Klosterhalfen, Anne: Two-level Eddy Covariance Measurements Improve Land-atmosphere Flux Exchange Estimates over a Heterogeneous Boreal Forest Landscape
1. TWO-LEVEL EDDY COVARIANCE MEASUREMENTS
IMPROVE LAND-ATMOSPHERE FLUX EXCHANGE
ESTIMATES OVER A HETEROGENEOUS BOREAL
FOREST LANDSCAPE
Anne Klosterhalfen, Jinshu Chi, Natascha Kljun, Anders Lindroth, Hjalmar Laudon,
Mats B. Nilsson, Matthias Peichl
ICOS Science Conference 2020 | September 17th, 2020
2. 2
MOTIVATION AND OBJECTIVE
• In its original theory, a homogeneous flux footprint area is a key
assumption for eddy covariance (EC) measurements. Still, the EC
technique is also applied over complex and non-homogeneous terrain.
→ What source area determines the observed fluxes?
• Footprint areas usually vary between day- and nighttime due to differing
atmospheric conditions. Thus, flux source areas and their relative
contribution to the net exchange differ, potentially resulting in a bias at
the diel scale.
→ How large is this potential bias in the surface fluxes due to diverging
day- and nighttime footprints?
3. 3
• 64° 15’ N, 19° 46’ E, 267 m a.s.l.
• heterogeneous boreal forest landscape
• managed forest dominated by Scots pine,
Norway spruce, and some birch
• clearcuts, peatlands, lakes, streams,
grasslands
• cold temperate humid climate
• 1.8°C mean annual Tair
• 614 mm mean annual P sum
SE-Svb – Svartberget Tall Tower
SITE DESCRIPTION
Krycklan catchment
A
4. 4
85 m - day
85 m - night 60 m - day
60 m - night
B
SE-Svb – Svartberget Tall Tower
SITE DESCRIPTION
2 tall tower eddy covariance systems
• uSonic-3 Omni sonic anemometer
(METEK), FGGA-24EP closed-path
greenhouse gas analyzer (LGR)
• CSAT3 3-D sonic anemometer,
EC155 closed-path gas analyzer
(Campbell)A
5. 5
Combination of 2-level eddy covariance measurements
METHODS
• study period:
01 September 2018 – 15 July 2019
processing of
half-hourly fluxes
(EddyPro®)
flux gap-filling,
source partitioning of CO2 fluxes
based on nighttime fluxes
(REddyProc, WUTZLER et al. 2018)
NEE, Reco, GPP, LE, H
single- and combined-
level flux data
first quality check,
adding storage terms,
filtering for advection (WHARTON et al. 2009)
10 Hz EC raw
data
profile, atmos.,
meteorol., soil
sensors data
data gap-filling
(4 steps)
data aggregation to different
time steps
(hourly, daily, monthly,
cumulated, diurnal)
combination of data from
60 and 85 m based on
day- and nighttime
Flux Footprint Prediction
(KLJUN et al. 2015)
single-level
flux measurements from
60 and 85 m
footprint estimates
for 60 and 85 m,
for day- and nighttime
85 m
60 m
daytime
nighttime
1
2
3
BA
6. 6
• footprint climatology of the entire study period
for day- and nighttime individually
→ difference between day- and nighttime footprint area (80% contour line):
• 85 m-level: ≈ 26 km2
• 60 m-level: ≈ 5 km2
• combined-level: ≈ -4 km2
85 m level
60 m level
combined-
level data
2-level combination based on day- and nighttime
FOOTPRINT SIMULATIONS
7. 7
• footprint climatology of each season for day- and nighttime individually
→ for summer and transition months the mismatch between day- and nighttime footprints was decreased
→ for winter months the day- and nighttime footprints were similar at one given height, and the combined-level approach increased the
footprint variability in the diel course
2-level combination based on day- and nighttime
FOOTPRINT SIMULATIONS
(b) transition
Mar, Apr / Sep, Oct
(a) winter
Nov - Feb
(c) summer
May - Jul
8. 8
(a) winter
Nov - Feb
(b) transition
Mar, Apr / Sep, Oct
(c) summer
May - Jul• diurnal course of
atmospheric stability,
footprint area, and area
fraction of each land
cover type for both
measurement heights
• for each season
individually
→ in the following results, we will
exclude the winter months!
2-level combination based on day- and nighttime
FOOTPRINT SIMULATIONS
85 m-level
60 m-level
forest
clearcuts
peatlands
lakes, grasslands,
residential areas
stable
convective
neutral
9. 9
Comparison of combined- and single-level data
GAP-FILLED FLUX DATA
85 m-level
combined-level
• fluxes in daily time steps and cumulated
transition summer transition summer
10. 10
Comparison of combined- and single-level data
GAP-FILLED FLUX DATA
• relative difference in fluxes between combined-level and 85 m-level data shown as
Δflux = (flux(85 m) - flux(combined-level)) / daily flux(combined-level) ⸱ 100
NEE GPP Reco LE H
smaller fluxes due to
diverging footprints
larger fluxes due to
diverging footprints
transitionsummer
11. 11
Comparison of combined- and single-level data
GAP-FILLED FLUX DATA
• difference in fluxes between combined-level and 85 m-
level data in daily time steps and cumulated sums
• Δflux = flux(85 m) - flux(combined-level)
combined-
level
85 m level 60 m level
NEE (g C m-2) -139.5 -156.3 -231.2
GPP (g C m-2) -536.0 -513.9 -633.2
Reco (g C m-2) 396.5 357.6 402.0
LE (MJ m-2) 392.5 375.3 434.6
H (MJ m-2) 435.6 431.3 524.2
+12%
-4%
-9%
-4%
-1%
cumulated sums for 01 Sep 2018 – 16 Jul 2019
(excluding Nov – Feb)
transition summer
-13% -0.4%
-4%
-11%
-5%
-2%
8%
-9%
-2%
-48%
12. 12
CONCLUSIONS AND OUTLOOK
• The combination of the 2-level measurements decreased the variability of the footprint areas in the diurnal
course. Also, the fractions of different land cover types in the footprint area matched better between day- and
nighttime.
• The 2-level data combination had an impact on finale fluxes. Cumulated sums changed by up to 12%.
→ Monthly and seasonal differences could be observed.
• combination of 2-level measurements based on day- and nighttime and based on atmospheric stability (at
least during winter)
• sensitivity analysis for Δflux regarding day length, seasons, flux magnitudes, or source area
• application of different source partitioning approaches
• implications for energy balance closure
• application to additional sites
13. 13
THANK YOU FOR YOUR ATTENTION!
We thank the staff from the Svartberget Field Station for their
continuous support, data acquisition, and instrument maintenance!