ICOS Atmospheric Stations: Spatial Characterization of CO2 Footprint Areas and Evaluating the Uncertainties of Modelled CO2 Concentrations

Abstract

The main purpose of this thesis is to present and analyze spatial characteristics of 31 different European atmospheric measurement stations in the ICOS (Integrated Carbon Observation System) station network. The characterization includes quantification of where air arriving at the stations can be expected to have come from, as well as a breakdown of what these areas cover with regards to land cover, point source emissions and population. A dataset regarding emissions of radiocarbon at nuclear power plants has also been processed because possible transports of radiocarbon from these facilities to the stations need to be accounted for when quantifying fossil fuel emissions based on measured ratio between 14C and 12C.

“Where the air arriving can be expected to have come from” for a specific date and time is synonymous with a station’s footprint. For a general characterization based on annual values, an average footprint based on all three-hourly footprints in the year is used. Each individual footprint has already been combined with data on anthropogenic emissions and a model that quantifies the biospheric component to estimate the CO2 concentration at the stations. The known sources that make up the total CO2 signal also allows for a breakdown into different categories of contribution. Averages of these are also part of the characterization. The annual averages for the different characterizations vary greatly between the stations: station Pallas’ anthropogenic contribution is only 2.2% of the estimated contribution at the station with the highest value, Heidelberg. Furthermore, there are large variabilities between the footprints at the individual stations that are used to generate annual averages. Individual footprints values are used in closer analysis exemplified for selected stations. However, years 2016 and 2017 show similar annual values for the different stations which indicate stability of the annual characterizations.

Dominant land cover classes in the model domain, including ocean, cropland, pastures and different types of forests, are found within the annual footprints of all stations. This is no surprise because of the large spatial extent of most footprints: on average 48% of the sensitivity is to the area within 300 km of the station, again with large differences between stations and between individual footprints of the stations. The large footprint extent is also why the spatial characterization is extra important: it is not possible to know what is within the footprint areas without it. Knowing what is in the area in close proximity to the stations is not enough.

Popular Abstract

ICOS (Integrated Carbon Observation System) is a measurement system for greenhouse gas observations around Europe. There are different station types including atmospheric, ecosystem and ocean. Atmospheric stations generally measure high above ground which make for big footprints. Footprints are generated by a transport model and indicate the area where travel of the air on the way to the station was close to the ground. When the air travels close to the ground its concentration of CO2 will potentially be influence by CO2 emissions as well as exchanges – fluxes –with the biosphere. In this project 31 different stations have been characterized in terms of their footprints. The average footprint areas for 2017 were used to establish the influence of regions close to the stations as opposed to further away. There are big differences between the stations, but the footprints are generally extensive with an average of 50% of the sensitivity within 300 km of the stations. There are also differences with regards to how much of the air traveled close to the ground. When there is essentially no travel in this area the station can be expected to measure the CO2 concentration of the clean background air.

Next, the footprint areas were analyzed in terms of the areas they cover:
• A land cover breakdown into 19 different land cover classes was established for each of the stations. On the whole, the highest ranked land cover the stations are sensitive to are 1) oceans, 2) staple crops except rice, 3) coniferous forests, 4) broad leaved forests. There are of course big differences between stations, with for instance the station located in the Mediterranean (Lampedusa) basically having no sensitivity to the different forest types.
• Population density was also analysed since it is a good indicator of the anthropogenic emissions influencing CO2 concentrations at the stations.
• Emissions from point sources such as power plants were also considered. Emission quantities within the footprint areas were translated into an average influence on the CO2 concentration from these.

There are already models that use the footprints in combination with emission statistics and a model that simulates the fluxes between the biosphere and atmosphere to estimate what the CO2 concentrations at the stations (or anywhere in the model domain) should be. This is called inverse modelling. Results from ICOS-CP’s inverse modelling were used to generate averaged influences from anthropogenic emissions and the biosphere for 2017 for the 31 stations. Uncertainties of the modelled CO2 concentrations are considered in the second part of the thesis. The difference between the ICOS CO2 measurements at the stations were compared to the modelled concentrations. The resulting differences were in turn quantified by hour and by month to expose potential trends in model performance and differences between stations. Generally, it more common that the model underestimates than that it overestimates the concentrations. Daytime estimates are generally closer than night-time estimates, and the same is true for the summer months compared to the winter months.

The differences between model and measured concentrations stem from uncertainties associated with the footprints, emissions statistics, and biospheric model. These can never be fully disentangled, but feedback from characterization of individual footprints or subsets of footprints can possibly help better pinpoint some of the likely sources of errors. An example of this is included in the thesis: footprints associated with high overestimates at the station located in Heidelberg could be related to a specific area in the model domain. In turn, our understanding of fluxes and emissions will hopefully increase and with it, our models will improve.

All the code scripts used to generate the results and figures of the thesis have been set up as user friendly Jupyter Notebooks. Additional stations can be analyzed given different date ranges, or given an individual footprint representing a specific time. Also, potential station location can be analyzed as a part of the decision making process.

• author: Storm, Ida
• supervisor: Ute Karstens; Harry Lankreijer
• organization: Dept of Physical Geography and Ecosystem Science
• alternative title: Characterizing atmospheric stations by their footprint areas and Evaluating the Uncertainties of Modelled CO2 Concentrations
• course: GISM01 20201
• year: 2020
• type: H2 - Master's Degree (Two Years)
• subject: Earth and Environmental Sciences
• keywords: Geography, GIS, ICOS, Carbon dioxide, Modelling
• publication/series: Master Thesis in Geographical Information Science
• report number: 116
• language: English
• id: 9007298
• date added to LUP: 2020-04-03 17:00:52
• date last changed: 2020-04-06 10:00:53