The long term sea level change is a crucial indicator of our climate, for example in response to increasing greenhouse gases. The spatial sampling offered by satellite altimetry and its continuity during the last 25 years are major assets to provide an improved vision of the mean sea level.

Data used

We use the gridded merged Ssalto/Duacs products, even if it is possible to compute the mean sea level directly from along-track monosatellite data like the Geophysical Data Records (GDRs). If you aim to process it with your own geophysical corrections, like tide or inverted barometer, we recommend you to use GDR-like datasets.

Methodology

Temporal analysis: the global Mean Sea Level (MSL) is directly computed on each individual map from an equi-area average of all the grid values. The MSL variations can then be plotted as a function of time. The MSL trend is estimated by the slope of the curve, after filtering higher frequency signals like 60-day and seasonal signals.
Geographical analysis: The sea level slope is computed at each individual grid point using a temporal analysis. The geographic distribution of the slopes can then be plotted on a map.

We propose to use the Ferret software combined with an Opendap access; Ferret is a free software widely used by oceanographers and meteorologists. Using the Ferret software allows us to perform a temporal analysis.

Temporal extraction

Launch Ferret, then type :
use “http://opendap.aviso.oceanobs.com/thredds/dodsC/dt_upd_global_merged_msla_h”
Now you are “connected” to the merged maps of sea level anomalies FTP directory. Within Ferret the command show data will provide file characteristics. The variable is named “grid_0001”, this is the parameter we are interested in.

Geographic extraction

No specific focus here, we are going to plot the mean sea level for the whole entire world.

Computing the mean sea level

Type the command let mean = GRID_0001[x=-180:180,y=-82:82@AVE]: this is to define a new variable “mean” that we are going to plot after.
Here Ferret has an interesting advantage: the average that Ferret is computing is already weighted with the cosine of the latitude. In ocean areas of lowest latitudes there are fewer points due to the Mercator grid, but of course they should not be less accounted for.

The command plot mean[01-DEC-1992:20-SEP-2006] now displays the temporal evolution of what we called “mean”, i.e the temporal series of each mean MSL estimation from December 1992 to September 2006.

 msla_mercator_sm

msla_regul_sm

fig 1. Example of sea level anomaly grids, September 20th, 2006. The maps are respectively on a Mercator grid and on a regular grid. To compute the global mean sea level, it would not make any sense not to attach the same weight for each ocean grid cell; that is why in considering the Mercator grid each value has to be weighted in latitude.

Results and comments

The mean sea level is dominated by several harmonics: :
– annual signal
– semi-annual signal
– 60 days signal.

Actually if we want to really focus on the sea level rise we have to filter out from these signals. Filtering these signals is a more complex procedure not implemented here, but the final result is already available in NetCDF format on the AVISO website (link). For example you can plot and superimpose the two curves to observe the differences between filtered and non filtered MSL estimations.

The rate of the mean sea level rise as seen by satellite altimetry appears to be about 3 mm/year. This information has to be further quantified to estimate the contribution of each climate component (ice regions, ocean atmosphere exchanges..).

 mean_sea_level_1992_2006_sm
fig 2. Mean sea level rise, computed using 14 years of altimetry data.