of 27
1
Supplementary Information
The weekly cycle of photosynthesis in Europe reveals the negative
impact of
particulate
pollution on ecosystem productivity
Liyin He
1
*
, Lorenzo Rosa
1
*
, David B. Lobell
2,3
, Yuan Wang
2
, Yi Yin
4
,5
, Russell Doughty
6
,
Yitong
Yao
4
,
Joseph A.
Berry
1
, Christian Frankenberg
4
,
7
1
Department of Global Ecology, Carnegie Institution for Science, Stanford,
CA
94305
,
USA
2
Department of Earth System Science, Stanford University, Stanford,
CA
94305
,
USA
3
Center on Food Security and the Environment, Stanford University, Stanford,
CA
94305
,
USA
4
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA
91125
, USA
5
Now at the
Department of Environmental Studies, New York University
,
New York, NY 10003
,
USA
6
College of Atmospheric and Geographic Sciences, University of Oklahoma, Norman, OK
73019,
USA
7
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA
91109
, USA
*Corresponding author:
lhe@carnegiescience.edu
and
lrosa@carnegiescience.edu
This PDF file includes:
Supplementary T
ext
S
1
Supplementary Figures 1 to
2
2
Supplementary Table
S
1
and S2
2
Supplementary Text
S
1
.
Quantifying Potential Increase in
Ecosystem Respiration due to
Temperature Rise under Reduced Aerosol Conditions
R
educed aerosol pollution can potentially lead to an increase in temperature, thereby enhancing
ecosystem respiration.
The key question is whether the increase in gross primary productivity
(GPP) outweighs the increase in respiration, resulting in
an
increase in net ecosystem productivity
(NEP), or vice versa. We address this question at both the flux tower scale and the continental
scale
in the following analysis
.
At the flux tower scale, we examine the changes in GPP and ecosystem respiration between
weekends and weekdays and find that, for most sites, both increase during the weekend. However,
the magnitude of the increase in GPP is larger than that of ecosystem r
espiration, leading to
an
increase in NEP (see
Supplementary Figure 15
).
At the continental scale, remote sensing techniques can only track GPP and not ecosystem
respiration. To approximate the change in ecosystem respiration due to elevated temperature, we
use a simplified
model
. We employ the well
-
established equation:
(
)
=
25
×
10
25
10
w
here
(
)
represents the
ecosystem
respiration at a specific temperature
,
25
denotes the
baseline respiration at
the
reference temperature of 25°C, and
10
represents
the factor by which
respiration
rate increases for every
10
°C
rise in temperature.
The ratio of ecosystem respiration between the weekend and weekday
is calculated
as:
(
푤푒푒푘푒푛푑
)
(
푤푒푒푘푑푎푦
)
=
10
푤푒푒푘푒푛푑
푤푒푒푘푑푎푦
10
The relative change in ecosystem respiration over the weekend compared to the weekday is
computed as:
(
(
푤푒푒푘푒푛푑
)
(
푤푒푒푘푑푎푦
)
1
)
×
100%
First, we calculate Q10 values using eddy covariance measurements at each flux tower site
(
Supplementary Table S1
). The derived Q10 values
spans
from a minimum of 1.0 to a maximum
of 3.0. Next, we employed an iterative approach to calculate the relative change in ecosystem
respiration over the weekend compared to the weekday. This involved exploring a range of Q10
values from 1.0 to 3.0 with a ste
p size of 0.2. The calculations were based on the weekend minus
weekday difference in air temperature derived from ERA5 daily air temperature.
The resulting Europe
-
averaged relative change in ecosystem respiration over the weekend was
plotted for each Q10 step
(
Supplementary Figure 22
)
. Notably, the analysis revealed a positive
correlation between Q10 values and ecosystem respiration. Specifically, when Q10 was set at 2,
ecosystem respiration increased by 0.2%, and when Q10 reached 3, the increase amounted to
0.35%. It is important to n
ote that these observed changes in ecosystem respiration are relatively
minor in comparison to the approximately
2.5% increase in SIF (or GPP) observed during the
weekend.
