Effect of environmental changes on vegetable and legume yields and nutritional quality

Significance

Environmental changes, including climate change, air pollution, and water scarcity and salinization, threaten global agricultural production, food security, and health. There is evidence that environmental change will reduce the yields of starchy staple crops, but impacts on (nonstaple) vegetables and legumes—important constituents of healthy diets—remain largely unknown. We systematically reviewed the available published evidence from experimental studies on the impact of environmental changes on yields and nutritional quality of (nonstaple) vegetables and legumes and found that environmental change would have a negative impact on yields without suitable responses from the agricultural sector. An enhanced understanding of the scale of environmental impacts on agricultural production is essential for the development of effective strategies to protect global population health.

Abstract

Environmental changes threaten agricultural production, food security, and health. Previous reviews suggest that environmental changes will substantially affect future yields of starchy dietary staples. To date, no comprehensive global analysis of the impacts of environmental change on (nonstaple) vegetables and legumes—important constituents of healthy diets—has been reported. We systematically searched for articles published between 1975 and 2016 on the effects of ambient temperature, tropospheric carbon dioxide (CO₂), and ozone (O₃) concentrations, water availability, and salinization on yields and nutritional quality of vegetables and legumes. We estimated mean effects of standardized environmental changes using observed exposure-response relationships and conducted meta-analyses where possible. We identified 174 relevant papers reporting 1,540 experiments. The mean (95% CI) reported yield changes for all vegetables and legumes combined were +22.0% (+11.6% to +32.5%) for a 250-ppm increase in CO₂ concentration, −8.9% (−15.6% to −2.2%) for a 25% increase in O₃ concentration,−34.7% (−44.6% to −24.9%) for a 50% reduction in water availability, and −2.3% (−3.7% to −0.9%) for a 25% increase in salinity. In papers with baseline temperatures >20 °C, a 4 °C increase in temperature reduced mean yields by −31.5% (−41.4% to −21.5%). Impacts of environmental changes on nutritional quality were mixed. In a business-as-usual scenario, predicted changes in environmental exposures would lead to reductions in yields of nonstaple vegetables and legumes. Where adaptation possibilities are limited, this may substantially change their global availability, affordability, and consumption in the mid to long term. Our results stress the importance of prioritizing agricultural developments, to minimize potential reductions in vegetable and legume yields and associated negative health effects.

Environmental changes, including climate change, land degradation, water scarcity, and biodiversity loss – that are predicted to become more profound in the 21st century – pose significant challenges to global agriculture, food security, and nutrition. The majority of previous research on environmental change and agriculture has focused on yields of staple crops, such as cereals. There is general consensus across projected climate scenarios that predicted future changes in temperature and rainfall will lead to significant reductions in the yields of many staple crops important for human populations, particularly in (sub)tropical areas (1). Some research has also explored the impact of changing environmental exposures on the nutrient content of staple crops (2–4).

In contrast, there has been comparatively little emphasis on the impact of environmental change on nutritionally important (nonstaple) vegetables and legumes, which appear to be relatively sensitive to environmental changes. For example, tomatoes and beans have lower failure point temperatures (the ambient temperature at which growth stops) than staple crops and are more vulnerable to heat stress (5). Furthermore, several vegetables and legumes are particularly vulnerable to develop visual injury (and hence marketability) due to environmental stress, notably small bleached spots due to high O₃ exposure (6), with legumes, leafy vegetables, and Solanaceae (including tomatoes) among the most sensitive crops (7). To date, there has been no overarching review of the global evidence of the impact of changing environmental exposures on the yields and nutritional quality of (nonstaple) vegetables and legumes.

Micronutrient deficiencies are a significant public health concern, affecting an estimated 2 billion people worldwide (8). Ensuring sufficient dietary intake of vegetables and fruit has been identified as critical in efforts to prevent and mitigate micronutrient deficiencies, as well as to tackle noncommunicable diseases (NCDs) such as cardiovascular disease (9, 10). According to the Global Burden of Disease Study, 1.5 million deaths per year globally are attributable to low vegetable consumption (11). Worldwide per capita consumption of vegetables and fruit is 20–50% below the minimum daily recommended level, although large regional differences exist (12). An understanding of the impact of potential changes in the availability of vegetables and legumes resulting from future changes in environmental exposures is important for both agricultural and public health policy planning. We present the results of a systematic review of the available published evidence on the impacts of changes in environmental exposures—in a standardized business-as-usual setting (i.e., no changes in agricultural practices, technologies, etc.)—on the yield and the nutritional quality of (nonstaple) vegetables and legumes. Our review focuses on experimental studies conducted in field and greenhouse settings and excludes desk-based modeling studies.

