Introduction
Surface ozone (O3) and atmospheric aerosols influence land ecosystem
carbon uptake both directly and indirectly through Earth system
interactions. O3 reduces plant photosynthesis directly through stomatal
uptake. The level of damage is dependent on both surface ozone
concentrations ([O3]) and the stomatal conductance (gs), the latter
of which is closely related to the photosynthetic rate (Reich and
Amundson, 1985; Sitch et al., 2007; Ainsworth et al., 2012). The impact of
aerosol pollution on vegetation is less clear. Atmospheric aerosols
influence plant photosynthesis through perturbations to radiation,
meteorology, and clouds. Observations (Cirino et al., 2014; Strada et
al., 2015) suggest that an increase in diffuse light partitioning in
response to a moderate aerosol loading can improve canopy light-use
efficiency (LUE) and promote photosynthesis, known as diffuse radiation
fertilization (DRF), as long as the total light availability is not
compromised (Kanniah et al., 2012). Atmospheric aerosols also reduce leaf
temperature (Steiner and Chameides, 2005; Cirino et al., 2014), but the
consequence for photosynthesis depends on the relationship between the local
environmental temperature and the photosynthetic optimum temperature of
approximately 25 ∘C. Aerosol-induced changes in evaporation and
precipitation are interconnected but impose opposite effects on
photosynthesis; less evaporation preserves soil moisture in the short term
but may decrease local rainfall (Spracklen et al., 2012) and lead
to drought conditions in the long term. Furthermore, aerosol indirect
effects (AIEs) on cloud properties can either exacerbate or alleviate the
above feedbacks.
China is currently the world's largest emitter of both carbon dioxide and
short-lived air pollutants (http://gains.iiasa.ac.at/models/).
The land ecosystems of China are estimated to provide a carbon sink (Piao
et al., 2009), but it remains unclear how air pollution may affect this sink
through the atmospheric influences on regional carbon uptake. O3
damages to photosynthesis, including those in China, have been quantified in
hundreds of measurements (Table S1 in the Supplement), but are limited to individual plant
species and specific O3 concentrations ([O3]). Previous regional
modeling of O3 vegetation damage (e.g., Ren et al., 2011; Tian et
al., 2011) does not always take advantage of valuable observations to
calibrate gross primary productivity (GPP)–O3 sensitivity coefficients for China domain and
typically the derived results have not been properly validated. The aerosol
effects on photosynthesis are less well understood. Most of the limited
observation-based studies (Rocha et al., 2004; Cirino et al., 2014;
Strada et al., 2015) rely on long-term flux measurements or satellite
retrievals, which are unable to unravel impacts of changes in the associated
meteorological and hydrological forcings. Modeling studies focus mainly on
the aerosol-induced enhancement in diffuse radiation (e.g., Cohan et al.,
2002; Gu et al., 2003; Mercado et al., 2009), but ignore other direct and
indirect feedbacks such as changes in temperature and precipitation.
Finally, no studies have investigated the combined effects of O3 and
aerosols or how the air pollution influences may vary in response to future
emission regulations and climate change.
Summary of models and simulations.
Model Name
Model class
Climate drivers
Number
Table
Purpose
of runs
index∗
ModelE2–YIBs
Coupled climate model
Online
24
2
Calculate ΔNPP by O3 and aerosolsat 2010 and 2030
YIBs
Vegetation model
MERRA
15
S2
Evaluate O3 damage scheme forChina PFTs
YIBs
Vegetation model
ModelE2–YIBs
30
S3
Isolate aerosol individual climaticimpacts on NPP
∗ Table index refers to the tables in the main text and
the Supplement.
In this study, we assess the impacts of O3 and aerosols on land carbon
uptake in China using the global Earth system model NASA GISS ModelE2 that
embeds the Yale Interactive Terrestrial Biosphere (YIBs) model. This
framework is known as NASA ModelE2–YIBs and fully couples the land
carbon-oxidant-aerosol-climate system (Schmidt et al., 2014; Yue and
Unger, 2015). The global-scale model accounts for long-range transport of
pollution and large-scale feedbacks in physical climate change. The coupled
Earth system simulations apply present-day and future pollution emission
inventories from the Greenhouse Gas and Air Pollution Interactions and
Synergies (GAINS) integrated assessment model (http://gains.iiasa.ac.at/models/). The simulations include process-based
mechanistic photosynthetic responses to physical climate change, O3
stomatal uptake, carbon dioxide (CO2) fertilization, and aerosol
radiative perturbations, but not aerosol and acid deposition (Table 1). The
O3 and aerosol haze effects on the land carbon cycle fluxes occur
predominantly through changes to GPP and net
primary productivity (NPP). Therefore, the current study focuses on GPP and
NPP impacts and does not address changes in net ecosystem exchange (NEE).
Methods
YIBs vegetation model
The YIBs model applies the well-established Farquhar and Ball–Berry models
(Farquhar et al., 1980; Ball et al., 1987) to calculate leaf
photosynthesis and stomatal conductance, and adopts a canopy radiation
scheme (Spitters, 1986) to separate diffuse and direct light for
sunlit and shaded leaves. The assimilated carbon is dynamically allocated
and stored to support leaf development (changes in leaf area index, LAI) and
tree growth (changes in height). A process-based soil respiration scheme
that considers carbon flows among 12 biogeochemical pools is included to
simulate carbon exchange for the whole ecosystem (Yue and Unger,
2015). Similar to many terrestrial models (Schaefer et al., 2012), the
current version of YIBs does not include a dynamic N cycle. Except for this
deficit, the vegetation model can reasonably simulate ecosystem responses to
changes in [CO2], meteorology, phenology, and land cover (Yue
et al., 2015). A semi-mechanistic O3 vegetation damage scheme (Sitch
et al., 2007) is implemented to quantify responses of photosynthesis and
stomatal conductance to O3 (Yue and Unger, 2014).
The YIBs model can be used in three different configurations: site-level,
global offline or regional offline, and online within ModelE2–YIBs (Yue and
Unger, 2015). The offline version is driven with hourly 1∘ × 1∘ meteorological forcings from either the NASA Modern
Era Retrospective analysis for Research and Applications (MERRA)
(Rienecker et al., 2011) or the interpolated output from ModelE2–YIBs.
The online YIBs model is coupled with the climate model NASA ModelE2
(Schmidt et al., 2014), which considers the interplay among meteorology,
radiation, atmospheric chemistry, and plant photosynthesis at each time
step. For both global and regional simulations, 8 plant functional types
(PFTs) are considered (Fig. S1). This land cover is aggregated from a
dataset with 16 PFTs, which are derived using retrievals from both the
Moderate Resolution Imaging Spectroradiometer (MODIS) (Hansen et al.,
2003) and the Advanced Very High Resolution Radiometer (AVHRR)
(Defries et al., 2000). The same vegetation cover with
16 PFTs is used by the Community Land Model (CLM) (Oleson et
al., 2010).
Both the online and offline YIBs models have been extensively evaluated with
site-level measurements from 145 globally dispersed flux tower sites,
long-term gridded benchmark products, and multiple satellite retrievals of
LAI, tree height, phenology, and carbon fluxes (Yue and Unger,
2015; Yue et al., 2015). Driven with meteorological reanalyses, the offline
YIBs vegetation model estimates a global GPP of 122.3 ± 3.1 Pg C yr-1, NPP of 63.6 ± 1.9 Pg C yr-1, and NEE of
-2.4 ± 0.7 Pg C yr-1 for 1980–2011, consistent with an ensemble of land models
(Yue and Unger, 2015). The online simulations with ModelE2–YIBs,
including both aerosol effects and O3 damage, yield a global GPP of
125.8 ± 3.1 Pg C yr-1, NPP of 63.2 ± 0.4 Pg C yr-1,
and NEE of -3.0 ± 0.4 Pg C yr-1 under present day conditions.
NASA ModelE2–YIBs model
The NASA ModelE2–YIBs is a fully coupled chemistry–carbon–climate model with
horizontal resolution of 2∘ × 2.5∘ latitude by
longitude and 40 vertical levels extending to 0.1 hPa. The model simulates
gas-phase chemistry (NOx, HOx, Ox, CO, CH4, and non-methane volatile organic compounds – NMVOCS),
aerosols (sulfate, nitrate, elemental and organic carbon, dust, and sea salt),
and their interactions (Schmidt et al., 2014). Modeled oxidants influence
the photochemical formation of secondary aerosol species (sulfate, nitrate,
secondary organic aerosol). In turn, modeled aerosols affect photolysis
rates in the online gas-phase chemistry (Schmidt et al., 2014).
Heterogeneous chemistry on dust surfaces is represented
(Bauer et al., 2007). The embedded radiation package includes
both direct and indirect (Menon and Rotstayn, 2006) radiative effects
of aerosols and considers absorption by multiple greenhouse gases (GHGs). Size-dependent
optical parameters of clouds and aerosols are computed from Mie scattering,
ray tracing, and T-matrix theory, and include the effects of non-spherical
particles for cirrus and dust (Schmidt et al., 2006). Simulated surface
solar radiation exhibits the lowest model-to-observation biases compared
with the other 20 IPCC-class climate models (Wild et al., 2013).
Simulated meteorological and hydrological variables have been fully validated
against observations and reanalysis products (Schmidt et al., 2014).
Emissions
We use global annual anthropogenic pollution inventories from the GAINS
integrated assessment model (Amann et al., 2011), which compiles historic
emissions of air pollutants for each country based on available
international emission inventories and national information from individual
countries. Intercomparison of present-day (the year 2010) emissions (Fig. S2) shows that the GAINS V4a inventory has similar emission intensity
(within ±10 %) in China as the IPCC RCP8.5 scenario (van Vuuren et
al., 2011) for most species, except for ammonia, which is higher by 80 %
in GAINS. The discrepancies among different inventories emerge from varied
assumptions on the stringency and effectiveness of emission control
measures. While the GAINS 2010 ammonia emissions from China are larger than
the RCP8.5 and HTAP emissions as shown in Fig. S2, they are close in
magnitude to the year 2010 emissions of 13.84 Tg yr-1 estimated by the
Regional Emission inventory in Asia (REAS, http://www.nies.go.jp/REAS/).
The GAINS inventory also projects medium-term variations of future emissions
at 5-year intervals to the year 2030. The current legislation emission
(CLE) scenario applies full implementation of national legislation affecting
air pollution emissions; for China, this represents the 11th 5-year
plan, including known failures. By 2030, in the CLE inventory, CO decreases
by 18 %, SO2 by 21 %, black carbon (BC) by 28 %, and organic
carbon (OC) by 41 %, but NOx increases by 20 %, ammonia by 22 %,
and NMVOCs by 6 %, relative to the
2010 emission magnitude in China. To account for potential rapid changes in
policy and legislation, we apply the maximum technically feasible reduction
(MTFR) emission scenario as the lower limit of future air pollution. The
MTFR scenario implements all currently available control technologies,
disregarding implementation barriers and costs. With this scenario, the 2030
emissions of NOx decrease by 76 %, CO by 79 %, SO2 by 67 %,
BC by 81 %, OC by 89 %, ammonia by 65 %, and NMVOC by 62 % in China,
indicating large improvement of air quality. Biomass burning emissions,
adopted from the IPCC RCP8.5 scenario (van Vuuren et al., 2011), are
considered as anthropogenic sources because most fire activities in China
are due to human-managed prescribed burning (Zhou et al., 2017).
Compared with the GAINs inventory, present-day biomass burning is equivalent
to < 1 % of the emissions for NOx, SO2, and
NH3; 1.6 % for BC; 3.0 % for CO; and 9.6 % for OC. By the
year 2030, biomass burning emissions decrease by 1–2 % for all pollution
species compared with 2010.
Summary of 24 online simulations with the ModelE2–YIBs model.
Simulations
Period
Emission
Emission
Ozone
Aerosol
inventories
sources
damage
indirect effect
G10NATNO3
2010
GAINSa
Natural
Null
No
G10ALLNO3
2010
GAINS
Alld
Null
No
G10ALLLO3
2010
GAINS
All
Low
No
G10ALLHO3
2010
GAINS
All
High
No
G30NATNO3
2030
GAINS CLEb
Natural
Null
No
G30ALLNO3
2030
GAINS CLE
All
Null
No
G30ALLLO3
2030
GAINS CLE
All
Low
No
G30ALLHO3
2030
GAINS CLE
All
High
No
M30NATNO3
2030
GAINS MTFRc
Natural
Null
No
M30ALLNO3
2030
GAINS MTFR
All
Null
No
M30ALLLO3
2030
GAINS MTFR
All
Low
No
M30ALLHO3
2030
GAINS MTFR
All
High
No
G10NATNO3_AIE
2010
GAINS
Natural
Null
Yes
G10ALLNO3_AIE
2010
GAINS
All
Null
Yes
G10ALLLO3_AIE
2010
GAINS
All
Low
Yes
G10ALLHO3_AIE
2010
GAINS
All
High
Yes
G30NATNO3_AIE
2030
GAINS CLE
Natural
Null
Yes
G30ALLNO3_AIE
2030
GAINS CLE
All
Null
Yes
G30ALLLO3_AIE
2030
GAINS CLE
All
Low
Yes
G30ALLHO3_AIE
2030
GAINS CLE
All
High
Yes
M30NATNO3_AIE
2030
GAINS MTFR
Natural
Null
Yes
M30ALLNO3_AIE
2030
GAINS MTFR
All
Null
Yes
M30ALLLO3_AIE
2030
GAINS MTFR
All
Low
Yes
M30ALLHO3_AIE
2030
GAINS MTFR
All
High
Yes
a GAINS is short for the v4a emission inventory of
Greenhouse Gas and Air Pollution Interactions and Synergies
(http://gains.iiasa.ac.at/models/index.html).b CLE is the emission scenario predicted based on current
legislation
emissions.c MTFR is the emission scenario predicted with maximum
technically
feasible reductions.d All emissions including both natural and anthropogenic sources.
For the detailed anthropogenic emissions, refer to Fig. S2.
The model represents climate-sensitive natural precursor emissions of
lightning NOx, soil NOx, and biogenic volatile organic compounds (Unger and Yue, 2014). Future 2030 changes in these natural
emissions are small compared to the anthropogenic emission changes.
Interactive lightning NOx emissions are calculated based on the climate
model's moist convection scheme that is used to derive the total lightning
and the cloud-to-ground lightning frequencies (Price et al., 1997;
Pickering et al., 1998; Shindell et al., 2013). Annual average lightning
NOx emissions over China increase by 4 % between 2010 and 2030.
Interactive biogenic soil NOx emission is parameterized as a function
of PFT-type, soil temperature, precipitation (including pulsing events),
fertilizer loss, LAI, NOx dry deposition rate, and canopy wind speed
(Yienger and Levy, 1995). Annual average biogenic soil NOx
emissions increase by only 1 % over China between 2010 and 2030. Leaf
isoprene emissions are simulated using a biochemical model that depends on
the electron transport-limited photosynthetic rate, intercellular CO2,
canopy temperature, and atmospheric CO2 (Unger et al., 2013). Leaf monoterpene
emissions depend on canopy temperature and atmospheric CO2 (Unger
and Yue, 2014). Annual average isoprene emission in China increases by 5 %
(0.39 Tg C yr-1) between 2010 and 2030 in response to enhanced GPP and
temperature that offset the effects of CO2 inhibition. Monoterpene
emissions decrease by 5 % (-0.25 Tg C) between 2010 and 2030 because
CO2 inhibition outweighs the effects of increased warming.
Simulations
NASA ModelE2–YIBs online
We perform 24 time-slice simulations to explore the interactive impacts of
O3 and aerosols on land carbon uptake (Table 2). All simulations are
performed in atmosphere-only configuration. In these experiments, [O3]
and aerosol loading are dynamically predicted, and atmospheric chemistry
processes are fully two-way coupled to the meteorology and the land
biosphere. Simulations can be divided into two groups, depending on whether
AIEs are included. In each group, three subgroups are defined with the
emission inventories of GAINS 2010, CLE 2030, and MTFR 2030 scenarios. In
each subgroup, one baseline experiment is set up with only natural emissions
(denoted by NAT). The other three implement all natural and anthropogenic
sources of emissions (denoted by ALL), but apply different levels of
O3 damage including none (denoted by NO3), low sensitivity (LO3), and
high sensitivity (HO3). To compare the differences between online and
offline O3 damage, we perform four additional simulations which do not
account for the feedbacks of O3-induced changes in biometeorology,
plant growth, and ecosystem physiology. Two simulations,
G10ALLHO3_OFF and G10ALLLO3_OFF, include both
natural and anthropogenic emissions. The other two, G10NATHO3_OFF and G10NATLO3_OFF, include natural emissions alone.
We use prescribed sea surface temperature and sea ice distributions
simulated by ModelE2 under the IPCC RCP8.5 scenario (van Vuuren et al.,
2011). For these boundary conditions, we apply the monthly-varying decadal
average of 2006–2015 for 2010 simulations and that of 2026–2035 for 2030
simulations. Well-mixed GHG concentrations are also adopted from the RCP8.5
scenario, with CO2 changing from 390 ppm in 2010 to 449 ppm in 2030, and
CH4 changing from 1.779 to 2.132 ppm. Land cover change projections
for this scenario suggest only minor changes between the years 2010 and
2030; for example, the expansion of 3 % for grassland is offset by the
losses of 1 % for cropland and 4 % for tropical rainforest. As a result,
we elect to apply the same land cover, which is derived from satellite
retrievals, for both present-day and future simulations (Fig. S1). We use
present-day equilibrium tree height derived from a 30-year spin-up procedure
(Yue and Unger, 2015) as the initial condition. All simulations are
performed for 20 years, and the last 15 years are used for analyses. For
simulations including effects of CO2 fertilization, climate change, and
O3 damages, GPP and NPP reach new equilibrium within 5 years while
those for NEE may require several decades due to the slow responses of the
soil carbon pools (Fig. S3). The full list of simulations in Table 2 offers
assessment of uncertainties due to interannual climate variability, emission
inventories (CLE or MTFR), O3 damage sensitivities (low to high), and
aerosol climatic effects (direct and indirect). Uncertainties calculated
based on the interannual climate variability in the model are indicated
using the format “mean ± 1 standard deviation”. Other sources of
uncertainty are explicitly stated.
Evaluation of simulated summertime carbon fluxes by ModelE2–YIBs.
Panels show GPP (top row) and NPP (bottom row) over China. Simulation
results (a, d) are the average of G10ALLHO3 and G10ALLLO3, which are
performed with the climate model ModelE2–YIBs using high and low ozone damage
sensitivities (Table 2). The correlation coefficients (r), relative biases
(b), and number of grid cells (n) for the comparisons are listed on the
scatter plots. Units: g C m-2 day-1.
YIBs offline with MERRA meteorology
We perform 15 simulations to evaluate the skill of the O3 damage scheme
for vegetation in China (Table S2). Each run is driven with hourly
meteorological forcings from NASA GMAO MERRA (Rienecker et al., 2011).
One baseline simulation is performed without inclusion of any O3
damage. The others, seven runs in each of two groups, are driven with fixed
[O3] at 20, 40, 60, 80, 100, 120, and 140 ppbv, using
either low or high O3 sensitivities defined by Sitch et al. (2007).
Thus, [O3] in these offline runs is fixed without seasonal and diurnal
variations to mimic field experiments that usually apply a constant level of
[O3] during the test period. We compare the O3-affected GPP with
the O3-free GPP from the baseline simulation to derive the damaging
percentages to GPP, which are compared with values for different PFTs from
an ensemble of published literature results (Table S1). All simulations are
performed for 1995–2011, and the last 10 years are used for analyses.
YIBs offline with ModelE2–YIBs meteorology
Using the offline YIBs vegetation model driven with meteorology simulated by general circulation model (GCM) ModelE2-YIBs, we perform 30 simulations to isolate the impacts of
aerosol-induced changes in the individual meteorological drivers on carbon
uptake (Table S3). Experiments are categorized into two groups, depending on
whether the GCM forcings include AIE or not. In each group, three subgroups
of simulations are set up with different meteorology for GAINS 2010, CLE
2030, and MTFR 2030 scenarios. Within each subgroup, five runs are designed
with the different combinations of GCM forcings. One baseline run is forced
with meteorology simulated without anthropogenic aerosols. The other four
are additionally driven with aerosol-induced perturbations in temperature,
PAR, soil moisture, or the combination of the above three variables. For
these simulations, the month-to-month meteorological perturbations caused by
aerosols are applied as scaling factors on the baseline forcing for each
month at each grid square. The differences of NPP between sensitivity and
baseline runs represent contributions of individual or total aerosol
effects. Each simulation is performed for 15 years, with the last 10 years
used for analyses. Uncertainties due to interannual climate variability in
the model are calculated using different time periods for the online (15 years, Table 2) and offline (10 years, Table S3) runs.
Evaluation of simulated summertime air pollution in China.
Evaluations shown include (a) aerosol optical depth (AOD) at
550 nm, (d) [O3] (units: ppbv), and (g) PM2.5
concentrations (units: µg m-3) with observations from
(b) the satellite retrieval of the MODIS (averaged for 2008–2012),
and (e, h) measurements from 188 ground-based sites (at the year
2014). Simulation results are from G10ALLNO3 performed with the climate model
ModelE2–YIBs (Table 2). The correlation coefficients (r), relative biases
(b), and number of sites or grids (n) for the comparisons are listed on the
scatter plots.
Results
Evaluation of ModelE2–YIBs over China
Land carbon fluxes: GPP and NPP
To validate simulated GPP, we use a gridded benchmark product for 2009–2011
upscaled from in situ FLUXNET measurements (Jung et al., 2009). For NPP, we
use a MODIS satellite-derived product for 2009–2011 (Zhao et al., 2005).
Both datasets are interpolated to the same resolution of 2∘ × 2.5∘ as ModelE2–YIBs. Simulated GPP and NPP reproduce
the observed spatial patterns with high correlation coefficients
(R=0.46–0.86, p < 0.001) and relatively low model-to-observation
biases (< 21 % on national scale) (Figs. 1 and S4). High
values of the land carbon fluxes are predicted in the east and the
northeast, where forests and croplands dominate (Fig. S1). For GPP,
the prediction in the summer overestimates by 6.2 % over the southern coast
(< 28∘ N), but underestimates by 23.7 % over the North
China Plain (32–40∘ N, 110–120∘ E). Compared with
the MODIS data product, predicted summer NPP is overestimated by
20.6 % overall in China (Fig. 1f), with regional biases of 40.0 % in the
southern coast, 51.2 % in the North China Plain, and 38.7 % in the northeast
(> 124∘ E).
Surface air pollution and AOD
For surface concentrations of PM2.5 and O3, we use ground
measurements available for 2014 from 188 sites operated by the Ministry of
Environmental Protection of China (http://www.aqicn.org/). In addition,
we use rural [O3] from published literature (Table S4) to evaluate the
model. For AOD, we use gridded observations of 2008–2012 from MODIS
retrievals. The model simulates reasonable magnitude and spatial distribution
of surface PM2.5 concentrations (Figs. 2 and S5). Predicted AOD also
reproduces the observed spatial pattern, but underestimates the high center
in the North China Plain by 24.6 % in summer. Long-term measurements of [O3] are very
limited in China. Comparisons with the 2014 1-year data from 188 urban
sites show that simulated [O3] reproduces reasonable spatial
distribution but overestimates the average concentration by
> 40 % (Figs. 2f and S5f). Such discrepancy is in part
attributed to the sampling biases, because urban [O3] is likely lower
than rural [O3] due to high NOx emissions (NOx titration) and
aerosol loading (light extinction) in cities. Based on “China Statistical
Yearbook for 2015” (http://www.stats.gov.cn), the total rural area
accounts for > 98 % of the domestic area. Evaluations at
rural sites (Table S4) show a mean bias of -5 % (Fig. 3). The magnitude
of such bias is much lower than the values of comparisons at urban-dominant
sites, where simulated [O3] is higher by 42.5 % for the summer mean
(Fig. 2f) and 55.6 % for the annual mean (Fig. S5f).
Evaluation of simulated [O3] at rural sites in China.
Simulation results are from G10ALLNO3 performed with the climate model
ModelE2–YIBs (Table 2). For the scatter plots, green, pink, and blue points
represent values in spring, summer, and autumn, respectively. The data
sources of all sites are listed in Table S4.
Evaluation of simulated radiation fluxes by ModelE2–YIBs. Panels
show summertime surface (a) total shortwave radiation (units:
W m-2) and (d) diffuse-to-total fraction with
(b, e) observations from 106 sites. Simulation results are from
G10ALLNO3 performed with the climate model ModelE2–YIBs (Table 2). The
correlation coefficients (r), relative biases (b), and number of sites
(n) for the comparisons are listed on the (c, f) scatter plots.
The blue points in the scatter plots represent sites located within the box
regions in southeastern China as shown in panel (a).
Shortwave radiation
We use ground-based observations of surface shortwave radiation and diffuse
fraction from 106 pyranometer sites managed by the Climate Data Center of the
Chinese Meteorological Administration (Xia, 2010). Site selection is
based on the availability of continuous monthly measurements during
2008–2012, resulting in 95 sites for the evaluation of total shortwave
radiation. For diffuse radiation, we select only the 17 sites that provide
continuous measurements during 2008–2012. Simulated surface shortwave
radiation agrees well with measurements at 106 sites for both summer (Fig. 4a–c) and annual (Fig. S6a–c) means. Simulated diffuse fraction
reproduces observed spatial pattern with high correlation coefficient (r=0.74 for summer and r=0.65 for annual, p < 0.01), though it is
larger than observations on average by 25.2 % in summer (Fig. 4d–f) and
35.2 % for the annual mean (Fig. S6d–f). Such bias is mainly attributed
to the overestimation in the north and northeast. For the southeastern
region, where high values of GPP dominate (Fig. 1), predicted diffuse
fraction is in general within the 10 % difference from the observations.
Ozone vegetation damage function
We adopt the same approach as Yue et al. (2016) by comparing
simulated GPP-to-[O3] responses (Table S2) with observations from
multiple published studies (Table S1). We aggregate these measurements
into six categories, including evergreen needleleaf forest (ENF), deciduous
broadleaf forest (DBF), shrubland, C3 herbs, C4 herbs, and a mixture of all
above species. We derive the sensitivity of GPP to varied [O3] (Fig. 5)
using the YIBs offline version. For most PFTs, simulated O3 damage
increases with [O3] in broad agreement with measurements. Predicted
O3 damage reproduces observations with a correlation coefficient of
0.61 (for all samplings, n=32) and in similar magnitudes (-17.7 % vs.
-20.4 %), suggesting that the damage scheme we adopted from Sitch et
al. (2007) is ready to use in China. For the same level of [O3],
deciduous trees suffer larger damages than evergreen trees because the
former species are usually more sensitive (Sitch et al., 2007) and have
higher gs (and therefore higher uptake) (Wittig et al., 2007). Flux-based
O3 sensitivity for C4 herbs is only half that of C3 herbs (Sitch et
al., 2007); however, concentration-based O3 damages, both observed and
simulated, are larger for C4 plants because of their higher uptake
efficiency following high gs (Yue and Unger, 2014).
O3 effects in China
We focus our study domain in eastern China (21–38∘ N,
102–122∘ E, including the North China Plain, the
Yangtze River Delta, and part of the Sichuan Basin), a region that suffers
from high levels of O3 and aerosols from anthropogenic pollution
sources (> 75 % contribution; Fig. S7). We estimate that
surface O3 decreases annual GPP in China by 10.3 % based on YIBs
offline simulations in the absence of feedbacks from O3 vegetation
damage to meteorology and plant growth. The damage is stronger in summer,
when the average GPP decreases by ∼ 20 % for both deciduous
trees and C3 herbs in the east (Fig. 6). In contrast, a lower average damage
to GPP of ∼ 10 % is predicted for evergreen needleleaf trees
(because of low sensitivity) and C4 herbs (because of the mismatched spatial
locations between C4 plants and high [O3]; Figs. S1 and 2d).
Comparison of predicted changes in summer GPP by O3 with
measurements. Simulations are performed using the offline YIBs vegetation
model (Table S2) and averaged for all grid squares over China weighted by the
area of a specific PFT. Black points show the simulated mean reductions with
error bars indicating damage range from low to high O3 sensitivity.
Solid squares with error bars show the results (mean plus uncertainty) based
on measurements reported in the literature (Table S1). Experiments performed
for vegetation types in China are denoted by blue symbols. The author
initials are indicated for the corresponding studies.
Predicted offline percentage damage to summer GPP caused by O3.
Panels show the damages to (a) ENF (evergreen needleleaf forest),
(b) DBF (deciduous broadleaf forest), (c) C3 herbs, and
(d) C4 herbs over China in the year 2010. Simulations are performed
with the climate model ModelE2–YIBs, which does not feed O3 vegetation
damages back to affect biometeorology, plant growth, and ecosystem
physiology. The results are averaged for the low and high damaging
sensitivities:
(12(G10ALLHO3_OFF + G10ALLLO3_OFF) / G10ALLNO3 - 1)
× 100 %. The average value over the box domain of
panel (a) is shown in the title bracket of each subpanel.
Significant changes (p < 0.05) are marked with black dots.
Changes in NPP over China due to combined and separate
effectsa of air pollution (units: Pg C yr-1).
2010
2030 CLE
2030 MTFR
O3 (mean)b
-0.59 ± 0.11 (-0.60 ± 0.13)
-0.67 ± 0.11 (-0.71 ± 0.16)
-0.29 ± 0.14 (-0.31 ± 0.10)
Low sensitivity
-0.43 ± 0.12 (-0.40 ± 0.13)
-0.43 ± 0.14 (-0.51 ± 0.16)
-0.22 ± 0.17 (-0.15 ± 0.10)
High sensitivity
-0.76 ± 0.15 (-0.80 ± 0.16)
-0.90 ± 0.13 (-0.92 ± 0.18)
-0.36 ± 0.16 (-0.46 ± 0.12)
Aerosol (total)c
0.20 ± 0.08 (-0.20 ± 0.09)
0.23 ± 0.14 (-0.09 ± 0.19)
0.16 ± 0.14 (0.04 ± 0.17)
Temperatured
0.03 ± 0.04 (0.01 ± 0.04)
0.04 ± 0.02 (0.02 ± 0.05)
0.03 ± 0.04 (0.00 ± 0.04)
Radiationd
0.09 ± 0.04 (-0.03 ± 0.04)
0.16 ± 0.06 (-0.01 ± 0.06)
0.11 ± 0.04 (-0.03 ± 0.03)
Soil moistured
0.07 ± 0.07 (-0.19 ± 0.10)
0.01 ± 0.09 (-0.09 ± 0.15)
0.03 ± 0.12 (0.00 ± 0.09)
O3+ aerosol (net)e
-0.39 ± 0.12 (-0.80 ± 0.11)
-0.43 ± 0.12 (-0.80 ± 0.10)
-0.12 ± 0.13 (-0.28 ± 0.14)
a Results shown are the averages ± 1 standard deviation. Simulations with both aerosol direct and indirect radiative
effects are shown in the brackets.b Mean O3 damages are calculated as half of differences in
ΔNPP between low and high sensitivities, e.g., present-day mean
O3
damage is 12(G10ALLHO3 + G10ALLLO3) - G10ALLNO3.c Combined aerosol effects are calculated with the ModelE2–YIBs
climate
model, e.g., present-day aerosol effect is G10ALLNO3 - G10NATNO3.d Separate aerosol effects are calculated with the offline YIBs
vegetation model driven with forcings from the climate model (Table S3).e The net impact of O3 damages and aerosol effects, for
example at present day, is calculated as
12(G10ALLHO3 + G10ALLLO3) - G10NATNO3.
The negative impact that O3 has on photosynthesis can influence plant growth. At the same
time, the O3-induced reductions in stomatal conductance (Fig. S8a) can
increase canopy temperature and inhibit plant transpiration, leading to
surface warming (Fig. S8b), dry air (Fig. S8c), and rainfall deficit (Fig. S8d). These biometeorological feedbacks may in turn exacerbate the dampening
of land carbon uptake. Application of ModelE2–YIBs that allows for these
feedbacks causes O3-induced damage to annual GPP of 10.7 %, a
similar level to the damage computed in YIBs offline. The spatial pattern of
the online O3 inhibition also resembles that of offline damages (not
shown). Sensitivity simulations with zero anthropogenic emissions show
almost no O3 damage (Fig. S9), because the [O3] exposure from
natural sources alone is usually lower than the threshold level of 40 ppbv,
below which the damage for most PFTs is limited (Fig. 5). Our results
indicate that present-day surface O3 causes strong inhibitions on total
NPP in China, ranging from 0.43 ± 0.12 Pg C yr-1 with low
sensitivity to 0.76 ± 0.15 Pg C yr-1 with high sensitivity (Table 3). The central value of NPP reduction by O3 is
0.59 ± 0.11 Pg C yr-1, assuming no direct impacts of O3 on plant respiration. About
61 % of such inhibition occurs in summer, when both photosynthesis and
[O3] reach maximum of the year.
Aerosol haze effects in China
Aerosols decrease direct solar radiation but increase diffuse radiation
(Fig. S10), the latter of which is beneficial for canopy photosynthesis. The
online-coupled model quantifies the concomitant meteorological and
hydrological feedbacks (Fig. 7) that further influence the radiative and
land carbon fluxes. Reduced insolation decreases summer surface temperature
by 0.63 ∘C in the east, inhibiting evaporation but increasing
relative humidity (RH) due to the lower saturation vapor pressure (Table S5). These feedbacks combine to stimulate photosynthesis (Fig. 8a), which,
in turn, strengthens plant transpiration (not shown). Atmospheric
circulation and moisture convergence are also altered due to the
pollution–vegetation–climate interactions, resulting in enhanced
precipitation (Fig. 7b) and cloud cover (Fig. 7d). Moreover, soil moisture
increases (Fig. 7f) with lower evaporation (Fig. 7e) and higher
precipitation (Fig. 7b). Inclusion of AIE results in distinct climatic
feedbacks (Fig. S11). Summer precipitation decreases by 0.9 mm day-1
(13 %), leading to a 3 % decline in soil moisture (Table S6). The AIE
lengthens cloud lifetime and increases cloud cover, further reducing
available radiation and causing a stronger surface cooling. Compared to
aerosol-induced perturbations in radiation and temperature, responses in
hydrological variables (e.g., precipitation and soil moisture) are usually
statistically insignificant on the domain average due to the large relative
interannual climate variability (Tables S5 and S6). The resulting
meteorological changes over China are a combination of locally driven
effects (such as changes in radiation and hence temperature) and
regionally to globally driven effects (such as changes in rainfall and hence soil
water).
Changes in summer meteorology due to direct radiative effects of
anthropogenic aerosols. All changes are calculated as the differences between
the simulations G10ALLNO3 and G10NATNO3. For (a) temperature,
(b) precipitation, and (c) relative humidity, we show the
absolute changes as G10ALLNO3 - G10NATNO3. For (d) middle cloud
cover, (e) evaporation, and (f) soil water content, we show
the relative changes as
(G10ALLNO3 / G10NATNO3 - 1) × 100 %. Significant changes
(p < 0.05) are marked with black dots.
Decomposition of aerosol-induced changes in summer NPP. Changes in
NPP are caused by aerosol-induced changes in (b) surface air
temperature, (c) photosynthetically active radiation (PAR),
(d) soil moisture, and (a) the combination of above three
effects. Simulations are performed with the offline YIBs vegetation model
driven with meteorological forcings simulated with the ModelE2–YIBs climate
model (Table S3). The NPP responses to PAR include the DRF effects. The maximum color
scale for panel (a) is 1.5 while the others (b–d) are 0.5. The average NPP
perturbation over the box domain in a is shown in the bracket of each title.
Only the significant changes (p < 0.05) are shown. Units:
g C m-2 day-1.
We separate the relative impacts due to aerosol-induced perturbations in
temperature, radiation, and soil moisture (Fig. 8). The overall impact of
the aerosol-induced biometeorological feedbacks on the carbon uptake depends
on the season and vegetation type. In the summer, the aerosol-induced
surface cooling brings leaf temperature closer to the photosynthetic optimum
of 25 ∘C, DRF enhances light availability of shaded leaves and LUE
of sunlit leaves, and the wetter soil alleviates water stress for stoma.
Consequently, aerosol-induced hydroclimatic feedbacks promote ecosystem NPP,
albeit with substantial spatiotemporal variability (Fig. 8a and Table 3).
Surface cooling enhances NPP in summer (Fig. 8b) but induces neutral net
impacts on NPP in spring and autumn (not shown), when leaf temperature is
usually below 25 ∘C, because the cooling-driven reductions of
photosynthesis are accompanied by simultaneous reductions in plant
respiration. We find strong aerosol DRF in the southeast and the northeast,
where AOD is moderate (Fig. 8c). Over the North China Plain and the
southwest, aerosol DRF is more limited. In these regions, the local
background aerosol layer and/or cloud are sufficiently optically thick
that the effect of anthropogenic aerosol pollution is largely to attenuate
direct sunlight and reduce NPP (Cohan et al., 2002). Aerosol-induced
cooling increases soil moisture over most of the east (Fig. 7f), but the
beneficial responses are confined to the central east (Fig. 8d), where C3
crops dominate (Fig. S1). These short-rooted plants are more sensitive to
short-term water availability than deep-rooted trees (Beer et al., 2010;
Yue et al., 2015).
In contrast, inclusion of AIE results in detrimental impacts on NPP (Table 3). Aerosol-induced drought strongly reduces regional NPP, especially over
the northeast and North China Plain (Fig. S12d), where cropland dominates
(Fig. S1). Meanwhile, the increases in cloud cover reduce available
radiation, leading to weakened aerosol DRF effects over the southeast while
strengthening NPP reductions in the southwest (Fig. S12c).
Combined effects of O3 and aerosol
Simultaneous inclusion of the aerosol effects on the land biosphere has
negligible impacts on O3 damage. The online O3 inhibition, which
is much stronger in magnitude than the aerosol effects, shows insignificant
differences relative to the offline values (10.7 % vs. 10.3 %). As a
result, we consider O3 and aerosol effects to be linearly additive. In
the year 2010, the combined effects of O3 and aerosols (Table 3)
decrease total NPP in China by 0.39 (without AIE) to 0.80 Pg C yr-1
(with AIE), equivalent to 9–16 % of the pollution-free NPP and 16–32 %
of the total anthropogenic carbon emissions (Liu et al., 2015).
Spatially, a dominant fraction (86 % without AIE and 77 % with AIE) of
the reduced carbon uptake occurs in the east, where dense air pollution is
co-located with high NPP (Figs. 1 and 2). Temporally, a dominant fraction
(60 % without AIE and 52 % with AIE) of the reduced carbon uptake occurs
in summer, when both NPP and [O3] reach maximum of the year.
Independently, O3 reduces NPP by 0.59 Pg C yr-1, with a large
range from 0.43 Pg C yr-1 for low damaging sensitivity to 0.76 Pg C yr-1 for high damaging sensitivity (Table 3). The sign of the aerosol
effects is uncertain. Without AIE, aerosol is predicted to increase NPP by
0.2 Pg C yr-1, because of regionally confined DRF effects and enhanced
soil moisture (Fig. 8). With inclusion of AIE, aerosol decreases NPP by 0.2 Pg C yr-1, mainly due to reduced soil moisture (Fig. S12). The
uncertainty of individual simulations, calculated from the interannual
climate variability, is usually smaller than that due to O3 damage
sensitivity and AIE (Table 3).
Future projection of pollution effects
Following the CLE scenario, by the year 2030, predicted summer [O3]
increases by 7 %, while AOD decreases by 5 % and surface PM2.5
concentrations decline by 10 % (Fig. 9). These changes are predominantly
attributed to changes in anthropogenic emissions, as natural sources show
limited changes. The reduction of AOD is related to the decreased emissions
of SO2, black carbon, and organic carbon (Fig. S2). In contrast, the
enhancement of [O3] results from the increased NOx emissions,
higher level of background CH4 (∼ 20 %), and higher air
temperature in the warmer 2030 climate. The moderate decline of aerosol
loading in the 2030 CLE scenario brings benefits to land ecosystems through
DRF effects (Table 3) because light scattering is often saturated in the
present-day conditions due to high local AOD and regional cloud cover.
Benefits from the aerosol pollution reductions are offset by worsening
O3 vegetation damage in the CLE future world (Fig. 10b). O3-free
([O3] = 0) NPP increases by 14 % in 2030 due to CO2
fertilization and global climate change. Despite [CO2] increases from
390 ppm in 2010 to 449 ppm in 2030 in the RCP8.5 scenario (van Vuuren et
al., 2011), which contributes to gs inhibition of 4 % on the country
level, the future O3-induced NPP damage in 2030 degrades to 14 % or
0.67 Pg C yr-1 (Table 3), with a range from 0.43 Pg C yr-1 (low
O3 sensitivity) to 0.90 Pg C yr-1 (high O3 sensitivity).
Predicted changes in summertime air pollution by 2030. Panels shown
are for (a, b) AOD, (c, d) [O3], and
(e, f) PM2.5 concentrations for the year 2030 relative to 2010
based on scenarios of (left) current legislation emissions (CLE) and (right)
maximum technically feasible reduction (MTFR). Results for the left panels
are calculated as (G30ALLNO3 - G10ALLNO3). Results for the right panels
are calculated as (M30ALLNO3 - G10ALLNO3). The average value over the box
domain of panel (a) is shown in the title bracket of each subpanel.
Only the significant changes (p < 0.05) are shown.
Impacts of air pollution on NPP in the whole of China. Results shown
are combined effects of aerosol and O3 on the summer NPP in
(a) 2010, (b) 2030 with CLE scenario, and (c) 2030
with MTFR scenario. Results for the top panels do not include aerosol
indirect effects (AIE) but do include the meteorological response to aerosol
direct radiative effects. The average NPP perturbation over the box domain in
panel (a) is shown in the bracket of each title. The perturbations
to annual total NPP by aerosol, O3, and their sum over the whole China
are shown in panels (d–f) for different periods, with (right) and
without (left) inclusion of AIE. Damages by O3 are averaged for low and
high sensitivities with error bars indicating ranges. The percentage changes
are calculated based on NPP without AIE. Simulations are performed with the
ModelE2–YIBs model. Only the significant changes (p < 0.05) are
shown in panels (a–c).
The MTFR scenario reflects an ambitious and optimistic future in which there
is rapid global implementation of all currently available technological
pollution controls. AOD decreases by 55 % and [O3] decreases by
40 % for this future scenario (Fig. 9). The model projects much lower
damage to NPP of only 0.12 Pg C yr-1, with a range from 0.06 Pg C yr-1 (low O3 sensitivity) to 0.20 Pg C yr-1 (high O3
sensitivity), mainly due to the 40 % reduction in [O3] (Fig. 10c).
Including both aerosol direct and indirect effects, O3 and aerosols
together inhibit future NPP by 0.28 Pg C yr-1, ranging from 0.12 Pg C yr-1 with low O3 sensitivity to 0.43 Pg C yr-1 with high
O3 sensitivity. As a result, The MTFR scenario offers strong recovery
of the land carbon uptake in China by 2030.
Discussion
Comparison with previous estimates
Previous estimates of O3 damages over the whole China region are very
limited. Two important studies, Tian et al. (2011) and Ren et
al. (2011), have quantified the impacts of surface O3 on carbon
assimilation in China. Both studies applied the dynamic land ecosystem model
(DLEM) with the O3 damage scheme proposed by Felzer et al. (2004),
except that Tian et al. (2011) focused on NEE while Ren et al. (2011) also investigated NPP. The Felzer et al. (2004) scheme calculates
O3 uptake based on stomatal conductance and the AOT40 (accumulated
hourly O3 dose over a threshold of 40 ppb). Yue and Unger (2014)
estimated O3-induced reductions in GPP over the USA using the scheme of Sitch et al. (2007) and found an average value of 4–8 % (low to high
sensitivity), consistent with the reduction of 3–7 % in Felzer et al. (2004). For this study, we estimate that present-day O3 decreases NPP
by 0.43–0.76 Pg C yr-1, higher than the 0.42 Pg C yr-1 calculated
by Ren et al. (2011). However, the percentage reduction of
10.1–17.8 % in our estimate is weaker than the value of 24.7 % in
Ren et al. (2011). The main reason for such discrepancy lies in the
differences in the climatological NPP. Combining all environmental drivers
(e.g., [CO2], meteorology, [O3], and aerosols), we predict an
average NPP of 3.98 ± 0.1 Pg C yr-1 for the year 2010
(uncertainties from AIE) with the ModelE2–YIBs model. This value is close to
the average of 3.35 ± 1.25 Pg C yr-1 for 1981–2000 calculated
based on 54 estimates from 33 studies (Shao et al., 2016). Using the DLEM,
Ren et al. (2011) estimated an optimal NPP of 1.67 Pg C yr-1
for 2000–2005 over China, which is only half of the literature-based
estimate.
In the absence of any previous studies of aerosol pollution effects on land
carbon uptake in China, our strategy is to compare separately the simulated
aerosol climatic feedback (climate sensitivity) and simulated NPP response
to climate variability (NPP sensitivity) with existing published results.
ModelE2–YIBs simulates an annual reduction of 26.2 W m-2 in all-sky
surface solar radiation over the east due to aerosols pollution (Table S5),
similar to the estimate of 28 W m-2 by Folini and Wild (2015). In response to this radiative perturbation, aerosol pollution causes
a widespread cooling of 0.3–0.9 ∘C in summer over the east (Fig. 7a), consistent with estimates of 0–0.9 ∘C by Qian
et al. (2003), 0–0.7 ∘C by Liu et al. (2009), and an
average of 0.5 ∘C by Folini and Wild (2015). Aerosol
pollution effects on regional precipitation patterns in China are not well
understood due to different climate model treatments of land–atmosphere
interactions and the interplay between regional and large-scale circulation.
In ModelE2–YIBs, without AIE, aerosol induces a “northern drought and
southern flood” pattern in agreement with Gu et al. (2006),
but different to Liu et al. (2009) who predicted widespread drought
instead. Including both aerosol direct and indirect effects, ModelE2–YIBs
simulates an average reduction of 0.48 mm day-1 in summer rainfall
over much of China (Fig. S11b), similar to the magnitude of 0.4 mm day-1 estimated with the ECHAM5–HAM model (Folini and Wild,
2015), but higher than the 0.21 mm day-1 predicted by the RegCM2 model
(Huang et al., 2007).
Sensitivity experiments with YIBs show that summer NPP increases following
aerosol-induced changes in temperature, radiation, and precipitation (Fig. 8). The cooling-related NPP enhancement (Fig. 8b) collocates with changes in
temperature (Fig. 7a), indicating that sensitivity of NPP to temperature is
negative over eastern China. Such temperature sensitivity is consistent with
the ensemble estimate based on 10 terrestrial models (Piao et al., 2013).
For the aerosol-induced radiative perturbation, many studies have shown that
moderate aerosol or cloud amount promotes plant photosynthesis through enhanced
DRF, while dense aerosol or cloud decreases carbon uptake due to light
extinction (Cohan et al., 2002; Gu et al., 2003; Rocha et al., 2004;
Alton, 2008; Knohl and Baldocchi, 2008; Mercado et al., 2009; Jing et al.,
2010; Bai et al., 2012; Cirino et al., 2014; Strada et al., 2015).
Theoretically, at each specific land location on the Earth, there is an AOD
threshold below which aerosol promotes local NPP. The threshold is dependent
on latitude, cloud or aerosol amount, and plant types. In a related study by
Yue and Unger (2017), we applied a well-validated offline radiation
model to calculate these AOD thresholds over China. We conclude that
present-day AOD is lower than the local thresholds in the northeast and
southeast but exceeds the thresholds in the North China Plain, explaining
why aerosol-induced diming enhances NPP in the former regions but reduces
NPP in the latter (Fig. 8c). On the country level, the NPP enhancement due
to aerosol DRF is 0.07 Pg C yr-1 in Yue and Unger (2017), very
close to the 0.09 Pg C yr-1 estimated with the ModelE2–YIBs model (Table 2).
Uncertainties
A major source of uncertainty originates from the paucity of observations.
For instance, direct measurements of aerosol pollution effects on NPP are
non-existent for China. The aerosol effects involve complex interactions
that challenge the field-based validation of the underlying independent
processes. Field experiments of O3 vegetation damage are becoming more
available, but their applications are limited by the large variations in the
species-specific responses (Lombardozzi et al., 2013), as well as
the discrepancies in the treatments of [O3] enhancement (Wittig et
al., 2007). Instead of equally using all individual records from multiple
literatures (Lombardozzi et al., 2013), we aggregate O3 damage
from each study, based on the seasonal (or growth-season) average. In
this way, the derived PFT-level GPP–[O3] relationships are not biased
towards the experiments with a large number of samplings. Such aggregation
also reduces sampling noise and allows construction of the quantified
GPP–[O3] relationships used for model assessment. Predicted [O3]
is largely overestimated at urban sites but exhibits reasonable magnitude at
rural sites (Figs. 2 and 3). Measurements of background [O3] in China
are limited in both space and time, restricting comprehensive validation of
[O3] and the consequent estimate of O3 damages on the country
level.
We have estimated O3 damages to NPP (instead of GPP), an optimal
indicator for net carbon uptake by plants. Our calculations assume no
impacts of O3 on autotrophic respiration. Yet, limited observations
have found increased plant respiration in response to O3 injury
(Felzer et al., 2007), suggesting that our calculation of
O3-induced NPP reductions might be underestimated. Current large
mechanistic uncertainties in the role of anthropogenic nitrogen (N)
deposition to China's land carbon uptake (Tian et al., 2011; Xiao et al.,
2015) have prohibited the inclusion of dynamic carbon–nitrogen coupling in
the Earth system model used here. Previous studies have suggested that
inclusion of N fertilization can relieve or offset damages by O3,
especially for N-limited forests (Ollinger et al., 2002). Relative to the
present day, atmospheric reactive N deposition increases by 20 % in the
CLE scenario but decreases by 60 % in the MTFR scenario, suggesting that
the stronger O3 damage in CLE might be overestimated, while the reduced
damage in MTFR might be too optimistic.
Our estimate of NPP responses to aerosol pollution is sensitive to modeling
uncertainties in concentration, radiation, and climatic effects. Simulated
surface PM2.5 is reasonable but AOD is underestimated in the North
China Plain (Fig. 2), likely because of the biases in aerosol optical
parameters. Using a different set of optical parameters, we predicted much
higher AOD that is closer to observations with the same aerosol vertical
profile and particle compositions (Yue and Unger, 2017). The model
overestimates diffuse fraction in China (Fig. 4), likely because of
simulated biases in clouds. Previously, we improved the prediction of
diffuse fraction in China using observed cloud profiles for the region
(Yue and Unger, 2017). Biases in simulated AOD and diffuse fraction
introduce uncertainties in the aerosol DRF, especially in the affected
localized model grid cells. Yet, averaged over the China domain, our
estimate of NPP change by aerosol DRF (0.09 Pg C yr-1) is consistent
with the previous assessment in Yue and Unger (2017) (0.07 Pg C yr-1). Aerosol-induced impacts on precipitation and soil moisture are
not statistically significant over the regionally averaged domain (Tables S5
and S6). However, for the 2010 and 2030 CLE cases with AIE, in two out of six
scenarios, the aerosol-induced impact on soil moisture dominates the total
NPP response (Table 3). Furthermore, the relatively coarse resolution of the
global model and usage of emission inventories may introduce additional
biases and exacerbate the total uncertainties.
Importantly, our estimate of NPP response to aerosol effects, with or
without AIE, is secondary in magnitude compared to the O3 vegetation
damage, suggesting that the net impact of current air pollution levels in
China is detrimental to the land carbon uptake there. Locally, this
pollution damage exerts a threat to plant health, terrestrial ecosystem
services, and food production. Globally, air pollution effects may enhance
planetary warming by decreasing the land removal rate of atmospheric
CO2. Our results show substantial benefits to the protection of plant
health and the regional land carbon sink in China from stringent air
pollution controls, especially for O3 precursors. Our analysis
highlights the complex interplay between immediate and more local pollution
issues, as well as longer-term global warming. Future air pollution controls
provide an additional co-benefit to human society: the offsetting of fossil-fuel CO2 emissions through enhanced land sequestration of atmospheric
CO2.