Smear: An open source, parallel python toolkit for spatial interpolation of large irregular two-dimensional scattered datasets

Geophysical datasets collected to study the Earth often come in the form of two-dimensional scattered points to which measurements are attached. These datasets often have highly irregular spatial distributions. Examples include bathymetric data, which are normally collected at high sample rates along a boat track but can have significantly large distances between each boat track. Rain gauge networks also tend to be very spatially irregular, where the locations of gauges are often determined by factors like funding and available cooperators. When working with these types of data, values for unsampled locations must be interpolated for use in visualization products, simulation models or aggregations to larger spatial units. The properties of the feature being modeled and the spatial distribution of measurements determine which interpolation methods are appropriate. For example, anisotropic systems like river channel beds where bathymetric variability is greater transverse to the flow direction, are usually more accurately modeled by methods that account for the anisotropy.

Most current spatial interpolation software are commercial and are often accessible only through a GUI interface and are not easily scriptable or extendable. Smear was developed to provide an extendable open source toolkit for spatial interpolation.

Smear utilizes open source python scientific and GIS libraries to implement several of these algorithms, including a channel-following anisotropic interpolation algorithm. Large datasets are efficiently managed through the use of k-dimensional trees for efficient neighborhood searches and a parallel implementation based on the IPython parallel architecture allows use of multicore systems and HPC clusters.

At the Texas Water Development Board, this toolkit has been used to improve lake volume and sediment estimates and to improve the representation of river and channel bathymetry within hydrodynamic models.