Season-long elevation of ozone concentration to projected 2050 levels under fully open-air conditions substantially decreases the growth and production of soybean

Authors

  • Patrick B. Morgan,

    1. Department of Plant Biology, University of Illinois, 379 Edwin R. Madigan Laboratory, 1201 West Gregory, Urbana, IL 61801, USA;
    Search for more papers by this author
  • Timothy A. Mies,

    1. Department of Crop Science, University of Illinois, AW 101 Turner Hall, 1102 S. Goodwin Avenue, Urbana, IL 61801, USA;
    Search for more papers by this author
  • Germán A. Bollero,

    1. Department of Crop Science, University of Illinois, AW 101 Turner Hall, 1102 S. Goodwin Avenue, Urbana, IL 61801, USA;
    Search for more papers by this author
  • Randall L. Nelson,

    1. USDA-Agricultural Research Service, Soybean/Maize Germplasm, Pathology and Genetic Research Unit, Urbana, IL 61801, USA
    Search for more papers by this author
  • Stephen P. Long

    1. Department of Plant Biology, University of Illinois, 379 Edwin R. Madigan Laboratory, 1201 West Gregory, Urbana, IL 61801, USA;
    2. Department of Crop Science, University of Illinois, AW 101 Turner Hall, 1102 S. Goodwin Avenue, Urbana, IL 61801, USA;
    Search for more papers by this author

Author for correspondence: Stephen P. Long Tel. +1 217 3332487 Fax: +1 217 2447563 Email: stevel@life.uiuc.edu

Summary

  • • Mean surface ozone concentration is predicted to increase 23% by 2050. Previous chamber studies of crops report large yield losses caused by elevation of tropospheric ozone, and have been the basis for projecting economic loss.
  • • This is the first study with a food crop (soybean, Glycine max) using free-air gas concentration enrichment (FACE) technology for ozone fumigation. A 23% increase in ozone concentration from an average daytime ambient 56 p.p.b. to a treatment 69 p.p.b. over two growing seasons decreased seed yield by 20%.
  • • Total above-ground net primary production decreased by 17% without altering dry mass allocation among shoot organs, except seed. Fewer live leaves and decreased photosynthesis in late grain filling appear to drive the ozone-induced losses in production and yield.
  • • These results validate previous chamber studies suggesting that soybean yields will decrease under increasing ozone exposure. In fact, these results suggest that when treated under open-air conditions yield losses may be even greater than the large losses already reported in earlier chamber studies. Yield losses with elevated ozone were greater in the second year following a severe hailstorm, suggesting that losses caused by ozone might be exacerbated by extreme climatic events.

Introduction

In industrialized countries of the northern hemisphere, tropospheric ozone concentration ([O3]) has risen by 1–2% per year (Chameides et al., 1994). Nearly one-quarter of the earth's surface is currently at risk from tropospheric ozone in excess of 60 parts per billion (p.p.b.) during mid-summer with even greater concentrations occurring locally (Fowler et al., 1999a,b; Ashmore, 2005; Karnosky et al., 2005). Western Europe, midwestern and eastern USA and eastern China are being exposed to some of the highest background [O3] (Prather et al., 2001). An economic assessment of air pollution effects on agriculture calculated that a 25% reduction in current tropospheric [O3] could benefit USA agriculture by $1–2 billion annually (Murphy et al., 1999). With the global human population currently over 6 billion and expected to exceed 8 billion by 2050 (Lutz et al., 2001), assessing the impacts of changing atmospheric [O3] on crops is crucial to understanding future food security. The Intergovernmental Panel on Climate Change (IPCC) Third Assessment Report used over 10 different atmospheric chemistry models to project future surface [O3], all suggested global and regional increases in background [O3]. Averaged across these models a global 23% increase by 2050 is projected with the IPCC A2 Emissions Scenario (Prather et al., 2001). Predicted increases for July over the next 100 yr are predicted to average 30 p.p.b. in midwestern USA and 50 p.p.b. for eastern China, both major soybean (Glycine max) growing areas (Prather et al., 2003).

Soybean occupies more land globally than any other dicotyledonous crop and was the fourth most important crop in 2002 with respect to harvested grain and area occupied (FAO-UN, 2003). It may serve as a model species for studying the responses of other annual plants in general to rising tropospheric ozone (Morgan et al., 2003). The soybean–maize agricultural rotation system occupied c. 63.2 million ha in the USA in 2004 – more than that of any other crop system – making it arguably the largest ecosystem in the contiguous 48 states (USDA, 2005). Among the major crops, soybean is one of the most susceptible to ozone with adverse effects apparent at [O3] as low as 40 p.p.b. (Fuhrer et al., 1997; Ashmore, 2002).

