Extensive plant damage due to tropospheric ozone was first observed during the Los Angeles smog episodes. In the early 1950s, reported that such plant damage could be reproduced in the laboratory by the reaction of organic trace gases or car exhaust with nitrogen oxides (NOx) in the presence of sunlight .

The influence of ozone on vegetation is dependent on the ozone dose and plant phenotype . Ozone enters leaves through plant stomata during normal gas exchange in the daylight hours and impairs plant metabolism, leading to yield reduction in agricultural crops .

In certain phenotypes, ozone exposure interferes with the hormone levels in plants and has been shown to lead to the accumulation of ethylene in the leaves. The presence of ethylene in the leaves interferes with the functioning of the hormone abscisic acid (ABA). ABA is a hormone which normally controls stomata closure and reduces water loss under drought conditions . Consequently, such plant phenotypes, when exposed to both drought and O3 stress, will continue to lose water despite the potential for dehydration. Ozone-related crop yield losses in such phenotypes may be enhanced in rain-fed regions, where kharif cops are frequently exposed to mid-season drought during the monsoon season. On the other hand, the yield of rice cultivars that show a healthy response to drought stress (i.e. close their stomatal aperture under water stress) could substantially benefit from the system of rice intensification (SRI) cultivation practice in areas with high ozone mixing ratios. Paddy fields under SRI cultivation are irrigated only when rice plots dry too much and the crop starts withering. A healthy response of rice plants to soil drying would reduce the ozone uptake. This could explain the higher yields frequently observed for SRI plots during field trials as well as the spatial variability in the yield difference between SRI plots and control treatments.

In phenotypes that are unable to control their stomata opening under ozone stress, O3 enters the leaf. It acts as a strong oxidant causing reactive oxygen stress (ROS) through hydrogen peroxide, superoxide, and hydroxyl radicals that alter the basic metabolic processes in plants . Ozone has been shown to destroy the structure and function of biological membranes leading to electrolyte leakage. This causes accelerated leaf senescence . Moreover, ozone can cause pollen sterility or induce flower, ovule, or grain injury and abortion . In such phenotypes ozone causes visible leaf injury, senescence, and abscission . By reducing the amount of healthy green leaf area available for photosynthesis, the accumulated damage eventually reduces crop yield, even if the exposure occurred at early vegetative stages of crop growth. Symptoms of ozone-associated leaf injury have been reported for 27 agricultural crops .

Certain other phenotypes respond to ozone stress by reducing their stomatal aperture . While this mechanism reduces the amount of ozone taken up by the plant and hence the oxidative stress inside the leaves, it also decreases CO2 uptake, leading to a reduction in photosynthesis. This affects the carbon transport to the roots, reduces nutrient and water uptake and, as a result of this, limits the storage of carbohydrates in the grains. Plants of this phenotype may show little to no visible leaf damage and often allocate significant resources to defences induced following ROS, but crop yields might be very sensitive to O3 stress during the grain filling stage. reported that for different wheat cultivars, the phenotypes with the least visible leaf damage were often the ones showing a maximum reduction in crop yield due to ozone.

The ozone-induced physiological damage such as lower yields and inferior crop quality lead to large economic losses .

Several large-scale programs targeted at assessing the impact of ozone on crop yields have resulted in a variety of different exposure metrics . The National Crop Loss Assessment Network (NCLAN) of the USA was the first systematic and large-scale study to assess the impact of O3 on crops in the world. It relied mainly on open-top field fumigation chambers (OTC) and used seasonal mean daytime exposure metrics (M7 and M12) to relate crop yield losses to ozone mixing ratios .

European researchers and policy makers focused on the critical-level concept as a tool to identify areas where the critical ozone levels are exceeded. The accumulated exposure over a threshold of 40 nmol mol-1 (AOT40) was adopted as a metric during a workshop in Kuopio, Finland, in 1996, and a set of critical-level values based on this index has been adopted for crops, forest trees, and semi-natural vegetation . AOT40 is the most widely used exposure plant response index. It is used by the United Nations Economic Commission for Europe (UNECE), the United States Environmental Protection Agency (USEPA), the World Meteorological Organization (WMO) and the World Health Organization (WHO) and is most frequently used in modelling studies targeted at assessing crop yield losses .

