ClimAG-Krigger: A new (paleo)Climatology-oriented toolbox for Anisotropic Global Kriging interpolation

COSENTINO, N. J.; LAMBERT, F.; OSSES, AXEL; VAN T' WOUT, ELWIN

Simposio; Blowing South: Southern Hemisphere Dust Symposium; 2021

IFAECI, Núcleo Mileno Paleoclima, AACS, CR2, UNLPam, INCITAP, CICTERRA, CESIMAR, GEOTRACES, ipam

Data-model comparisons are common when addressing (paleo)climate questions. Many applications requirederiving continuous surface fields of scalar variables from a set of irregularly distributed data points,typically for model validation against data or data-derived model input as initial or boundary conditions.While various interpolation techniques and interfaces exist, few can simultaneously: (1) interpolate acrosslocal to global spatial scales, (2) perform anisotropic interpolation using the spatial structure derived fromthe data instead of an assumed one, and (3) explicitly derive uncertainty in the interpolated fields due to bothdata density and measurement error. We present a standalone interpolation toolbox including a graphicaluser interface (GUI), which is aimed at the general earth science community. It uses a kriging algorithmwhose distance metric is the geodesic on an oblate spheroid, be it the WGS-84 reference ellipsoid for applications on the surface of the Earth, or an equivalent ellipsoid with varying radii for interpolation onvertical levels above the surface. While kriging algorithms exist that perform interpolation on such non-Euclidean distances, they do not provide a check for conditionally negative semi-definiteness (CNSD) of thevariogram matrix, which is a requisite for the kriging method. Since mathematical theory of kriging onspheroids or ellipsoids has not yet provided a set of authorized variance-distance functions, we incorporateda numerical check for CNSD condition for each data realization and variance-distance modeling scheme.The GUI will allow the user a high degree of customization. Preliminary results are promising, with robustresults for isotropic interpolation. The derivation of CNSD variogram matrices for anisotropic interpolationremains the major challenge of the project. When completed, ClimAG-Krigger will provide the communitywith an easy-to-use, robust tool for anisotropic global kriging that will be specifically tailored for(paleo)climate applications. In particular for this symposium, applications to global dust deposition since theLast Glacial Maximum will be discussed.