LUNDIsim: open-data faulted model meshes with porosity/permeability, for scientific data processing and flow simulation benchmarks.

@inproceedings{Duval2026RM,
 abstract = {Scientific data volumes, produced in numerical simulation workflows, increasingly become more massive. Huge numerical data sizes raise concerns in computability, interpretability, and sustainability. Earth science (geology, meteorology, oceanography, astronomy…) is notably impacted. We were motivated by a reservoir flow simulation context, with respect to model mesh compression. There: a) on the one hand, engineers often use upscaling to accelerate processing, from visualizing to computations yielding meaningful outputs, like water-cut curves; b) on the other hand, refining either mesh grid resolution or data precision was shown to drastically shrink model mesh storage or memory footprints. To evaluate HexaShrink [PEYROT-2019], a refinable scheme producing a hierarchy of meshes of increasing levels of details, we tailored the open dataset LUNDIsim [DUVAL-2026]. We named it after the Icelandic name (lundi) of the gentle Atlantic puffin, a nod to SKUA and Petrel simulation software suites, called after more competing seabirds. LUNDIsim represents a faulted geological mesh, inspired by the SPE10 comparative Challenge. It was conceived to ponder the impact of lossy data compression on simulation outcomes, on four distinct subsurface environments (Nearshore0, Nearshore1, Nearshore2, and Fluvial), provided in a MWE (Minimal Working Example), with ancillary data for reuse. It is also suitable for other advanced mesh processing workflows in geology and reservoir engineering, from visualization to machine learning, to help bridge objective data metrics to subjective result evaluation [DUVAL-2025].},
 author = {Duval, Laurent and Astart, Laurent and Preux, Christophe and Payan, Frédéric},
 booktitle = {2026 {RING} meeting},
 language = {en},
 pages = {316},
 publisher = {ASGA},
 title = {LUNDIsim: open-data faulted model meshes with porosity/permeability, for scientific data processing and flow simulation benchmarks.},
 year = {2026}
}