Survey
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
A Hindcast Model of the Mid-Atlantic Ocean Region: a comparison with observation currents 1. Introduction This paper provides a comparison of a three-dimensional model simulation to four observation datasets in the Mid-Atlantic Bight (MAB). The observations were done mainly in the continental shelf regions, from near Cape Hatteras to southwestern portion of Georges Bank (Figure 1). Each observation dataset has its own specific time period from a few months to a few years. A detail description is in section 2. These observations provide an excellent test of the model’s ability to reproduce the seasonal as well as sub-tidal variations in ocean currents. An outline of the paper is as follows. In section 2, we described the observation datasets from four field studies. Section 3 presents the model setup. The comparison of the numerical simulation with observations is described in section 4. Section 5 is a discussion and summary. 2. Observation data 2.1 MMS moorings The Minerals Management Service (MMS) funded a mooring array near Cape Hatteras (Figure 1) from March 1992 through February 1994. 15 deployments of the moorings were located along three cross-shelf lines (A, B, and C) and one along-shelf line at the shelf edge (D). The offshore most moorings C4, B4, and A5 were situated both upstream and downstream from the Gulf Stream separation point from the continental shelf. Table 1 shows each depth for measurement at all mooing stations. A technical report (Berger et al. 1995) summarized the conclusion of the field project. The velocity data is used to validate the simulation results in the study. 2.2 CMO-Coastal Mixing and Optics Mooring The Coastal Mixing and Optics project was funded by the Office of Naval Research. The experiment was designed to examine the mixing of ocean water on a continental shelf and the effects of mixing on water column optical properties. The site of the CMO experiment was the "Mud Patch" of the Mid-Atlantic Bight (MAB) continental shelf, the southwestern portion of Georges Bank (GB). The site is located about 110 km south of Martha's Vineyard, Cape Cod, Massachusetts, U.S.A. in approximately 70 m of water. The field experiment was conducted between September 26, 1996 and June 9, 1997. The acoustic Doppler current profiler (ADCP) measurements are used to compare with the model simulation. 2.3 COBY A bottom mounted ADCP of 600-kHz was deployed at 32 m of depth at the station of station 5 on the COBY transect. The instrument measured two-period currents, first from Jan 28 through Jul 25, 2006 and the second from Oct 8, 2006 through Jun 23, 2007. 2.4 OMP-NC Ocean Margins Program The Department of Energy designed and implemented a field study called the Ocean Margins Program (OMP) to examine carbon cycling in the continental margin of the western North Atlantic Ocean. The OMP mooring array was located on the shelf and upper slope between the Chesapeake Bay and Cape Hatteras (Figure 1). A total of 26 moorings were deployed during the first deployment, between 5 February and 16 May 1996, while 23 moorings were deployed during the second deployment from 23 June through 14 October 1996. This field program resulted in the most extensive, multidisciplinary set of water column and seabed observations ever obtained over an ocean margin. The observed currents are used to compare with the model simulation. 3. Numerical model description The three-dimensional model used in the study is based on the Princeton Ocean Model (POM; Mellor, 2002). The model solves a hydrostatic, primitive Navier Stokes equation on a horizontal orthogonal curvilinear Arakawa “C” grid and a vertical terrain-following sigma coordinate. The ocean is assumed to be incompressible, and Boussinesq approximation is used. We carried out a nested-grid simulation from the northern portion of the South Atlantic Bight to the south and the Gulf of Maine to the north, with x y 5 km and 25 vertical sigma levels (Figure 1 lower panel). The simulation is named ‘fine grid’ simulation hereafter. The nested domain is embedded inside a larger-scale parent-grid domain of half the horizontal resolution x y 10 km but the same 25 sigma levels. The large domain mainly covers the entire northwestern Atlantic Ocean (98W~55W, 6N~50N, Figure 1 lower panel). The large domain simulation is named ‘coarse grid’ simulation hereafter. The coarse grid simulation is integrated from the end of 1992 through 2008 and includes daily discharges from 17 major rivers and estuaries along the U.S. and Canadian (the St. Lawrence River) eastern coast. The model is forced by six-hourly Cross-Calibrated, Multi-Platform (CCMP) ocean surface wind velocity and also by M2 tide. Real-time sea surface temperature and sea surface height anomalies from satellite are assimilated into the model. The monthly temperature-salinity (T/S) climatology is used to specify initial and open boundary conditions. The fine grid simulation has completed the integration from the end of 1992 to the end of 1999. 