This paper addresses the issue of indexing and querying spatiotemporal raster data. This is done by developing a prototype named \apptitle{} that allows users to make queries on historical precipitation data, e.g., to determine the origin of water masses that has caused a flooding. Three functionally equivalent implementations have been developed. One is based on the GeoRaster component for Oracle Database, providing a performance baseline. Another is based on storing single pixel values along with their corresponding coordinates in a table. The last one is similar to the latter, but uses the Hilbert space-filling curve for indexing two-dimensional raster space. Among the key findings is that the Hilbert space-filling curve is not well-suited when raster data is sparse, nor when data is queried using small, non-rectangular search windows. Furthermore, the GeoRaster-based implementation has a storage requirement that is four times bigger compared to the other two, and a query performance that is slower by several orders of magnitude.