The Madagascar seasat dataset presents a problem where data are collected along crossing tracks. These tracks are not straight, and appear to be irregular in the model space. Previous methods assumed that the data were regularly sampled in the model space coordinate system. I warp the model space to look more like the data space, so that prediction-error filters can be estimated in the more regularly-sampled data space. I try two different approaches to this problem on the warped space, one where the data is single-valued and fixed, while the other is multivalued and allowed to vary. The former method works very well, while the second one works well.