The turbulent motions in ocean interact in a broad continuous range of the spatial scales. To study the quantitative structures at multiple spatial scales in oceanic dynamics, the multi-level dissipation element (ML-DE) analysis is proposed. Considering a sea surface temperature (SST) field $\theta (x,y,t)$, from any spatial point, one can inevitably reach a local maximum temperature point and a local minimum one along the ascending and descending directions of the gradient of $\theta (x,y,t)$, respectively. A dissipation element (DE) is thus the region consisting of the points whose gradient trajectories share the same extremal point pair. The characteristic parameters are the temperature difference $\Delta \theta$ and the distance $\ell$ between the maximum and minimum points of each DE. ML-DE, which is an extension of DE analysis to any scale level $R$, can efficiently capture the geometrical features of the turbulent flow at different scale levels, which improves the understanding of the physics in ocean turbulence. In this work, ML-DE is applied to the SST field obtained for the South China Sea. The statistical results have been compared with those from the conventional analysis, e.g. the Fourier spectrum and structure function. The difference of the dynamic behaviors of SST in the winter and summer time is also addressed.