National geodatabase of ocean current power resource in USA

Xiufeng Yang a,n, Kevin A. Haas a, Hermann M. Fritz a, Steven P. French b,Xuan Shi c, Vincent S. Neary d, Budi Gunawan d

a School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA, USAb Center for Geographic Information Systems, Georgia Institute of Technology, Atlanta, GA, USAc Department of Geosciences, University of Arkansas, Fayetteville, AR, USAd Wind and Water Power Technologies, Sandia National Laboratories, Albuquerque, NM, USA

a r t i c l e i n f o

Article history:Received 6 February 2014Received in revised form17 November 2014Accepted 4 January 2015Available online 20 January 2015

Keywords:Ocean currentsHydrokinetic energyGulf Stream systemResource assessment

a b s t r a c t

Ocean currents represent an alternative source of clean energy given their inherent reliability, per-sistence and sustainability. The general ocean circulation is characterized by large rotating ocean gyresresulting in rapid ocean currents along the western boundaries because of the Coriolis Effect. The GulfStream system is formed by the western boundary current of the North Atlantic Ocean flowing along theeast coast of the United States, and is of particular interest as a potential energy resource for the UnitedStates. This study presents a national database of ocean current kinetic energy resource derived fromseven years of numerical model simulations to help advance awareness and market penetration forocean current energy. A web based GIS interface is provided for dissemination of the national energyresource data: http://www.oceancurrentpower.gatech.edu/. The website includes GIS layers of computedmonthly and yearly mean ocean current speed and associated power density along the coastlines of theUnited States, as well as joint and marginal probability histograms for current velocities at a variablehorizontal resolution of 4–7 km. Various tools are provided for viewing, identifying, filtering anddownloading the data from this website. The Gulf Stream system, especially the Florida Current,concentrates the highest kinetic power density (42000 W=m2). The majority of the kinetic power andits variability are only present in relatively shallow water given the strong correlation with the surfacewind stress. The kinetic energy flux in the Florida Current is estimated over 30 years to provide temporalvariability of the undisturbed kinetic energy with high statistical significance. Available power ofapproximately 5 GW associated with the undisturbed natural flow condition from the Gulf Streamsystem is predicted based on hypothetical turbine parameters. Successful development of renewableenergy generation requires further studies to account for more precise technical, economic andenvironmental constraints.

& 2015 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4972. Methodology – creation of the database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4983. Methodology – dissemination of the database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 498

3.1. Data layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4983.2. Identify tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4993.3. Filter tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499

4. Results – variability of the Florida Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5004.1. Data validation for the Florida Current. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5004.2. Spatial variability in the Florida Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5014.3. Temporal variation of the kinetic energy flux in the Florida Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501

5. Results – available power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5046. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/rser

Renewable and Sustainable Energy Reviews

http://dx.doi.org/10.1016/j.rser.2015.01.0071364-0321/& 2015 Elsevier Ltd. All rights reserved.

n Corresponding author: Tel.: þ1 912 433 2758.E-mail address: xfyang@gatech.edu (X. Yang).

Renewable and Sustainable Energy Reviews 44 (2015) 496–507

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505Appendix A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506

1. Introduction

There is a growing interest in renewable energy around the world.Concerns about environmental pollution and climate change areamong the driving forces for seeking affordable and environmentallyfriendly energy alternatives [1]. The decrease in fossil fuel reservesmakes renewable energy attractive to both the government and theindustry as the importance of renewable energy is increasinglyrecognized. In the past three decades global energy consumptionalmost doubled [2], while some predict the global fossil fuel reservesmay be exhausted within a century [3]. Some countries are alsoseeking energy independence by reducing fossil fuel imports fromforeign regions and developing domestic alternatives of clean energy.Renewable energy has great benefits compared to fossil fuels, includ-ing environmental improvement, fuel diversity, and national security ifsupplies cover a significant portion of the country's energydemands. Furthermore, the investment into the renewable energyindustry will most likely be spent on materials and infrastructurerather than on energy imports, thereby spurring local economiesthrough the creation of jobs [4,5].

The world's oceans cover more than 70% of the earth's surfaceand are a promising reservoir of alternative energy resources.Energy production from the ocean presently constitutes a negli-gible portion of our daily energy supply, while the worldwideelectricity produced by ocean based devices is predicted to reachmore than 7% by 2050 [6]. In countries with ocean access, coastalareas concentrate a wealth of natural and economic resources andare typically among the most developed areas in a country.Renewable energy from coastal and offshore regions is wellpositioned to supply countries' most populated areas if efficientharvesting is feasible.

Ocean currents are the continuous flow of ocean water incertain directions, and can vary greatly in terms of dominatingdriving forces, spatial locations, temporal and spatial scales. Fastmoving ocean currents are rich in hydrokinetic energy. Since wateris about 800 times denser than air, ocean currents of about 1/9 thewind speed carry comparable kinetic power density to wind. Themajor driving forces for large scale currents (on the order of1000 km length-scale) are wind stress, and temperature andsalinity differences (or associated density differences). Besides,meso-scale (on the order of 100 km length-scale) ocean currentscan also be driven by tides, river discharge and pressure gradients(generated by sea surface slope setup from coastal long waves, forexample). Among these forces, only the astronomical tidal forcingis deterministic, and thus allows for accurate forecasting. Tidalforcing, however, has a negligible contribution on the Gulf Streamocean current system [7,8]. Therefore, this study only examinesthe non-deterministic forces, among which the most importantare wind and density differences, and uses a probabilisticapproach to define the ocean currents.

