Assigned Session: #AGM28: Generic Meeting Session
Gaussian Process Regression for ICESat-2 Point-Cloud Interpolation
Abstract ID: 28.7300 | Accepted as Poster | Poster | 2025-02-27 13:00 - 14:30 | Ágnes‐Heller‐Haus/Small Lecture Room
Thorsten Seehaus (0)
Seehaus, Thorsten (1), Gardner, Alex (2)
Thorsten Seehaus (1)
Seehaus, Thorsten (1), Gardner, Alex (2)
1
(1) Friedrich-Alexander-Universität Erlangen-Nürnberg, Wetterkreuz, Erlangen, Germany
(2) NASA Jet Propulsion Laboratory, Pasadena, USA
(2) NASA Jet Propulsion Laboratory, Pasadena, USA
This study investigates the application of Gaussian Process Regression (GPR) for interpolating ICESat-2 ATL11 land ice height observations. ICESat-2 provides high-quality but spatially sparse measurements of ice surface elevation. GPR, with its ability to model complex spatial dependencies, offers a promising approach for interpolating these sparse observations and generating high-resolution maps of ice surface elevation. Two study sites, the Larsen-B embayment on the Antarctic Peninsula and a region in Southern Svalbard with numerous surging glaciers, were selected to test the performance of GPR for glacier volume change analysis. Various predictor sets and GPR kernels were evaluated and compared to contemporaneous surface elevation change measurements from TanDEM-X over a period of several years. Additionally, yearly surface elevation change maps were derived from ICESat-2 point measurements to investigate the temporal evolution of glacier volume changes. Preliminary results demonstrate the potential of GPR to accurately interpolate ICESat-2 ATL11 data, enabling more comprehensive and spatiotemporally continuous monitoring of ice sheet and glacier balances.
![]() |
||||||||
|