From previous studies, elevated ground-level [O3] has been shown to have a deleterious effects on soybean dry matter production and yield (Heagle et al. 1998; Miller et al. 1998). These studies comprise a considerable volume of literature, but until the current study, the effect of ozone on soybean or any other crop species has not been examined using free-air gas concentration enrichment (FACE) technology. A recent meta-analytic summary of 53 peer-reviewed journal articles on the effect of elevated [O3] on soybean, showed that above-ground dry mass at maturity was decreased by 40% on average, by a mean [O3] of 60 p.p.b., when compared with plants grown in ozone-free air (Morgan et al., 2003). Where mean [O3], in this context, is the average concentration for the 7–12 daylight hours in which ozone was added (Ashmore, 2002; Morgan et al., 2003). Studies elevating mean [O3] for 7 h d−1 within open-top chambers suggest that increase from 30 p.p.b. to 60 p.p.b. would decrease soybean seed yield by 16% (calculated from Ashmore, 2002). Such elevations of surface [O3] are common in rural areas in the USA (Morgan et al., 2003; Prather et al., 2003), suggesting that soybean yields likely are already reduced by [O3]. As noted above, background surface [O3] is expected to rise a further 30 p.p.b. over this century in the American Midwest. Extrapolating from Ashmore (2002), this increase would lower soybean yield by an additional 16%. Such losses under open-air field conditions, when combined with losses caused by simultaneous climate change, could lower global food supply during this century more than previously anticipated (Long et al., 2005).

Although these previous studies provided a comprehensive assessment of the elevated [O3] effects on soybean, all studies were conducted in chambers such as growth cabinets, glasshouses or open-top chambers. In the absence of fully open-air field-scale treatment, there is uncertainty as to whether these effects will be realized under open-field conditions. The meta-analysis of soybean chamber studies suggested that photosynthetic rates of the topmost leaves were decreased by 20% when [O3] was elevated from c. 60 to 75 p.p.b. (Morgan et al., 2003). However, it was recently demonstrated that soybean grown in [O3] elevated to this level under open-air field conditions failed to show any decrease in the photosynthetic rate of topmost leaves upon completion of expansion (Morgan et al., 2004). A small cumulative [O3] effect on leaf photosynthesis occurred as leaves, which were formed during pod-filling, approached senescence (Morgan et al., 2004). Dermody et al. (2006) similarly found no effect of elevation of ozone on leaf area index during early vegetative growth, but did find a pronounced effect in late grain filling. As photoassimilates are the basis for dry matter accumulation, these different photosynthetic and leaf area responses to elevated [O3] likely lead to differences in plant growth between chamber and FACE studies. The forced ventilation of chambers might increase ozone effects by improved coupling of the concentration in the atmosphere with the leaf surface. Elagoz & Manning (2005) in a detailed analysis of another grain legume showed substantial differences between plants grown within open-top chambers and those grown outside in ambient atmospheres. Such findings suggest that alterations in microclimate and rainfall within open-top chamber could modify plant responses to elevated ozone.

Current FACE technology, provides a means to examine the impact of elevated [O3] with minimal alteration of microclimate and the soil–plant–atmosphere continuum (McLeod, 1995; McLeod & Long, 1999; Long et al., 2004). It also allows treatment of a much larger area of crop than is possible within chambers. Soybean provides a real-world agronomic system that is well suited for field studies of elevated [O3] at the low replication (n = 3 or n = 4) typical of FACE systems. As an inbred crop, cultivated soybean on homogeneous agricultural soils will show minimal phenotypic variation (Gizlice et al., 1994; Cui et al., 2001; Li et al., 2001). The homogenous nature of the crop ecosystem in combination with a fully replicated, randomized complete-block experimental design increases the likelihood of detecting subtle elevated [O3] effects even in the field. Because of the large size of the field plots, frequent destructive sampling was possible to determine both the timing of elevated [O3] effects and changes in dry matter partitioning.

This study examined the effect of elevated [O3] on soybean above-ground net primary production (ANPP) and yield under fully open-air field conditions. The primary objectives were to determine: (1) whether the large decreases in production caused by growth under elevation of [O3] to the levels anticipated for 2050 in chambers are realized under fully open-air field conditions; (2) the growth stage at which production losses occurred; and (3) the effects on partitioning within the plant.

