A Simulation for Deployment of Argo Floats

A Simulation for Deployment of Argo Floats

OKA Eitarou*1

Argo project started in 2000. In this international project, approximately 3000 profiling floats are deployed in the world ocean to build a real-time monitoring system of the upper and middle layers for long-term weather forecasts and climate predictions. Japan will deploy approximately 400 Argo floats in the western part of the tropical and North Pacific and a part of the tropical Indian Ocean. In order to deploy the floats efficiently and as uniformly as possible using limited cruises available, a numerical simulation of float deployment using the particle-tracking method and velocity data of an ocean general circulation model is performed. Method and application of this simulation are briefly shown in this report.

keywords: Argo, float deployment, simulation

*1:Climate Variations Observational Research Program,
Frontier Observational Research System for Global Change


1. Introduction

Argo project started in 2000 (Mizuno, 2000; Roemmich and Owens, 2000; Saiki, 2000). This is an international project to build a real-time monitoring system of upper and middle layers of the ocean using profiling floats for long-term weather forecasts and climate predictions. In several years, approximately 3000 profiling floats, called Argo floats, will be deployed over the world ocean with average spacing of 300 km under cooperation of the participating countries.

Argo floats drift at a specified depth, called a parking depth (usually 2000m), and rise up to the sea surface every specified cycle (usually ten days) by increasing their volume and changing the buoyancy. During the ascent, they measure temperature, salinity, and pressure with the CTD sensor. They stay at the sea surface for about a half day, transmitting the CTD data and their positions to the land via the ARGOS system, and then return to the parking depth. They make approximately 150 CTD observations during the lifetime of about four years.

As one of the participating countries of Argo, Japan has charge of the float deployment in the western part of the tropical and North Pacific and a part of the tropical Indian Ocean (Fig.1). Approximately 430 floats are needed to fill these areas with average spacing of 300km. The Frontier Observational Research System for Global Change (FORSGC) and the Japan Marine Science and Technology Center (JAMSTEC), the main organizations to implement Japan ARGO project, have launched 17 floats by the end of the fiscal year 2000. They are going to launch about 80 floats during 2001 and 100 floats during each of 2002-2004. By the end of the fiscal year 2004, they will have launched approximately 400 floats, which are almost sufficient to fill the Japanese areas of deployment with the target density.

The Japanese floats are launched on cruise tracks of R/V Mirai of the JAMSTEC, research vessels of the related agencies (Japan Meteorological Agency, Japan Coast Guard, and Japan Fisheries Agency) and universities, and other voluntary ships. The locations of launch are quite limited, because these cruise tracks are usually determined regardless of the implementation of Japan ARGO. After the launch, the floats drift alternately at the parking depth and at the sea surface, and gradually disperse over the Japanese areas of deployment. They are deployed over the Japanese areas in this way.

In order to deploy the floats efficiently and as uniformly as possible using limited cruise tracks, the best points to launch the floats on each cruise track should be presumed. For this purpose, the FORSGC/JAMSTEC are performing a numerical simulation of float drifts using the particle-tracking method and velocity data of an ocean general circulation model (OGCM). Method and application of this simulation are briefly shown in this report.

The OGCM and the simulation are described in Section 2. Characteristics of long-term drifts of Argo floats in the Japanese areas of deployment obtained by the simulation are shown in Section 3. Drifts of the Japanese floats launched in the fiscal year 2001 are simulated in Section 4. Summary is given in Section 5.

Fig.1 Areas of float deployment by Japan in Argo project. Number of floats needed to fill each area with average spacing of 300km is shown.

