**4.3. Remarks**

In spite of the scarcity of the log data, the proposed method achieved extremely good results. In case of data abundance, if the data is not well calibrated it could compromise the quality of the convergence and the influence of the not calibrated wells in the final model would be noticeable.

In this case, both synthetic seismic data and acoustic impedance cube captured the main geologic features of these complex reservoirs, noticeable in the correlation coefficients between the seismic and the synthetic amplitudes. The quality of the seismic data takes a minor role since the method overcomes the situation of imposition of artificial correlations as it happens in the standard methods;

Since the co-simulation of the impedances uses a local coefficient correlation, it is possible to compute the local uncertainty associated to the seismic acoustic impedances;

The uncertainty of the seismic acoustic impedance could be used to access the uncertainty associated with the porosity model, as presented in [24].

Tests prove that the variation of the final correlation coefficient is about 2% with the modification of the initial seed, it means that others parameters such as the number of layers and the size of it, as other parameters can be optimized to produce better results. But these results are more difficult to reach when the complexity of the geology and the structural model became more elaborated.

To handle different geology scenarios such channels or specific shape reservoirs, an adapted approach is proposed in the next part of this work.
