In this paper, we develop a novel observation-driven methodology to explore spatio-temporal dependencies between satellite-derived Sea Surface Height (SSH) and Sea Surface Temperature (SST) fields. Level-set-based and registration-based critera are defined to evaluate and detect spatial links between SSH and SST anomaly fields. The method is applied to one-year SST and SSH time series in the highly dynamical Aghulas return current region. The analysis evidences a seasonal variation of the overall correlation between SST and SSH fields. As further revealed, the coldest SST anomalies are reported to efficiently trace the lowest SSH anomalies for all seasons, while the warmest SST anomalies solely match the largest SSH anomalies during winter. The second criterion relies on the registration on SSH and SST anomaly fields. The registration energy is shown to corresponds to the seasonal influence of the mixed layer depth revealed here by atmospheric forcing. These results show us that the SST-derived SSH reconstruction using the surface quasi-geostrophic (SQG) approximation should take into account stratification effects especially during summer. As discussed, the proposed methodology can enforce the reconstruction of sea surface current from a joint analysis of satellite altimetry data and high-resolution satellite-derived SST data.