Surface ocean currents are generally wind driven and developtheir typical clockwise spirals in the northern hemisphere andcounter-clockwise rotation in the southern hemisphere because ofthe imposed wind stresses. The Gulf Stream system exemplifieswind driven currents in the northern hemisphere intensified atthe western boundary of the Atlantic Ocean because of the Corioliseffect. Beginning in the Caribbean and ending in the northernNorth Atlantic, the Gulf Stream is one of the world's most intenselystudied ocean current systems. On average, the Gulf Stream is

approximately 90 km wide and 1000 m deep. The current speed isfastest near the surface with the maximum speed typicallyexceeding 2 m/s [9–11]. The Gulf Stream system's time scales varyfrom seasonal (stronger in the fall and weaker in the winter) toweekly [12–14]. Stronger meandering occurs primarily down-stream of Cape Hatteras, North Carolina.

An ocean current energy converter extracts and converts themechanical energy of the current into a transmittable energy formsuch as electricity. A variety of conversion devices have beenproposed or are currently under active development. Such devicesinclude water turbines similar to scaled wind turbines driving agenerator via a gearbox, and oscillating hydrofoils driving ahydraulic motor. The available in-stream kinetic power per unitarea, or power density Pstream from ocean currents, is calculatedusing the following equation:

Pstream ¼ 12ρV3 ð1Þ

where ρ is the density of water and V is the magnitude of thecurrent velocity. This equation represents the power available atthe individual device level.

The total power extraction potential from ocean currents,however, does not simply correspond to the superposition ofindividual power densities from multiple devices. The dynamicsof ocean circulation and accumulative effects of converters need tobe considered. A number of ocean current energy assessmentshave been performed in the past to evaluate the power potential ofthe Gulf Stream system. The earliest systematic studies date backto the 1970s. A research project named “Coriolis Program” pre-dicted that an amount of about 10 GWof hydrokinetic power couldbe extracted from the Gulf Stream using turbines [15]. A moreconservative prediction suggested an amount of up to 1 GW ofkinetic energy can be extracted from the Gulf Stream by turbinearrays without seriously disrupting climatic conditions [16]. How-ever neither study elucidated on the details of their resourceestimates. A recent study by Duerr and Dhanak [17] considered afraction of the undisturbed kinetic power density in the GulfStream as equivalent to the available power potential. Theyestimated approximately 20–25 GW of available hydrokineticpower in the Florida Current based on the computer model HybridCoordinate Ocean Model (HYCOM) data. Duerr and Dhanak [17]further stated that the power potential reduces to 1–4 GW if someoperational constraints are applied. Yang et al. [18] concluded thatkinetic power extraction potential from ocean currents should notbe considered equivalent to the undisturbed power density, andthe accumulative effect of power extraction on the flow itselfneeds to be considered. By using a simplified ocean circulationmodel and including the cumulative effect from power extraction,Yang et al. [19] estimated the upper limit of the theoretical powerpotential from the Florida Current portion of the Gulf Streamsystem to be approximately 5 GW for average flow conditions.

The present study provides the characterization of oceancurrents along the coasts of the United States from a probabilisticperspective with an emphasis on the energetic Gulf Streamsystem. A GIS database with a web interface disseminates datato interested parties and the general public. The kinetic energyflux is considered a primary indicator of the kinetic energyreserve, and therefore 30 years of kinetic energy flux time seriesare predicted and used to provide estimates of the temporal

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507 497

variation of the kinetic energy transport in the Florida Currentwith statistical significance. The available power in the undis-turbed natural flow condition from the Florida Current is approxi-mated based on assumed turbine parameters.

2. Methodology – creation of the database

The creation of the GIS database relies on numerical model dataproviding high spatial and temporal resolution as well as statisti-cally significant duration (�7 years). The US coastal areas havebeen divided into a number of sub-regions. For regions withmultiple model data available, observational drifter data (⟨http://www.aoml.noaa.gov/phod/dac/index.php⟩) is used to validate andselect the best model data for the database creation. Severaldifferent ocean model data from various sources were obtainedand compared with ocean surface drifter data in terms of theirstatistical agreement. The data with the highest statistical agree-ment with the surface drifter data were selected for each area. Themethod and procedure of model selection are documented ingreat detail in [20]. The model selection results are summarized inTable 1 and visually presented in Fig. 1. Among all the availableocean model data, Hybrid Coordinate Ocean Model (HYCOM) andNavy Coastal Ocean Model (NCOM) model data have statisticallybetter agreement with the drifter data and therefore were selectedfor this study. HYCOM is a data-assimilative, hybrid isopycnal-sigma-pressure, primitive equation ocean circulation model thatevolved from the Miami Isopycnic-Coordinate Ocean Model(MICOM) [21,22]. In the HYCOM online data server (http://www.hycom.org), two different sets of real-time modeling data arepublicly available. HYCOM–NCODA Global Analysis (HYCOM Glo-bal) provides a global coverage, while HYCOM–NCODA Gulf ofMexico Analysis (HYCOM GoM) covers the Gulf of Mexico area atfiner resolution. NCOM is primarily based on Princeton OceanModel (POM) and the Sigma/Z-level Model (SZM). NCOM has freesurface and is based on primitive equations and the hydrostatic,Boussinesq, and incompressible approximations [23]. The NCOMdata of the U.S. east coast for 2009–2010 were obtained from theNaval Research Laboratory. As shown in Fig. 1, on the west andAlaska coasts (blue color), 7 years of global HYCOM data were

selected. On the east coast (magenta color), a combination of2 years of NCOM and 5 years of global HYCOM data were selected.On the Gulf and Florida coasts (red color), 7 years of HYCOM GoMdata were selected. The database stores the joint velocity magni-tude and direction probability histograms (Pi;j ) computed from7 years of data as follows:

