Skip to main content
U.S. flag

An official website of the United States government

Publication Date
16 January 2024

Truly Conserving with Conservative Remapping Methods

A procedure for correcting and applying weights obtained from remapping packages ensures that area means and integrals are preserved even when grid cells are partially or fully masked.
Print / PDF
Powerpoint Slide

Conservative mapping of data from one horizontal grid to another (“conservative regridding”) may be essential when exchanging mass and energy between coupled model components and in monitoring the mass of trace constituents or a model’s energy imbalance. Most existing algorithms commonly relied on to remap data from one horizontal grid to another conservatively may for certain grids, however, be unable to preserve the true integral or mean properties.  This is true if an algorithm incorrectly reconstructs grid cell shapes.  The errors can be unacceptably large, at least for relatively low-resolution grids.  A calculational procedure and a method for correcting the errors in the weights are proposed to remedy this common shortcoming of existing regridding packages.  


The guidance provided in this paper will enable those requiring conservative remapping to eliminate difficulties in the successful application of the weights that are generated by existing regridding packages. Application of the described procedure successfully preserves the mean or integral property of interest even when grid cells are partially or fully masked and when the weights are based on flawed reconstructions of grid cell shapes.


We have shown that despite any inaccuracies in a remapping package’s reconstruction of grid cell shapes, careful application of the weights, unmasked fractions, and (when needed) the application of a correction can preserve the true global integrals or means of interest with only minor distortion of the original field.  The treatment of partially or fully masked grid cells enables the application of a single set of pre-determined weights to fields where masks may vary over time, thereby avoiding their computationally expensive recalculation.  It is hoped that future enhancements to remapping packages will obviate the corrective step currently required in the treatment of certain grids.

Point of Contact
Karl E. Taylor
Lawrence Livermore National Laboratory
Funding Program Area(s)