3
Supplementary Figure 1. Map of global
weekend minus weekday TROPOMI NO
2
during
2019, 2020 and 2021.
Spring refers to March, April, and May. Summer refers to June, July, and
August. Autumn refers to September, October, and November. Winter refers to December, January,
and February.
4
Supplementary Figure 2. Mean normalized TROPOMI NO
2
on different days of the week in
typical countries, including (a)
Europe,
(b)
the Middle East,
(c) Asia, and (d) countries in the
Americas, Africa
,
and Australia, during 2019, 2020, and 2021.
The values have been
normalized by dividing them with their corresponding mean values.
5
Supplementary Figure
3
. Map of weekend minus weekday (A) TROPOMI NO
2
and (B)
Apple
mobility index
(
B)
in Europe.
TROPOMI NO
2
is
taken in
2019 and 2021, while the
mobility data
covers the period from
Mar
ch
2020 to Mar
ch
2022. Apple mobility index
represents the traffic volume on the day
compared to a baseline period of January 13th, 2020
and
is taken from
https://github.com/ActiveConclusion/COVID19_mobility/
.
6
Supplementary Figure 4. Mean normalized Apple mobility index on different days of the
week in typical countries, including (a)
Europe
, (b)
the Middle East
, (c) Asia, and (d)
countries in the Americas,
Africa,
and Australia, during
March
2020 to Mar
ch
2022.
The
values have been normalized by dividing them with their corresponding mean values.
The data is
taken from
https://github.com/ActiveConclusion/COVID19_mobility/
.
7
Supplementary Figure 5. The difference of TROPOMI solar induced fluorescence, aerosol
optical depth, and absorbed photosynthetic active radiation in Europe during the COVID
pandemic (2020
)
between weekend and weekday.
The insert histograms show the distribution
of the corresponding variable.
8
Supplementary Figure 6. Percent changes between weekend and weekday of TROPOMI
solar induced fluorescence, aerosol optical depth and absorbed photosynthetic active
radiation in Europe during 2018, 2019 and 2021.
The insert histograms show the distribution
of the corresponding
variable,
and dashed black lines represent the median values
.
9
Supplementary Figure
7
. Map of weekend minus weekday TROPOMI NO
2
in Europe
during 2019 and 2021.
The left panel
shows
the raw values, and the right panel shows the
percentage change. The insert histograms illustrate the distribution of the
corresponding
variable
,
and dashed black lines represent the median values
.
10
Supplementary Figure
8
. Map of the difference in MODIS cloud optical thickness between
weekends and weekdays from 2018 to 2021.
11
Supplementary Figure
9
. The difference of ERA5 air temperature, vapor pressure deficit,
and soil
moisture
between weekend and weekday in Europe during 2018, 2019 and 2021.
The
top panels display the raw values, and the bottom panels show the percentage change. The insert
histograms show the distribution of the corresponding variable
, and dashed black lines represent
the median values
.
12
Supplementary Figure
10
.
The estimated standard error of the
S
ensitivity of solar induced
fluorescence to aerosols
corresponding to Figure 3.
The standard error is calculated through
regression coefficient analysis using the
statsmodels
package in Python.
13
Supplementary Figure
1
1
.
Relative changes of TROPOMI solar induced fluorescence
(unit: %)
responding
to 0.1 increase in aerosol optical depth in Europe
derived from
weekly
patterns, which refers to weekend minus weekday signals, during 2018, 2019 and 2021.
The
grids marked with black dots indicate that the regression coefficient is significant with p
-
value <
0.05. The histogram shows the distribution of the derived sensitivity
and the black dashed line
represent the median
. The results correspond to those s
hown in
Figure 3
, with normalized by the
95th percentile of daily solar induced fluorescence in each pixel and 0.1 scaling of aerosol optical
depth.
14
Supplementary Figure 1
2
. The correlation between daily TROPOMI solar
-
induced
fluorescence (SIF) and eddy covariance gross primary productivity (GPP)
for days when
both measurements are
available.
