Initial tools developed for grid administrators and users have built on the technology and representational techniques of large parallel systems. Like their predecessors, grid tools must cope with extreme variations in scale, rapidly evolving hardware and software environments, and the competing demands of operating systems and middleware. Computational grids present several unique challenges, however, that go well beyond the lessons we have learned from parallel and distributed tools: the volatile nature of grid resources, their extreme heterogeneity, and the lack of coordinated management. Because they define a new and unfamiliar computing environment, there is a significant human challenge as well. Grid users will be extremely diverse, including resource providers, resource managers, users of data and derived data products, etc., as well as application developers. The future usability of the grid will depend on how well grid tools can capture information on grid resources and synthesize a higher-level perspective that helps users make sense of this complex new environment. This article identifies the tool requirements that will have most impact on the usability. It then explores recent advances in information visualization, demonstrating that many of the techniques grid tools will need already exist in preliminary form. A series of examples illustrates how those techniques can be applied to portray grid landscapes (graphical representations of grid resources, activities, behavior, costs, etc.) in useful and meaningful ways.