together with their uncertainties. Relationships such as between the subpolar wind stress curl and transports in the Nordic Sea support the DAMOCLES objective to improve prediction capabilities. Additional benefit will be drawn by an analysis of storm track shifts in response to varying ice cover (SMHI/RC).
- DAMOCLES data products and improved parameterization schemes from core themes 1,2 and 3 (see table 4.1) will be used to improve prediction capabilities of NWP tools by developing data assimilation procedures for temperature, humidity and sea ice and applying improved parameterizations (from activity 1 and the atmospheric core theme 2) in the operational models of atmosphere (HIRLAM) and ocean (HIROMB). This should lead to better boundary and start conditions for NWP. The NWP groups of met.no and SMHI will carry out this activity jointly. The SMHI/NWP group will focus on sea ice assimilation into the oceanic model HIROMB, while the met.no/NWP group will focus on improved assimilation of temperature-related radiances from space born microwave instruments into HIRLAM and the development of better forward modelling capabilities, i.e. utilizing direct satellite observations in combination with physical relations. Satellite microwave temperature channels are probably the single most important source of information available on the Arctic atmosphere. One of the main challenges to the better use of these data over the Arctic is understanding the surface contribution to the satellite signal, in particular the microwave emissivities of sea ice. This in turn is related to the recent history of atmospheric conditions at the location. Thus, there is a close integration between use of the satellite data for atmospheric retrievals and the research on ice property retrieval (core theme 1) and atmospheric influence on the radiances (atmospheric core theme 2). The basis for this type of integrative work has been established in the present EU project IOMASA. The improvement of NWP models such as the HIRLAM/HIROMB system serves the long-term DAMOCLES purpose of a better monitoring and observation system by providing better weather forecasts for the Arctic and Europe. This in turn directly enables an improved reanalysis capability, which will be used in overarching activity 1 (production of reanalyses) and will be available in the future. Reanalysis products are very useful for climate model validation not only during the DAMOCLES period. In addition, improved parameterizations for NWP models support the development of climate model parameterization. Often, parameterizations are identical.
- The impact of assimilating in-situ data from a Fram Strait array and acoustic tomography (ocean core theme 3) will be assessed in twin experiments (with and without assimilation) using synthetic acoustic measurements. For that purpose, the Fram Strait model system will be gradually updated to include assimilation of ice concentration and ice drift using schemes developed under the MERSEA IP. Furthermore, the assimilation techniques (EnKF and EnOI) of the NERSC operational data and model system will be refined to incorporate data from acoustic thermometry - and the drifting Arctic tomography array. Synthetic acoustic data will be created by feeding ocean states from ice-ocean model to acoustic propagation models/inversion modules. Assimilation of synthetic data will be tested both for the high resolution Fram Strait model and the operational TOPAZ system for synthetic data. When real acoustic data start to be delivered after August 2007, the refined assimilation techniques and schemes will be used to combine ice-ocean model output, acoustic data and data from sub-surface moorings to provide an improved representation of oceanographic fields across the Fram Strait. This will provide input to the ocean theme 3 for further acoustic inversions and for calculation of flux estimates in activity 3.
- Improved predictive skill of the ocean data assimilation system NAOSIMDAS (developed with a focus on data enhancement in activity 3) will be quantified by hindcast experiments. The model will be applied to an observational period, split in two parts. Only the data from the first period are assimilated, and the initial condition and the boundary condition for the first period simulation are optimised. Starting from there, a prediction for the second period is performed. The observations from the second period are used to quantify the skill of the prediction. This skill is compared to the skill of a control run, in which no assimilation has taken place.
Methods for enhancement of observational data sets by data assimilation and analysis
Observations from core themes 1-3 are combined with regional NWP (HIRLAM/HIROMB) and ocean-sea ice climate (MI-IM/MIPOM, TOPAZ or Fram Strait model, NAOSIM) models using different data assimilation techniques (OI, SEIK, EnOI or EnKF, 3D Var, 4D Var). The goal is to allow the production of sets of consistent atmospheric, oceanic and sea ice fields suitable for domain-wide investigation of key parameters and processes, and for improved climate model validation. Impact studies (impact themes) build on these fields. Prototype applications for DAMOCLES data are developed and tested in this theme, while the final production, utilizing all available data, is carried out as overarching activity 1. The individual activities are:
The operational coupled ocean-sea ice forecast model MI-IM/MIPOM is embedded in a SEIK (singular evolutive interpolated Kalman filter) assimilation scheme. The scheme uses OSI-SAF ice concentration and hydrographical observations and delivers analysed fields of ice concentration and thickness. Observations of ice thickness made available during the project are used to validate the model and to further develop the assimilation scheme so as to include such data.
Temperature/sound speed profiles or direct acoustic measurements from the Fram Strait array will be assimilated close to real time. The model and the assimilation method depend on the insights from activity 2: Either ensemble optimal interpolation (EnOI) or the ensemble Kalman filter (EnKF) will be applied to the higher resolution Fram Strait model or to the coarser TOPAZ. Independent data from the different DAMOCLES moorings and drifting arrays are used for validation. The analysed ocean state will be used as input to the task of estimating algorithms of heat-, mass and fresh water fluxes in the Fram Strait which are required by ocean core theme 3.
The feasibility of a regional Arctic atmosphere-ocean-sea ice analysis is demonstrated with a combined optimal interpolation (OI)/3D variational assimilation (3DVar) scheme for the coupled modelling system HIRLAM/HIROMB used for operational NWP. The combination OI/4D variational assimilation (4DVar) will be evaluated on short assimilation periods.
Building on the adjoint ocean-sea ice model ADNAOSIM generated in this themes’ activity 1, a prototype of a 4D variational data assimilation system will be set up. New observations from core themes 1-3 will be assimilated into the model. 4DVar will be applied, as the approach is best-suited to assimilate different kinds of observations and permits the assimilation of observations that are unevenly distributed in space and time (such as glider profiles) thus avoiding the need to interpolate and extrapolate observations. The approach achieves the fit to the data by varying the model's initial conditions as well as its lateral and atmospheric boundary conditions and thus guarantees full consistency with the dynamics of the underlying model.