Results

Screening.

The initial database searches yielded 73,613 titles. After screening titles and abstracts and reading full texts, 237 papers (including one paper identified through consulting experts in the field and one paper identified by reference screening) were found to be relevant and were assessed for quality. Sixty-three papers (27%) did not meet the four quality criteria and were excluded from further analysis. A total of 174 papers (1,540 experiments) were included in the final analysis, of which 148 reported on yields, 49 reported on nutritional quality, and 23 reported on both (SI Appendix). Twenty-four papers (216 experiments) reported confidence limits and were available for inclusion in the meta-analysis (Fig. 1). Eighty-six papers reported on field studies, 89 papers reported on greenhouse studies, and one paper reported on both. Each paper included one or more experiments (comparison of yield and/or nutritional quality between baseline and exposed crops), covered one or more environmental exposures, and evaluated one or more crop types.

Fig. 1.

PRISMA chart showing the number of papers in each search stage. *Combined with a systematic review on fruits; the systematic review on fruits to be published elsewhere. **Two papers analyzed both fruits and vegetables/legumes.

Geographic Locations.

The experiments reported in the included papers were conducted in 40 different countries (Fig. 2), with the majority conducted in southern Europe, North America, and southern Asia. Of the 86 field studies, 25 were conducted in tropical countries, 36 in subtropical countries, 24 in temperate countries, and 1 in a boreal country.

Fig. 2.

Overview of field and greenhouse studies per country. Field studies are divided into those assessing the impact of environmental changes on nutritional quality (blue), yield changes (green), or both (yellow).

Impact of Single Environmental Exposures.

Ambient temperature.

We included 13 papers (30 experiments; 1 field study and 12 greenhouse studies) assessing the impact of ambient temperature change, all of which reported on yield changes (SI Appendix). The effect of a standardized 4 °C increase in temperature was mixed (mean yield change, −4.9%; 95% CI, −47.6% to 37.8%). There was clear heterogeneity of effect depending on baseline temperature (SI Appendix); experiments with a baseline temperature above 20 °C (n = 18) showed a mean yield change of −31.5% (95% CI, −41.4% to −21.5%), whereas experiments with a baseline temperature equal to or below 20 °C (n = 12) showed a mean yield change of +34.9% (95% CI, −47.9% to +117.6%). None of the included papers reported uncertainty estimates, and a meta-analysis could not be performed. None of the included papers reported the impact of increased ambient temperature on the nutritional quality of vegetables or legumes.

CO₂.

We included 44 papers reporting on the impact of changing atmospheric CO₂ levels (201 experiments; 14 field studies and 30 greenhouse studies). Yield changes (35 papers; 80 experiments) resulting from a standardized 250-ppm increase in CO₂ concentration were positive (mean yield change, +22.0%; 95% CI, +11.6% to +32.5%) (Fig. 3A). This finding was supported by a meta-analysis of the available data (7 papers; 18 experiments), which suggested an overall positive impact on yields (pooled effect size, +13.6%; 95% CI, +9.7% to +17.5%) that appears to be greater for legumes than for leafy vegetables (pooled effect size, +28.1%; 95% CI, +21.3% to +34.8% vs. +9.0%; 95% CI, −1.1% to +19.1%) (Fig. 3B). Heterogeneity across papers was characterized as mild for legumes and moderate for leafy vegetables, and the corresponding funnel plots suggested some publication bias (SI Appendix). Marginal yield increases per standardized increase of 250-ppm CO₂ exposure appeared to be substantially attenuated when the evaluated range of CO₂ concentrations (experimental–baseline) exceeded 400 ppm (SI Appendix).

Fig. 3.

