The database research community's appetite for new applications has led to increased interest in the data management needs of scientists. This area encompasses a huge range of applications, extending from public repositories of observational data such as the popular Sloan Digital Sky Survey to one-of-a-kind runs of simulation codes crafted by individual scientists. In this talk, we will survey the most common data management needs found in the hard sciences, describe the new database research challenges that arise from these needs, and outline ways to address some of these challenges.