Initialisation and Forecasting

There are still some significant challenges in the data-assimilation techniques themselves, and one can expect to see significant improvement there. The assimilation of observations into present-day ocean models is still far from being optimal. Improved estimates of the state of the physical ocean, marine ecosystems and ocean-atmosphere interactions will rely upon new cross-cutting research directions in terms of both methods and operational implementations.

In meteorology (the history of which predates the evolution of ocean forecasting), the implementation of data assimilation methodology has followed a progressive pace starting with optimal interpolation, followed by sequential approaches and today most larger NWP centres are increasingly investing in 4D-VAR variational approaches with a noticeable increase in interest in ensemble approaches. Operational oceanography is today at the stage of applying sequential approaches but variational methodologies are on the verge of being used, at least for seasonal forecast applications. Because of the specificities of oceanography (e.g. the meso-scale non-linearities) it is still unclear whether 4D-VAR is fully applicable (Luong et al. 1998) and further research must be undertaken in this direction. A promising way might be the hybridisation between variational and sequential approaches thus combining advantages of both methodologies (Robert et al. 2006). However, 4D-VAR systems have not been comprehensively tested for highly non-linear applications. For instance, as we move to higher resolution and longer predictive time scales, the assumptions that underpin VAR systems (e.g. linearity in tangent-linear models) become less valid.

The development of data assimilation into physical coastal ocean models has lagged behind its development in basin-scale models, and is still in its infancy. Current methods need to be tested and enhanced for coastal applications. Data assimilation in coastal models has a vital role to play, not only as a tool to provide short-term forecasts, but more importantly for the rigour it brings to the analysis of model error, and to the design of observing systems (see the CSSWG White Paper, De Mey et al. 2007 for a detailed account).

Biogeochemical modelling and data assimilation are much less mature than physical modelling. Consequently, there is a strong need for both on-going development and validation of biogeochemical and, ultimately, ecosystem models.

The impact of the physical models on the accuracy of the ecosystem models is of particular importance (e.g. Berline et al. 2007). High horizontal and vertical resolution physical models are required to resolve the physical features that are critical to the ecosystem. Errors in physical models are problematic and can render outputs from ecosystem models meaningless. Vertical velocities are a particular example as they are critical for nutrient transport. In coastal areas correct representation of optical depth is also critical for primary production (Fig. 1.7). This requires accurate suspended sediment concentrations. These requirements for accuracy present a challenge for physical models.

A major trend in environmental research in the coming decade will see the development of the next generation weather, climate, and Earth system monitoring, assessment, data-assimilation, and prediction systems (Shapiro et al. 2008). These

Fig. 1.7 MODIS image of ocean colour off Australian NW shelf. The figure illustrates the complex processes acting in the coastal zone due to blending of different time/ space scales (e.g. ocean-shelf topographic interaction). Forecasting systems operating in such complex environments require sophisticated multi-scale (nested) models and scale-sensitive observing systems for accurate initialization (Courtesy: CSIRO Marine and Atmospheric Research)

systems will no longer focus on individual components of the Earth system (such as the oceans) but aim at treating the complex physical and biogeochemical components as one system. Coupled data assimilation means that observations in one medium impact the state of the other medium. In 4D-VAR fully coupled assimilation means simultaneous minimization of the cost function of the component models, e.g. atmosphere and ocean. An example of a less complex system is coupled ocean-atmosphere modelling. Ultimately truly coupled physical-biogeochemical initialization systems need to be developed, whereby the ocean, sea-ice, land surface, and atmosphere are initialized in unison. Consequently, a key challenge in data assimilation over the next decade will be the development of data assimilation techniques for Earth system modelling that are fit-for-purpose for a wide range of applications, including ocean-atmosphere weather forecasting, seasonal-to-decadal and climate change prediction.

Was this article helpful?

0 0
Guide to Alternative Fuels

Guide to Alternative Fuels

Your Alternative Fuel Solution for Saving Money, Reducing Oil Dependency, and Helping the Planet. Ethanol is an alternative to gasoline. The use of ethanol has been demonstrated to reduce greenhouse emissions slightly as compared to gasoline. Through this ebook, you are going to learn what you will need to know why choosing an alternative fuel may benefit you and your future.

Get My Free Ebook


Post a comment