The titles give the site name and corresponding land cover
defined in
Supplementary Table
S
1
.
15
Supplementary Figure 1
3
. The averaged difference between weekend and weekday
measurements of (A) gross primary productivity estimated from eddy covariance
measurements, (B) percent changes in gross primary productivity estimated from eddy
covariance measurements, (C) TROPOMI sol
ar
-
induced fluorescence at the 0.25
-
degree grid
cell where the eddy covariance tower is located, and (D) the comparison between the two
during 2018, 2019, and 2021.
Only days with both measurements available are considered.
Different
colors in (D) represent different land cover types defined in
Supplementary Table
S
1
.
16
Supplementary Figure 1
4
. Annual gross primary production increases through aerosol
pollution reduction in two pollution mitigation scenarios.
The increase in country
-
level annual
net carbon uptake under pollution mitigation scenarios in Europe, with aerosol level reduced to (A)
the average of weekly minimum 3
-
days and (B) COVID
-
19 period, represented by year 2020.
Green, blue, and yellow bars r
epresent the increase of annual carbon uptake by forest,
grasslands/savannas/shrublands and cropland, respectively.
To estimate the range of estimated
values, we consider the uncertainties associated with SIF sensitivities to AOD, the conversion
factor of SIF to GPP, the conversion factor of GPP to NEE, and the definition of the growing
season based on the fraction of p
hotosynthetically active radiation (fPAR). We employed a
bootstrap approach, resampling the data 1000 times. The central estimates are represented by the
median, while the upper and lower bounds correspond to the 95th and 5th percentiles, respectively.
17
Supplementary Figure 15. The averaged difference between weekend and weekday
measurements of (A) gross primary productivity, (B) ecosystem respiration and (C) net
ecosystem productivity estimated from eddy covariance measurements during 2018, 2019,
and 202
1.
Only days with both measurements available are considered.
18
Supplementary Figure 1
6
. The effect of cloud filtering on the difference of TROPOMI solar
induced fluorescence between weekend and weekday in Europe during 2018, 2019 and 2021
using (A) 60% and (B) 40%.
The spatial pattern is similar compared to
Figure 1 where cloud
filtering of 80% was used.
The insert histograms show the distribution of the corresponding
variable,
and the dashed black lines represent the median values.
19
Supplementary Figure 1
7
. The difference in TROPOMI phase angles between weekend and
weekday in Europe during 2018, 2019, 2021 and the three
-
year averages.
20
Supplementary Figure 1
8
.
Weekend minus weekday relative
TROPOMI solar induced
fluorescence in Europe during 2018, 2019 and 2021.
The relative SIF represents SIF normalized
by the continuum level NIR
-
reflected radiance. The spatial pattern is consistent with Figure 1,
indicating that the widespread decrease in SIF is predominantly associated with reductions in
absorbed photosynthetic
ally active radiation (APAR), rather than being influenced by signal
attenuation caused by aerosols.
The insert histogram show
s
the
distribution,
and the dashed black
line represent
s
the median values.
21
Supplementary Figure
1
9
. Difference
between weekend and weekday (2018, 2019 and 2021)
of
TROPOMI
using the TROPOSIF product
produced by Guanter et al.
(1)
The spatial
pattern is consistent with the
Caltech
TROPOMI SIF
by
Köhler et al.
(2)
used in this study.
The
insert histogram show
s
the distribution, and the dashed black line represent
s
the median values.
22
Supplementary Figure
20
.
The relationship
among solar
-
induced fluorescence (SIF),
absorbed photosynthetically active radiation (APAR), and aerosol optical depth
at 550nm
(AOD
550
).
A
ll available weekly difference data of SIF, APAR, and AOD observations from
2018 to 2021 (excluding 2020)
are used here
. The linear relationship observed between SIF and
AOD suggests that the impact of AOD on SIF
at
ambient AOD levels
can be adequately captured
using a linear
regression
framework.
23
Supplementary Figure
2
1
. FLUXCOM annual
net ecosystem change (
NEE
)
vs.
gross
primary productivity (
GPP
)
for each
land cover
type in Europe.
