Facing the Forecaster’s Dilemma: Reflexivity in Ocean System Forecasting

Unlike atmospheric weather forecasting, ocean forecasting is often reflexive; for many applications, the forecast and its dissemination can change the outcome, and is in this way, a part of the system. Reflexivity has implications for several ocean forecasting applications, such as fisheries management, endangered species management, toxic and invasive species management, and community science. The field of ocean system forecasting is experiencing rapid growth, and there is an opportunity to add the reflexivity dynamic to the conventional approach taken from weather forecasting. Social science has grappled with reflexivity for decades and can offer a valuable perspective. Ocean forecasting is often iterative, thus it can also offer opportunities to advance the general understanding of reflexive prediction. In this paper, we present a basic theoretical skeleton for considering iterative reflexivity in an ocean forecasting context. It is possible to explore the reflexive dynamics because the prediction is iterative. The central problem amounts to a tension between providing a reliably accurate forecast and affecting a desired outcome via the forecast. These two objectives are not always compatible. We map a review of the literature onto relevant ecological scales that contextualize the role of reflexivity across a range of applications, from biogeochemical (e.g., hypoxia and harmful algal blooms) to endangered species management. Formulating reflexivity mathematically provides one explicit mechanism for integrating natural and social sciences. In the context of the Anthropocene ocean, reflexivity helps us understand whether forecasts are meant to mitigate and control environmental changes, or to adapt and respond within a changing system. By thinking about reflexivity as part of the foundation of ocean system forecasting, we hope to avoid some of the unintended consequences that can derail forecasting programs.

High-resolution stochastic downscaling method for ocean forecasting models and its application to the Red Sea dynamics

High-resolution modelling of a large ocean domain requires significant computational resources. The main purpose of this study is to develop an efficient tool for downscaling the lower-resolution data such as those available from Copernicus Marine Environment Monitoring Service (CMEMS). Common methods of downscaling CMEMS ocean models utilise their lower-resolution output as boundary conditions for local, higher-resolution hydrodynamic ocean models. Such methods reveal greater details of spatial distribution of ocean variables; however, they increase the cost of computations and often reduce the model skill due to the so called “double penalty” effect. This effect is a common problem for many high-resolution models where predicted features are displaced in space or time. This paper presents a stochastic–deterministic downscaling (SDD) method, which is an efficient tool for downscaling of ocean models based on the combination of deterministic and stochastic approaches. The ability of the SDD method is first demonstrated in an idealised case when the true solution is known a priori. Then the method is applied to create an operational Stochastic Model of the Red Sea (SMORS), with the parent model being the Mercator Global Ocean Analysis and Forecast System at 1/12∘ resolution. The stochastic component of the model is data-driven rather than equation-driven, and it is applied to the areas smaller than the Rossby radius, within which distributions of ocean variables are more coherent than over a larger distance. The method, based on objective analysis, is similar to what is used for data assimilation in ocean models and stems from the philosophy of 2-D turbulence. SMORS produces finer-resolution (1/24∘ latitude mesh) oceanographic data using the output from a coarser-resolution (1/12∘ mesh) parent model available from CMEMS. The values on the fine-resolution mesh are computed under conditions of minimisation of the cost function, which represents the error between the model and true solution. SMORS has been validated against sea surface temperature and ARGO float observations. Comparisons show that the model and observations are in good agreement and SMORS is not subject to the “double penalty” effect. SMORS is very fast to run on a typical desktop PC and can be relocated to another area of the ocean.

Combining Argo and Satellite Data Using Model-Derived Covariances: Blue Maps

Blue Maps aims to exploit the versatility of an ensemble data assimilation system to deliver gridded estimates of ocean temperature, salinity, and sea-level with the accuracy of an observation-based product. Weekly maps of ocean properties are produced on a 1/10°, near-global grid by combining Argo profiles and satellite observations using ensemble optimal interpolation (EnOI). EnOI is traditionally applied to ocean models for ocean forecasting or reanalysis, and usually uses an ensemble comprised of anomalies for only one spatiotemporal scale (e.g., mesoscale). Here, we implement EnOI using an ensemble that includes anomalies for multiple space- and time-scales: mesoscale, intraseasonal, seasonal, and interannual. The system produces high-quality analyses that produce mis-fits to observations that compare well to other observation-based products and ocean reanalyses. The accuracy of Blue Maps analyses is assessed by comparing background fields and analyses to observations, before and after each analysis is calculated. Blue Maps produces analyses of sea-level with accuracy of about 4 cm; and analyses of upper-ocean (deep) temperature and salinity with accuracy of about 0.45 (0.15) degrees and 0.1 (0.015) practical salinity units, respectively. We show that the system benefits from a diversity of ensemble members with multiple scales, with different types of ensemble members weighted accordingly in different dynamical regions.

International Quality-Controlled Ocean Database (IQuOD) v0.1: The Temperature Uncertainty Specification

Ocean temperature observations are crucial for a host of climate research and forecasting activities, such as climate monitoring, ocean reanalysis and state estimation, seasonal-to-decadal forecasts, and ocean forecasting. For all of these applications, it is crucial to understand the uncertainty attached to each of the observations, accounting for changes in instrument technology and observing practices over time. Here, we describe the rationale behind the uncertainty specification provided for all in situ ocean temperature observations in the International Quality-controlled Ocean Database (IQuOD) v0.1, a value-added data product served alongside the World Ocean Database (WOD). We collected information from manufacturer specifications and other publications, providing the end user with uncertainty estimates based mainly on instrument type, along with extant auxiliary information such as calibration and collection method. The provision of a consistent set of observation uncertainties will provide a more complete understanding of historical ocean observations used to examine the changing environment. Moving forward, IQuOD will continue to work with the ocean observation, data assimilation and ocean climate communities to further refine uncertainty quantification. We encourage submissions of metadata and information about historical practices to the IQuOD project and WOD.

