NOAA/AVHRR infrared satellite imagery is well suited to the study of mesoscale dynamic structures such as vortices or temperatures fronts. Since data are delivered on a daily basis, the problem of modeling structures in these images is an important matter. We use the spline vector field technique to compute a dense vector field on an image such that specific orbits of the field correspond to the boundary of fronts and vortices. This allows for a unique mathematical model able to characterize many complex structures in interaction. Moreover, the model is well suited for motion tracking. In this study we develop the image processing tools needed to handle such satellite image sequences, introduce the spline vector field technique, and show its ability to model complex structures on specific examples.