Jiping XIE and Jiang ZHU
Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China
1.Background
Since observations of surface trajectories of ARGO floats have
location errors from 150m to 1000m, the direct estimation of
trajectory-mean surface currents from observations could have errors of
5.4 cm s-1 for u- and v- component in Pacific on
average. Based on Kalman Filter, a new method to estimate the surface
trajectories of ARGO floats by combining surface trajectory
observations and a trajectory prediction model is proposed by Xie and
Zhu (2008), as shown in Fig.1. The method aims to improve surface
velocity estimation via reducing positioning errors of surface
trajectories. Theoretical errors of predicted trajectories are about 1
km that is consistent with innovations (about 1.1 km). By optimal
combining information from predictions and observations with their
error statistics, the estimated trajectory positioning errors are
reduced significantly. The estimated surface velocities have errors of
4.4 cm s-1 for u- and v- component on average.
Furthermore, the surface currents from the Argo floats are
compared with the surface drifter-derived currents and the Tropical
Atmosphere Ocean program (TAO) measurements. The comparisons show good
agreement for both the current amplitude and the direction of surface
currents. Results indicate the feasibility of obtaining ocean surface
currents from the Argo array and of combining the surface currents from
Argo and the ocean surface drifters for in situ mapping of the global surface currents (ref., Xie and Zhu, 2009).
A dataset of surface current vectors with error estimate from
1999 to 2010 is derived from the trajectories of the Array for
Real-time Geostrophic Oceanography (Argo) drifting on surface over the
global ocean.
Fig.1 Sketch of analyzing the Argo trajectory by Kalman filter (ref. Xie and Zhu, 2008)
2.Method and data
The raw data of the Argo floats in delayed mode were obtained from
the global data centers (ftp://usgodae1.usgodae.org/pub/outgoing/ARGO
and ftp.ifremer.fr/ifremer/ARGO) from 1999 to 2007 in the global
oceans. For surface current estimations, many factors that can cause
potential errors such as positioning abnormality, miscommunication,
etc. should be considered. The Quality Control (QC) measures emphasize
on positioning and interval-time check. In the positioning check, we
assume that all of the satellite fixes in a surface trajectory should
be close to each other. If any fix has a distance to the others of over
200 km, the fix will be deleted. The time between a pair of adjacent
fixes in a surface trajectory usually is usually several minutes to
several hours. In the interval-time check, if an interval-time is
shorter than 10 min, the late fix will be deleted. In addition, if the
distance between a pair of fixes in a surface trajectory is shorter
than the total positioning error, the late fix is deleted. If each
velocity vector between a pair of adjacent fixes is more than 2 m s-1,
the trajectory will be neglected.
The method used in this study was proposed by Xie and Zhu
(2008). Briefly, the surface current estimation from the surface
trajectory of each Argo float is improved by using the Kalman filter
technique.
Table 1 Specific of the variables in Sur_disc.dat
Variable |
Definition |
Argo float ID |
WMO float identifier. |
cycle number |
A profiling float performs cycles. The float cycle number is related with this current vector. |
fixed points |
The valid locations in trajectory drifting on surface. |
first location time |
Julian time of the beginning
location. The integer part represents the day, the decimal part
represents the time of the location in a day. The Julian day is
relative to Jan 1 1950. Example: |
last location time |
Julian time of the last location, but its mean is same with the above mentioned. |
latitude of first location |
Latitude of the fist location in trajectory. (Unit: degree north) |
longitude of first location |
Longitude of the fist location in trajectory. (Unit: degree east) |
latitude of last location |
Latitude of the last location in trajectory. (Unit: degree north) |
longitude of last location |
Longitude of the last location in trajectory. (Unit: degree east) |
u component |
U component of surface current. (Unit: eastward, cm/s) |
RMS error of u |
Evaluation of the uncertainty of u. (unit: cm/s) |
v component |
V component of surface current. (Unit: northward, cm/s) |
RMS error of v |
Evaluation of the uncertainty of v. (unit: cm/s) |
3.Dataset description
This dataset version2 of surface currents contains 2 sections:
The discrete surface current vectors (total: 562652) which are averaged along their trajectories. The vector file (Sur_disc.dat)
is ASCII, and includes 13 variables: Argo float ID, cycle number,
fixed points in trajectory, first location time, last location time,
latitude of first location, longitude of first location, latitude of
last location, longitude of last location, u component, RMS error of u,
v component, RMS error of v, as listed in Table 1.
Based on the discrete current vectors, we further derive the climatology of annual mean surface currents (Sur_grid.dat) in normal grid of 1°x1° and monthly surface currents (Sur_grid_mon.dat)
in normal grid of 2°x2°. For annual mean currents, the latitude is
increased from 69.5°S by 1° interval, and the longitude from 24.5°E by
1° interval. For monthly mean currents, the latitude begins from 69°S
to 69°N, and the longitude begins from 24.5°E eastwardly by the 2°
interval. These two unformatted files includes 5 variables: zonal
velocity, meridional velocity, uncertainty of zonal velocity,
uncertainty of meridional velocity, sampling number. Their details are
listed in Table 2.
Table 2 Specific of the variables in averaging data
Variable |
Definition |
U |
Zonal currents averaged in little bins. (Unit: eastward, cm/s). The default value: 32767. |
V |
Meridional currents averaged in little bins. (Unit: northward, cm/s). The default value: 32767. |
Eu |
Evaluation of the uncertainty of zonal currents. (unit: cm/s) |
Ev |
Evaluation of the uncertainty of meridional currents. (unit: cm/s) |
Kn |
Numbers of the averaged current vectors in each bin |
For example, to read the averaging file the code of fortran90 is following:
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
N0=180; M0=70; NN=12;
open(2,file=’Sur_grid_mon.dat’, form='unformatted', access='direct', recl=N0*M0, convert=’little_endian’)
l=0
do k=1,NN
l=l+1; read(2,rec=l) ((U(i,j,k),i=1,N0),j=1,M0)
l=l+1; read(2,rec=l) ((V(i,j,k),i=1,N0),j=1,M0)
l=l+1; read(2,rec=l) ((Eu(i,j,k),i=1,N0),j=1,M0)
l=l+1; read(2,rec=l) ((Ev(i,j,k),i=1,N0),j=1,M0)
l=l+1; read(2,rec=l) ((real(Kn2(i,j,k)),i=1,N0),j=1,M0)
end do
close(2)
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
If to read the annual mean current file, you can adjust the first line into:
N0=360; M0=140; NN=1;
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
4.General feature
In this dataset, the surface current vectors nearly cover the global ocean. Figure 2 and Figure 3 respectively show the spatial distributions of the sampling number and the surface currents averaged in 1°x1° bins.
Fig.2 Numbers of surface current vectors in each 1°x1° bin from Argo trajectory (1999-2010).
Fig.3 Annual mean surface current speed in 1°x1° bin from Argo trajectory
(1999-2010, unit: cm s-1).
Citation: