slide 63

The Gauss-Marx model is explained in this information. It involves using a reference epoch and a design matrix to relate unknowns to observations. The model can be solved using the least square method, but it only works if there are no gaps in the interferogens. In a real case, if there are subsets of interferogens that are not linked, the model has a rank deficit and the solutions at certain epochs are zero, which is not realistic. Various methods have strategies to address this problem. Here is an example of the Gauss-Marx model. Phi t3, t4 means the method image is at t3. Slash image is at t4. X is the phase change related to the reference epoch. Here, the reference epoch is said to be first epoch. The design matrix can relate the unknowns to observations. Once having the model, the unknowns can be solved by the least square method. It looks straightforward. However, there is only workable if interferogens cover all saw acquisition time. In other words, it has no gap in the interferogens. In real case, you may have a scenario like you have two subsets of interferogens and two subsets are not linked. Under this scenario, it has rank deficit. Although we still can solve the Gauss-Marx model by some methods such as SPD, the solution has some problems as the epochs between two subsets of interferogens. The solutions at this epoch are zero, which does not make sense in real world. Many methods of SPS has their own strategies to address this problem. You can read the papers for more details.