Emerging Skill in Multi-Year Prediction of the Indian Ocean Dipole

The Indian Ocean Dipole is a leading phenomenon of climate variability in the tropics, which affects the global climate. However, the best lead prediction skill for the Indian Ocean Dipole, until recently, has been limited to ~6 months before the occurrence of the event. Here, we show that multi-year prediction has made considerable advancement such that, for the first time, two general circulation models have significant prediction skills for the Indian Ocean Dipole for at least 2 years after initialization. This skill is present despite ENSO having a lead prediction skill of only 1 year. Our analysis of observed/reanalyzed ocean datasets shows that the source of this multi-year predictability lies in sub-surface signals that propagate from the Southern Ocean into the Indian Ocean. Prediction skill for a prominent climate driver like the Indian Ocean Dipole has wide-ranging benefits for climate science and society.

The Regional Ice Ocean Prediction System v2: a pan-Canadian ocean analysis system using an online tidal harmonic analysis

Canada has the longest coastline in the world and includes diverse ocean environments, from the frozen waters of the Canadian Arctic Archipelago to the confluence region of Labrador and Gulf Stream waters on the east coast. There is a strong need for a pan-Canadian operational regional ocean prediction capacity covering all Canadian coastal areas in support of marine activities including emergency response, search and rescue, and safe navigation in ice-infested waters. Here we present the first pan-Canadian operational regional ocean analysis system developed as part of the Regional Ice Ocean Prediction System version 2 (RIOPSv2) running in operations at the Canadian Centre for Meteorological and Environmental Prediction (CCMEP). The RIOPSv2 domain extends from 26∘ N in the Atlantic Ocean through the Arctic Ocean to 44∘ N in the Pacific Ocean, with a model grid resolution that varies between 3 and 8 km. RIOPSv2 includes a multivariate data assimilation system based on a reduced-order extended Kalman filter together with a 3D-Var bias correction system for water mass properties. The analysis system assimilates satellite observations of sea level anomaly and sea surface temperature, as well as in situ temperature and salinity measurements. Background model error is specified in terms of seasonally varying model anomalies from a 10-year forced model integration, allowing inhomogeneous anisotropic multivariate error covariances. A novel online tidal harmonic analysis method is introduced that uses a sliding-window approach to reduce numerical costs and allow for the time-varying harmonic constants necessary in seasonally ice-infested waters. Compared to the Global Ice Ocean Prediction System (GIOPS) running at CCMEP, RIOPSv2 also includes a spatial filtering of model fields as part of the observation operator for sea surface temperature (SST). In addition to the tidal harmonic analysis, the observation operator for sea level anomaly (SLA) is also modified to remove the inverse barometer effect due to the application of atmospheric pressure forcing fields. RIOPSv2 is compared to GIOPS and shown to provide similar innovation statistics over a 3-year evaluation period. Specific improvements are found near the Gulf Stream for all model fields due to the higher model grid resolution, with smaller root mean squared (rms) innovations for RIOPSv2 of about 5 cm for SLA and 0.5 ∘C for SST. Verification against along-track satellite observations demonstrates the improved representation of mesoscale features in RIOPSv2 compared to GIOPS, with increased correlations of SLA (0.83 compared to 0.73) and reduced rms differences (12 cm compared to 14 cm). While the RIOPSv2 grid resolution is 3 times higher than GIOPS, the power spectral density of surface kinetic energy provides an indication that the effective resolution of RIOPSv2 is roughly double that of the global system (35 km compared to 66 km). Observations made as part of the Year of Polar Prediction (2017–2019) provide a rare glimpse at errors in Arctic water mass properties and show average salinity biases over the upper 500 m of 0.3–0.4 psu in the eastern Beaufort Sea in RIOPSv2.

Implementing a Parallel Version of a Variational Scheme in a Global Assimilation System at Eddy-Resolving Resolution

Recent advances in global ocean prediction systems are fostered by the needs of accurate representation of mesoscale processes. The day-by-day realistic representation of its variability is hampered by the scarcity of observations as well as the capability of assimilation systems to correct the ocean states at the same scale. This work extends a 3DVAR system designed for oceanic applications to cope with global eddy-resolving grid and dense observational datasets in a hybridly parallelized environment. The efficiency of the parallelization is assessed in terms of both scalability and accuracy. The scalability is favored by a weak-constrained formulation of the continuity requirement among the artificial boundaries implied by the domain decomposition. The formulation forces possible boundary discontinuities to be less than a prescribed error and minimizes the parallel communication relative to standard methods. In theory, the exact solution is recovered by decreasing the boundary error toward zero. In practice, it is shown that the accuracy increases until a lower bound arises, because of the presence of the mesh and the finite accuracy of the minimizer. A twin experiment has been set up to estimate the benefit of employing an eddy-resolving grid within the assimilation step, as compared with an eddy-permitting one, while keeping the eddy-resolving grid within the forecast step. It is shown that the use of a coarser grid for data assimilation does not allow an optimal exploitation of the present remote sensing observation network. A global decrease of about 15% in the error statistics is found when assimilating dense surface observations, and no significant improvement is seen for sparser observations (in situ profilers).