Dot plot (A) and forest plot (B) showing the available experimental evidence of yield changes in vegetables and legumes resulting from a standardized increase of 250 ppm CO₂ concentration. More details of the forest plot are provided in SI Appendix.

Nine papers (102 experiments; 3 field studies and 6 greenhouse studies) reported the impacts of increased CO₂ on nutritional quality of vegetables (SI Appendix) and due to limited data, analysis was restricted to leafy vegetables. A standardized 250-ppm increase in CO₂ concentration had no overall impact on mean concentrations of nutritional quality parameters in leafy vegetables. These findings were supported by meta-analyses that could be performed for iron (three papers; seven experiments), vitamin C (three papers; five experiments), flavonoids (three papers; five experiments) and antioxidants (three papers; six experiments). The available evidence suggests that a standardized 250-ppm increase in CO₂ levels had no impact on iron, vitamin C, or flavonoid concentration (pooled effect, +17.0%; 95% CI, −18.3% to +52.2%; +3.2%; 95% CI, −12.6% to +19.1%; and +3.8%; 95% CI, −23.3% to +31.0%, respectively). However, a meta-analysis suggested an increase in antioxidant concentrations (pooled effect, +27.5%; 95% CI, +1.18% to +53.9%). Heterogeneity across papers was characterized as severe for all quality parameters, and the corresponding funnel plots suggested possible publication bias, especially related to the results for iron and antioxidants (SI Appendix).

O₃.

We identified 21 papers that reported on the impact of tropospheric ozone concentration (122 experiments; 15 field studies and 6 greenhouse studies). Yield changes (18 papers; 76 experiments) resulting from a standardized 25% increase in O₃ were negative (mean yield change, −8.9%; 95% CI, −15.6% to −2.2%) (Fig. 4A). This finding was supported by the meta-analysis of the available data reporting measures of uncertainty (3 papers; 15 experiments, legumes only), which suggested substantial yield decreases (pooled effect size, −18.5%; 95% CI, −22.8% to −14.2%). Heterogeneity across papers was severe (Fig. 4B), and the corresponding funnel plot suggested minor publication bias (SI Appendix).

Fig. 4.

Dot plot (A) and forest plot (B) showing the available experimental evidence of yield changes in vegetables and legumes resulting from a standardized 25% increase in O₃ concentration (plot B: legumes only). More details of the forest plot are provided in SI Appendix.

Scatterplots of the available evidence suggested that the (negative) incremental effect of the standardized increased O₃ concentration on yields was greatly reduced when the evaluated range (experimental–baseline) exceeded 25 ppb (SI Appendix).

Four papers (39 experiments; 4 field studies) reported the impacts of changes in O₃ concentration on the nutritional quality of leafy vegetables (37 experiments) and Solanaceae (2 experiments). The available evidence is relatively limited, and the overall effect of 25% increases in O₃ concentrations on leafy vegetables were mixed and varied largely by quality parameter (SI Appendix). The available evidence consistently suggests that higher O₃ concentrations increase vitamin C concentrations in leafy vegetables (2 papers; 13 experiments), but no pooled analysis could be performed.

Water availability.

We identified 65 papers (511 experiments; 41 field studies, 25 greenhouse studies, and 1 combined field and greenhouse study) that reported on the effect of reduced water availability. Yield changes (55 papers; 334 experiments) resulting from a 50% reduction in water availability were negative (mean yield change, −34.7%; 95% CI, −44.6% to −24.9%) (Fig. 5). None of the included papers reported uncertainty estimates.

Fig. 5.

Dot plot showing the available experimental evidence of yield changes in vegetables and legumes resulting from a standardized 50% reduction in water availability.

Fifteen papers (177 experiments; 8 field studies and 7 greenhouse studies) reported on the effect of water stress on nutritional quality. The overall effects were mixed and varied substantially by crop group; leafy vegetables appeared to be positively affected, while the effects on legumes were largely null (SI Appendix). The impacts on Solanaceae were mixed, with positive changes reported for vitamin C concentrations (8 papers; 18 experiments; mean concentration change, +37.6%; 95% CI, +11.7% to +63.5%), but no significant changes reported in mean concentrations of carotenoids and antioxidants (5 papers, 28 experiments: mean concentration change, +51.2%; 95% CI, −88.8% to +192.7% and 4 papers, 10 experiments: mean concentration change, +8.2; 95% CI, −38.0% to +54.4%, respectively). Meta-analysis evaluating the impact of water stress on vitamin C concentrations in Solanaceae supported the findings of the crude analysis (4 papers, 10 experiments: pooled effect, +28.5%; 95% CI, +15.3% to +41.7%) (SI Appendix). Heterogeneity across papers was severe, and the corresponding funnel plot suggested possible publication bias (SI Appendix).