2. Method of simulation

@ Drifts of Argo floats were simulated using the particle-tracking method and velocity data of an OGCM. The OGCM is the High-Resolution OGCM run by the Ocean Research Department of the JAMSTEC. It is based on MOM2 (Modular Ocean Model version 2) developed at the NOAA/GFDL. It covers the global ocean between 75S and 75N and has a resolution of 1/4 degrees in the horizontal and 55 levels in the vertical (See Ishida et al. (1997) for details of the model description). The surface boundary conditions are based on climatological datasets. For the heat and freshwater fluxes, temperature and salinity at the top level are linearly restored toward the climatological data of Levitus (1982). For the momentum, wind stress data of Hellerman and Rosenstein (1983) are used. A bi-harmonic operator is used for the horizontal dissipation, and the coefficients are set to -1 x 1019 cm4/sec for both momentum and tracers. The formulation of Pacanowski and Philander (1981) is used for the vertical dissipation.

The first two years of the model run were the initial spin-up stage; the annual averages of the climatologies were used for the surface boundary conditions, and the model was driven from the initial state at rest. Then, a seasonal variability experiment, in which the model was driven by the monthlyclimatologies, was performed for 18 years initiated from the last stage of the spin-up run.

The horizontal velocity output was stored every five days during the last year of the seasonal variability experiment, namely the 20th year of the model run. This output is used as the seasonal velocity data in the particle-tracking computation. The velocity data at Level 1 (top layer at 5-m depth) and Level 43 (2045m), which are closest to the sea surface and the parking depth of 2000m, respectively, are used. The velocity data at Level 39 (1523m) are also used because some floats have a parking depth of 1500m.

The particle-tracking computation is made with a time step of two hours. At every time step, horizontal velocities at grid points are linearly interpolated with those at every five days immediately before and after. Then, velocity at the position of a particle, corresponding to an Argo float, is interpolated with the velocities at grid points surrounding the particle. Using the velocity at the particle, the position of the particle at the next time step is computed.

The Argo floats which the FORSGC/JAMSTEC are currently operating in the ocean have an observational cycle of ten days, including a period at the sea surface for about 11 hours. Correspondingly, the particle-tracking computation is made on the condition that particles stay alternately at Level 43 for 114 time steps (9 days and 12 hours) and at Level 1 for 6 steps (12 hours). The ascent and descent of floats, which take about six hours each, are neglected in the computation. In case that a float enters a region shallower than 2000m while staying at the sea surface and touches the sea bottom during the following descent, the float is supposed not to move at the sea bottom for about nine days until it starts rising up again. Correspondingly, it is assumed in the computation that if a particle exists in a region shallower than Level 43 at the end of its stay at Level 1, it stops for the following 114 time steps at Level 43 and then starts moving again at Level 1.

In the simulation, 25 or 100 particles are tracked for each float to be launched in order to examine drift of the float statistically. These particles are initially distributed uniformly in a 1 deg. x 1 deg. box centered by the launch point of the float.

3. Characteristics of float movement in Japanese areas for deployment

Drifts of Argo floats for three years in the western part of the tropical and North Pacific were simulated (Fig.2). The long-term drifts are toward different directions among three latitudinal ranges. They are mainly southwestward between 20S and 0, northwestward between 0 and 25N, and southeastward between 30N and 50N. These drifts strongly reflect the current pattern at the sea surface; the southwestward (northwestward) drifts in the tropics are mainly due to the westward South (North) Equatorial Current and poleward Ekman flows driven by the Trade Winds, and the southeastward ones in mid-latitudes are to eastward currents such as the Kuroshio Extension, the Subarctic Front, and the North Pacific Current and southward Ekman flows driven by the westerlies.

These drifts result in areas of convergence and divergence of floats. The floats converge and their number increases between 15N and 35N, particularly at 20-30N where the number increases by more than 50% during three years (Fig.3). On the other hand, the floats diverge poleward and their number decreases by more than 50% at 5S-10N around the equator. The number also decreases significantly north of 35N, because the floats drift eastward and leave the areas of launch.

According to this simulation, the floats drift for a comparable distance at the sea surface and at the parking depth of 2000m, because currents are much weaker but the floats stay much longer at the parking depth than at the sea surface. The reason that the currents at the parking depth affect the long-term drifts of floats little is that they are insignificant on the long-term average in most areas and are nearly non-divergent.