Recently stomatal-flux-based critical levels were proposed. These address concerns that the AOT40-based critical levels are based on the concentration of ozone in the atmosphere, whilst the ozone-related damage depends on the amount of the pollutant reaching the sites of damage within the leaf. Models using stomatal uptake of O3 (flux; F) or its cumulative value (dose; D) have significantly improved the prediction of plant injury. In particular, they have addressed the asynchronicity of maximum stomatal conductance (gsto) and peak ozone in plants that close their stomata when temperatures or the water vapour pressure deficit around the leaves are too high . The stomatal flux of ozone is modelled using a multiplicative algorithm adapted from . This algorithm incorporates the effects of air temperature, vapour pressure deficit of the air surrounding the leaves, light, soil water potential, plant phenology and ozone concentration on the maximum stomatal conductance, i.e. the stomatal conductance under optimal conditions. The exposure–yield relationships based on this algorithm consider the accumulated stomatal flux over a specified time interval as PODY (the phytotoxic ozone dose over a threshold flux of Y nmol O3 m-2 projected leaf area (PLA) s-1, with Y ranging from 0 to 9 nmol O3 m-2 PLA s-1; ). Studies evaluating the PODY-based exposure–yield relationship for a wide range of climate zones have emphasized the need for a local parametrization of the stomatal-flux model . To the best of our knowledge, no parametrization for southern Asian wheat and rice has been reported in the peer-reviewed literature. The wheat parametrization has been developed using European cultivars , and for rice the parametrization has been developed using only one Japanese rice cultivar, Koshihikari , which is known for its ozone resistance . Despite the fact that the stomatal-flux-based model is recommended by the UNECE CLRTAP (Convention on Long-range Transboundary Air Pollution) for ozone risk assessment in Europe , exposure–yield relationships have so far been internationally agreed upon only for a limited number of crops .

In the present study, we present new ozone exposure crop yield relationships for Indian rice, wheat and maize cultivars derived through a review of the peer-reviewed literature of open-top chamber studies on southern Asian cultivars.

We verify these new relationships using ozone monitoring data from the atmospheric chemistry facility in Mohali and yield data from a number of relay seeding experiments conducted in Punjab and Haryana. In these experiments crops were unintentionally exposed to different ozone levels by virtue of their sowing date being shifted, but the relevant studies were not conducted to investigate the effect of ozone on yields and consequently they did not include on-site ozone monitoring or clean-air control treatments.

We subsequently use a high-quality data set of in situ ozone measurements at a regionally representative suburban site called Mohali and the newly derived exposure–yield functions to assess ozone-related crop yield losses for wheat, rice, cotton and maize for Punjab and the neighbouring state Haryana for the years 2011–2013. Crop yield loss estimates calculated using two different exposure metrics, AOT40 and M7, are intercompared for a number of sowing dates and exposure–yield functions for the two major crop growing seasons of kharif (June–October) and rabi (November–April).

All ozone measurements were performed at the IISER (Indian Institute of Science Education and Research) Mohali atmospheric chemistry measurement facility (30.67∘ N, 76.73∘ E; 310 m a.s.l.; Fig. ). The measurement site is regionally representative and located in the north-west Indo-Gangetic Plain (NW IGP). Ozone measurements from several other sites located in the IGP and the adjoining mountain regions (Fig. ) will be discussed in detail in Sect.  to demonstrate that the measurements obtained at the facility are, indeed, regionally representative.

The measurement site is located inside a residential campus of around 1.25 km2 with 800–1000 residents. Local influence is expected to be significant only at low wind speeds (< 1 m s-1), which occur only rarely . The predominant daytime wind direction is west to north-west during winter, summer and in the post-monsoon season and south to south-east during the monsoon season. The “fetch” region of air masses arriving at the site is dominated by irrigated cropland (marked in light blue in Fig.  in the state of Punjab, north-west of the site). During the monsoon season, south-easterly winds bring air masses from a fetch region covering irrigated cropland in the state of Haryana, south-east of the site.

At the measurement site, inlets and meteorological measurements are co-located atop the ambient air quality station (AAQS) about 20 m above ground. A comprehensive description of the site and its representativeness for the north-west Indo-Gangetic Plain can be found in , and a thorough description of the meteorology of the site for all seasons can be found in .

