This article summarizes some topics of modeling as they impinge on the future development of scientific data visualization. The benefits from visualization techniques in analyzing data are well established, but to build on these pioneering efforts, one must recognize modeling as a distinct structural component in the larger context of visualization and problem-solving systems. Volume modeling is the entry way to this arena of future development, and model-based rendering describes how scientists will view the results. Important side developments such as multiresolution modeling and model-based segmentation will contribute structural capability to these systems. All of these components ultimately depend on the mathematical foundations of scattered data modeling and on model validation and standards to incorporate this modeling methodology into effective tools for scientific inquiry.