Materials and Methods

Site description

The soybean FACE (SoyFACE) facility consists of a 32 ha field (80 acre; South Farms, University of Illinois at Urbana-Champaign, USA; 40°03′21.3′ N, 88°12′3.4′ W, 230 m elevation; http://www.soyface.uiuc.edu). Soybean (G. max (L.) Merr.) and maize (Zea mays L.) are each planted in half of the field and follow an annual rotation typical of contemporary Illinois and Corn Belt agriculture. SoyFACE elevated [O3] treatments consisted of eight octagonal plots within the 16 ha planted with soybean (for extended site and operation descriptions see Ainsworth et al., 2004; Rogers et al., 2004; Dermody et al., 2006). The plots were separated by 100 m to avoid cross-contamination. The ozone fumigation system was based on the CO2 FACE design of Miglietta et al. (2001) and adapted to release ozone (for a full description see Morgan et al., 2004). The ozone concentration in control plots was continually monitored, and enrichment in the treated plots was proportional to the concentration in the controls such that enrichment tracked the natural pattern of ozone formation (Morgan et al., 2004). Ambient concentrations and concentrations in the treatment plots were continuously monitored (model 49C O3 analyser; Thermo Environmental Instruments, Franklin, MA, USA; calibration USA EPA Equivalent Method EQQA-0880–047, range 0 to 0.05–1.0 p.p.m.). Increase in future [O3] is assumed to result from increased background concentrations of precursors and climatic conditions more favourable to ozone formation (Prather et al., 2003). It was therefore assumed that weather conditions favouring ozone formation today will be those favouring formation tomorrow. Rather than a fixed increase, ozone was added as a fixed proportion of the instantaneous background [O3] recorded in the control plots. The treatments were current [O3] (control) and elevated [O3] (current × 1.23) arranged in a randomized complete block design (n = 4) to control for small differences in topography and soil across the field. The target for elevated [O3] was c. 23% above current concentrations, during daylight hours, based on the IPCC projected mean increase in tropospheric [O3] for 2050 (Prather et al., 2001). Ozone was not added when leaves were wet. This resulted in several hours and some days when [O3] was not elevated. To achieve a season-long 23% elevation it was therefore necessary to augment fumigation by elevating to a higher level on dry days. The maximum increase on any one day was limited to 50% above the concentration in the control plots.

Ozone fumigation began 20 d after planting and continued daily during daylight hours through the growing season until harvest. The maximum 8-h average [O3] was 62 p.p.b. and 50 p.p.b. in the control plots for 2002 and 2003, respectively, and 75 p.p.b. and 63 p.p.b. in the elevated [O3] plots. The effective increase in [O3] in the elevated plots was 21% and 25% for 2002 and 2003, respectively. Based on 1-min average concentrations, the achieved elevation was within ±10% of the set point concentration for 74% of the time, and within ±20% of the set point 90% of the time. Weather was similar in both years (Fig. 1), although total rainfall was slightly higher in 2003 (approx. 370 vs 430 mm, in 2002 and 2003, respectively). However, in 2003 a severe hailstorm in July removed many of the leaves in all plots. During the 2003 pod-filling developmental stage, a depression in daily 8-h maximum average [O3] coincided with decreased temperature and light that likely resulted from cloud cover.

Figure 1.

Weather at the SoyFACE facility during the 2002 and 2003 growing seasons. Rainfall is reported as daily (small closed bars) and cumulative monthly (large open bars) for both years. The minimum and maximum temperature, and daily integral of photosynthetic photon flux density (PPFD) are given for each day from planting to final harvest.

Planting

An indeterminate soybean cultivar 93B15 (Pioneer Hi-Bred, Des Moines, IA, USA) of maturity group III was planted on June 1 (day of year, DOY 152) in 2002 and May 27 (DOY 147) in 2003. The crops were planted using a mechanical seed planter to a field density of c. 200 000 plants per hectare. Row spacing was 0.38 m and cultivation followed standard agronomic practice. Rows within the plots were oversown by hand on the day of planting and thinned after crop emergence to approximately 20 plants per meter row length to ensure uniform density. In line with most soybean cultivation, the crop was not irrigated. July–August precipitation was 180 mm in 2002 and 250 mm in 2003; the 110 yr mean for this period in central Illinois is 180 mm. The crop was treated with a pre-emergent herbicide (Boundary; Syngenta Crop Protection Canada, Inc., Guelph, ON, Canada) in both years and again post-emergence in 2003 on DOY 175. A general insecticide (Sevin; Bayer CropScience, Research Triangle Park, NC, USA) was used to control an infestation by Japanese beetle (Popillia japonica Newman) in 2002 before defoliation could reach a level that would have affected yield.

