Abstract

The effect of atmospheric aerosols and regional haze from airpollution on the yields of rice and winter wheat grown in China isassessed. The assessment is based on estimates of aerosol opticaldepths over China, the effect of these optical depths on the solarirradiance reaching the earth’s surface, and the response of rice andwinter wheat grown in Nanjing to the change in solar irradiance. Twosets of aerosol optical depths are presented: one based on a coupled,regional climate/air quality model simulation and the other inferredfrom solar radiation measurements made over a 12-year period atmeteorological stations in China. The model-estimated optical depthsare significantly smaller than those derived from observations, perhapsbecause of errors in one or both sets of optical depths or because thedata from the meteorological stations has been affected by localpollution. Radiative transfer calculations using the smaller,model-estimated aerosol optical depths indicate that the so-called“direct effect” of regional haze results in an ≈5–30%reduction in the solar irradiance reaching some of China’s mostproductive agricultural regions. Crop-response model simulationssuggest an ≈1:1 relationship between a percentage increase (decrease)in total surface solar irradiance and a percentage increase (decrease)in the yields of rice and wheat. Collectively, these calculationssuggest that regional haze in China is currently depressing optimalyields of ≈70% of the crops grown in China by at least 5–30%.Reducing the severity of regional haze in China through air pollutioncontrol could potentially result in a significant increase in cropyields and help the nation meet its growing food demands in the comingdecades.

China as a Case Study.

To examine whether regional haze can affect crop yields, we adopt acase study approach and focus on China. We do this for two reasons. Inthe first place, observations suggest that regional haze over China isespecially severe (22,23), and thus it represents an opportune regionfor an initial assessment of regional-haze effects.

Secondly, agricultural productivity is generally recognized to be acritical factor in determining the future economic trajectory of China.China is the most populous nation in the world, with one of the mostrapidly developing economies (24). The question of whether China willbe able to feed its growing population and at the same time sustain arapid pace of economic development has been the subject of considerabledebate (25–27). It is generally agreed that China’s food demand willincrease by ≈1%⋅yr⁻¹ over the next twodecades. Less certain is whether China will be able to meet thisgrowing demand internally or will have to import increasing amounts offoodstuffs from other nations. One aspect of resolving this question isunderstanding the extent to which air pollution affects the yields ofcrops grown in China and how this effect may change in the comingdecades.

Given China’s heavy reliance on coal, its burgeoning industrialsector, and its growing use of automobiles (28,29), regional airpollution may already be affecting crop yields in China. In fact,recent analyses of non-urban air quality data from China and regionalair quality model simulations indicate that agricultural areas in Chinaare exposed to potentially harmful amounts of phyotoxic pollutants suchas O₃ and acidic precipitation (30,31). However,because of a lack of data on the effect of these pollutants on cropsgrown in China, it has not yet been possible to quantitatively assesstheir impact on agricultural productivity.

In contrast to phyotoxic pollutants, the amount of solarradiation reaching the earth’s surface is a standard input variable inso-called crop-response models, used to simulate the yields of specificcrops as a function of environmental conditions (32). Moreover, thesemodels have been tested and applied to agricultural systems throughoutthe world, including those in China (33–37). Thus, these models can beused to calculate the effect of changes in surface solar radiation onthe yields of specific crops grown in China. By combining thesecalculations with estimates of the effect of regional haze on solarradiation in China, it should be possible to derive a rough lower limitestimate of the impact of regional air pollution on China’sagricultural productivity.

Atmospheric Aerosols and Solar Radiation.

The amount of solar radiation reaching the earth’s surface isquantified here in terms ofI_s(λ), the surfacesolar irradiance as a function of wavelength, λ.I_s has units ofW⋅m⁻²⋅μm⁻¹and is defined as the number of watts of solar radiation havingwavelengths between λ and λ + dλ (in units of micrometers) thatimpinge on a square meter of the earth’s surface. The total surfacesolar irradiance,I_s^tot, has units ofW⋅m⁻² and is obtained by integratingI_s over the solar spectrum:

1

In general,I_s can be divided into twocomponents: (i)I_s^dir, the directirradiance representing the direct beam of light from the sun; and(ii)I_s^diff, the diffuseirradiance or skylight representing the radiation from the sun thatreaches the surface after having been scattered by atmospheric gases,aerosols, and/or clouds.

