M. Rixen

Enhanced ocean temperature forecast skills through 3‐D super‐ensemble multi‐model fusion

[1]xa0An innovative multi-model fusion technique is proposed to improve short-term ocean temperature forecasts: the three-dimensional super-ensemble. In this method, a Kalman Filter is used to adjust three-dimensional model weights over a past learning period, allowing to give more importance to recent observations, and take into account spatially varying model skills. The predictive performance is evaluated against SST analyses, CTD casts and gliders tracks collected during the Ligurian Sea Cal/Val 2008 experiment. Statistical results not only show a very significant bias reduction of this multi-model forecast in comparison with the individual models, their ensemble mean and a single-weight-per-model version of the super-ensemble, but also the improvement of other pattern-related skills. In a 48-h forecast experiment, and with respect to the ensemble mean, surface and subsurface root-mean-square differences with observations are reduced by 57% and 35% respectively, making this new technique a suitable non-intrusive post-processing method for multi-model operational forecasting systems.

Uncertainty forecast from 3-D super-ensemble multi-model combination: validation and calibration

Measurements collected during the Recognized Environmental Picture 2010 experiment (REP10) in the Ligurian Sea are used to evaluate 3-D super-ensemble (3DSE) 72-hour temperature predictions and their associated uncertainty. The 3DSE reduces the total Root-Mean-Square Difference by 12 and 32% respectively with reference to the ensemble mean and the most accurate of the models when comparing to regularly distributed surface temperature data. When validating against irregularly distributed in situ observations, the 3DSE, ensemble mean and most accurate model lead to similar scores. The 3DSE temperature uncertainty estimate is obtained from the product of a posteriori model weight error covariances by an operator containing model forecast values. This uncertainty prediction is evaluated using a criterion based on the 2.5th and 97.5th percentiles of the error distribution. The 3DSE error is found to be on average underestimated during the forecast period, reflecting (i) the influence of ocean dynamics and (ii) inaccuracies in the a priori weight error correlations. A calibration of the theoretical 3DSE uncertainty is proposed for the REP10 scenario, based on a time-evolving amplification coefficient applied to the a posteriori weight error covariance matrix. This calibration allows the end-user to be confident that, on average, the true ocean state lies in the −2/+2 3DSE uncertainty range in 95% of the cases.

Quantifying, predicting, and exploiting uncertainties in marine environments

1 OverviewFollowing the scientific, technical, and field trial initiativesongoing since the Maritime Rapid Environmental Assess-ment (MREA) conferences in 2003, 2004, and 2007, theMREA10 conference provided a timely opportunity to re-view the progress on various aspects of MREA, with aparticular emphasis on marine environmental uncertaintymanagement. A key objective of the conference was toreview the present state of the art in quantifying, predicting,and exploiting marine environmental uncertainties. The in-tegration of emerging environmental monitoring and mod-eling techniques into data assimilation streams and theirsubsequent exploitation at an operational level involves acomplex chain of nonlinear uncertainty transfers, includinghuman factors. Accordingly, the themes for the MREA10conference were selected to develop a better understandingof uncertainty, from its inception in the properties beingmeasured and instrumentation employed to its eventual im-pact in the applications that rely upon environmentalinformation.Contributions from the scientific community were en-couraged on all aspects of environmental uncertainties: theirquantification, prediction, understanding, and exploitation.Contributions from operational communities, the consumersof environmental information who have to cope with uncer-tainty, were also encouraged. All temporal and spatial scaleswere relevant: tactical, operational, and strategic, includinguncertainty studies for topics with long-term implications.Manuscripts reporting new technical and theoretical devel-opments in MREA, but acknowledging effects of uncer-tainties to be accounted for in future research, were alsoincluded.The response was excellent with 87 oral presentations (11of which were invited keynote speakers) and 24 posterpresentations during the conference. A subset of these pre-sentations was submitted to this topical issue, and 22 manu-scripts were published by Ocean Dynamics. The followingsection includes an overview of the conference themes andsummary of the published manuscripts.2 Conference themes and findings2.1 Quantify: review and quantify contributions to marineenvironmental uncertaintyThis theme considered sensor and platform related uncer-tainties from: satellite remote sensing, coastal and marineradars,autonomous vehicles (gliders, AUVs), representationerrors, and calibration/validation.Pleskachevsky et al. (2011) present a new approach forbathymetry estimation from combined optical and syntheticaperture radar data covering two different depth domains.The overlapping range from 20 up to 10 m provides someperspectives on fusing the results from the two approaches.Underwater topography is derived using the shallow water