Karstens, Ute: Assessment of regional atmospheric transport model performance using ²²²Radon observations
1. Assessment of regional atmospheric transport model
performance using 222Radon observations
Ute Karstens1, Ingeborg Levin2, Michel Ramonet3, Christoph Gerbig4, Alessandro Capuana2, Sebastién
Conil6, Julian Della Coletta2, Arnoud Frumau7, Maksym Gachkivskyi2, François Gheusi8, Victor Kazan3,
Dagmar Kubistin5, Matthias Lindauer5, Morgan Lopez3, Lars Maurer2, Nikos Mihalopoulos9, Jean-Marc
Pichon10, Gerard Spain11, and Scott Chambers12
Correspondence: ute.karstens@nateko.lu.se
ICOS Science Conference 2020 Theme 4: Innovation and uncertainty in observation, analysis and budgeting techniques 15-17 September 2020
1: ICOS Carbon Portal, Lund University, Lund, Germany; 2: Institut für Umweltphysik, Universität Heidelberg, Germany; 3: Laboratoire des Sciences du Climat et de
l’Environnement, Gif sur Yvette, France; 4: Max Planck Institute for Biogeochemistry, Jena, Germany; 5: Deutscher Wetterdienst, Observatorium Hohenpeißenberg,
Hohenpeißenberg, Germany; 6: DRD/OPE, Andra, Bure, France; 7: Netherlands Organisation for Applied Scientific Research (TNO), Petten, The Netherlands; 8:
Observatoire Midi-Pyrénées, Toulouse, France; 9: Department of Chemistry, University of Crete, Greece; 10: CNRS, Clermont Ferrand, France; 11: National University of
Ireland, Galway, Ireland; 12: ANSTO Environmental Research, Lucas Heights, Australia
2. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
• Inversions provide a way of estimating trace gas fluxes and emissions from
measurements of their atmospheric concentrations
• Inversions rely on realistic simulations of atmospheric transport
• Inversions require quantitative knowledge of transport model uncertainties
➔ Assessment of transport model performance
➔ Radon as tracer for atmospheric transport
2
Introduction
3. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
• Radon has a fairly homogeneous surface source (compared to greenhouse gas fluxes)
• Radon is a decay product of Uranium, a component of all natural soils
• Exhalation rate of Radon depends on soil type and permeability
• Oceanic Radon fluxes are orders of magnitude smaller than fluxes from soils
• Radon has a well-defined sink
• Radioactive decay with a half-life of 3.82 days
• longer than turbulent time scales (~conservative tracer in atmospheric boundary layer)
• short enough to build up a vertical gradient in the troposphere
• Due to these characteristics, Radon is applied in atmospheric mixing studies and also as
tracer for the evaluation of transport model performance.
However, this requires knowledge of the continental Radon exhalation rate.
Radon as tracer for atmospheric transport
3
4. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
• Climatological Radon flux map based on Karstens et al. (2015)
• Radon source ~ Uranium content
• Soil texture, porosity
• Soil moisture (land surface reanalysis products)
Radon flux map for Europe
222Rn exhalation rate January 222Rn exhalation rate July
spatial variations
temporal variations
The red line indicates a constant Radon flux
(1 atom m-2
s-1
) that is often used in global
transport modelling studies.
4
Average 222Rn exhalation
from European land surface
5. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
Radon measurements in the ICOS
atmospheric station network
Trainou
KIT
Cabauw
Gartow
• Recommended measurement component
• Four tower stations selected here for model
comparison:
• Cabauw (200 m a.l.g., 222Rn: ANSTO)
• Gartow (134 m a.g.l., 214Po: HRM)
• Trainou (180 m a.g.l, 222Rn: ANSTO)
• KIT Karlsruhe (200 m, 100 m, 214Po: HRM
and 30 m 214Po: HRM and 222Rn: ANSTO)
• All HRM 214Po data corrected for line loss (Levin et
al., 2017) and disequilibrium and brought to ANSTO
scale (Schmithüsen et al., 2017)
*HRM = Heidelberg Radon Monitor
5
6. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
• Stochastic Time-Inverted Largrangian Transport model
• covers Europe on 1/12° x 1/8° grid (10 km x 10 km)
• driven by ECMWF operational analysis
• routinely used in regional inversions, e.g. in Jena CarboScope inversion system
• available as online tool at ICOS Carbon Portal https://stilt.icos-cp.eu/viewer
6
Transport model example: STILT
7. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
Summer and winter episodes of atmospheric Radon
black line: Observations
blue line: STILT model simulation results using the climatological Radon flux map
7
Variations on different time scales:
Diurnal cycle
• vertical mixing processes in the
atmospheric boundary layer
• amplitude sensitive to strength of
local Radon fluxes
Synoptic scale
• variations in air mass origin, i.e.