Harvesting

Within each octagonal plot, an area of approx. 18 m−2 was used for biweekly harvests to assess the growth of the crop and a second area of approx. 29 m−2 used to assess the final yield at maturation. Biweekly over both growing seasons, two small subplots consisting of one section of row (60 cm long in 2002, 100 cm long in 2003) were selected within the 18 m−2 area within each octagonal plot. The row width was 38.1 cm. These subplots were selected by fully randomized design with the restriction that they could not be within 40 cm of a previously sampled subplot or the edge of the sampling area. Plants were sampled between 09 : 00 h and 11 : 00 h. Before cutting, basal diameter at the soil surface was measured using vernier callipers for all plants within the subplot. Soybean stems were then cut within 5 mm of the soil surface and all plants were placed in paper bags. The paper bags were immediately placed over ice in insulated storage containers to minimize wilting and dry mass losses through respiration. Sampling, counting and measuring of all subsamples were completed within the day of collection to minimize mass losses as a result of respiration or decay. Litterfall was captured in 60 × 38 cm galvanized steel mesh baskets placed between rows. The baskets were supported 3 cm above the soil to minimize decay. Litter was removed from the baskets biweekly and dried as described later.

Developmental stage was reported according to the system described by Ritchie et al. (1997). For each plant within each subplot, the numbers of leaves, branches, pods and nodes were recorded. The following criteria were applied to determine whether an organ should be included in a count and dry mass for a given category. Leaves were counted if they were open trifoliates following Ritchie et al. (1997) for which at least one complete leaflet remained and ≥ 50% of the area was green. If < 50% of the leaf was green, it was included in litter. Unopened trifoliates were included in stem dry mass although their contribution was minimal (< 5 g m−2). Pods were included in the determination of number and the dry mass assessment if they contained at least one seed that was more that 3 mm in length (Ritchie et al., 1997), otherwise their mass was included as part of the stem. Branches were counted provided they had at least one node with an attached leaf or a petiole scar. Node number was counted and plant height measured from the cut base to the topmost node bearing an open trifoliate. Normally, the primary shoot was the tallest, but following destruction of most apical meristems on the primary stems in the 2003 hailstorm (DOY 198), node number was counted and plant height was measured to the topmost node bearing an open trifoliate on the highest branch (apical replacement). After counting, the shoot of each plant was separated into stems and petioles, leaf tissue (based on the above criteria), and pods. These were then placed in drying bags and dried in a forced-air oven at 65°C to constant mass, requiring at least 4 d. Above-ground net primary production (ANPP) was calculated as the sum of the dried leaf, stem, pod and cumulative litterfall mass. Immediately following the 2003 hailstorm, fallen material was collected from a 4 m2 area within each octagonal plot, washed to remove soil, and dried to determine loss.

In 2003, once pods had appeared, random subsamples of 100 pods from each small subplot were taken to assess the number of seeds per pod. After drying, subplot samples were mechanically threshed to separate pod husks from seeds. The seeds were redried for 1–2 d to constant weight and seed mass was determined. From the threshed seeds, a subsample of 200 seeds from each subplot was weighed to calculate the individual seed mass.

At crop maturity, final yield was assessed by harvesting a large (29 m−2) subplot within each octagonal plot. No plant material had been removed from this area at any stage during the growing season. All shoot dry mass was harvested by hand and threshed mechanically to separate seed from all other shoot components, the ratio of these two giving the harvest index (HI).

Statistical analysis

Effects of elevated [O3] on seed yield, HI and cumulative litterfall were assessed with a randomized complete block mixed model analysis of variance (PROC MIXED, SAS v8.01; SAS Institute, Cary, NC, USA) with treatment as the fixed effect (α = 0.1). For all other parameter comparisons, a repeated measures mixed model analysis of variance (PROC MIXED) was used with DOY, treatment, and the DOY by treatment interaction as fixed effects (α = 0.1). Years were analysed separately because of the defoliation caused by hail in 2003. For all dry mass and plant growth parameters that could be recorded throughout the growing season (e.g. leaf mass and number), the effect of sampling dates within a year were analysed. For dry mass and plant parameters that developed in later stages (e.g. seed mass and pod number), the analysis was conducted from the first sampling date on which they were observed to final harvest. The best-fit variance/covariance matrices were chosen for each variable using Akaike's information criterion to correct for inequality of variance between the sampling dates (Keselman et al., 1998; Littell et al., 1998, 2000). A priori pairwise comparisons were made between treatments within sampling dates (α = 0.05).

Results

Weather conditions at the experimental site were typical of the region in 2002, but slightly wetter and cooler than average in 2003 (Fig. 1). Ozone fumigation began 20 d after planting and continued daily during daylight hours through the growing season until harvest. The maximum 8-h average [O3] was 62 p.p.b. and 50 p.p.b. in the control plots for 2002 and 2003, respectively, and 75 p.p.b. and 63 p.p.b. in the elevated [O3] plots (Fig. 2). The maximum growing-season 8-h average [O3] for the preceding 5 yr was 59 p.p.b. Therefore, 2002 was close to this average, but 2003 significantly below average. The lower [O3] in 2003 resulted in lower accumulated exposure over a threshold of 40 p.p.b. (AOT40) and sum of hourly average [O3] greater than or equal to 60 p.p.b. (SUM06) in both control and elevated treatment plots (Fig. 2), but also resulted in greater relative increases in the treatment plots (130% increase in AOT40 in 2003 vs 90% in 2002).

Figure 2.

Ozone treatment for 2002 and 2003 growing seasons. In 2002, ozone fumigation began approx. 20 d after planting of the crop. The daily maximum 8-h average concentration is reported for current (black line) and elevated (grey circles) [O3] measured at crop height. Where the maximum 8-h average is the mean concentration of the 8-h period within each day with the highest ozone concentration, most commonly 09 : 00–17 : 00 h. The cumulative accumulated exposure over a threshold of 40 p.p.b. (AOT40) and sum of hourly average [O3] greater than or equal to 60 p.p.b. (SUM06) ozone indices for the season are calculated from the 1-h average of daytime concentrations for current (open circles) and elevated (grey circles). Where SUM06 = inline image(units = p.p.m.h) andAOT40 = inline image (units = p.p.m.h).

Elevated [O3] decreased seed yield from 467 g m−2 in the controls to 397 g m−2 (15%) in 2002 and from 287 g m−2 in the controls to 215 g m−2 (25%) in 2003 (Fig. 3; statistical significance of yield and biomass changes are given in Table 1). Although the absolute decrease in yield was the same in both years, the relative decrease was much greater in 2003. This likely resulted from a hailstorm in July that severely damaged the crop, removing approx. 45 g m−2 of shoot dry mass, and from which the elevated [O3] plants recovered more slowly. In 2002, reduced yield was entirely attributable to a significant decrease in individual seed weight, the number of pods per plant and seeds per pod being unaffected (Fig. 3 and Table 1). However, yield losses in 2003 resulted from both lighter individual seeds (8%) and four fewer pods per plant (17% decrease, statistical significance given in Table 2). Elevated [O3] had similar impacts on both yield and shoot dry mass at maturity. Although a slight increase in harvest index was indicated, it was not significant (Table 1).

Figure 3.

Seed yield and the average weight per seed at final harvest in 2002 and 2003 for soybean (Glycine max cv. Pioneer 93B15) grown under current (open bars) or elevated [O3] (closed bars) in the field. Mass of individual seeds was calculated from a subsample of 200 seeds for each subplot for each growing season. Each reported value is the least square mean of four replicate plots and the error bars indicate ± standard error of the difference between means. Stars denote significant (α = 0.1) treatment effects.

Table 1.  The probability of significance of crop-level responses to [O3] treatment (Treat) and day of year (DOY) in two growing seasons
 2002 TreatDOYTreat × DOY2003 TreatDOYTreat × DOY
  1. Mixed model analysis of variance of fixed effects on measured growth and dry mass of field-grown soybean (Glycine max cv Pioneer 93B15) in current and elevated [O3] over the growing season. Cumulative above-ground net primary production (ANPP) is the sum of shoot dry mass and cumulative litterfall per unit ground area. Small plot P-values were calculated from a repeated measures, mixed model analysis of variance of a randomized complete block with DOY, treatment and the subsequent interaction as fixed effects. Bold values indicate significance at α < 0.1. NA, not applicable.

Large Subplots
Seed yield (g m−2)0.023NANA0.056NANA
Harvest index0.118NANA0.390NANA
Small subplots
ANPP (g m−2)0.020< 0.00010.005< 0.0001< 0.00010.0001
Shoot dry mass (g m−2)0.010< 0.00010.0004< 0.0001< 0.00010.112
Stem dry mass (g m−2)0.015< 0.00010.0007< 0.0001< 0.0001< 0.0001
Leaf dry mass (g m−2)0.002< 0.00010.0120.0004< 0.00010.247
Pod dry mass (g m−2)0.0002< 0.00010.1920.0009< 0.00010.006
Litterfall (g m−2)0.979< 0.0001< 0.00010.015< 0.00010.050
Mass per seed (g)0.002< 0.00010.0090.027< 0.00010.670
Stem proportion0.871< 0.00010.2620.496< 0.00010.793
Leaf proportion0.737< 0.00010.0680.365< 0.00010.038
Pod proportion0.571< 0.0001< 0.00010.966< 0.00010.498
Table 2.  The probability of significance of individual plant-level responses to [O3] treatment (Treat) and day of year (DOY) in two growing seasons
 2002 TreatDOYTreat × DOY2003 TreatDOYTreat × DOY
  1. Mixed model analysis of variance of fixed effects on plant dimensions and organ numbers of field-grown soybean (Glycine max cv Pioneer 93B15) in current and elevated [O3] over the growing season. P-values were calculated from a repeated measures, mixed model analysis of variance of a randomized complete block with DOY, treatment and the subsequent interaction as fixed effects. Bold values indicate significance at α < 0.1; Dashes, missing data.

Individual plants
Basal diameter (mm)0.0002< 0.00010.6170.633< 0.00010.719
Height (cm)< 0.0001< 0.00010.0040.004< 0.00010.038
Branches (number)0.060< 0.00010.6230.308< 0.00010.182
Nodes (number)0.440< 0.00010.8970.046< 0.00010.180
Leaves (number)0.003< 0.00010.2500.145< 0.00010.828
Pods (number)0.432< 0.00010.8140.038< 0.00010.281
Seeds per pod (number)0.289  0.5680.318

Elevated [O3] decreased shoot dry mass and the mass of the constitutive organs in both growing seasons, but to a greater extent following the hailstorm in 2003 (Table 1; Fig. 4). Decreases in shoot dry mass of plants grown in elevated [O3] relative to controls appeared late in the 2002 growing season and leaf, stem and pod dry mass all reflected this late season loss. However, decreases were apparent throughout 2003, notably in stem dry mass, possibly reflecting a weakened capacity for recovery following the hailstorm (Fig. 4). The 2003 soybean crop was 60% defoliated on a leaf area basis by hail in both treatments (O. Dermody, unpublished) reducing shoot dry mass by approximately 21% on DOY 198 (Fig. 4), however, the crop recovered and redeveloped a closed canopy. More rapid recovery in the control [O3] was reflected in increased node formation, which also corresponded to increased plant height in the controls (Table 2). Loss of total leaf dry mass with senescence of the crop was accelerated in elevated [O3], which coincided with two to three fewer leaves per plant during pod-fill in 2002 (Table 2). The decreased production in elevated [O3] in 2003 following the hail may also explain the lower biweekly litterfall (Table 1; Fig. 5). Cumulative litterfall for the season was significantly decreased in 2003 by elevated [O3], but not in 2002 (Fig. 5). In 2002 litterfall was therefore a greater proportion of total biomass in elevated [O3], probably reflecting earlier senescence of leaves.

Figure 4.

Shoot dry mass over the two growing seasons for soybean (Glycine max) under current (open symbols) or elevated [O3] (closed symbols) in the field. Shoot, plant, leaf, stem and pod dry mass per unit ground area (g m−2) are reported. Developmental stages are represented by the horizontal bar as follows vegetative (open), flowering (R1, cross-hatched), and pod-filling (R4, closed). Each point is the least square mean of four replicate plots and the error bars indicate ± standard error of the difference between means on each date. Significant (α = 0.05) and marginally significant (α = 0.1) pairwise linear contrasts between treatments within each date are denoted by stars and bracketed stars, respectively. The arrow in the upper panel for 2003 indicates the date of the major hailstorm which partly defoliated the crop.

Figure 5.

Cumulative leaf, petiole and stem litterfall from soybean (Glycine max) grown under current (open bars) or elevated [O3] (closed bars) in the field. Dry mass was collected throughout the growing season. Stars denote significant (α = 0.1) treatment effects.

Partitioning of shoot dry mass to stem was not significantly affected by elevated [O3] (Table 1). Dry mass partitioning to leaves in elevated [O3] was significantly less during late pod-filling. This was likely caused by increased senescence and leaf loss (P = 0.0117, DOY 259) in the 2002 growing season, which indirectly increased the proportion of shoot mass that was comprised of pods. Following the hailstorm in 2003, the proportion of shoot mass represented by leaves significantly increased in elevated [O3] relative to controls reflecting a decreased rate of stem growth probably resulting from this atypical growth season (Fig. 4).