As noted above, atmospheric aerosols can reduceI_s via the direct effect, whereby solar photonsare scattered and absorbed by the aerosols themselves, and the indirecteffect, whereby aerosols enhance the ability of clouds to scatter andabsorb solar photons (2,3). Because the magnitude of the indirecteffect is far more uncertain than that of the direct effect (4,5),only the direct effect is considered here.

The direct effect of aerosols onI_s can bedescribed in terms of the three unitless parameters: the aerosoloptical depth, τ_a, the aerosol singlescattering albedo,w_a, and the aerosolasymmetry factor,g_a. LikeI_s, all three of these parameters vary withwavelength, λ. The optical depth is given by

2

where TOA is used to represent the top of the atmosphere, andσ_ep is the aerosol (or particulate) extinctioncoefficient: i.e., the sum of σ_ap andσ_sp, the aerosol absorption and scatteringcoefficients, respectively. These coefficients have units ofm⁻¹ and represent the inverse of the e-foldinglength for attenuation of an incident beam of radiation by aerosols dueto absorption and/or scattering. Note that sinceτ_a is obtained by integrating a quantity havingunits of m⁻¹ over height, it is unitless.

Values for τ_a are usually reported for λ= 550 nm. Extrapolation of this value to other wavelengths is oftenmade by using an empirically derived parameter,a, referredto as the Angstrom exponent (38):

3

When most of the aerosol scattering is due to submicron-sized fineparticles, the Angstrom exponent in the visible range of the spectrumis typically ≈1. However, values as large as 2 or more can apply ifthe scattering is due to superfine mode aerosols and as small as 0 (oreven slightly <0) if the scattering is due to fine particles >1 μmand/or coarse particles (32,33). What little data from China that isavailable suggests that an Angstrom exponent of ≈1 is appropriate foranthropogenic aerosols (see, for example, ref.41).

The physical meaning of the optical depth can be understood in terms ofits relationship toI_s^dir. For example, in acloudless atmosphere,

4

whereI_o(λ) is the solar irradiance atwavelength λ entering the top of the atmosphere, θ is the solarzenith angle, and τ_g is the optical depth dueto (Rayleigh) scattering by atmospheric gases. Thus we see that theoptical depths, τ_a andτ_g, are the values needed in the exponent ofEq.4 to calculate the direct surface irradiance when thesun is directly overhead. The “cos θ” term is used to correctfor the longer path length needed to traverse the atmosphere when thesun is not overhead.

From Eq.4, it follows that the magnitude ofI_s^dir relates exponentially andinversely to τ_a. By contrast,I_s^diff tends to increase withτ_a. (This occurs because an increase inτ_a increases the amount of light scattered and, hence,the amount of diffuse radiation reaching the surface.)I_s^diff also depends onw_a andg_a. The single scattering albedo,w_a, is the ratio of scattering tototal extinction by the aerosols: i.e.,

5

The asymmetry factor,g_a, is usedto define the fraction of scattered radiation that is scattered in theforward direction. This fraction can be approximated by (1 +g_a)/2; wheng_a = 1, all radiation is scattered inthe forward direction, and wheng_a =−1, all is scattered in the backward direction.

I_s^diff tends to increase with increasingw_a andg_a, as well asτ_a. However, because of the possibility ofmultiple scattering (by gases, cloud droplets, and aerosols), as wellas the fact that σ_sp,σ_ep,w_a, andg_a can vary with height,I_s^diff is a complicated function of therelevant parameters, and numerical simulations are generally requiredto calculate its magnitude.