The study region is defined
as in
Figure 1
. The spatial resolution of FLUXCOM is 0.083 degrees. The
land cover
map is based
on MODIS IGBP classification. Deciduous needleleaf forest and closed shrubland are excluded
due to their limited coverage in the study region. Snow and ice are also excluded in the analysis.
24
Supplementary Figure 2
2. Europe
-
averaged relative change in ecosystem respiration over
the weekend compared to weekday with different Q10 values.
25
Supplementary Table S1. The location and IGBP land cover classification of eddy covariance
sites used in the study.
ENF, CSH, MF, GRA, DBF, WET, and OSH represents evergreen
needleleaf forests, closed shrublands, mixed forests, grasslands, deciduous broadleaf forests,
permanent wetlands and open shrublands, respectively.
Site
Lon
Lat
IGBP
Q10
BE
-
Bra
4.52
51.31
ENF
1.63
BE
-
Maa
5.63
50.98
CSH
1.77
BE
-
Vie
6.00
50.30
MF
1.86
CH
-
Dav
9.86
46.82
ENF
2.00
DE
-
Gri
13.51
50.95
GRA
1.81
DE
-
Hai
10.45
51.08
DBF
1.98
DE
-
HoH
11.22
52.09
DBF
2.32
DE
-
RuR
6.30
50.62
GRA
2.13
DE
-
Tha
13.57
50.96
ENF
1.89
DK
-
Gds
9.33
56.07
ENF
1.15
DK
-
Skj
8.40
55.91
WET
2.15
FI
-
Hyy
24.29
61.85
ENF
2.02
FI
-
Ken
24.24
67.99
ENF
2.41
FI
-
Let
23.96
60.64
ENF
2.32
FI
-
Sii
24.19
61.83
WET
2.26
FI
-
Var
29.61
67.75
ENF
1.97
FR
-
Bil
-
0.96
44.49
ENF
1.94
FR
-
Fon
2.78
48.48
DBF
1.81
FR
-
LGt
2.28
47.32
WET
2.16
FR
-
Tou
1.37
43.57
GRA
1.09
IT
-
BFt
10.74
45.20
DBF
1.45
IT
-
Cp2
12.36
41.70
MF
0.99
IT
-
Lsn
12.75
45.74
OSH
1.95
IT
-
SR2
10.29
43.73
ENF
1.64
IT
-
Tor
7.58
45.84
GRA
2.96
SE
-
Deg
19.56
64.18
WET
2.57
SE
-
Htm
13.42
56.10
ENF
2.55
SE
-
Nor
17.48
60.09
ENF
2.04
SE
-
Svb
19.77
64.26
ENF
2.26
26
Supplementary Table S
2
.
The regression coefficient and associated standard error
corresponding to
Supplementary Figure 21
.
The regression model used is
as follows:
푁퐸퐸
=
푁퐸퐸
퐺푃푃
×
퐺푃푃
+
0
where NEE represents the
n
et
e
cosystem
e
xchange and GPP represents
g
ross
p
rimary
p
roductivity.
IGBP
푵푬푬
푮푷푷
standard error
of
푵푬푬
푮푷푷
standard error
of
0
evergreen needleleaf forest
-
0.353
0.002
0.355
0.006
evergreen broadleaf forest
-
0.410
0.023
0.909
0.108
deciduous broadleaf forest
-
0.691
0.005
1.603
0.019
mixed forest
-
0.417
0.002
0.664
0.007
open shrublands
-
0.587
0.008
0.679
0.008
woody savannas
-
0.349
0.001
0.438
0.003
savannas
-
0.305
0.001
0.442
0.003
grasslands
-
0.286
0.001
0.356
0.002
permanent wetlands
-
0.346
0.012
0.393
0.029
croplands
-
0.190
0.001
0.182
0.003
urban and builtup
-
0.214
0.005
0.278
0.014
crop natural vegetation mosaic
-
0.278
0.004
0.428
0.014
barren or sparsely vegetated
-
0.343
0.014
0.331
0.010