Lagrangian data in a high-resolution numerical simulation of the North Atlantic: I. Comparison with in situ drifter data

https://doi.org/10.1016/S0924-7963(01)00015-XGet rights and content

Abstract

A model/data comparison was performed between simulated drifters from a high-resolution numerical simulation of the North Atlantic and a data set from in situ surface drifters. The comparison makes use of pseudo-Eulerian statistics such as mean velocity and eddy kinetic energy, and Lagrangian statistics such as integral time scales. The space and time distribution of the two data sets differ in the sense that the in situ drifters were released inhomogeneously in space and time while the simulated drifters were homogeneously seeded at the same time over a regular 1° grid. Despite this difference, the total data distributions computed over the complete data sets show some similarities that are mostly related to the large-scale pattern of Ekman divergence/convergence.

Comparisons of eddy kinetic energy and root mean square velocity indicate that the numerical model underestimates the eddy kinetic energy in the Gulf Stream extension and in the ocean interior. In addition, the model Lagrangian time scales are longer in the interior than the in situ time scales by approximately a factor of 2. It is suggested that this is primarily due to the lack of high-frequency winds in the model forcing, which causes an underestimation of the directly forced eddy variability. Regarding the mean flow, the comparison has been performed both qualitatively and quantitatively using James' statistical test. The results indicate that over most of the domain, the differences between model and in situ estimates are not significant. However, some areas of significant differences exist, close to high-energy regions, notably around the Gulf Stream path, which in the model lies slightly north of the observed path, although its strength and structure are well represented overall. Mean currents close to the buffer zones, primarily the Azores Current, also exhibit significant differences between model results and in situ estimates. Possibilities for model improvement are discussed in terms of forcings, buffer zone implementations, turbulence and mixed layer parameterizations, in light of our model/data comparison.

Introduction

Numerical ocean circulation models are becoming more accurate in their simulations of ocean flow, primarily due to the increasing computational power that allows higher grid resolution and more complex dynamics. Despite this improvement, as well as the availability of more accurate data for initial conditions, boundary conditions, and forcing fields, the question of how “realistic” high-resolution numerical model simulations of oceanic flow really are is still open. A quantitative assessment of model skills requires a comparison of the model results with in situ ocean data on space/time scales comparable to those of the model. For large-scale basin models, this implies the use of observational data with extensive horizontal, vertical, and temporal coverage, so that both large- and synoptic-scale quantities can be evaluated. In this paper, we take a first step in this direction by comparing simulated surface (mixed layer) drifters from a high-resolution (1/12°, 6-km average grid spacing) numerical simulation of the North Atlantic to in situ near-surface drifters. Lagrangian data have been previously used in model/data comparisons (e.g. Davis, 1996), especially using pseudo-Eulerian estimates such as the mean flow (e.g. Acero-Schertzer et al., 1997, Stutzer and Krauss, 1998). In the following sections, a quantitative comparison of the mean flow is performed together with a qualitative comparison of other statistics, both Eulerian and Lagrangian.

The ocean numerical model is the Miami Isopycnic Coordinate Ocean Model (MICOM), configured with realistic topography and stratification, a Kraus–Turner mixed layer parameterization, and forcing by monthly climatology from the Comprehensive Ocean-Atmosphere Data Set (COADS). A set of numerical drifter trajectories was computed in the uppermost layer of the model (mixed layer), and their statistics are compared with those of in situ near-surface drifter data, processed and archived at the WOCE/CLIVAR Drifter Data Assembly Center at NOAA-AOML. The in situ drifters are drogued with a 10-m holey sock centered at a depth of 15 m. Trajectories of these satellite-tracked near-surface drifters used in our study span the time period between 1989 and 1998.

The comparison is performed using two main types of statistics: (a) pseudo-Eulerian statistics, in which maps of mean flow kinetic energy (MKE), eddy kinetic energy (EKE), root mean square (r.m.s.) fluctuating velocity magnitude, and mean horizontal flow U(x,y)=(U(x,y), V(x,y)) are computed from both the model and in situ drifters, considered here as moving current meters; and (b) Lagrangian velocity statistics, in which Lagrangian integral time scales are directly computed from the drifter trajectories.

