3.3.6. Improvement in depth estimation.

The goal of the following experiments was to study the standard deviation of depths (in a test-area in the Mediterranean used in earlier investigations) estimated from gravity, when the gravity is affected by errors according to the following error models:

 (i)   EGM96,
 (ii)  GOCE,
 (iii) GOCE scaled by a constant scale factor and
 (iv) GOCE scaled linearly.

The method used for the optimum inversion of the gravity data was the least squares collocation inversion method developed by (Barzaghi et all., 1992; Knudsen 1993; Tscherning et al., 1994).

A two -layers model was used in these experiments. The mean depth of the first layer was 3 km and the density contrast to the previous layer was 1.6 gr/cm3. The mean depth of the second layer was 18 km and the density contrast was 0.6 gr/cm3. These values are realistic according to earlier studies in the same test-area (Arabelos, 1998).

Gravity anomalies were computed from the expansion of the geopotential model EGM96 to degree 200, adding to the coefficients random errors. This procedure was done using the "Monte-Carlo" simulator described previously. For each of the four error models mentioned above, 50 gravity 5'x 5' grids were produced. The inversion was done for each grid selecting model covariance functions to be in agreement with the empirical covariance functions of the gravity anomalies.

For each of the 50 grids of each of the cases (i)...(iv) the solution (i.e. the depths of the shallow and the deeper layer) was obtained after three iterations. In this way the differences between "observed" and computed gravity anomalies (i.e. the response of the two layers) was generally below 0.5 mGal in the case of EGM96 and 0.3 mGal in the case of the test models .

Table Variation in standard deviatioin of the depth estimated from gravity anomalies with respect to variations of the standard deviation of the gravity anomalies according the 4 error models.
(mGal) Depth layer 1
(m) Depth layer 2
EGM96 4.5 - 7.5 160 - 320 300 - 900
GOCE 0.6 - 1.1 15 - 75 40 - 210
GOCE (const. scaling) 0.4 - 0.9 15 - 60 30 - 160
GOCE (linear scaling) 1.0 - 1.9 30 - 125 70 - 290


From the results of table it could be concluded that the EGM96 error model is too pessimistic, the corresponding for GOCE too optimistic and the scaled GOCE maybe more realistic.

In most cases are the results from GOCE up to a factor 10 better than the results from EGM96. This is significant for depth-estimation in areas' where the depth estimation from ERS-1 geodetic mission altimeter data is influenced by sea-surface topography. Here the bathymetry may be improved, and thereby aid in the hydro-dynamic modelling.