Pi;j ¼ni;j

N; ði¼ 1;2;…;K; j¼ 1;2;…; LÞ ð2Þ

where i and j are the indices, and K and L are the number of binsfor velocity magnitudes and directions. ni;j is the number ofobservations in each bin and N is the total number of observations.The joint probability histogram can be calculated based on theentire 7 years of data (N� 2555) or data from one particularmonth (N� 210). The marginal probability histograms (Pi; Pj) canbe derived by summing Pi;j up along either velocity magnitudes ordirections:

Pi ¼XLj ¼ 1

Pi;j ð3Þ

Pj ¼XKi ¼ 1

Pi;j: ð4Þ

3. Methodology – dissemination of the database

An interactive and web-based GIS system has been developedusing the selected model data to facilitate dissemination of theocean current power information to interested users includingelectric power utilities, policy makers, regulators, and turbinemanufacturers. The webpage can be accessed at http://www.oceancurrentpower.gatech.edu/. The GIS tools allow the user tointeract with the ocean current database. The ocean current data isstored in a geodatabase that enables the search query function viaa rich internet application (RIA) supported by ArcGIS server. Userscan interact with the map using the pull down menus on the topor widgets on the right side of the page. Besides the mapnavigation functions, the RIA also enables users to identify thesource data, retrieve the ocean current information for a givenlocation, and export the selected data. In summary, this systemprovides the following capabilities:

� View GIS layers and map display of monthly and yearly meansurface currents and power densities.

� View and download velocity probability distribution along theU.S. coastline at a number of different water depths.

� Download monthly and yearly mean surface current velocitiesand power densities for various regions.

This website provides functionality similar to the tidal energywebsite (http://www.tidalstreampower.gatech.edu/) documentedin Defne et al. [24]. Key components were modified as describedbelow to further enhance the technical capabilities.

3.1. Data layers

The web page consists of multiple layers (one data points layerand a set of color mapped raster layers) that can be switched onand off from the data layers widget. The color mapped raster layersinclude maps of the water depth, the mean current speed for eachmonth and the total, and the mean kinetic power density. Theselayers are generated by interpolating the model results fromcomputational grids onto an ArcGIS raster grid and serve for arapid visual examination. The data point layer contains moredetailed information that corresponds to actual model grid points

Fig. 1. Map showing different colors representing different combinations ofselected model data used to create the database for different regions. (Forinterpretation of the references to color in this figure legend, the reader is referredto the web version of this article.)

Table 1Ocean model data selected for different areas along the United States coast.

Location Selected model (time-span)

East coast HYCOM Global (5 years) and NCOM (2 years)Florida Strait HYCOM GoM (7 years)Gulf of Mexico HYCOM GoM (7 years)West, Alaska and Hawaii coasts HYCOM Global (7 years)

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507498

and can be queried through the interactive tools, and is thereforemore suited for in-depth analyses.

Fig. 2 shows an example of the web interface (http://www.oceancurrentpower.gatech.edu) that has the map of the UnitedStates in the background with the map of the mean surface currentspeed along the coastline on top. The web interface has fourinteractive buttons (“Map”, “Navigation”, “Tools” and “Help”) onthe top and multiple widget windows on the right. A pull downmenu will appear once the mouse pointer hovers on the buttonicon. On the right of the screen is the area of interactive widgets.Corresponding widget windows will appear in this area uponactivation of different functions. The database also providesdifferent widget windows: “Overview map”, “Data layers”, and“Show legend”. The “Overview map” widget shows the location ofthe current map view in the context of the larger geographicalarea. The “Data layers” widget toggles between 15 different datalayers on the map to display. The “Show legend” widget shows thelegend for data from all different layers.

3.2. Identify tool

The Identify tool under the “Tools” menu shown in Fig. 2 isused to identify a single data point either by clicking on the map orby specifying a longitude and latitude. The Identify tool returns themodel data of the selected point. Both the joint and marginalhistograms from the database for ocean current velocity can beplotted for any specific month or the entire year according to Eqs.(2)–(4). These histograms facilitate an overview of the probabilisticnature of the ocean current resource at a selected location.Similarly, the vertical variation of the current speed at correspond-ing locations can also be plotted with the identify tool. Once a datapoint is identified, the point will be highlighted in green, andinformation including longitude, latitude, water depth, meancurrent speed, mean power density and speed standard deviationwill show in the widget window. Further instructions will also begiven in the widget window to show plots of “Vertical speedprofile”, “Joint probability histogram”, “Marginal probability histo-gram” or “Download spreadsheet”. For example, Fig. 3 shows an

example of the vertical structure of current speed profiles for alocation approximately 20 mile offshore from West Palm Beach,Florida (26.912 N, 79.680 W) generated by the database. Fig. 3ashows the monthly variability of the vertical structure of thevelocity and Fig. 3b shows the variability in the vertical structurein terms of the standard derivation around the mean value. Fig. 3illustrates how the vertical structure of the current speed at thisparticular location remains similar with mostly an increase invelocity magnitude for higher flow conditions.