Under unpolluted conditions over continents,τ_a is generally ≈0.05 or less,w_a is >0.9, andg_a is ≈0.7 (9). By comparison,τ_g is ≈0.06. Thus, in clean air, most of thescattering of visible radiation from the sun is generally caused byRayleigh scattering. As conditions become more polluted and the fineparticle concentrations in the lower 1- 2 km of the atmosphereincrease, τ_a increases. If the particlescontain a significant amount of elemental carbon,w_a decreases. As a result, aerosolscattering and absorption can dominate over Rayleigh scattering in thepolluted atmosphere. This is, in fact, the case over much of theeastern half of the United States, where annually averaged values ofτ_a generally range from ≈0.2 to 0.5 (1,42).At a site in Oklahoma, for example, observations during 1998 yielded aτ_a of 0.2 ± 0.14 (see Fig.1).

Model-Calculated Aerosol Optical Depths Over China.

Fig.2 depicts annually averaged values forτ_a (at 550 nm) over China derived from a12-month (August, 1994–July, 1995) simulation of the regionaldistribution of anthropogenic sulfate aerosols in East Asia using thecoupled, regional climate/air quality model of Chameidesetal. (30). (Also depicted in Fig.2 are τ_avalues derived from surface solar irradiance measurements at 35 sitesin China; these will be discussed later.) In Fig.3,model-based τ_a values are illustrated for 4months of the year, each representative of one of the four seasons.

The model-based τ_as in Figs.2 and3 werecalculated from Eq.2, with the aerosol extinctioncoefficient given by

6

where [SO₄²⁻] is themodel-calculated aerosol sulfate concentration (ing⋅m⁻³), α = 5.3m²⋅g⁻¹ is thespecific extinction coefficient for dry sulfate aerosols (4),f(RH) ≥ 1 and accounts for the increase in scatteringas relative humidity,RH, increases due to deliquescence(5), andfrac is the ratio of the extinction due to thesulfate portion of the aerosols to the total aerosol extinction.Measurements in urban-source regions in northeastern China indicatethat sulfate is responsible for ≈ 50% of the total aerosolscattering (44). We therefore assumed a value of 0.5 forfrac.

Inspection of Fig.2 reveals annually averagedτ_as ranging from 0.05 or less in western China(where anthropogenic emissions are relatively small) to almost 0.7 inthe Sichuan area. In the eastern part of the country, annually averagedoptical depths tend to be largest in the central portion of the nation(i.e., Sichuan and the Yangtze Delta) relative to areas lying to thenorth and south. From Fig.3, we see that the region of maximum opticaldepths tends to move northward and southward with the seasons,reflecting the changing wind patterns and actinic fluxes. It should benoted that the springtime is the period when China (especially northernChina) experiences significant loadings of wind-blown dust. However,since our estimates were derived from model simulations ofanthropogenic sulfate aerosol, the influence of wind-blown dust is notreflected in the figures.

Comparison with Aerosol Optical Depths Inferred from RadiationMeasurements.

The τ_a values derived from radiationmeasurements in Fig.2 are from Zhouet al. (23). Theseinvestigators retrieved aerosol optical depths from data collected overa 12-year period (1979–1990) at meteorological stations located nearthe outskirts of various cities throughout China. Theτ_as were derived by using the method of Qiu(45), in which the total extinction of the direct beam of sunlight byaerosols is inferred from measurements at each site ofI_s^dir under cloud-free conditions, thenumber of sunshine hours, and surface water vapor partial pressure, aswell as the appropriate ozone column abundance from the Total OzoneMeasurement Satellite (TOMS) Version 7. Zhouet al. (23)reported τ_as for a wavelength of 750 nm. InFig.2, we have scaled these values to a wavelength of 550 nm (which isa more conventional wavelength for reportingτ_a), assuming an Angstrom exponent of 1.However, this scaling does not introduce any additional uncertaintysince Zhouet al. assumed the same value for the Angstromexponent in their original derivation.

Inspection of Fig.2 reveals some qualitative consistency between themodel-based and measurement-based τ_as; forexample, both have maximum values in the Sichuan area and generallylower values to the north and south of the Yangtze Delta region.However, there are significant quantitative differences, with themeasurement-based τ_as being larger than themodel-based values at all locations. The smallest relative differenceis found in the Sichuan area, where the measurement-basedτ_as are ≈30–40% larger than the model-basedvalues. In southeastern China, they differ by a factor of ≈2, innortheastern China the difference is as much as a factor of 5, and tothe west of Sichuan the difference is a factor of 10 or more.