Water salinity.

We identified 45 papers (465 experiments; 18 field studies, 27 greenhouse studies) on the effect of water salinity. Yield changes (36 papers; 200 experiments) resulting from a 25% increase in salinity of irrigation water were negative (mean yield change, −2.3%; 95% CI, −3.7% to −0.9%) (Fig. 6). None of the included papers reported uncertainty estimates.

Fig. 6.

Dot plot showing the available experimental evidence of yield changes in vegetables and legumes resulting from a standardized 25% increase in water salinity.

Thirteen papers (252 experiments; 8 field studies, 5 greenhouse studies) reported the impacts of increased water salinity on nutritional quality of leafy vegetables and Solanaceae. The overall effect was mixed with no dominant direction. In Solanaceae, carotenoid concentrations appeared to be predominantly positively affected by increased salinity (SI Appendix).

Combined Impact of Multiple Environmental Exposures.

Fifteen papers (50 experiments; 2 field studies, 13 greenhouse studies) assessed the combined impact of changes in environmental exposures on vegetable or legume yields. All papers evaluated the impact of elevated tropospheric CO₂ concentrations in combination with a change in another environmental exposure. There was little methodological standardization across papers, and analysis was limited to reporting the direction of impact on yield in the included papers (SI Appendix). Experiments that included combined environmental stressors (including 15 experiments on the combined impact of elevated CO₂ concentration and temperature) largely resulted in null or negative impacts on yields. Two papers (24 experiments) assessed the effect of elevated tropospheric CO₂ and O₃ concentrations on nutritional quality and reported significantly decreased concentrations of zinc, iron, calcium, and magnesium in root vegetables. Due to the limited number of papers, no pooled analysis could be performed.

Discussion

Results in Context.

Our systematic review synthesizes the available published evidence from experimental studies on the impact of critical changes in environmental exposures on yields and nutritional quality of legumes and nonstaple vegetables under a business-as-usual scenario. The available evidence base is relatively large but fragmented and heterogeneous; however, some consistent results were found. Our review suggests that, in the absence of adaptation strategies, increasing ambient temperature in (sub)tropical areas, tropospheric O₃, water salinity, and decreasing water availability would all negatively affect vegetable and legume yields. As has been previously demonstrated for other crop types, our review also identified that increasing CO₂ concentrations will have a positive impact on vegetable and legume yields, although these increases might be substantially attenuated in the presence of other environmental stressors (namely raised tropospheric O₃ and increased ambient temperatures) and may level off at CO₂ concentration increases above baseline of >400 ppm. The suggested reductions in positive yields impacts resulting from elevated CO₂ concentrations in the presence of other environmental exposures may be particularly important in future impact assessments, given that several of the evaluated environmental exposures—most notably increases in CO₂ concentrations and in ambient temperature—are likely to occur concomitantly in the future. A relatively limited evidence base further suggests that environmental changes also may affect the nutritional quality of vegetables and legumes, although findings are heterogeneous.

Previous reviews have identified that, in the absence of appropriate adaptation strategies, increasing ambient temperature will affect major staple crop yields, and that these impacts will differ by climatic zone, with yield declines in tropical zones and some yield increases in temperate zones (13–17). While we were not able to disaggregate our estimates by climatic zone due to data limitations, our findings for nonstaple vegetables and legumes similarly showed that when baseline temperatures were high (above 20 °C), increases in ambient temperature resulted in substantial declines in yields. These yield declines were not evident at low baseline temperatures where some yield increases were reported. Our work extends previous reviews by assessing the impact on yields of changes in multiple environmental exposures both individually and in combination. A previous review identified the presence of negative impacts of increased tropospheric CO₂ concentrations on the nutritional quality (i.e., zinc and iron concentrations) of staple crops (2). This phenomenon, also referred to as the “dilution effect” (18), is hypothesized to be related to reduced canopy transpiration or to changes in metabolite or enzyme concentrations whereby concentrations of micronutrients in the edible product decline (e.g., ref. 19). Furthermore, it has been associated with increased photosynthesis resulting in larger crops, but unaltered (and hence diluted) micronutrient content (e.g., ref. 2). Here we examined the available evidence of the effect of different environmental exposures on nutritional quality and found that the direction and scale of impact varied by environmental exposure and crop type.