Fig. 4a and b shows the annual joint and marginal probabilityhistograms of the surface current velocity according to Eqs. (2)–(4)for the same location (26.912N, 79.680W) respectively. Fig. 4aprovides the distribution of current velocity in both magnitudeand direction in a polar coordinate system, and helps usersidentify the dominant current speed and direction. The distancefrom the origin to the bright red area in the joint histogram inFig. 4a indicates the dominant velocity magnitude (�1.8 m/s) andthe angle indicates the dominant velocity direction (�901). Thedistribution is calculated based on 7 years of daily snapshot datasubject to uncertainty, which is addressed with a certain con-fidence interval level. A brief discussion on the confidence intervalfor the probability distributions provided from this database isgiven in the Appendix.

3.3. Filter tool

The filter tool also under the “Tools” menu is used to downloaddata at selected grid points. A single point or multiple points canbe selected using the filter tool by drawing a rectangular windowor polygon. Data in the selected area can be filtered based on thewater depth, mean current magnitude or mean power densityspecified by the user or a combination of them prior to down-loading. This provides the user with the option to focus on theareas meeting certain criteria such as a minimum depth or aminimum speed. A SQL Server database was created to store acopy of the non-spatial tabular data in the geodatabase to improvethe performance of the export functions. A REST Web servicewas developed through Visual Studio.NET that executes a SQL

Fig. 2. The GIS map of the mean surface ocean current speed with pull down menus at the top and interactive widgets on the right (⟨http://www.oceancurrentpower.gatech.edu⟩).

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507 499

transaction to implement the search query over the SQL Serverdatabase to generate the output spreadsheet. This approachprovides a rapid response time. Once an area is selected by theuser, a button will appear to prompt the user to download aspreadsheet. The spreadsheet includes geographical coordinates,the modeled depth, the mean surface current speed for eachmonth and one entire year, the current speed standard deviation,the mean power density and the name of the region for each gridpoint in the selected region.

4. Results – variability of the Florida Current

The GIS map of the mean surface current power density showsthat the Florida Current has the highest power density(42000 W=m2) along the United States coast. Therefore extract-ing renewable energy from ocean currents in this region isparticularly attractive for local electricity needs because of theproximity of the Florida Current to the large population in south-eastern Florida metropolitan area (o50 km). This section, there-fore, presents an in-depth analysis of the variability of ocean

currents from the Florida Current in the context of ocean currentkinetic energy extraction.

4.1. Data validation for the Florida Current

Model validation is conducted for this region based on twoadditional measurement datasets: submarine cable data andAcoustic Doppler Current Profiler (ADCP) data. The telecommuni-cation cables running almost perpendicularly across the FloridaStrait from West Palm Beach, FL to Eight Mile Rock, GrandBahamas Island (http://www.aoml.noaa.gov/phod/floridacurrent/)can be used to measure the volume transport through the FloridaStrait. This measurement is based on the working principle thatthe flow through the earth's magnetic field can induce a voltage inthe cable providing a measure of the volumetric flow aftercalibration [25]. The volume flux obtained from the submarinecables can be used to explore correlations between volume fluxand kinetic energy flux in the Gulf Stream. The long cable datameasurement record (from 1982 to present) provided by theAtlantic Oceanographic and Meteorological Laboratory of NOAAis extremely valuable. Fig. 5 shows the comparison of the 30-dayrunning averaged volume flux in the Florida Current between

Fig. 3. An example of vertical structure of the current speed at (26.912N, 79.680 W): (a) monthly variation and (b) annual average plus or minus the standard deviation(STD); red dots and stars highlight the annual mean velocity profiles. (For interpretation of the references to color in this figure legend, the reader is referred to the webversion of this article.)

Fig. 4. An example of the (a) joint and (b) marginal probability histograms of surface current velocity at (26.912N, 79.680W). (For interpretation of the references to color inthis figure, the reader is referred to the web version of this article.)

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507500

HYCOM GoM model data and the submarine cable measurementfrom 2004 to 2010. The mean volume flux is 31.6 Sv from HYCOMGoM and 31.3 Sv from the cable measurement. In addition, thecorrelation coefficient between the two time series is approxi-mately 0.77, indicating strong agreement between two datasets.This verifies the quality of the model data in predicting the bulkflow properties in the Florida Current.

In Fig. 6, the vertical structure of the current speed averagedover one-year period of record from HYCOM GoM data is com-pared with ADCP data provided by the Southeast National MarineRenewable Energy Center at Florida Atlantic University (⟨http://snmrecwebext.eng.fau.edu/resource-measurement-modeling/available-data/⟩) for the same period of time and location. Fig. 6 showsthe comparison in terms of the mean, minimum and maximumcurrent speed. The model data accurately predicts the meanvertical speed profile as well as the minimum and maximumspeed for this selected location. This verifies the quality of themodel data in reproducing realistic vertical flow structures withinthe Florida Current.

A more detailed validation of the HYCOM model is provided byNeary et al. [26] with emphasis on the high power density regionin the Florida Strait. Their aim was to examine the deviation ofHYCOM GoM outputs from the HYCOM GLOBAL model, and thosebased on three independent observation sources: NOAA's sub-marine cable transport data, Florida Atlantic University's (FAU)ADCP data at a high power density location, and the SoutheastCoastal Ocean Observing Regional Association's (SECOORA) HFradar data in the high power density region of the Florida Strait.Comparisons with these three independent observation sets, andHYCOM GLOBAL outputs, indicate discrepancies with HYCOM

model outputs. Overall the HYCOM GoM model can provide a best-practical assessment of the ocean current hydrokinetic resource inhigh power density regions like the Florida Strait. However, predic-tions may be improved through advanced data assimilation andmodelforcing for periods with less accurate predictions of temporal transportvariation. Large scale ocean currents such as the Gulf Stream systemare primarily driven by winds. However other factors could also affectthe characteristics of the current velocity, such as the rotation of theearth, density differences, tides, river discharge, surface air pressuredifferences, bottom friction, and ocean topography. Therefore thefluctuation of current velocities varies with different time scales fromhours to weeks to years. Tides cannot be resolved in our analysis sincetime series of daily velocities are used in this study. Further additionalindependent observational data sources may be included to improvepredictions.

