IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 53, NO. 5, MAY 2015 2875
Empirical Forecasting of HF-Radar Velocity
Using Genetic Algorithms
Alejandro Orfila, Anne Molcard, Juan M. Sayol, Julien Marmain, Lucio Bellomo, Celine Quentin, and Yves Barbin
Abstract—We present a coastal ocean current forecasting system using exclusively past observations of a high-frequency radar (HF-Radar). The forecast is made by developing a new approach based on physical and mathematical results of the nonlinear dynamical systems theory that allows to obtain a predictive equation for the currents. Using radial velocities from two HF-Radar stations, the spatiotemporal variability of the fields is first decomposed using the empirical orthogonal functions. The amplitudes of the most relevant modes representing their temporal evolution are then approximated with functions obtained through a genetic algorithm. These functions will be then combined to obtain the hourly currents at the area for the next 36 h. The results indicate that after 4 h and for a horizon of 24 h, the computed predictions provide more accurate current fields than the latest available field (i.e., persistent field).
Index Terms—Empirical modeling, high-frequency radar (HF-Radar), operational oceanography.
KNOWLEDGE of coastal currents either for forecastingor nowcasting purposes is a relevant scientific and technological issue that has been receiving increasing attention in the last decades due to the large importance that shelves have in the economy, the biogeochemical cycles, or in the engineering activities. The preservation of the water quality and the conservation of coastal areas are two important objectives of developed societies essential to guarantee the sustainable management of coasts. By contrast, the degradation of shallow waters associated with accidental or illegal spills is directly related with the losses of habitats being among the largest threats in marine environments , .
Manuscript received March 18, 2014; revised August 20, 2014; accepted
October 26, 2014. This work was supported by the TOSCA MED Project under Grant G-MED09-425, by the MEDESS-4MS MED Project under Grant 2S-MED11-01, and by the MICINN Project under Grant CGL2011-22964.
The work of J. M. Sayol was supported by the Ph.D. CSIC-JAE program cofunded by the European Social Fund and CSIC. This work was partially performed while A. Orfila was a Visiting Scientist at the Mediterranean Institute of Oceanography at Universite de Toulon-Sud through a MECD fellowship (PRX12/00454).
A. Orfila and J. M. Sayol are with the Department of Marine Technologies and Operational Oceanography, IMEDEA (CSIC-UIB), Balearic Islands, 07190, Spain (e-mail: email@example.com; firstname.lastname@example.org).
A. Molcard, J. Marmain, L. Bellomo, C. Quentin, and Y. Barbin are with the Mediterranean Institute of Oceanography, Université de ToulonVar, Aix-Marseille Université, CNRS, IRD, MIO UM 110, 83957 La Garde,
France (e-mail: email@example.com; firstname.lastname@example.org; email@example.com; firstname.lastname@example.org; email@example.com).
Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TGRS.2014.2366294
Coasts are the transition areas between the open ocean and the land. The hydrodynamic in these areas is the result of a complex interaction of processes where the energetic inputs from the atmosphere (mainly through wind and heat fluxes) and those from the ocean boundaries (density gradients and energy provided by waves) balance with dissipation by coast and the bottom boundary layer and with the input of mass through rivers run-off. In addition, the complexity of the dynamics of coastal seas is increased by the nonlinear interactions among the physical processes that sometimes generate well-defined locally intense currents. Therefore, coastal areas have specific dynamical features with different spatial and temporal resolution from the open ocean that need to be constantly monitored in order to have an accurate representation of the physical mechanisms driving the dynamics.
The forecasting of coastal currents constitutes one of the most important challenges in geosciences. This relevance is given by the important role that coasts play in human-related activities. Maritime traffic, search and rescue operations, environmental control, military operations, etc., are some examples of the activities that require accurate and continuous forecast of currents in coastal seas. Unfortunately, the strong space–time variability of these areas together with the complexity of the processes inherent there make the prediction a difficult task.
Although numerical modeling is the most common approach to forecast the ocean, it requires continuous data support, which sometimes is hardly achieved. On the other hand, numerical model performance is quickly degraded if only surface data is provided. In addition to the numerical modeling approach, any dynamical system such as the ocean can be empirically modeled by expressing time-evolving measurements in a suitable functional form.
As a consequence of this importance, the establishment of coastal observing systems (COS) has been recently identified as an important component of the marine strategy by the European Commission (2010–2013), as well as for most advanced countries with economically significant coastal areas . In this respect, a significant effort has been devoted in the last years to different initiatives worldwide regarding COS. These observatories, such as the Integrated Marine Observing System in Australia; the Ocean Observatories Initiative; and the different regional components from the Integrated Ocean Observing
System in USA, Neptune, and Venus in Canada, Cosyna in
Germany, SOCIB (Balearic Islands Coastal Observing & Forecasting System) in Spain, and MOOSE (Mediterranean Ocean
Observing System for the Environment) in France among others, routinely acquire coastal ocean variables presenting them 0196-2892 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.