4. Model-data comparisons The comparison aims at validating the model simulation. The current data are used for comparisons. Unless otherwise stated, all variables have been daily averaged to remove the tidal oscillations prior to any analysis. 4.1 MMS moorings and fine grid simulation The comparison is carried out between MMS mooring currents and fine gird model from January 1993 through February 1994. Figure 2 compares mean and ellipses of currents at depth 2 (referred to table 1) for fine-grid model with those computed from observations at the 15 MMS moorings. Mean currents are shown as arrows at each station. From observations, the mean currents are northeastward at the southern stations of Cape Hatteras, and southward at the northern stations except at A5. The mean currents are increasing offshore along the three cross-shore lines. The amplitudes of the current ellipses reflect the current standard deviations. In the shelf region, the standard deviations are larger than the mean, whereas the mean currents are stronger in the further offshore region. This is caused by the influence of persistent northeastward Gulf Stream. The current ellipses are mainly aligned with the local isobath in both observations and model simulation. The model reproduced the general circulation pattern similar to the observed, though the model ellipses are larger than the observations. To compare the seasonal velocity variations, an empirical orthogonal function (EOF) analysis for MMS mooring currents and model currents are carried out. Fifteen-day low pass of the currents at depth 2 (Table 1) of 9 mooring stations after removal of stations with missing data were used to do the analysis. The first EOF mode accounts for 58% of the total velocity variance for the observations and 70% for the model, and the second mode accounts for 23% of the variance for the observations and 12% for the model (Figure 3 and 4). The EOF 1 from both observations and the model describes seasonal current variations in the shelf of the Cape Hatteras. The currents mainly flow southward in winter and spring, and northward in summer and fall. And the EOF 2 shows the currents have an along-shore convergence at the Cape Hatteras. The EOF modes are in good agreement between observations and the model. Table 1 indicates the complex correlation for depth-averaged MMS currents and simulation currents. In general, the correlation coefficient is relatively high close to the shore and low near the shelf break. The maximum coefficient is about 0.7 at station A1. The rotation angle indicates that the modeled current is rotated counterclockwise from MMS current. The minimum angle is about 30 at C1. Figure 5 compares the mean (left panel) and standard deviation (right panel) of depth-averaged principal-axis currents of the MMS moorings and the model. The fine-grid simulation produced the mean currents close to the observations at the inshore stations (e.g. A1, A2, C1, B2, D2, etc.), however, the currents in the offshore station are over predicted (e.g. A4, A5, and C4). The simulation produced higher standard deviation than the observed at all stations. The bias of model currents is small (<0.1 m s-1) at inshore stations and increasing offshore (Figure 6). The maximum bias is at B4, about -0.5 m s-1. Figure 7 shows the correlation coefficient of depth-averaged principal axis currents at all MMS mooring stations. Of the 15 stations, 6 stations give correlations of > 0.5. The maximum coefficient is larger than 0.8 at C3, while the minimum is less than 0.1 at A3. The skill assessment evaluates quantitative agreement between model results and observations. According to Wilmott (1981), the skill is defined as Skill 1 m o2 m o o o (1) 2 where m is the model variable, and o is the observed variable, and the time mean denotes as < >. The value of skill is in the range of 0~1, from a large discrepancy to a perfect agreement between model and observations. In the study, the depth-averaged currents along principal axis are used to evaluate the model skill. The simulated currents exhibited comparatively high skill (>0.4) at many stations, while low skill (< 0.4) at stations near the shelf break (Figure 7). Also according to Oke et al. (2002), there is a relationship between the root of mean squared error, RMSE deviation error, cross-correlation 1 CC S m S o 1 m m o 2 , mean bias, SDE S m S o error m o o m i m 2 CCE 2Sm So (1 CC) . The relation is BIAS m o , the standard o i . o 2 The , and the cross-correlation RMSE 2 BIAS 2 SDE 2 CCE 2 (2). Figure 8 shows the each item in Equation (2) at MMS mooring stations. The RMSE is relatively low (less than 0.05 m-2 s-2) at most of the stations. The large RMSE (> 0.05m-2 s-2) is mainly attributed to the southeastward bias at stations A4, B3, B4, and C4. At stations C2 and C3, all three error subcomponents contribute to the RMSE. 4.2 CMO mooring and fine grid simulation Figure 9 compares the depth-averaged fine-grid model currents with CMO data. The complex correlation produces the coefficient about 0.6 and the angle about 14 degrees between the two (Table 2). The bias for current speeds is -0.03 m s-1 and the absolute error is 0.1 m s-1. The mean currents from model and CMO are in good agreement, while the modeled current ellipse is smaller than the CMO one. Also, the CMO current ellipse is more aligned with the local isobath. The model yields a skill of 0.52. 