4.2. Spatial variability in the Florida Current

The spatial variation of the Florida Current is investigated byexamining the distribution of the mean and standard deviation(STD) of the current speed on the ocean surface as well as in avertical cross-section plane shown in Figs. 7 and 8. The portion ofthe Gulf Stream system in the Florida Strait (i.e. the FloridaCurrent) is shown to be predominantly flowing northward. Thecore of the current where the flow is the strongest is concentratedwithin about 100 m of the surface layer and spans about half of thechannel width. The core of the Florida Current is slightly offsetwestwards of the channel centerline (Figs. 7a and 8a) reducing thepotential cost of transmitting extracted power to shore, assumingextraction devices are likely to be deployed close to the core of thecurrent flow. Figs. 7b and 8b show the level of variability of thecurrent speed in the Florida Current. The highest temporalvariability in the Florida Current is located on the edge of thestrongest current facing the Florida shoreline. The variability of thecore of the current is relatively weak. Comparing daily andmonthly snapshots of the current speed distribution shows thathigh variation usually occurs near the edge as a direct result of themeandering and seasonal broadening of the core of the currentflow. The ratio of standard deviation to the mean value, alsoknown as the coefficient of variation, is plotted in Figs. 7c and 8c,which show a low level of variability inside the core of the currentand relatively high variability outside the core. According to theGIS maps from the database, as the current flows northward pastthe Florida Strait, the variability increases partly because of thedecrease of geographical constraint from the bathymetry.

4.3. Temporal variation of the kinetic energy flux in theFlorida Current

The kinetic energy flux is a primary indicator of the undis-turbed hydrokinetic energy reserve in ocean currents. The kineticenergy flux Ef in the Florida Current can be integrated fromHYCOM data as follows:

Ef ¼12ρZ

j V!j 2 V!Ud A! ð5Þ

where ρ is the water density, V!

is the velocity vector and d A!

isthe cross-sectional vertical area. The mean level of energy flux inthe Florida Current is calculated from 7 years of model data to beapproximately 22.6 GW with variability at multiple time scalesfrom weeks to years.

The change of energy flux at different levels in the watercolumn is examined by integrating the energy flux density acrossthe channel width but not over depth. Fig. 9 shows the kineticenergy flux density (GW/m) as a function of the depth for4 different months and the annual mean. The general shape of

Fig. 6. Comparison of one-year averaged vertical current speed profiles betweenADCP measurement and HYCOM GoM data at approximately (26.07N, 79.84W).

Fig. 5. Comparison of 30-day running averaged volume flux in the Florida Currentfrom cable measurement and from HYCOM GoM model data.

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507 501

different curves in Fig. 9 is similar to the vertical profile of thecurrent speed, strongest near the surface and weakest near thebottom. Fig. 9 also shows that more than half of the total kineticenergy flux is concentrated in the upper 200 m of thewater column inthe Florida Current. Monthly variability becomes negligible below100m water depth, implying that the monthly variability mostlyresults from seasonal variation of surface momentum flux such assurface wind stress. The summer season features the highest level ofenergy flux while winter has the lowest. February and May have fluxlevels comparable to the annual mean.

The kinetic energy flux is of great interest in this study sincethe goal is associated with estimating power potential from oceancurrents. The cable measurements provide 30 years of volume fluxin the Florida Current; however, there is no measurement of thekinetic energy flux, and the time series of kinetic energy flux datais limited to the 7 years of model data. Therefore it is meaningfulto seek a solid relationship between volume flux and kinetic

energy flux so that longer and statistically more significant recordof kinetic energy flux can be projected based on 30 years ofhistorical volume flux data from the cable measurement. Toestablish a statistical relationship between volume flux and kineticenergy flux, the 7 years of HYCOM data is divided into two groups,2004 to 2006 and 2007 to 2010. The data from 2007 to 2010 isused to establish an empirical statistical relationship betweenvolume flux and energy flux in the Florida Current. The timeseries of the volume and kinetic energy fluxes of 2007–2010 areplotted in Fig. 10a. The empirical statistical relationship betweenvolume and kinetic energy fluxes is then established with a leastsquare fit in the following form:

Ef ¼ aQn ð6Þ

with coefficients a and n to be determined. The fitted curve is shownin Fig. 10b with coefficients a¼ 0:001592, n¼ 2:766 and R2 ¼ 0:72.

Fig. 7. Florida Current: (a) annual mean surface current speed, (b) standard deviation, and (c) the coefficient of variation.

Fig. 8. Cross-sectional distribution of (a) the annual mean current speed, (b) the standard deviation, and (c) the coefficient of variation in Florida Current at the latitude of26.6264N.

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507502

To test the robustness of this empirical relationship, the volumeflux data of 2004–2006 from the cable measurement is used asinput in the above equation. A time series of the predicted kineticenergy flux for 2004–2006 is therefore generated as the output.Fig. 11 shows the time series of kinetic energy flux from both theHYCOM model data and from the prediction by the cable data for2004–2006. A 30-day running average is applied to both signalsfor smoothing. Some months of cable data are missing, andtherefore result in a number of gaps in the data predicted by thecable measurement. The actual time series (blue) and the predic-tion (red) in Fig. 11 have reasonable agreement featuring acorrelation coefficient of approximately 0.70.

