En ny metode til kortlægning af snedybder med satellitdata
Session
Studerende har ordet
Abstract
Viden om sæsonbestemt sneophobning er nøglen til forvaltning af vandressourcer, især i bjergområder. Imidlertid er nøjagtige målinger af snedybde eller SWE med en høj spatiotemporal opløsning sparsomme.
I dette indlæg undersøger vi effektiviteten af en multi-satellit-tilgang til på stor skala at kortlægge snedybden af den dybe sne i de sydskandinaviske bjerge i Norge. Først måles snedybder ved hjælp af geolokaliserede fotoner fra ICESat-2-satellitten. Derefter oprettes et indeks, der korrelerer med ændringer i dybden af dyb sne fra "C-bånd SAR"-udstyret på Sentinel-1-satellitterne.
Vi har udviklet en snedybdemodel ved hjælp af polynomieregression baseret på forholdet mellem de ICESat-2-afledte snedybder og Sentinel-1-indekset. Den resulterende model er i stand til at kortlægge snedybder ved en opløsning på 500 m hver sjette dag med en RMSE på 36 cm - sammenlignet med lokale vejrstationer.
Målgruppe
Alle deltagere ved Kortdage 2022 med interesse for remote sensing, kryosfæren, og masser af satellitdata.
Yderligere uddybning af abstract
The presentation by Rasmus P. Meyer and Mathias P. Schødt is the result of a 60 ECTS Geography & Geoinformatics master’s thesis project, developed during the academic term of 2021–2022 at the University of Copenhagen. Initially, the project’s point of departure was taken on the premise of deriving snow depths in mountainous regions using spaceborne LiDAR from the ICESat-2 satellite. This was later expanded by investigating the relationship between snow depths and radar polarimetry from the Sentinel-1 satellite. The reason being, that if we could combine the two satellite data sources, we might be able to model snow depths at a much higher spatiotemporal resolution. This train of thought resulted in the following two questions on which we based most of the research:
- How can snow depths be derived from ICESat-2 data and how accurate are they?
- How can accurate snow depth maps with a high spatiotemporal resolution be produced by combining ICESat-2-derived snow depths with Sentinel-1 PolSAR using regression modeling?
As part of the model validation process, we spent five days in the mountainous regions near the Hardangervidda plateau in southern Norway. The field excursion resulted in the collection of 300 snow depth measurements, and gave us a better understanding of our study area and the spatial variability of snow depths, which was crucial for the model development. We will present and discuss the results and methods used to develop our snow depth model.