There are several systematic differences in the nature of the model and observed data that have to be taken into account when performing the comparison. First, the in situ ocean data are from a specific time interval, 1989–1998, while the model-simulated data describe a perpetual year since the model is forced with the same monthly winds and fluxes from 1 year to the next. The in situ drifter data therefore contain inter-annual and high-frequency fluctuations that are absent from the model data. Second, the in situ drifter data are representative of the 15-m deep velocity field, while the model results provide a bulk representation of the variable mixed layer depth, generally between 20 and 100 m, but occasionally reaching 2000 m in winter at high latitudes. For these reasons, complete agreement between the model and the observed data should not be expected. Nevertheless, the comparison can be meaningful in regions where the motion is more related to internal instabilities than to direct atmospheric forcing, as well as in large-scale current structures. The differences between model results and data should provide useful indications for improving further implementations of the model in terms of forcing fields and parameterizations of subgrid scale and mixed layer processes.

The paper is organized as follows: In Section 2, the characteristics of MICOM are discussed. The simulated and in situ drifter data sets are presented in 3 The simulated model drifter data set, 4 The in situ drifter data set, respectively. The comparison between the data sets is presented in Section 5. A summary and conclusions are provided in Section 6.

Section snippets

The numerical ocean circulation model

The Miami Isopycnic Coordinate Ocean Model (MICOM) is well documented in the literature. For a review, the reader is referred to Bleck et al. (1992) and Bleck and Chassignet (1994). The fundamental reason for modeling ocean flow in density coordinates is that this system suppresses the diapycnal component of numerically caused dispersion of material and thermodynamic properties (temperature, salinity, etc.). This characteristic allows isopycnic models to prevent the warming of deep water

The simulated model drifter data set

At the beginning of year 14, a total of 25,000 numerical particles were launched at the surface in the mixed layer and at depths of 400, 1000, 1500, and 3000 m (5000 particles at each level). The trajectories and diagnostics were computed for a 2-year period, with the particle positions being saved every 12 h and the Eulerian velocity fields every day. The numerical particles were launched in a regular 1°×1° grid, initially motivated by a study on ARGOS optimal design (Roemmich et al., 1999).

We

The in situ drifter data set

The complete AOML/NOAA drifter data set for the region 98°W–17°E, 33°S–65°N, during the time period 1989–1998, is used for our analysis. The amount of in situ drifter data is 221,336 buoy-days (number of drifters multiplied by number of days, with drifter positions available every 6 h), approximately 1/20 of the size of the simulated drifter data set (∼3,600,000 buoy-days). As can be seen in Fig. 2a, the in situ drifter sampling of our analysis domain greatly increases during 1992 and 1993 and

Comparison of statistical quantities

In this section, a comparison between model and in situ drifters is performed by considering the following statistical quantities: the mean flow U(x,y), the mean flow kinetic energy (hereafter MKE), the eddy kinetic energy (hereafter EKE), or equivalently mean velocity magnitude and eddy r.m.s. velocity, and the Lagrangian velocity time scale denoted by T=(Tu,Tv). Analysis and results are presented in the following, first discussing the Eulerian quantities and then the Lagrangian.

Summary and concluding remarks

In this paper, a comparison has been presented between the statistics of in situ and simulated drifter trajectories in the North Atlantic. The space and time distribution of the two data sets are different. Model drifters were homogeneously seeded in a regular 1° grid, having a constant lifetime of approximately 2 years. The in situ drifters, on the other hand, have inhomogeneous initial conditions, and lifetimes with an average of 238 days and maximum values of approximately 1400 days. Despite

Acknowledgements

The authors wish to thank E. Ryan for help in the technical development, E. Zambianchi for several useful discussions, L. Smith for her careful editing of the manuscript, and three anonymous reviewers for their comments. The authors acknowledge the PIs who collected the North Atlantic drifter data and generously made them available through NOAA-AOML. This research was supported by the National Science Foundation through Grant OCE-9811358 and by the Office of Naval Research through Grants ONR

References (28)

  • da Silva, A.M., Young, C.C., Levitus, S., 1994. Atlas of surface marine data. Technical Report. National Oceanic and...
  • R.E Davis

    Observing the general circulation with floats

    Deep-Sea Res.

    (1991)
  • R.E Davis

    Comparison of autonomous Lagrangian circulation explorer and fine resolution Antarctic model results in the South Atlantic

    J. Geophys. Res.

    (1996)
  • R.E Davis

    Preliminary results from directly measuring middepth circulation in the tropical and South Pacific

    J. Geophys. Res.

    (1998)
  • Cited by (0)

    View full text