Strengths and Limitations.

Our review has several strengths. We conducted a thorough and systematic search of the published literature in multiple languages using seven databases and screened papers for important markers of research quality. We included only experimental studies (not modeled analyses) and standardized the environmental impacts in our analysis. We presented the totality of available data in dot plots and calculated crude mean impacts to give an indication of the direction of effect and performed meta-analyses when possible. We identified studies conducted on five continents, but few included papers were conducted in Central and South America, Africa, and southeast Asia.

Our review has some limitations, related largely to the design, methods, and reporting of included papers and our standardization and pooling of results. Many of the included papers were primarily designed to investigate mechanisms to enhance the yields and quality of vegetables and legumes, or to explore exposure-resistant varieties; thus, changes in environmental exposure levels were not always within realistic ranges of environmental change. Differences in study objectives also limited the representativeness of vegetable and legume cultivars under investigation. For example, to explore salinization adaptation strategies, agricultural researchers often conducted research on salt-tolerant cultivars. Similarly, studies investigating the impact of reduced water availability mimicked water stress by applying a substantial but stable reduction in watering throughout all phenological stages of plant growth, yet the sensitivity of vegetables and legumes to reduced water availability varies by growth stage. Experimental crop variety selection may also have changed over the study period (from 1975 onward), which might have affected the yield response of crops to environmental exposures.

Due to the variety of study methods, evaluated ranges, crop types, and outcome measures, only linear relationships between environmental exposures and outcomes were analyzed. To explore critical potential nonlinear trends, threshold analyses were performed for ambient temperature and tropospheric CO₂ and O₃ concentrations; for temperature, this enabled a stratified analysis for experiments with lower and higher baseline temperatures. Ideally, further regional analysis would have been conducted to explore differences in impact on yield by climatic zone, but this was not possible due to data scarcity. The comparative analysis and pooling of results required standardization of environmental change exposure levels. We used the Intergovernmental Panel on Climate Change Fifth Assessment Report (IPCC-AR5) forecasts to guide this standardization, but some of the changes evaluated are large and likely to relate to longer-term impacts.

The possibilities for meta-analysis were relatively limited, since only a small percentage of papers (14%) reported precision estimates. Nonetheless, whenever possible, it was used to support crude analysis carried out on all studies. The representativeness of papers included in the meta-analyses is unclear, and the reduced study numbers restricted weighted analysis of the effect of each environmental exposure on vegetable and legume yields and nutritional quality. The funnel plots corresponding to the various meta-analyses conducted in this review show evidence of some publication bias (SI Appendix).

A number of papers, especially those published more than a decade ago, considered baseline levels of CO₂ below current atmospheric levels (400–410 ppm) (20). Finally, several papers could not be included due to reporting issues that limited possible data extraction.

Possible Health Effects and Wider Impacts.

The identified challenges for nonstaple vegetable and legume production should be considered within the wider context of global public health. As worldwide vegetable consumption levels are already below recommended guidelines (21), the potential health impacts of further reductions in vegetable and legume consumption might be substantial; low vegetable consumption could increase risk of several noncommunicable diseases, such as coronary heart disease and stroke, and the risk of different types of cancers (11, 22, 23). As the primary source of some essential nutrients, such as fiber, folate, and several vitamins, reduced vegetable and legume consumption could also lead to nutrient deficiencies that may be hard to overcome through substitution with other foods. Adequate consumption of nonstaple vegetables and legumes is a fundamental recommendation in all national and international food-based dietary guidelines (24). Therefore, ensuring sufficient availability of, and access to, vegetables and legumes represents an urgent global nutrition and public health challenge.

