regridding

Regridding, also called remapping in some contexts, is the process of transforming data defined on one computational grid to another grid with a different resolution, orientation, or topology. It is widely used in numerical weather prediction, climate modelling, ocean modelling, and geographic information systems to enable data exchange between models or datasets that use incompatible grids. Regridding can operate on scalar fields (such as temperature or precipitation) or on vector and tensor fields, and may be one-way or bidirectional during model coupling.

Grids may be regular (such as latitude–longitude grids), curvilinear, cubed-sphere, or unstructured, and they may be designed to be equal-area or to preserve other geometric properties. The main objectives of regridding are to preserve important physical quantities (for example, conserving total mass or energy), minimize interpolation artefacts, and maintain stability and accuracy of the coupled system. The choice of method depends on the field and the physical constraints.

Two broad families of regridding methods are interpolation-based and conservative remapping. Interpolation maps values at destination grid points from nearby source values using nearest-neighbor, bilinear/trilinear, or higher-order schemes. Conservative remapping assigns values by integrating the source field over the exact overlapping areas between source and destination grid cells, producing area- or flux-conserving transfers. In practice, remapping weights are computed and applied, sometimes with masks for land/sea, and libraries such as ESMF Regridding or SCRIP provide implementations.

Challenges include numerical diffusion, potential loss of positivity, and spurious oscillations near sharp gradients or irregular boundaries. Regridding is computationally intensive if high accuracy is required and overlaps are complex. It is a central tool in model coupling and in processing, downscaling, and assimilation workflows, enabling data integration across disparate grid systems.