WP1 Climate variability and change (Leader: L. Svendsen)

We will first investigate relevant physical mechanisms through analysis of the NorCPM reanalysis over the period of 1950-2010 produced for the CMIP6 DCPP (the first phase of WP1), including comparisons with atmosphere reanalyses and observations. Then the important mechanisms found in the CMIP6 NorCPM reanalysis will be revisited in the NorCPM 20th-century reanalysis produced by CoRea (the second phase of WP1), including a comparison with the 20th-century atmospheric reanalysis products (e.g. ERA-20C) and the coupled reanalysis CERA-20C. The analysis methods will include standard statistical analysis such as empirical orthogonal function analysis, regression, composite and time series analysis, and standard dynamical diagnostics such as wave activity flux analysis to identify important teleconnection patterns. Furthermore, we will study climate change since 1850 and investigate how the assimilation of oceanic data improves the sea ice state in a strongly coupled DA manner

WP2 Data assimilation for climate reanalysis (Leaders: P. Raanes and Y. Wang)

To propagate information from the contemporary observation networks backwards in time and limit discontinuities caused by changes in the observation network, we will first develop a multitimescale smoother DA method based on the EnKS. In order to separate out the different timescales before applying the EnKS, two design strategies are immediately clear: (1) using temporal averaging, similar to (Berre et al., 2010), and (2) applying empirically-determined spectral decompositions, similar to (Menetrier et al., 2015). To guide, improve, and test these theoretical developments, we will run experiments with a toy model of increasing complexity, starting with the simplest spatially-distributed model exhibiting chaos at two timescales (Lorenz, 1996), and proceeding to a reduced-order coupled ocean-atmosphere model (MAOOAM, De Cruz et al., 2016). To avoid the introduction of spurious variability in the vertical direction, several advanced localisation techniques such as scale-dependent localisation (Buehner et al., 2015), local spatial averaging (Berre et al., 2010) and linear filtering (Menetrier et al., 2015) will be tested in similarly idealised frameworks.

WP3 Implementation and reanalysis production (Leader: Y. Wang)

The multi-timescale EnKS with advanced vertical localisation will be implemented in NorCPM. We will design a twin experiment to study the time lag sensitivity, compare the multiscale EnKS with the EnKF and the traditional EnKS, and estimate the benefit of vertical localisation. We will produce a NorCPM reanalysis from 1850 to 2010 (named NorCPM-20CR) with the optimally tuned DA settings, using 30 ensemble members. The NorCPM-20CR product will be validated against independent observation datasets and compared with other ocean reanalyses (e.g. CERA-20C and ORA-20C). We will maintain a strong focus on the continuity of the NorCPM-20CR product and on the reliability of its ensemble.