The ability of nations to respond to changing environmental conditions will be an important determinant of population health and economic impacts. Negative impacts are most likely to occur in poorly functioning markets and among poor rural and urban populations where environmental changes may both directly and indirectly affect the availability, affordability and consumption of vegetables and legumes. In addition to exposures evaluated in this review, yields could be affected by an increased frequency of extreme events—such as floods, cyclones and heat waves—that are more difficult to overcome by adaptation strategies.

Several additional challenges could occur. The increased visible bruising of vegetables caused by raised tropospheric O₃ concentration may reduce market value and lead to lost agricultural revenue (25). Heat stress could also affect producers directly and could cause reduced labor productivity (26), further compounding the effects of increasing temperature on crop yields. The identified impacts of environmental exposures may complicate proposed shifts toward more sustainable and healthy diets that are typically characterized by high consumption of vegetables and legumes.

Conclusions

Improved reporting of methodological details and study results from agricultural experiments is essential to tackle the gaps in the evidence base identified in this review. Ideally, a standardized list of environmental impacts, both single and in combination, would be used in these experiments, and this would enable much greater harmonization of data and comparison of findings in formal meta-analyses. Clear reporting of sample sizes, effect sizes, and uncertainty intervals are critical elements for comprehensive pooled analysis, and these were frequently missing for the current analysis.

Despite the inherent limitations of conducting systematic reviews in this field, our analysis identifies the potential for substantial impacts from environmental change on global nonstaple vegetable and legume yields. Our findings also demonstrate the value of connecting research in the environmental, food systems, and health sectors to identify previously unquantified challenges for agricultural production and food systems to deliver diverse and healthy diets for all in the future.

Methods

Literature Search Strategy.

This review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (27). We sought to identify all field and greenhouse studies published between January 1, 1975, and September 30, 2016 (including online ahead of print publications) that examined the relationship between a single or combination of environmental exposures and yields and/or nutritional quality of vegetables and legumes. (For this review on vegetables and legumes, we also included crops such as tomatoes, cucumbers, peppers, avocados, courgettes, pumpkins, and aubergines that are typically consumed as vegetables.) Our exposures were changes in the following five major environmental factors: ambient temperature; tropospheric carbon dioxide (CO₂) concentration; tropospheric ozone (O₃) concentration; water availability, and; water salinity. Our primary outcomes were change (baseline vs. exposure) in (i) vegetable or legume yield, defined as a percentage (%), and (ii) nutritional quality, defined as the concentration of nutritionally relevant substances in vegetables and legumes (28). We included all nutritionally relevant substances reported in included papers, namely fiber, flavonoids, ascorbic acid (vitamin C), carotenoids, phenolic compounds, antioxidants (including antioxidant activity), vitamin E, zinc, potassium, calcium, iron, magnesium, and manganese.

Seven databases were searched between October 17 and November 30, 2016: OvidSP MEDLINE, OvidSP Embase, EBSCO GreenFILE, Web of Science Core Collection, Scopus, Ovid SP CAB Abstracts, and OvidSP AGRIS. The search was conducted separately for yield and nutritional quality of vegetables and legumes using search terms for each environmental exposure and the 20 most common, nonstaple vegetables and legumes based on global food supplies estimated in Food and Agriculture Organization food balance sheets (12). Search strategies were paired with a second systematic review evaluating the impact of environmental change on fruit (to be published separately). The search strategy (SI Appendix) was first developed in OvidSP MEDLINE and then adjusted as necessary for other databases. The search strategy was complemented with examining reference lists of included papers and contacting subject experts (n = 4).

Selection Criteria and Data Extraction.

We included experimental studies conducted in greenhouse or field settings and excluded modeling studies. Papers were included if written in English, French, Spanish, German, or Dutch. Titles were screened for relevance by two reviewers (P.F.D.S. and F.A.B.). Relevant abstracts were assessed for inclusion by two reviewers (P.F.D.S. and F.A.B.), and any disagreements were resolved in discussion with a third reviewer (H.T.).