The above experiment verifies the robustness of the empiricalstatistical approach in predicting kinetic energy flux from volumeflux by the cable measurement. Therefore a more dependablerelationship between volume flux and kinetic energy flux fromlonger duration of data in the Florida Current is computed basedon 7 years of HYCOM data with R2 ¼ 0:75:

Ef ¼ 0:001598 Q2:764 ðGWÞ ð7Þ

The 30-year time series of kinetic energy flux computed usingEq. (7) with 30-day running average is then calculated and plottedin Fig. 12 together with 95% confidence interval. The predicted

mean kinetic energy flux from 30 years of cable data is about22.8 GW, and the standard deviation is about 5.4 GW.

Figs. 13 and 14 show both monthly and yearly variations ofthe mean kinetic energy flux computed based on 30 years of data.

Fig. 9. The vertical kinetic energy flux density in the Florida Current for arbitrarymonths (February, May, July, and November) and the annual mean.

Fig. 10. (a) Time series of the volume and kinetic energy flux (2007–2010) from HYCOM model data, and (b) the least square fit of the relationship between volume andkinetic energy flux with 95% confidence interval based on data from 2007 to 2010.

Fig. 11. Comparing 30-day running-averaged kinetic energy flux prediction from2004 to 2006 from cable data with the energy flux calculated from HYCOM modeldata. (For interpretation of the references to color in this figure, the reader isreferred to the web version of this article.)

Fig. 12. Projected 30-year time series of 30-day running averaged kinetic energyflux using historical cable data with 95% confidence interval.

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507 503

The 95% confidence interval for the monthly mean energy flux isalso plotted in Fig. 13 as error bars to resolve the uncertainty. Themuch longer duration of data record (30 years) provides anarrower confidence interval demonstrating the improved levelof confidence. The mean kinetic energy flux is the highest in thesummer, particularly in July when the peak occurs (�25.57 GW).The lowest mean energy flux occurs in November (�20.30 GW),which is in agreement with previous findings. The kinetic energyflux also shows very strong year to year variability as seen inFig. 14. The annual mean power reaches a high 27 GW in 2002, anda low 18 GW in 1991.

5. Results – available power

From a practical point of view, it is helpful to quantify theundisturbed kinetic power in terms of hypothetical turbine arrays.Although this approach is neglecting the effects of extraction, ithelps to determine the approximate size and capacity of turbinearrays necessary to extract a certain amount of power. To examinethe undisturbed kinetic power from hypothetical turbine arrays, itis assumed that turbines are uniformly deployed 50 m below thesea surface in the Gulf Stream and current velocities from thedatabase are used to calculate the power. The principle velocity

component in the Florida Current is northward along the channel,and the undisturbed kinetic power (Pk) from this turbine array isestimated using the following equation:

Pk ¼Σ12ρ Vj j3Ef AsAcN ð8Þ

where V is the current speed at the assumed turbine depth, ρ isthe water density (1025 kg=m3Þ, Ef is the assumed efficiency(40%), As is the swept area of device (1600 m2Þ, Ac is the surfacearea of computation cell (� 16 km2Þ and N is the assumed numberof devices per unit surface area (1=4km2Þ corresponding to 2 kmspacing between devices. Open ocean turbine technology is notyet fully developed and tested; therefore it is not possible toobtain all the technological details of turbines to be used for theFlorida Current. Estimates are based on assumed turbine para-meters, but any modification would produce a correspondinglinear change in the total power estimate. Eq. (8) also does notaccount for operational availability or downtime for maintenanceand repairs.

The turbine region is specified within a box area covering thewater area between Florida to the Bahamas shown in Fig. 15.Turbines are assumed to be uniformly deployed in this area withthe specified parameters. From a technological point of view, mostdevices require a minimum “cut-in” flow speed at which deviceswill start producing power. Therefore devices should only bedeployed in areas where the mean speed exceeds a certainthreshold. For this analysis the threshold is set to 1 m/s, whichreduces the surface area of the turbine region to the area markedby black dots shown in Fig. 15 (approximately 2:0� 104 km2).

Fig. 16 shows the average monthly variation of the powergeneration from this hypothetical turbine array with error barsrepresenting the level of uncertainty. The peak power is shown tooccur in July and reaches almost 7 GW and the lowest poweroccurs in November and is about 4.3 GW. The annual mean kineticpower from this hypothetical turbine array is about 5.2 GWcorresponding to a mean power per device of approximately1.1 MW based on approximately 4500 devices installed. In theanalytical analysis of the theoretical power potential from theFlorida Current by Yang et al. [19], the theoretical maximum powerpotential from the region with comparable surface area represent-ing the Florida Current is approximately 5.1 GW, comparable to theestimated Pk from Eq. (8).

Comparison between the estimated Pk and the theoreticalpower potential from the Florida Current shows the feasibility ofachieving the theoretical power limit by the commonly usedapproach based on undisturbed power density when typical, but

Fig. 13. Monthly variation of mean kinetic energy flux with a 95% confidenceinterval based on projected 30 years of kinetic energy flux.

Fig. 14. Yearly variation of mean kinetic energy flux in the Florida Current based onprojected 30 years of kinetic energy flux (cable data in the year of 1999 is missing).

Fig. 15. Map with the box showing the original turbine area and black dotsrepresenting the area with mean current speed exceeding 1 m/s.