4.3 COBY data and coarse grid simulation Figure 10 compares the depth-averaged coarse-grid model currents with COBY data. A good agreement between model and observations is present. The model yields a skill of 0.83. The complex correlation coefficient is about 0.7 and the angle is -6 degrees (Table 2). The bias and absolute error are about 0.01 m s-1 and 0.09 m s-1, respectively. The speed value of modeled mean current is about 0.003 m s-1, and the observed is 0.01 m s-1. The modeled current ellipse is more aligned with the shoreline. 4.4 OMP moorings and coarse grid simulation The mean and ellipses of depth-averaged currents from the fine-grid model and the OMP moorings are compared in Figure 11. North of Cape Hatteras, the modeled mean currents mainly flow southward, while the mean currents tend to flow northeastward at the southernmost stations. The magnitude of currents is increasing offshore. The maximum mean current (> 0.2 m/s) appears at station 26. The modeled ellipses are mainly aligned along local isobath, in general agreement with the OMP observations. Table 2 shows the complex correlation coefficients and rotation angles. The maximum correlation is about 0.5 at stations 7 and 11, and the minimum angle is about -50 at stations 12 and 14. Figure 12 compares the time mean and standard deviation of the depth-averaged currents along the principal axis. The simulation produces higher mean values at stations near shelf edge, except for station 26, where the OMP observation is higher. The simulation also produces higher standard deviation in offshore stations except for station 26. The bias of the simulation is small (< 0.1 m/s), at stations in the shelf, but large (>0.1 m/s) at stations close to the shelf break. The model skill assessment from Equation 1 yields relative high skills (> 0.5) at stations in the shelf, and relative low skills (<0.5) at the shelf break stations. For example, the maximum skill (>0.6) is obtained at stations 3 and 4. Figure 14 evaluates the contributions of BIAS, SDE and CCE to RMSE at OMP mooring stations. The RMSE is relatively low (less than 0.2 m-2 s-2) at most of the stations. The largest RMSE (> 0.35m-2 s-2) is at station 26. The CCE mainly attributes to RMSE. The contribution of SDE is negligible. 5. Discussion The comparison between the model and observations shows the model reproduces the seasonal as well as sub-tidal variability of currents fairly well in the MAB. For all stations, the model skill is relative high (>0.5) in the shelf regions, but getting low at the shelf break. In general, the correlation coefficient shows similar trend that it is relative high (>0.5) in the shelf but decreasing at the shelf break. In the shelf region north of Cape Hatteras, the EOF mode 1 of MMS and model indicate a clear seasonal cycle of current anomalies. It is southward in winter and spring and northward in summer and fall. References: Berger, T. J., P. Hamilton, R. J. Wayland, J. O. Blanton, W. C. Boicourt, J. H. Churchill, and D. R. Watts, (1995), A physical oceanographic field program offshore North Carolina, Minerals Management Service Tech. Rep., OCS Study MMS 94-0047, U.S. Department of the Interior, New Orleans, LA, 345 pp. Wilmott, C.J. (1981), On the validation of models, Physical Geography, 2, 184-194. Table 1: MMS moorings: station and depth Station A1 A2 A3 A4 A5 B1 B2 B3 B4 C1 C2 C3 C4 D2 D1 Depth 1 5 5 5 100 60 5 5 5 100 5 5 5 100 5 5 Depth 2 16 20 30 300 300 14 20 30 300 14 20 30 300 30 30 30 55 800 800 30 55 800 30 55 800 55 55 Depth 4 1200 1200 1200 1200 Depth 5 1900 1900 1900 1900 Depth 3 Depth 6 2900 Tabel 2. Complex correlation coefficients of depth-averaged MMS, OMP, CMO, and COBY currents with depth –averaged fine-grid simulation currents. The rotation angle indicates that the modeled current is rotated counterclockwise from the observed current. Θ Station R Θ Station R (deg) Θ Station R (deg) (deg) MMS A1 0.7 -10 OMP 1 0.1 33 OMP 18 0.3 -54 A2 0.3 -29 2 0.3 -88 19 0.2 -48 A3 0.01 -40 3 0.4 11 20 0.2 164 A4 0.6 -13 4 0.4 11 22 0.2 92 A5 0.5 -36 5 0.2 -44 23 0.3 -14 B1 0.4 -20 7 0.5 9 24 0.03 99 B2 0.4 4 8 0.4 -22 25 0.2 65 B3 0.6 -20 9 0.1 -151 26 0.4 -23 B4 0.4 -114 11 0.5 -123 27 0.2 -136 C1 0.4 3 12 0.4 -5 COBY 0.7 -5.8 C2 0.4 4 13 0.3 16 CMO 0.6 14 C3 0.6 7 14 0.3 -5 C4 0.3 -20 15 0.3 12 D1 0.01 15 16 0.1 68 D2 0.03 -172 17 0.3 171