Data extraction was performed by a single reviewer (P.F.D.S., F.A.B., or H.T.), and a random sample of 10% was checked by a second reviewer (P.F.D.S., F.A.B., or H.T.). Extracted data included location, publication year, experiment year, study design (greenhouse or field study), air temperature (minimum, maximum, average), baseline and experimental levels of the environmental exposure under study, crop type and cultivar, yields at baseline and under experimental conditions, and nutritional quality parameters at baseline and under experimental conditions.

Study Quality.

Papers were assessed for quality using a modified checklist derived from the Critical Appraisal Skills Program (CASP) for randomized controlled trials (29) (SI Appendix). Criteria relating to randomization and blinding were removed from the checklist, as they were infrequently used in the assessed papers. Study quality was assessed by two reviewers (P.F.D.S. and F.A.B.) and included in the review if they met the following quality criteria: (i) clear description of study design; (ii) clear description of methods; (iii) appropriate comparison group; and (iv) rigorous and clearly described analysis. Only papers that reported precision estimates of measured effects (i.e., confidence intervals and/or SDs) were included in the meta-analysis.

Quantitative Data Synthesis.

Included papers were grouped by environmental exposure and further subdivided by vegetable or legume group: Solanaceae (including tomato, aubergine, and pepper), root vegetables, leafy vegetables, Cucurbitaceae (including cucumber and courgette), and legumes (Fabaceae). Due to the varied nature of ambient conditions under which experiments were conducted, greenhouse and field studies were combined in analysis. Sensitivity analysis identified that the direction and scale of study findings were similar in the two study designs.

Outcomes from individual experiments described in included papers were standardized to a fixed change in environmental exposure level (SI Appendix) guided by two factors: (i) the IPCC-AR5 forecasts for mid to late 21st century for each exposure (30, 31) and (ii) the range of exposures evaluated in the included papers (SI Appendix). For tropospheric O₃, salinity, and water availability, the standardized difference was defined as percent change from baseline, and for tropospheric CO₂ and ambient temperature, an absolute increase was used to accommodate papers that reported “ambient” as a baseline value without providing actual temperatures. For salinity, we evaluated papers that specifically assessed water salinity (not soil salinity) through flooding, saline ground water, or saline irrigation water. For experiments evaluating multiple environmental exposures, we included actual reported changes in our analysis.

The reported impacts of standardized changes in environmental exposures on vegetable and legume yields and nutritional quality from all included papers were displayed visually in dot plots, and crude summary impact estimates (“mean changes”) with 95% CIs were calculated. The Huber (sandwich) estimate of variance (32) was used to adjust for the clustered nature of the data using each paper as a cluster unit. Data from papers that provided estimates of precision (13.8% of all included papers) were used to calculate pooled effects using meta-analysis. The results of meta-analyses were used as a sensitivity mechanism to check and further quantify the crude summary data but were not used as stand-alone results due to the low percentage of papers that could be included. A minimum number of three papers was required for pooled analysis. We performed a random-effects meta-analysis to account for assumed between-study heterogeneity in true effects.

For each environmental exposure, initial analysis was performed combining all crop groups. Further exploratory analysis by crop group was conducted if a minimum of three papers were available for a specific crop group. Potential environmental “tipping points” were analyzed by visual examination of scatterplots in which evaluated ranges and baseline conditions were displayed against yield or nutritional quality effects of the standardized exposure. Three apparent tipping points were explored: ambient baseline temperature above 20 °C, tropospheric CO₂ concentration increases >400 ppm from baseline, and tropospheric O₃ concentration increases >25 ppb from baseline.

Risk of publication bias was assessed by visual inspection of funnel plots of the meta-analysis and by performing an Egger test (33). Heterogeneity across papers in each funnel plot was assessed with the I² statistic and labeled as mild, moderate, and severe in terms of heterogeneity (with cutoff values of <25%, ≥25%–≤50%, and >50%, respectively). Crude summary impact estimates were conducted for papers reporting the combined effect of multiple environmental exposures with the aim of examining the direction of interaction between multiple environmental exposures. Analyses were performed for all vegetables and legumes combined and for each crop group. Each nutritional indicator was analyzed separately for each crop group and environmental exposure.

All data and coding will be made available through the LSHTM data repository (LSHTM Data Compass).