Evaluation of the Operational CMEMS and Coastal Downstream Ocean Forecasting Services During the Storm Gloria (January 2020)

Storm Gloria was the 10th named storm in Europe for the 2019–2020 winter season, and it severely affected Spain and France. This powerful storm represents an excellent study case to analyze the capabilities of the different ocean model systems available in the Spanish Mediterranean coasts to simulate extreme events, as well as to assess their suitability to enhance preparedness in maritime disasters with high impacts on coastal areas. Five different operational ocean forecasting services able to predict the storm-induced ocean circulation are evaluated. Three of the systems are delivered by the Copernicus Marine Service (hereafter CMEMS): the CMEMS global scale solution (GLO-1/12°), the specific Mediterranean basin scale one (MED-1/24°), and the regional solution for the Atlantic façade (IBI-1/36°), which includes also part of the western Mediterranean. These CMEMS core products are complemented with two higher resolution models focused on more limited areas, which provide operational forecasts for coastal applications: the WMOP system developed at the Balearic Islands Coastal Observing and Forecasting System (SOCIB) with a horizontal resolution of roughly 2 km and the Puertos del Estado (PdE) SAMOA systems with a 350-m resolution that cover the coastal domains of the Spanish Port Authorities of Barcelona, Tarragona, Castellón and Almeria. Both the WMOP and SAMOA models are nested in CMEMS regional systems (MED and IBI, respectively) and constitute good examples of coastal-scale-oriented CMEMS downstream services. The skill of these five ocean models in reproducing the surface dynamics in the area during Gloria is evaluated using met-ocean in situ measurements from numerous buoys (moored in coastal and open waters) and coastal meteorological stations as a reference to track the effects of the storm in essential ocean variables such as surface current, water temperature, and salinity throughout January 2020. Furthermore, modeled surface dynamics are validated against hourly surface current fields from the two high-frequency radar systems available in the zone (the SOCIB HF-Radar system covering the eastern part of the Ibiza Channel and the PdE one at Tarragona, which covers the Ebro Delta, one of the coastal areas most impacted by Gloria). The results assess the performance of the dynamical downscaling at two different levels: first, within the own CMEMS service (with their regional products, as enhanced solutions with respect to the global one) and second in the coastal down-streaming service side (with very high-resolution models reaching coastal scales). This multi-model study case focused on Storm Gloria has allowed to identify some strengths and limitations of the systems currently in operations, and it can help outlining future model service upgrades aimed at better forecasting extreme coastal events.

High resolution stochastic downscaling method for ocean forecasting models and its application to the Red Sea dynamics

High-resolution modelling of a large ocean domain requires significant computational resources. The main purpose of this study is to develop an efficient tool for downscaling the lower resolution data such as available from Copernicus Marine Environment Monitoring Service (CMEMS). Common methods of downscaling CMEMS ocean models utilize their lower resolution output as boundary conditions for local, higher resolution hydrodynamic ocean models. Such methods reveal greater details of spatial distribution of ocean variables; however, they increase the cost of computations, and often reduce the model skill due to the so called double penalty effect. This effect is a common problem for many high-resolution models where predicted features are displaced in space or time. This paper presents a Stochastic Deterministic Downscaling (SDD) method, which is an efficient tool for downscaling of ocean models based on the combination of deterministic and stochastic approaches. The ability of the SDD method is first demonstrated in an idealised case when the true solution is known a priori. Then the method is applied to create an operational eddy-resolving Stochastic Model of the Red Sea (SMORS) with the parent model being the eddy-permitting Mercator Global Ocean Analysis and Forecast System. The stochastic component is data-driven rather than equation-driven and applied to the areas smaller than the Rossby radius, where distributions of ocean variables are more coherent. The method, based on objective analysis, is similar to what is used for data assimilation in ocean models, and stems from the philosophy of 2D turbulence. The SMORS model produces higher resolution (1/24th degree latitude mesh) oceanographic data using the output from a coarser resolution (1/12th degree mesh) parent model available from CMEMS. The values on the high-resolution mesh are computed under condition of minimisation of the cost function which represents the error between the model and true solution. The SMORS model has been validated against Sea Surface Temperature and ARGO floats observations. Comparisons show that the model and observations are in good agreement and SMORS is not subject to the ‘double penalty’ effect. SMORS is very fast to run on a typical desktop PC and can be relocated to another area of the ocean.

Uncertainty Quantification of Trajectory Clustering Applied to Ocean Ensemble Forecasts

Partitioning ocean flows into regions dynamically distinct from their surroundings based on material transport can assist search-and-rescue planning by reducing the search domain. The spectral clustering method partitions the domain by identifying fluid particle trajectories that are similar. The partitioning validity depends on the accuracy of the ocean forecasting, which is subject to several sources of uncertainty: model initialization, limited knowledge of the physical processes, boundary conditions, and forcing terms. Instead of a single model output, multiple realizations are produced spanning a range of potential outcomes, and trajectory clustering is used to identify robust features and quantify the uncertainty of the ensemble-averaged results. First, ensemble statistics are used to investigate the cluster sensitivity to the spectral clustering method free-parameters and the forecast parameters for the analytic Bickley jet, a geostrophic flow model. Then, we analyze an operational coastal ocean ensemble forecast and compare the clustering results to drifter trajectories south of Martha’s Vineyard. This approach identifies regions of low uncertainty where drifters released within a cluster predominantly remain there throughout the window of analysis. Drifters released in regions of high uncertainty tend to either enter neighboring clusters or deviate from all predicted outcomes.