Hi Lev, Have you looked at the Ferret functions "curv_to_rect_map" and "curv_to_rect"? I don't know in what way your files have broken specifications, but see whether they have the information needed for these calls. The map function creates a set of weights from a source
curvilienar grid to a destination rectilinear grid that then can
be saved and reused to map different data fields on the same
source grid to the destination grid. The map function is slow to
run, but then applying it is fast. For finding a subset of a curvilinear grid see the ideas in this
FAQ:
http://ferret.pmel.noaa.gov/Ferret/faq/plotting-subsets-of-data-on-a-curvilinear-grid The "curv_to_rect_map" and "curv_to_rect" functions use a good conservative mapping method. There are newer and faster methods, and we have one implemented in PyFerret which requires a Python regridding module. Please see this page if you're interested: http://ferret.pmel.noaa.gov/Ferret/documentation/pyferret/ferret-functionality#PYFERRET_REGRID_FUNCTIONS It is on our long list of future improvements to implement the ESMF regridding functionality as a faster method for regridding, and implementing it in the "auxiliary regridding" syntax already available for regridding along single 1-D axes, so that it would appear as a regular regridding operation within the core of Ferret and PyFerret instead of function calls. Ansley On 4/24/2018 9:33 AM, Lev Tarasov
wrote:
This is unfortunately a too frequent problem at least for data sources I'm using. The source files tend to be on curvilinear grids (eg DEM, or obscure GCM/RCM grid or some other curvilinear projection) and the internal net-cdf grid specs are broken so that simple regridding doesn't work. However, the dat set includes grid-center lat and lon for |