There are a number of possible explanations for thediscrepancies. For example, the measurement-based values may beessentially correct and the model-based values too low, either becauseof an under-prediction in SO₄²⁻or an overestimation infrac. There is in fact some evidencein support of this view. The measurement-basedτ_as are about a factor of 2 larger than thosetypically observed in the eastern United States (i.e., 0.5–0.9 inChina and 0.2–0.5 in the United States). This is qualitativelyconsistent with the particle emission inventories for the two countriesestimated by Wolf and Hidy (46): 46 × 10¹²g⋅year⁻¹ in China and ≈22 ×10¹² g⋅year⁻¹ in theUnited States. The measurement-based τ_as arealso about a factor of 2 larger than those inferred from similarmeasurements made between the late 1950s and 1980 (47). This is alsoqualitatively consistent with emissions inventories, which indicatethat anthropogenic emissions in China increased by a factor of ≈2from the 1970s to the 1980s (48).

It is also possible that the measurement-basedτ_as are too large, perhaps because of animproper filtering of cloud-influenced data. The rather largeτ_a values reported by Zhouet al.(23) in the western part of China, where anthropogenic emissions arerelatively low, suggest that this may have occurred for at least someof the sites.

A third possibility is that both the measurement-based and model-basedτ_as are correct but reflect different spatialscales. Since the radiance measurements were made in suburbanlocations, these data may have been affected by local pollution sourcesand thus are not representative of the surrounding region. The modelresults, on the other hand, were obtained by using a 60- × 60-km gridand are, by definition, regional in scale.

In the calculations presented below, we use the model-estimatedτ_a values to evaluate the effect of regionalhaze in China on surface solar irradiance and, in turn, on crop yields.Since the model-estimated τ_a values are thesmaller of the two sets of τ_as, this approachprovides us with a more conservative measure of the regional-hazeeffect.

Calculations of I_s Over China.

To estimate the direct effect of the aerosol loadings discussed aboveon solar radiation, a broad-band, one-dimensional radiative transfermodel (49) was used to calculate the surface solar irradiance as afunction of wavelength. The solar spectrum from 200 nm to 4 μm (i.e.,the wavelengths that encompass ≈99% of the total solar irradiancereaching the top of the atmosphere) was divided into 15 bands. Theradiative transfer equation for each band was then solved by using theΔ four-stream approximation, an approach that has been shown to bewell suited for calculating solar radiative fluxes and heating rates inan atmosphere containing aerosols and clouds (50).

The magnitude of direct effect on the surface solar irradiance for agiven value of τ_a was obtained by carrying outtwo sets of calculations—one with τ_a equal tothe model-estimated value and another with τ_aset to zero—and then calculatingΔI_s^tot(τ_a), therelative change in the total surface irradiance, where

7

Note that, given the above definition, a positive value inΔI_s^tot denotes a reduction in the surfaceirradiance due to the aerosol direct effect.

The calculations were made by using July-averagedτ_as and July 15 solar zenith angles appropriatefor each location. We have chosen summertime conditions because,although not shown here, the discrepancies between model-estimated andmeasurement-based τ_as tended to be smallestduring July. The profiles for atmospheric temperature, pressure, watervapor, and ozone used in the model calculations were taken from themid-latitude summer atmospheric conditions reported by McClatcheyet al. (51). A flat surface with an albedo of 20% at 0 kmabove sea level was also assumed for all calculations. In addition tothe above, the model calculations require the specification of a numberof input parameters. These are discussed below and summarized in Table1.