Drifts of floats were also simulated for the Indian Ocean (Fig.4). The long-term drifts of floats are mainly southwestward between 25S and 5S due to the westward South Equatorial Current and poleward Ekman flows. This is similar to the western Pacific. In contrast, the long-term drifts north of 5S in the monsoon region are quite different from those at corresponding latitudes in the western Pacific. They are eastward between 5S and 3N due to the eastward Equatorial Jet appearing in the monsoon transitions, and are mainly southeastward in the Arabian Sea and the Bay of Bengal. It should be noted that the drift patterns shown in Figs.2 and 4 almost do not change if these experiments are started from a date other than 1 January, even in areas where currents have large seasonal variability.

The long-term drifts in the Japanese areas indicate a degree of difficulty in the float deployment in each area. In the Pacific, the deployment is relatively easy at 30-50N east of Japan, because the floats launched just east of Japan drift eastward and spread over the area. The deployment is more difficult south of 25N in the tropics where the floats drift westward and hence need to be launched in the eastern part of the area far from Japan. The deployment in the tropics is particularly difficult in two regions. One is east of 165E where cruises are infrequent. The other is near the equator where the floats diverge poleward and hence need to be supplemented regularly. As for the Indian Ocean, the floats need be launched in the western part of the Japanese area north of 5S and in the northeastern part south of 5S for the efficient deployment.

Fig.2 Drifts of Argo floats for three years after the launch, simulated for the western part of the tropical and North Pacific. Starting points of the drifts are shown by dots, and are connected with the end points by colored lines. The colors of lines indicate the direction of drift; red is for between the north and the east, blue the east and the south, light blue the south and the west, and orange is for between the west and the north. Each drift shown is an average for 25 floats which were launched uniformly in a 1 deg. x 1 deg. box centered by the dot on 1 January. Green and brown regions indicate the land and areas shallower than 2000m in the OGCM, respectively. Thick black lines show the Japanese areas of float deployment in Argo project.

Fig.3 Ratio of number of Argo floats existing in each 5 deg. x 10 deg. area three years after the launch to that at the time of the launch, calculated for the computation shown in Fig.2. Green and brown regions indicate the land and areas shallower than 2000m in the OGCM, respectively. Thick black lines show the Japanese areas of float deployment in Argo project.

Fig.4 Same as Fig.2 but for the Indian Ocean.

4. Drifts of the Japanese floats launched in the fiscal year 2001

Japan ARGO project started in the fiscal year 2000 (starting in April 2000). In 2000, the FORSGC/JAMSTEC launched 15 floats tentatively in the southeast region of Japan. In 2001, they are going to launch 68 floats in various areas in the western part of the tropical and North Pacific (Fig.5) and 10 floats around 5S, 90E in the tropical Indian Ocean. Eight cruises of R/V Mirai of the JAMSTEC, R/V Keifu-maru and R/V Ryofu-maru of the Japan Meteorological Agency, S/V Takuyo of the Japan Coast Guard, and R/V Hakuho-maru of the Ocean Research Institute at the University of Tokyo are to be used for the launch.

The launch points on each cruise track were determined on the basis of a simulation of float drifts for the track. As an example of such simulations, one for MR02-K02 cruise of R/V Mirai in February-March is shown in Fig.6. In this cruise 2002, 15 floats are to be launched at latitudes between 37-20N and 26N on the track from Hachinohe to a TRITON buoy at 8N, 156E. Three years after the launch, the floats spread zonally between 29N and 37N, particularly to the east of the track. This is due to the eastward Kuroshio Extension flowing at a latitude of about 34N in the OGCM (Fig.7). To the south at 24-29N, the floats spread to the west of the track due to the counter current. Some of the floats may drift beyond the Izu-Ogasawara Ridge and enter the Shikoku Basin. Thus, the simulation forecasts that the floats launched on the meridional cruise track will be dispersed efficiently by the zonal currents.

