The Aliasing of Mesoscale Spatial Variability Into Time-series Data: Satellite Based Estimates for the JGOFS-BATS Site

David M. Glover and Scott J. McCue
Dept. Marine Chemistry & Geochemistry
Woods Hole Oceanographic Inst.
Woods Hole, MA 02543
508-289-2656
dglover@whoi.edu

Scott C. Doney
Climate & Global Dynamics
National Center for Atmospheric Research
Boulder, CO 80307
303-497-1639
doney@ncar.ucar.edu

Robert H. Evans
Rosentiel School of Marine and Atmospheric Science
University of Miami
Miami, FL 33149
305-361-4799
bob@miami.rsmas.miami.edu

Introduction

Glover and Doney (1996) presented an analysis of the potential for using AVHRR data to explain discrepancies between models of limited dimensionality and actual observations. The analysis was limited by the use of ``compositing'' to fill in the temporal gaps present in the AVHRR PathFinder data set. This problem is examined using objectively analyzed AVHRR (Mariano and Brown, 1992) and TOPEX/Poseidon data. The objective analysis of the satellite data produces equally spaced time slices of anisotropically corrected, best estimates for each pixel or data point. The comparison of an one-dimensional, synoptically forced model of the JGOFS time series station near Bermuda with actual observations (Doney, 1996) underscores the difficulty in determining the difference between model errors and unresolved three-dimensional effects. The key to apportioning the progenitor of these differences lies with understanding how the spatial variability changes with time and how the anomalies from a climatology can be separated into longer scale trend surface backgrounds and shorter scale variabilities and how they can be linked to mesoscale influences. A blend of surface satellite data (AVHRR) and upper water column integrated satellite data (TOPEX/Poseidon) is used to constrain, in an ensemble sense, what we perceive as mesoscale spatial variability. The sub-orbital tracks of the altimeter can only bound a control volume around BATS and the satellite derived SSTs do not necessarily represent at-depth water mass structures. But when used together the temporal and spatial variability of water mass balance and SST provide important insights into the amount of mesoscale variability recorded by and aliased into the time series station record.

Mathematical Basis

If satellite data coverage were as continuous as model output, we could readily use it to help explain discrepancies between models of limited dimensionality and actual observations. Unfortunately cloud cover, satellite down time and the orbital mechanics of polar orbiting satellites are just a few of the processes by which the continuity of satellite data is interrupted. For low resolution (monthly) time-series data, key events may thus be missed by both the ship and the satellite. In equation 1 we state the traditional way of expressing objectively analyzed data. The satellite view of SST is composed of a climatology and an anomaly field. But the anomaly field can be further decomposed into a larger scale trend surface and its residuals as in equation 2. We can then associate these components with a global, regional and local scale variability respectively (eqn. 3). In this fashion, for sea surface temperature, we can turn to satellite data (AVHRR) to diagnose the role of mesoscale variability at least in some mean sense.



 Our study area contains sub-orbital tracks 50D, 202D, 217A and 115A from the TOPEX/Poseidon satellite and a 10o by 10o window of objectively analyzed AVHRR Pathfinder SSTs. The box defined by the TOPEX/Poseidon sub-orbital tracks is used as a ``control volume'' and the AVHRR data is examined for scale effects from approximately 50 to 1000 km centered on the BATS station. Figure 1 displays objectively analyzed AVHRR and TOPEX/Poseidon data from week 28 of 1993.
Equation 4 gives, as an example, the manner in which the spatial mean SSTs are calculated. The analysis begins at the BATS site and works its way out in ever increasing scales of influence (X). The scale X determines the number of pixels (N) included in the spatial analysis (e.g. a scale of 90 km involves a 5 by 5 box of nominally 18 km pixels around Bermuda). The spatial variability calculations are performed in an analogous fashion.



Results and Discussion

Figure 2 shows the spatial mean for each of the components of SST studied here (AVHRR, climatology, anomaly, regional and local scales) and spatial variability for the anomaly, regional and local components at three different scales of influence (54 km in green, 522 km in blue, and 1062 km in red). Additionally in Fig. 2a the SST from Doney (1996) is plotted in black and SST measurements from BATS in magenta circles. In the spatial variability plots the nominal noise level (0.44oC) and revised noise level (0.2oC) are represented by black dot-dash and dotted lines respectively. (To see the full spatial and temporal variability at the local scale click here . WARNING: this mpeg file is 6.2 MB).

These results clearly show that the spatial mean temperature is largely independent of the scale of influence, except for short local scales. But as the scale of influence is increased the amount of spatial variability decreases in all cases except the local scale. Furthermore an analysis of the spatial variabilities shows that the total amount of variability (AVHRR) is dominated by sources from the climatology. The remaining variability (anomaly) is almost entirely represented by regional scale sources. The mismatch between in-situ data and model results can be seen as a mismatch between the scales to which each are responding (click here to see a movie of the 11 year time series of sea surface temperatures around BATS, CAUTION: this mpeg file is 1.6 MB). The model, entirely locally synoptically forced, does not include the effects from larger scale lengths that the in-situ data appears to be responding. The implications for modeling processes that may be influenced by mesoscale rectification (e.g. CO2 exchange) need to be carefully examined. As one increases spatial scale lengths the power shifts from regional to local (i.e. when the scale of influence is small most of the mesoscale variability is taken up in the trend surface).

 In order to check these results at depth, we are in the process of examining TOPEX/Poseidon data as a depth integrated spatial variability signal. Treating the box circumscribed by these orbits around Bermuda as a control volume should allow us to examine the largest scale lengths in the AVHRR data. The seasonally varying spatial variability of the fluxes into and out of the volume (Fig. 3) should be in phase with the same seasonally varying spatial variability observed in SST anomalies.

References

Doney, S.C., 1996. A synoptic atmospheric surface forcing data set and physical upper ocean model for the U.S. JGOFS Bermuda Atlantic Time-Series Study (BATS) site. J. Geophys. Res., Oceans, 101, 25,615-25,634.

Glover, D.M. and S.C. Doney, 1996, The aliasing of mesoscale spatial variability into time-series data: Satellite based estimates for the JGOFS-BATS site, Trans. Am. Geophys. Union, 77(46): F402-F403.

Mariano, A.J. and O.B. Brown, 1992, Efficient objective analysis of dynamically heterogeneous and nonstationary fields via the parameter matrix, Deep-Sea Res., 39(7/8): 1255-1271.


The text, graphics, and other materials contained in this webpage and attached documents are intended solely for scholarly use by the scientific and academic community. No reproduction, re-transmission or linking of this page to any other page without the author's expressed written permission is permitted.
© 1998 -- David M. Glover, WHOI --