The model calculations have a relatively weak dependence on thealtitude profile for the aerosol scattering and absorption. This wasdefined by assuming that the aerosol scattering and absorption werelinearly proportional to the total aerosol mixing ratio (with analtitude independent proportionality constant) and, in turn, specifyingan altitude profile for the aerosol mixing ratio. In all calculations,the aerosols were assumed to be well mixed from the surface to the topof the boundary layer, which was set to a height of 2 km, a valueappropriate for summer clear sky conditions. [Fine mode aerosols, aswell as particles in the smaller-sized fraction of the coarse mode, areexpected to be well mixed in the boundary layer because theiratmospheric residence times are significantly longer than the few-hourmixing time of the boundary layer. Aircraft measurements of aerosolconcentrations tend to support this inference (52).] Above theboundary layer, aerosol mixing ratios were assumed to decreaseexponentially with altitude using a scale height of 1 km, similar tothat of Liuet al. (8).

Clouds are highly variable in space and time and have a stronginteraction with solar radiation. Thus our assumptions concerningclouds may have a significant effect on our model calculations. Toassess the uncertainty in our calculations that might arise fromclouds, we have carried out calculations with three differentassumptions concerning cloud cover and the transmissivity of theclouds. In our base model calculations, we assume cloud freeconditions. In addition, sensitivity calculations are presented for a“Thin Cloud Case” in which 100% cloud cover is assumed with acloud transmissivity at all wavelengths of 80% and for a “ThickCloud Case” in which 100% cloud cover is assumed with a cloudtransmissivity of 40%. In each of these cases,ΔI_s^tot was calculated from Eq.6 by using the value forI_s^tot(0)appropriate to that case.

The optical properties of the aerosols (i.e.,w_a,g_a) depend on the chemical compositionand size distribution of the aerosols over China (which have yet to becharacterized and most likely vary with time and location). The modelresults are relatively insensitive to the asymmetry factor,g_a, and thus the specification of thisparameter is not critical to our conclusions (53,54). We have adopteda value of 0.67 and assumed that the angular distribution of thescattering could be represented with a Henyey-Greenstein phase function(55,56).

Unlike the asymmetry factor, the model calculations are quite sensitiveto the single scattering albedo,w_a.Recall from Eq.6 that this factor decreases as theabsorptivity of the aerosols increases. Absorption of solar radiationby aerosols is generally due to the presence of elemental carbon (e.g.,in soot) or mineral aerosols. In the United States, aw_a of ≈0.9 is often observed (14).However, in China, where emissions of soot and dust may very likely besubstantially higher, a value forw_aof 0.9 may be too high. Recent measurements made continuously over a2-week period in Beijing by one of the authors (M.B.) indicated a meanw_a of 0.80 ± 0.06. To bracketthe possible range in the single scattering albedo, we have carried outmodel calculations using two values forw_a: a “Low Absorption Case” withw_a = 0.95 and a “High AbsorptionCase” withw_a = 0.75.

Effect of Reductions in Surface Irradiance on “Optimal” CropYields Grown in China.

To assess the likely impact of reductions inI_s^tot on crop yields in China, we carriedout a series of sensitivity calculations using a crop response modelfor both rice and winter wheat. Before describing the specifics ofthese calculations, a brief discussion of the role of sunlight inlimiting photosynthesis rates in agricultural systems is useful. Ingeneral, photosynthesis depends on a variety of inputs, including waterand nutrients as well as sunlight (58). Any one of these inputs canlimit the rate at which green plants carry out photosynthesis and storecarbon. In natural or unmanaged ecosystems, nutrients and/or waterare often limiting, and thus reductions in surface solar irradiance maynot have a significant impact on photosynthesis rates. In managedecosystems such as those in cultivation for food crops, on the otherhand, conditions are often manipulated to maximize crop yields throughirrigation and the application of fertilizers. Thus the possibilitythat surface irradiance can affect net yields of crops is far greater.

In the specific case of rice, field observations tend to confirm thatyields are, in fact, affected by the amount of solar radiation receivedby the crops (15–17,20). For example, Fig.5illustrates data gathered on rice grown in commercial and experimentalfields in Texas. In both datasets, there is a statistically significantpositive correlation between the rice yields and the cumulative surfacesolar irradiance received during a 40-day period when rice has itsgreatest sunlight requirement. On the basis of these and other data,Stansel and Huke (15) posited a strong linear relationship betweensolar radiation and the yields of rice. Wheat yields are more oftenlimited by moisture or temperature than by solar radiation. However,when these conditions are not limiting, yields can be affected byvariations in solar radiation receipt (18).