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507504

still hypothetical, turbine parameters are used. However theestimate of Pk carries uncertainty and could vary significantly byadjusting turbine parameters. In addition, the estimate of Pk basedon undisturbed power density is only meaningful when theaccumulative effect of power extraction on the flow field isrelatively small. The flow speed will be reduced as an increasingamount of energy is extracted and consequently a larger numberof turbines will be needed to extract the same amount of power.Therefore the estimate of Pk based on the undisturbed velocityfield is only useful for providing an order of magnitude for thenumber of devices, and is not recommended to be used solely fordetermining the maximum available power.

6. Conclusions

The ocean currents along the U.S. coastline were characterizedby using 7-year ocean model data collected from various sources.The selected ocean model data were used to develop a geodata-base that maps the ocean current energy resource distribution forthe entire U.S. coastline, and also provides a database of the jointvelocity magnitude and direction probability histograms for off-shore areas within approximately 200 mile of the U.S. coast. Theocean model data were validated using submarine cable and ADCPmeasurements, and results indicate a reasonable accuracy of themodel data allowing a best-practical prediction. The database isaccessible through a web interface http://www.oceancurrentpower.gatech.edu, which provides rapid visual examination ofthe resource distribution. The web interface also provides toolsto view, identify, query and download data from the database. TheGIS maps illustrate that the Florida Current has the highest meansurface power density among all ocean current sites within the U.S., with a mean surface power density exceeding 2000 W=m2,while most of the U.S. offshore waters have a mean surface powerdensity lower than 100 W=m2. The majority of the kinetic energyin the Florida Current is concentrated in the upper 200 m of thewater column.

The Florida Current features characteristic seasonal variation,with the current flow being stronger in the summer time andrelatively weaker in the winter time. Strong seasonal variability isprevalent in the upper layer of the water column (approximatelyupper 100 m), while the variation in deeper water is negligible,indicating an apparent correlation between the seasonal variabil-ity of the transport in the Florida Current and the seasonal

variation of the surface wind stress. As the Gulf Stream flowsnorthward past Cape Hatteras into the wide open ocean, thevariability greatly increases.

High kinetic power density is a good indicator of the richkinetic energy potential from ocean currents. A 30-year kineticenergy flux series was predicted in this study to estimate thetemporal variability of the Florida Current with high statisticalsignificance. The available power associated with the undisturbednatural flow condition from the Gulf Stream system is estimatedusing assumed turbine parameters purposefully selected to matchthe total theoretical power estimate of approximately 5 GW from[19]. Even if only 30% of this power is recovered, the powergeneration will be more than that from a typical nuclear reactor,enough to support a million homes based on the informationpublished by the U.S. Energy Information Administration [27].

Successful development of renewable energy generationrequires approaches that are technically feasible, economicallyviable, socially acceptable, and environmentally compatible [28].This may lead to a number of additional constraints. Powerextraction from ocean currents could have significant effects onthe local and global flow hydrodynamics, and, therefore couldpossibly change the water quality, sediment transport, and causeother ecological impacts. These issues need to be addressed whendeveloping an ocean current energy site.

Acknowledgments

This study was supported by the U.S. Department of Energy,Wind and Hydropower Technologies Program award number DE-EE0002661. Any opinions, finding, and conclusions or recommen-dations expressed herein are those of the authors and do notnecessarily reflect the views of the Department of Energy. Theauthors also want to sincerely thank the HYCOM consortium forsharing HYCOM data. The authors also thank Charlie Barron andLucy Smedstad from the Naval Research Laboratory for sharingNCOM data.

Appendix A

The probability distribution of the current speed from the GISdatabase (http://www.oceancurrentpower.gatech.edu/) shows theprobability of a current speed residing in particular intervals. Thelevel of confidence interval of the probability distribution quanti-fies the uncertainty associated with the distribution. In general,the estimator of proportion in each interval p̂ is given by thefollowing equation:

p̂¼ Xn

ðA:1Þ

where X is the number of elements in the interval, and n is thetotal number of elements. When n is large, the sample proportionp̂ is well approximated as normal with mean p̂ and standard

deviationffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1� p̂� �

=nq

[29]. Therefore a confidence interval for p̂ is

given by the following equation:

p̂�zα=2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1� p̂� �

=nq

; p̂þzα=2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffip̂ 1� p̂� �n

s0@

1A ðA:2Þ

where zα=2 denotes the upper α=2 point of the standard normaldistribution. For a 95% confidence interval, α¼ 0:05, zα=2 ¼ 1:96.Using 7 years of daily data, n� 7� 365¼ 2555: Defining

γ ¼ zα=2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffip̂ 1� p̂� �

=nq

, γ is computed as a function of probability p̂,

as shown in Fig. A.1a. For example, Fig A.2a shows the example of

Fig. 16. Monthly variation of total kinetic power in the Florida Current with 95%confidence interval.

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507 505

the probability of surface ocean current speed at (26.912E,79.600W). For a certain speed value (e.g. �1.52 m/s), there is acorresponding probability p0 (0.06) as marked by the dash line inFig A.2a. In Fig. A.1a, p0 ¼ 0:06 corresponds to a value ofγ0 � 0:0095. Therefore the confidence interval for p0 ¼ 0:06 isapproximately p0 �γ0; p0 þγ0

� �or p0ð1:52 m=sÞ ¼ 0:0670:0095.

The dashed curves in Fig A.2a show the 95% confidence intervalfor the probability distribution of the current speed. Similarly, forthe monthly probability histograms estimated from 7 years ofdata, the total number of elements n� 7� 30¼ 210. The proce-dure to obtain 95% confidence interval of the probability is thesame, but a different curve for the confidence interval as shown inFig. A.1b, needs to be used as illustrated by the example providedin Fig A.2b. Clearly the uncertainty for the monthly distribution ismuch higher due to the reduction in the number of data points.

References

[1] Vanek F, Albright L, Angenent L. Energy systems engineering: evaluation andimplementation. New York: McGraw-Hill; 2008.

[2] ExxonMobil, 2012 the outlook for energy: a view to 2040. Technical Report,ExxonMobil; 2012.

[3] Shafiee S, Topal E. When will fossil fuel reserves be diminished? Energy Policy2009;37:181–9.

[4] AWEA. AWEA U.S. wind industry annual market report: year ending 2011.Technical Report. Washington, D.C.: American Wind Energy Association; 2012.

[5] Navigant. Job creation opportunities in hydropower. Technical Report.Washington, D.C.: National Hydropower Association; 2009.

[6] Esteban M, Leary D. Current developments and future prospects of offshorewind and ocean energy. Appl Energy 2012;90:128–36.

[7] Leaman KD, Molinari RL, Vertes PS. Structure and variability of the Florida Currentat 271N: April–1982 July 1984. J Phys Oceanogr 1987;17:565–83.

[8] Shay LK, Lee TN, Williams EJ, Graber HC, Rooth CGH. Effects of low-frequencycurrent variability on near-inertial submesoscale vortices. J Geophys Res1998;103:18691–714.

[9] Stommel H. The Gulf Stream: a physical and dynamical description. second ed..Berkeley, CA: University of California Press; 1965.

[10] Richardson PL. Average velocity and transport of the Gulf-Stream near 55w.J Mar Res 1985;43:83–111.

[11] Fratantoni DM. North Atlantic surface circulation during the 1990s observedwith satellite-tracked drifters. J Geophys Res Oceans 2001;106:22067–93.

[12] Kelly KA, Gille ST. Gulf-Stream surface transport and statistics at 691W fromthe Geosat Altimeter. J Geophys Res Oceans 1990;95:3149.

[13] Zlotnicki V. Sea-level differences across the Gulf-Stream and Kuroshio Exten-sion. J Phys Oceanogr 1991;21:599–609.

[14] Hogg NG, Johns WE. Western boundary currents. Rev Geophys 1995;33:1311–34.

[15] Lissaman PBS. Coriolis program. Oceanus 1979;22:23–8.

Fig. A.2. An example of the probability distribution and 95% confidence interval of surface current speed at (26.912E, 79.600W) estimated from (a) daily data of 7 years and(b) daily data of only December..

Fig. A.1. Confidence interval profiles for probability estimation based on (a) daily data of 7 years, and (b) daily data for an individual month of 7 years..

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507506

[16] Von Arx W, Stewart H, Apel J. The Florida current as a potential source ofusable energy. In: proceedings of the Mac Arthur Workshop Feasibility ofextracting usable energy from the Florida Current; 1974. p. 91–101.

[17] Duerr AES, Dhanak MR. An assessment of the hydrokinetic energy resource ofthe Florida Current. IEEE J Ocean Eng 2012;37:281–93.

[18] Yang X, Haas K, Fritz H. Theoretical assessment of ocean current energypotential for the Gulf Stream system. Mar Technol Soc J 2013;47.

[19] Yang X, Haas KA, Fritz HM. Evaluating the potential for energy extraction fromturbines in the gulf stream system. Renewable Energy 2014;72:12–21.

[20] DOE. Assessment of energy production potential from ocean currents alongthe United States Coastline. Technical Report. Washington, D.C.: U.S. Depart-ment of Energy; 2013.

[21] Bleck R. An oceanic general circulation model framed in hybrid isopycnic-Cartesian coordinates. Ocean Model 2002;4:55–88.

[22] Halliwell GR. Evaluation of vertical coordinate and vertical mixing algorithms inthe HYbrid-coordinate Ocean Model (HYCOM). Ocean Model 2004;7:285–322.

[23] Barron CN, Kara AB, Martin PJ, Rhodes RC, Smedstad LF. Formulation,implementation and examination of vertical coordinate choices in the GlobalNavy Coastal Ocean Model (NCOM). Ocean Model 2006;11:347–75.

[24] Defne Z, Haas KA, Fritz HM, Jiang LD, French SP, Shi X, et al. Nationalgeodatabase of tidal stream power resource in USA. Renew Sustain Energy Rev2012;16:3326–38.

[25] Larsen JC, Sanford TB. Florida Current volume transports from voltagemeasurements. Science 1985;227:302–4.

[26] Neary VS, Gunawan B, Ryou AS. Performance evaluation of HYCOM-GOM forhydrokinetic resource assessment in the Florida Strait. Technical Report:ORNL/TM-2012/221. Oak Ridge National Laboratory, Oak Ridge, TN; 2012.Available at: ⟨https://wiki.ornl.gov/sites/publications/Files/Pub36784.pdf⟩.

[27] EIA. State Electricity Profiles 2010. Technical Report. Washington, D.C.: U.S.Energy Information Administration; 2012.

[28] NRC. An evaluation of the U.S. Department of Energy's marine and hydro-kinetic resource assessments. The National Academies Press; 2013.

[29] Johnson RA. Statistics: principles and methods. 5th ed.. Hoboken, New Jersey:Wiley & Sons; 2006.

X. Yang et al. / Renewable and Sustainable Energy Reviews 44 (2015) 496–507 507