The depression of cumulative ANPP caused by elevated [O3] increased as the growing season progressed, as illustrated by the very significant interactions with DOY in both years (Fig. 6; Table 1). In 2002, significant differences (pairwise linear contrasts) developed late in the growing season and persisted throughout the remainder of the soybean lifecycle (Fig. 6). The cumulative effect of elevated [O3] over the 2002 season decreased ANPP by 11% compared with controls. In 2003, ANPP of the control was 50% lower than in 2002 and the impact of elevated [O3] was greater, decreasing ANPP by 23% (Fig. 6).

Figure 6.

Cumulative above-ground net primary production (ANPP) over two seasons for soybean (Glycine max) grown under current (open symbols) or elevated [O3] (closed symbols) in the field. Developmental stages are represented by the horizontal bar as follows vegetative (open), flowering (R1, cross-hatched), and pod-filling (R4, closed). Each point is the least square mean of four replicate plots and the error bars indicate ± standard error of the difference between means on each date. Significant (α = 0.05) and marginally significant (α = 0.1) pairwise linear contrasts between treatments within each date are denoted by stars and bracketed stars, respectively. The arrow in the upper panel for 2003 indicates the date of the major hailstorm which partly defoliated the crop.

Discussion

Ashmore (2002) showed, from over 30 independent chamber studies of season long fumigation of soybean, that there was a linear decline in yield with increase in the 7 h. mean ozone concentration from c. 30–130 p.p.b. of 0.53% per p.p.b. Applying this decline to the 8-h mean ozone concentrations in the present experiment implies that elevation from the ambient 62 p.p.b. to 75 p.p.b. in 2002 and the ambient 50 p.p.b. to 63 p.p.b. in 2003 would have both resulted in an 8% yield loss in each year. However, the observed yield loss was 15% in 2002 and 25% in 2003 (Fig. 3). This was an unexpected result since increased coupling of atmosphere to vegetation in chambers has been suspected of increasing ozone uptake (McLeod, 1995; McLeod & Long, 1999) and thereby amplifying yield loss. This finding therefore not only confirms the large decreases in yield with rising [O3] predicted from chamber studies but suggests that the real loss might be even greater.

One explanation for the greater yield losses could be the fact that since there was no fumigation on wet days, fumigation on dry days could be up to 50% above ambient [O3] (Fig. 2). If yield loss was nonlinear, this variation around the cumulative season-long average increase in [O3] of 23% could exaggerate the loss and explain the greater yield decrease seen here relative to open top chamber experiments. However, Ashmore (2002) shows that the decline in soybean yield across a range of open-top chamber studies is remarkably close to linear between 30 p.p.b. and 130 p.p.b. Since our treatment concentrations remained within this range variation in the daily increase should not have altered the overall response. This is further substantiated by the difference between 2002 and 2003. If the larger than average increases of up to 50% on some days caused damage disproportionate to the numerical increase, this effect would have been expected greatest in 2002 when average ambient [O3] was higher. In fact, the proportionate loss of yield was considerably higher in 2003 (Fig. 3). The fumigation approach used here was to increase [O3] as a proportion of the instantaneous background [O3]. This approach increases both the mean concentration and peak concentration. If a future elevated [O3] is achieved without increase in peak concentrations, then our treatment regime will have exaggerated the maxima. However, as noted above, even the highest peaks (Fig. 2) were within the range in which soybean yield appears to respond linearly to [O3], implying that yield losses should reflect the cumulative average increase regardless of the higher peak concentrations.

Another explanation for the larger than expected yield losses might be that the cultivar used in this experiment is relatively sensitive to ozone. However, compared with 22 different lines covering a broad range of germplasm scored for decreased biomass in the adjacent subplots (R. L. Nelson, unpublished), the Pioneer 93B15 cultivar used here was average in response. This cultivar is a recent selection, which, like most modern cultivars, will have been selected with an already high current [O3]. Nevertheless, like the other cultivars tested, it is clearly very vulnerable to increases in [O3]. Since the HI did not decline, it suggests that the effect of ozone on yield resulted largely from decreased assimilation rather than decreased reproductive potential. This ozone-induced loss in net assimilation would result from significant decreases in photosynthetic capacity (Morgan et al., 2004) and leaf area (Dermody et al., 2006).

