In this talk I present a prototype system for the management of earth science data which is novel in that it takes a DBMS centric view of the task. Our prototype -- called BigSur -- is shown in the context of its use by two geographically distributed scientific groups with demanding data storage and processing requirements. BigSur currently stores about 1 Terabyte of data, about one thousandth of the volume required by the NASA EOSDIS system. We claim that the design principles embodied in BigSur provide sufficient flexibility to achieve the difficult scientific and technical objectives of Mission to Planet Earth. The talk discusses the design of Big Sur, its current status and lessons we have learned in the process.