continental versus marine air
• regional differences in Radon fluxes
Seasonal cycle
• seasonal variations in atmospheric
transport patterns
• Radon fluxes smaller in winter and
early spring due to higher soil
moisture
8. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
Seasonal cycle of observed and modelled daytime Radon
KIT tall tower
at 200, 100, and 30 m
black line: Observations
blue line: STILT model simulation results using the climatological Radon flux map
red line: STILT model simulation results using a constant Radon flux from land surfaces
daytime (12-16 UTC ≙ 13-17 local time) data was used
8
Cabauw tall tower at 200 m
Trainou tall tower at 180 m
Gartow tall tower at 132 m
9. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
Seasonal cycle of observed and modelled daytime Radon
black line: Observations
blue line: STILT model simulation results using the climatological Radon flux map
red line: STILT model simulation results using a constant Radon flux from land surfaces
daytime (12-16 UTC ≙ 13-17 local time) data was used
9
Climatological Radon flux
➔ systematic underestimation
➔ shape of seasonal cycle fits
better
Constant Radon flux
➔ better fit in winter
➔ underestimation in summer
10. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
Seasonal cycle of observed and modelled daytime Radon
black line: Observations
blue line: STILT model simulation results using the climatological Radon flux map
dashed blue line: STILT simulations with a scaled climatiological Radon flux map
red line: STILT model simulation results using a constant Radon flux from land surfaces
daytime (12-16 UTC ≙ 13-17 local time) data was used
10
Scaling experiment
Doubling of Radon flux
improves overall agreement
➔ Deficiencies in the flux map ?
Climatological Radon flux
➔ systematic underestimation
➔ shape of seasonal cycle fits
better
Constant Radon flux
➔ better fit in winter
➔ underestimation in summer
11. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
Diurnal cycle of observed and modelled Radon in winter
black line: Observations
blue line: STILT model simulation results using the climatological Radon flux map
dashed blue line: STILT simulations with a scaled climatiological Radon flux map
11
Winter
• atm. Radon fairly constant
• simple flux scaling
improves agreement at
highest level (at most sites)
• model shows diurnal cycle
at KIT 30m
y-axis scales are different !
12. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
Diurnal cycle of observed and modelled Radon in summer
black line: Observations
blue line: STILT model simulation results using the climatological Radon flux map
dashed blue line: STILT simulations with a scaled climatiological Radon flux map
12
Summer
• pronounced diurnal cycle
• simple flux scaling
- better agreement during
daytime
- stronger overestimation
during night
• phase shift between model
and observations
➔ Deficiencies in the vertical
mixing processes in the
model ?
y-axis scales are different !
13. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
Vertical Radon profiles at KIT tower and Cabauw tower
13
sunrise sunset sunrise sunset
modeled Rn at 20 and 200 m
observed Rn at 20 and 200 m
modeled Rn at 30, 100 and 200 m
observed Rn at 30, 100 and 200 m
No flux scaling in order to focus on potential explanations for the mismatch in the model’s process representation
8/2018 8/2018
14. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
Vertical Radon profiles at KIT tower and Cabauw tower
14
sunrise sunset sunrise sunset
modeled Rn at 20 and 200 m
observed Rn at 20 and 200 m
modeled Rn at 30, 100 and 200 m
observed Rn at 30, 100 and 200 m
No vertical gradient during daytime ➔ vertical mixing probably too strong in the model
At KIT overstimation of accumulation close to the surface ➔ vertical mixing probably too weak during night
8/2018 8/2018
15. ICOS ERIC Carbon Portal, Lund, Sweden | www.icos-cp.eu
Radon observations can help in the assessment of atmospheric transport models
but we need to disentangle deficiencies in model transport and deficiencies in Radon fluxes
• Preliminary results point to some important deficiencies in the STILT model:
• Vertical mixing starts too early in the morning
• Vertical mixing seems slightly too strong during daytime, not maintaining a (small) vertical
gradient in the lower boundary layer
• Vertical mixing during night can be too much suppressed leading to too high activity
concentrations accumulating in the lowest boundary layer
➔ Boundary layer parameterizations in the transport model and/or in the atmospheric models
that produce the driving meteorological fields need to be improved
• The climatological Radon flux map generally underestimates real fluxes from soils in Europe
➔ An improved Radon flux map will be developed and validated with Radon flux
measurements in the EMPIR project 19ENV01 traceRadon
15
Conclusions