Not only were the yield decreases seen in both years greater than previous chamber studies, but the biological basis appeared to differ. The meta-analysis of previous chamber studies suggested that decreased net photosynthesis was solely responsible for decreased production at moderate elevations of [O3], as used in this experiment, with leaf loss only becoming a contributory factor at higher [O3] levels (Morgan et al., 2003). However, toward the end of the growing season, significant decrease in leaf area index and individual leaf duration were found in 2002 (Dermody et al., 2006). Similarly, photosynthetic analyses of the crop in SoyFACE showed no difference in leaf photosynthesis over much of the growing season (Morgan et al., 2004). Decreases in leaf photosynthesis were only found as leaves neared senescence (Morgan et al., 2004), which was accelerated in elevated [O3] (Dermody et al., 2006). This increased leaf senescence was reflected the significantly more rapid leaf dry-mass losses (Fig. 4) and decreased leaf number (Table 2) during 2002. Accelerated senescence induced by elevated [O3] would limit canopy photosynthesis (Pell et al., 1997; Miller et al., 1998) and account for the larger decreases in ANPP and yield observed here, despite a smaller decrease in leaf photosynthesis than observed in chamber studies.

The defoliation resulting from the 2003 hailstorm provided a unique, albeit unplanned, test of the effect of elevated [O3] on the ability of a field crop to recover from an extreme climatic event. Parallel situations might include outbreaks of defoliating insects, high winds, as well as hail, all of which could increase with global climate change. Destruction of leaf area (60%; O. Dermody unpublished) and dry matter (21%) by the storm did not differ between treatments but growth and mass accumulation were suppressed in elevated [O3] following the storm (Fig. 4). Over the following month, average daily dry mass gain was 19% greater in current [O3] than in the elevated [O3] treatment. Soil cores taken from the same subplots used here to assess above-ground biomass showed no significant effects of [O3] treatment on root mass in either of these two years (Rodríguez, 2004). The faster recovery of the controls would be explained if reserves in shoots, roots and the remaining leaves (e.g. stored nonstructural carbohydrates) were greater in the controls, allowing faster leaf regrowth. Another possibility is that decreased translocation of carbohydrates (Grantz, 2003) increased foliar accumulation of nonstructural carbohydrates (Mulchi et al., 1992; Meyer et al., 1997), which would have been lost in the defoliation, thereby limiting regrowth in the plants in elevated [O3]. Regardless, limited carbohydrate availability causes increased flower and pod abortion (Bruening & Egli, 2000) suggesting that a decrease in the resource pool would limit regrowth and reproductive production in elevated [O3]. In the 2003 growing season, a decreased number of pods per plant was consistent with carbohydrate limitations to yield that were compounded in elevated [O3] (Table 2). Further evidence that carbohydrate was probably limiting in the hail-damaged crop is given by the significant reduction in average seed weight in both current and elevated [O3] in 2003 relative to 2002 (Fig. 3). These factors likely explain why yield loss caused by [O3] in 2003 was 10% more than in 2002.

Conclusions

Previous chamber studies suggested that the elevation of ozone levels projected over the next 50 yr would result in large losses of net primary production and crop yield, particularly in soybean in the USA. This first FACE treatment of any annual crop with elevated [O3] not only confirmed these losses but suggested the losses to be potentially even greater under fully open-air field treatment. It also showed that the losses may be further exacerbated when elevated [O3] is combined with an extreme event, as occurred in 2003. This confirmation has important implications for future global food supply, since large increases in tropospheric [O3] are projected for some of the major food-producing areas of the globe, including eastern China and the American Midwest (Prather et al., 2001; Prather et al., 2003). However, atmospheric [CO2] is also increasing and is expected reach 550 p.p.m. by the middle of this century owing to anthropogenic emissions (Prentice et al., 2001). Often predictions of future food supply under global atmospheric change have suggested that while increasing temperature and decreased soil moisture may lower yields, this will be more than compensated for by the direct effects of rising [CO2], particularly in the northern temperate zone (Izaurralde et al., 2003; Parry et al., 2004). These projections have not considered the direct effects of the simultaneous increase in [O3], which, on the basis of this first replicated open-air treatment of a food crop to elevated [O3], could offset any positive response to rising [CO2]. Together, these results suggest that if the IPCC projections of future tropospheric [O3] (Prather et al., 2001) are correct, then current projections of future food security (Izaurralde et al., 2003; Parry et al., 2004) are overestimates. This study reveals a clear need for more extensive examination of the impacts of rising tropospheric [O3] on food crops under fully open-air conditions.

Acknowledgements

We thank O. Dermody for helping design the sampling protocol, F. G. Dohleman, E. A. Heaton and V. E. Wittig for assistance in organizing the collection and processing harvested material and C. P. Chen, D. R. Ort, A. D. B. Leakey, E. A. Heaton and S. L. Naidu for comments on drafts of this manuscript. Funding was provided by the Illinois Council for Food and Agricultural Research (C-FAR), Archer Daniels Midland Company (ADM), Pioneer Hi-Bred, Argonne National Laboratory, and USDA-ARS.

Ancillary