Synergistic exploitation of the methane product from Sentinel-SP for applications in the Arctic (STEPS)

The main goal of this feasibility study was to evaluate the potential of adding value to the Sentinel 5P TROPOMI methane product over Norway and the Arctic through the synergistic use of relevant observations from other Sentinel satellites and machine learning. We assessed the data availability of E...

Full description

Bibliographic Details
Main Authors: Stebel, Kerstin, Kylling, Arve, Schneider, Philipp, Ytre-Eide, Martin
Format: Report
Language:English
Published: NILU 2022
Subjects:
Online Access:https://hdl.handle.net/11250/2998039
Description
Summary:The main goal of this feasibility study was to evaluate the potential of adding value to the Sentinel 5P TROPOMI methane product over Norway and the Arctic through the synergistic use of relevant observations from other Sentinel satellites and machine learning. We assessed the data availability of ESA operational and research-based WFMD XCH4 products over the Northern hemisphere, the Nordic countries and the Arctic/Northern latitudes. ESA’s XCH4 data have poor coverage over Norway. Seeing the two datasets as complementary, seems to be the most reasonable approach for utilization them. Furthermore, we investigated potential synergies between satellite products from different platforms. A random forest (RF) machine learning algorithm was implemented. It shows the importance of daytime land surface temperature (LST) as predictor variable for CH4. Our results indicate that the RF-model has a very good capability of filling small gaps in the data. Hovedmålet med denne mulighetsstudien var å evaluere potensialet for å tilføre verdi til Sentinel 5P TROPOMI-metanproduktet over Norge og Arktis gjennom synergistisk bruk av relevante observasjoner fra andre Sentinel-satellitter og maskinlæring. Vi vurderte datatilgjengeligheten til ESAs operasjonelle og forskningsbaserte WFMD XCH4-produkter over den nordlige halvkule, de nordiske landene og de arktiske/nordlige breddegrader. ESAs XCH4-data har dårlig dekning over Norge. Å se de to datasettene som komplementære, ser ut til å være den mest fornuftige tilnærmingen for å bruke dem. Videre undersøkte vi potensielle synergier mellom satellittprodukter fra ulike plattformer. En maskinlæringsalgoritme (Random forest, RF) ble implementert. Den viser viktigheten av jordoverflatetemperatur (LST) på dagtid som variabel for CH4. Våre resultater indikerer at RF-modellen har en meget god evne til å fylle små hull i dataene. publishedVersion