In the crop response model calculations presented here, we assume asufficiency of water and nutrients for the crops so that they are neversubjected to water and/or nutrient stress. For this reason, ourresults represent the effect of reductions inI_s^tot on optimal crop yields as opposed tocrop yields grown under a specific set of conditions during a specificyear. Depending on the actual conditions under which the crops aregrown in China, these optimal yields may or may not reflect the actualyields.

Implications for Agriculture in China: An Opportunity for IncreasedCrop Yields.

If the crop yield model calculations discussed in the previoussection for rice and wheat grown in Nanjing are generally applicable toChinese agriculture, then our calculations suggest that air pollutionand regional haze in China are having a significant impact on theoptimal yields of crops grown in the nation. The predicted reduction inyields for China’s highly productive eastern agricultural regionsranges from a little less than 5 to more than 30% and depends mostcritically on the absorptive properties of the aerosols. However, it islikely that this range represents a lower limit because (i)the indirect effect of aerosols on solar radiation was neglected;(ii) the radiative transfer calculations were carried usingthe smaller, model-estimated τ_as;(iii) the possible reduction in surface irradiance by thepollutant nitrogen dioxide was not considered (61); and (iv)the effects on crops of phytotoxic air pollutants typically associatedwith regional haze such as ground-level ozone (30) were not considered.

One implication of our results is that mitigation of regional haze overChina could have the benefit of significantly increasing the optimalyields of crops grown in the nation. Translating these optimal-yieldincreases into actual increases in agricultural productivity would ofcourse require sufficient supply of water and nutrients. However, givenChina’s plans to boost agricultural productivity by ≈30% over thenext three decades (27), this would not appear to be a problem.

Conclusion.

A rudimentary assessment of the direct effect of atmospheric aerosolson agriculture in China suggests that optimal crop yields aresignificantly affected by regional-scale air pollution and itsassociated haze. This in turn implies that the mitigation of thispollution could help boost crop yields in China. Our calculationssuggest that, under optimal growing conditions, crop yields in easternChina could be enhanced by ≈5–30%, possibly more if the indirecteffect by aerosols and other air pollutants also significantly affectcrop yields.

Whether such a scenario is feasible and economically realistic isa question whose answer will require a good deal of furtherstudy—including more detailed measurements of the chemical and opticalproperties of aerosols over agricultural areas of China. Nevertheless,given the projections of a rapidly rising food demand in China in thecoming decades, and concerns about whether this rising demand can bemet internally through enhanced agricultural productivity, furtherstudy may prove to be a worthwhile endeavor.

More generally, it should be noted that regional haze is not unique toChina. It occurs in virtually all heavily populated regions, those thatare developing economically as well as those that are alreadyeconomically developed (4,14). While a good deal of effort is beingmade to characterize the climatic impact of this reduction, furtherinvestigation of the direct ecological impacts, especially thoseoccurring within agricultural systems, may also prove to be worthwhile.

Acknowledgments

We thank Robert Dickinson at Georgia Tech, Jim Jones at theUniversity of Florida, and Susan Solomon of the NOAA AeronomyLaboratory for acting as reviewers and providing helpful comments andsuggestions, as well as the instrument mentor R. Halthore at BrookhavenNational Laboratory for helpful discussions related to the data. Thiswork was supported in part by funds from the U.S. National Aeronauticsand Space Administration under Grant NAG5–3855 for the China-MAPresearch project. Cinel sunphotometer data were obtained from theAtmospheric Radiation Measurement (ARM) Program sponsored by the U.S.Department of Energy, Office of Energy Research, Office of Health andEnvironmental Research, Environmental Sciences Division. Cinelsunphotometer is part of AERONET, a network of sunphotometers managedby B. N. Holben (NASA/GSFC).

Footnotes

References