Resumen de Estudio de la geostrofia de los océanos desde el espacio
The ocean geostrophic currents are a factor of upmost importance in determining the climate pattern. To be able to determine them has long been a quest of climatologist, oceanographers, geophysicist, and modelers alike, as well as practitioners of fishery and maritime commercial enterprises.
The Surface Geostrophic Currents (SGC) can be determined from the gradients of the Dynamic Topography (DT); previously determined from data of the Sea Surface Height (SSH) and a Earth¿s geoid (N). Altimeters provide SSH data with resolutions of around 10 km along-track. This high resolution makes having an accurate (and with spatial resolution as high as possible) N the key requirement in order to improve the DT accuracy and resolution. One attempt is the GOCE satellite mission which has as the major goal to acquire an estimation of N as accurate as 1-2 cm for space-scales shorter than 100 km. The GOCE mission plays a central role in this thesis being the target of study in chapters 1,4,5. Since the geoid is being improved through a currently updating GOCE¿s geoid, we decided to study the seasonal and inter-seasonal variability of the SGC from SSH anomalies referred on long-terms mean SSH (chapters 2,3). In the first chapter, a first assess of GOCE data is presented. The SGC are estimated and studied from a MDT determined by using a geoid from data of the first 61-day cycle of GOCE. In order to evaluate the improvement provided by GOCE, we determined also the SGC as obtained by using the state-of-the-art GRACE (instead of GOCE) geoid model ITG-Grace2010S, to determine the MDT. Because, in a principle, in-situ observation might be closer to the real data, we used drifter buoys measurements to validate the results. In order to compose a more general vision about the different techniques available to estimate the SGC we add a comparison with velocities estimated by simulations of an ocean general circulation model, the ECCO model.
Here we have proved how GOCE provides a significant improvement when compared with previous (geodetic) determinations of the SGC. Here we used a Gaussian filter of 83 km (full potential in this first release of the GOCE¿s gravity model) which keeps the Equator still masked by noise. Nevertheless, the middle and higher latitudes are perfectly depicted being close similar to in-situ measurements in both their geographical delimitation and velocities magnitude. Although for the equator band a higher degree of filtering would be preferred (leading to attenuated velocities for the middle-higher latitudes) this first approach to the GOCE product proves how the improvement provided by this new satellite mission makes the geodetic determination of the SGC start to be comparable with in-situ measurements.
In chapter 2, a detailed study on the seasonal variations of the SGC is provided. For that, we have use altimetric sea level anomalies to determine a weekly climatology for the SGC. In this way, it was found a strong seasonal cycle of around 40 cm/s of amplitude for all the Equatorial band and the major currents areas for the zonal component. This seasonal cycle is hardly reduced for the meridional component which reach values not higher than 15-20 cm/s for most of Equator band. Empirical Orthogonal Function (EOF) and complex EOF (CEOF) analysis were applied in order to determine the annual and semi-annual harmonics of the seasonal cycle. It was found that, for a global analysis those two harmonics accounts for around the 70% and the 57% of the variance for the zonal and the meridional components of the currents, respectively.
In chapter 3 a similar study is carried out for the inter-seasonal variations. This time we use sea level anomalies to determine monthly maps of SGC for more than 18 year, 1992-2010. Then we removed from the data the annual and semi-annual harmonics and the lineal trend component by least-squares fitting. Through an EOF analysis the main variations of the SGC were related with the sea surface pressure gradient which is in turn related with El Niño event. Areas of major influence are the equatorial Pacific and Indian Oceans. It was found that for La Niña phase the currents anomalies at these areas run toward Australia, and viceversa. That implies that during the positive phase of El Niño event, the anomalies of the SGC at the equatorial Pacific are eastward accelerated toward the South American coast. This area reached up to 40 cm/s of anomaly during the strong El Niño in 1998.
In chapter 4 we adapt an anisotropic filter (call Edge Enhancing Diffusion, EED) from the field of the image processing. The geostrophic flow has an anisotropic behavior (emphasized when the current is stronger) being composed of a direction where the current flow along and a direction across to the flow. The standard Gaussian filter is isotropic (weighting all directions equally) leading to significant signal attenuation specially in sharp areas of the MDT. The Perona and Malik filter (PMF) is a non-linear isotropic filter that reduces the weights of locations whose gradient could be identified as an edge of the image. In this sense, the PMF provide better results than a Gaussian filter since sharp areas of the MDT are now kept and therefore the SGC velocities are not attenuated. Nevertheless, the PMF is still limited since only the magnitude of the diffusion flux can be controlled and not its direction. That implies that the noise in the nearest neighbors of edges is not attenuated since the small flux perpendicular to edges prevents the filtering parallel to the edge. The EED is a filtering strategy based on the directional information of the gradient and consequently it can be used to force the filtering process to the direction along the flow.
Here we have used a zoom in the North-Western Pacific ocean from a geodetically estimated MDT and its derived SGC. The main conclusion of this study is that the EED should clearly be preferred to both the PMF and the Gaussian filter, particularly in cases with a noisy signal. For a easy case (by using a spectral-wise MDT) both the EED and the PMF provide similar results. When simulating noisy surfaces (by using a point-wise MDT) the EED showed similar results to the easy case. Contrary to that, the PMF produced noisy results (or too attenuated if we want to remove the noise).
Finally, in chapter 5, a comprehensive analysis of the 8 Earth¿s Gravity Models (combining the releases 01,02,03 with the three methodologies to determine them: the direct (DIR), spectral-wise (SPW) and time-wise (TIW) methods. SPW solutions is not available for the REL03) already available from GOCE is provided. We selected the North-Western Pacific area for that purpose since the presence of a strong gravity signal (due to the tectonic plates confluence) as well as of three strong currents: the Kuroshio (KC), the North Equatorial (NEC) and the Equatorial Counter (ECC) Currents. When comparing all models up to a common maximum degree (Nmax=200, resolution of ~100 km) it can be observed how the SPW and the REL01-TIW solutions are depicted with the noisier induced SGC. This noise is increased when the models are raised up to their maximum potential (maximum available harmonic degree). The REL01-DIR solution is depicted as the less noisy surface since it uses altimetry for a background model. In general, when the MDTs are filtered (by using the EED filter), all the surfaces looks similar. Main differences appeared in those areas with stronger gravity signal (Philippine-Pacific plates joint) and a spot-like signal off the south-east Japanese coast. Here we used the Maximenco¿s (combining GRACE and drifters), RIO09¿s (synthetic solution) and Chambers¿ (DNS08MSS minus EGM08) MDTs to compare the results. Comparison is carried out by using buoys measurements as the reference value. GOCE induced results (purely geodetic) are shown to be close to the Maximenco and RIO09 solutions and much better than those by Chambers. Some discrepancies are found at the ECC that could be related with the possible underestimation of the velocities by the drifter in such area (due to a poor coverage).