The Regional Ice Ocean Prediction System v2: a pan-Canadian ocean analysis system

Canada has the longest coastline in the world and includes a diversity of ocean environments, from the frozen waters of the Canadian Arctic Archipelago to the confluence region of Labrador and Gulf Stream waters on the East Coast. There is a strong need for a pan-Canadian operational regional ocean prediction capacity covering all Canadian coastal areas, in support of marine activities including emergency response, search and rescue as well as safe navigation in ice-infested waters. Here we present the first pan-Canadian operational regional ocean analysis system developed as part of the Regional Ice Ocean Prediction System version 2 (RIOPSv2) running in operations at the Canadian Centre for Meteorological and Environmental Prediction (CCMEP). The RIOPSv2 domain extends from 26° N in the Atlantic Ocean through the Arctic Ocean to 44° N in the Pacific Ocean, with a model grid-resolution that varies between 3 and 8 km. RIOPSv2 includes a multi-variate data assimilation system based on a reduced-order extended Kalman filter together with a 3DVar bias correction system for water mass properties. The analysis system assimilates satellite observations of sea level anomaly and sea surface temperature, as well as in situ temperature and salinity measurements. Background model error is specified in terms of seasonally varying model anomalies from a 10-year forced model integration allowing inhomogeneous anisotropic multi-variate error covariances. A novel online tidal harmonic analysis method is introduced that uses a sliding-window approach to reduce numerical costs and to allow time-varying harmonic constants, necessary in seasonally ice-infested waters. As compared to the Global Ice Ocean Prediction System (GIOPS) running at CCMEP, RIOPSv2 also includes a spatial filtering of model fields as part of the observation operator for sea surface temperature. In addition to the tidal harmonic analysis, the observation operator for sea level anomaly is also modified to remove the inverse barometer effect due to the application of atmospheric pressure forcing fields. RIOPSv2 is compared to GIOPS and shown to provide similar innovation statistics over a 3-year evaluation period. Specific improvements are found in the vicinity of the Gulf Stream for all model fields due to the higher model grid-resolution, with smaller root-mean-squared (RMS) innovations for RIOPSv2 of about 5 cm for SLA and 0.5 °C for SST. Verification against along-track satellite observations demonstrates the improved representation of meso-scale features in RIOPSv2 compared to GIOPS, with increased correlations of SLA (0.83 compared to 0.73) and reduced RMS differences (12 cm compared to 14 cm). While the RIOPSv2 grid resolution is 3 times higher than GIOPS, the power spectral density of surface kinetic energy provides an indication that the effective resolution of RIOPSv2 is roughly double that of the global system (35 km as compared to 66 km). Observations made as part of the Year of Polar Prediction (2017–19) provide a rare glimpse at errors in Arctic water mass properties and show salinity biases of 0.3–0.4 psu in the eastern Beaufort Sea in RIOPSv2.

An improved variational Data Assimilation method for ocean models with limited number of observations.

<p>Data assimilation (DA) is a critical component for most state-of-the-art ocean prediction systems, which optimally combines model data and observational measurements to obtain an improved estimate of the modelled variables, by minimizing a cost function. The calculation requires the knowledge of the background error covariance matrix (BECM) as a weight for the quality of the model results, and an observational error covariance matrix (OECM) which weights the observational data.</p><p>Computing the BECM would require knowing the true values of the physical variables, which is not feasible. Instead, the BECM is estimated from model results and observations by using methods like National Meteorological Centre (NMC) or the Hollingsworth and Lönnberg (1984) (H-L). These methods have some shortcomings which make them unfit in some situations, which includes being fundamentally one-dimensional and making a suboptimal use of observations.</p><p>We have produced a novel method for error estimation, using an analysis of observations minus background data (innovations), which attempts to improve on some of these shortcomings. In particular, our method better infers information from observations, requiring less data to produce statistically robust results. We do this by minimizing a linear combination of functions to fit the data using a specifically tailored inner product, referred to as an inner product analysis (IPA).</p><p>We are able to produce quality BECM estimations even in data sparse domains, with notably better results in conditions of scarce observational data. By using a sample of observations, with decreasing sample size, we show that the stability and efficiency of our method, when compared to that of the H-L approach, does not deteriorate nearly as much as the number of data points decrease. We have found that we are able to continually produce error estimates with a reduced set of data, whereas the H-L method will begin to produce spurious values for smaller samples.</p><p>Our method works very well in combination with standard tools like NEMOVar by providing the required standard deviations and length-scales ratios. We have successfully ran this in the Arabian Sea for multiple seasons and compared the results with the H-L (in optimal conditions, when plenty of data is available), spatially the methods perform equally well. When we look at the root mean square error (RMSE) we see very similar performances, with each method giving better results for some seasons and worse for others.</p>