In this simulation, a technique was used in the particle-tracking computation in order to correct unrealistic drifts caused by the OGCM velocity structure. In the OGCM, a northeastward flow constantly exists along the east coast of Japan between 37N and 42N, where the Oyashio flows in the opposite direction in the real ocean (Fig.7). In the simulation, this northeastward flow carries many of the floats existing in a western part of the mixed-water region to the Subarctic region, causing an unrealistic float distribution. In order to remove the effect of this flow, the floats which enter the region at 41-43N, 142-150E within four years after the launch were removed in the computation.

Drifts of the floats launched in the western part of the tropical and North Pacific in the fiscal year 2001 are simulated. Density distribution of these floats at the end of the fiscal year 2003 are shown in Fig.8. The float density is close to the target one (approximately one float in each 3 deg. x 3 deg. grid) at 25-40N, 140-150E southeast of Japan, at 15-25N, 130-140E south of Japan, and in some areas in mid-latitudes and the tropics, but much lower than the target density or zero in most areas in the subarctic and the tropics. The FORSGC/JAMSTEC are going to deploy floats so as to fill these low-density areas in 2002 and the following years.

Fig.5 Launch points (circles) of Argo floats planned by Japan in the western part of the tropical and North Pacific in the fiscal year 2001. Size of circles shows number of floats launched at each point. Black lines show the Japanese areas of float deployment in Argo project.

Fig.6 Distribution of Argo floats (circles) three years after the launch, simulated for MR02-K02 cruise of R/V Mirai in February-March 2002. In the simulation, 100 floats (colored dots) were launched uniformly in each 1 deg. x 1 deg. box centered by the real launch point on 25 February, and were tracked for four years. The floats which entered the region at 41-43N, 142-150E surrounded by red lines within four years after the launch were removed (See the text for details). Green and brown regions indicate the land and areas shallower than 2000m in the OGCM, respectively. Black lines show the Japanese areas of float deployment in Argo project.

Fig.7 Distribution of the annual-mean OGCM velocity at Level 1 around Japan. Dots show starting points of the velocity vectors. Green and brown regions indicate the land and areas shallower than 2000m in the OGCM, respectively. See the text for the region at 41-43N, 142-150E surrounded by red lines.

@

Fig.8 Density distribution of Argo floats for 3 deg. x 3 deg. grids at the end of the fiscal year 2003, simulated for the floats (circles) launched by Japan in the western part of the tropical and North Pacific in the fiscal year 2001. In the simulation, 100 floats were launched uniformly in each 1 deg. x 1 deg. box centered by the real launch point on the scheduled date, and were tracked for four years. The velocity data at Level 39 corresponding to the parking depth of 1500m were used for the floats launched south of 8N along 165E and at 2S-0, 175-160W. The density distribution was obtained by dividing number of floats existing in each 3 deg. x 3 deg. grid by 100. Black lines show the Japanese areas of float deployment in Argo project.

5. Summary

The FORSGC/JAMSTEC are performing a numerical simulation using the particle-tracking method and velocity data of the High-Resolution OGCM of the JAMSTEC in order to deploy Argo floats efficiently and as uniformly as possible. Method and application of this simulation were briefly shown.

This simulation has some points to be improved. One of them is that since the velocity data for a single year (the 20th year of the model run) are repeatedly used in the particle-tracking computation for more than one years, the resulting float drifts tend to reflect strongly features of the velocity data of that year such as distribution of meso-scale eddies. In order to improve this point, we are preparing to add the velocity data during the 21st-23rd year, and the new velocity data will span four years which are equal to the lifetime of Argo floats.

The evaluation of this simulation is now being made using the position data of floats (Davis, 1998) kindly supplied by Prof. R. E. Davis.

Acknowledgments

The author would like to thank A. Ishida in the Ocean Research Department of the JAMSTEC for kindly supplying the velocity data of the High-Resolution OGCM and giving him lectures on the particle-tracking computation. He also thanks the Computer and Information Division of the JAMSTEC for use of the super computer NEC SX-5.

References