Ozone was measured using UV absorption photometry at a time resolution of one measurement every minute, with an accuracy that is better than 3 % and an overall uncertainty of less than 6 %. Quality assurance of the large data set was accomplished by regular calibrations using a NIST traceable ozone primary standard generator and frequent zero drift calibrations. Over the time span reported in this paper, zero drift always remained below ±0.5 nmol mol-1 between two subsequent zero drift calibrations. The drift of the calibration factor during span calibrations was usually less than ±3 % and always below ±8 %, even after preventive maintenance. A detailed description of the ozone measurements and the supporting meteorological measurements can be found in .

We use two metrics to investigate the ozone exposure for crops in Punjab and Haryana and derive southern-Asia-specific exposure–yield relationships for wheat, maize and rice. These are the mean daytime surface ozone (M7) and accumulated exposure over a threshold of 40 nmol mol-1 (AOT40).

The Mx metric is defined as the mean daytime 7 (M7) and 12 h (M12) surface ozone concentrations during daylight hours, i.e. 09:00–15:59 and 08:00–19:59 LT respectively, in the crop growing season . M7=1n∑i=1nO3ifor09:00–15:59LT

AOT40 is defined as the sum of differences between the hourly ozone concentrations and 40 nmol mol-1 during the crop growing season for [O3] > 40 nmol mol-1. AOT40=∑i=1nO3i-40forO3>40nmolmol-1

Of these parameters M7 gives equal importance to all measurements and accounts for the yield losses due to ozone concentrations of less than 40 nmol mol-1, while AOT40 gives a higher weight to high ozone mixing ratios . Hence, the former will perform better while evaluating plant damage and yield losses at low ozone concentration, while the latter will capture the effect of events with very high O3 mixing ratios on plant physiology and yields better .

For any long-term data set, gaps in the data are inevitable due to preventive maintenance, calibrations and technical problems that arise from time to time. The total number and percentage of missing hourly average ambient data for each month from October 2011 to November 2013 are listed in Table . For calculating AOT40 and M7, continuous and complete daytime data are required, since any missing value can potentially lead to an underestimation of the real ozone exposure. Hence, missing values need to be filled in. For short data gaps of ≤ 3 h arising due to zero drift calibration or span calibrations we interpolated the measurements before and after the gap for filling in the missing values. Most gaps in the time series are due to calibrations. For longer data gaps we calculated the average diel ozone profile for the respective month and for each missing hour filled in the monthly average ozone value of the respective hour. In most months less than 5 % of the total hours were filled in. Only during the monsoon season does the requirement to occasionally purge the system with dry zero air lead to longer data gaps, and up to 21 % of the hourly averages had to be filled using the method described above.

Rabi (winter season) and kharif (summer monsoon) are the two main crop-growing seasons in northern India. In Punjab, kharif crops include rice, cotton, maize, sugarcane and vegetables . During rabi season wheat is grown in almost all of Punjab (> 90 % of the area).

In Haryana, kharif crops include rice, cotton and sugarcane and, in most of the unirrigated areas of Haryana, pearl millet and sorghum . Major rabi crops in Haryana include wheat, gram, sugarcane and mustard .

The most popular crop rotation systems in Punjab include rice–wheat (> 70 %) and cotton–wheat (∼ 20 %) as well as maize–wheat crop rotation systems. In Haryana rice–wheat (∼ 40 %) and cotton–wheat (∼ 20 %) rotation is popular in the north, but in the dryer parts of Haryana, pearl-millet–mustard and pearl-millet–wheat rotations are preferred . Maize is currently not very popular but heavily promoted as an alternative to rice when a deficient monsoon is anticipated.

The present study investigates crop yield losses for wheat and maize (rabi) and rice, maize and cotton (kharif). In Supplement S1, we discuss the growth stages during which these crops are potentially sensitive to ozone-related yield losses, as well as the time periods during which the plants reach those growth stages in the northern Indo-Gangetic Plain. To summarize briefly, different rice cultivars take between 90 to 140 days to reach harvest maturity after the ∼ 20–30-day-old seedlings have been transplanted into the fields. In this study we calculate the accumulated and average ozone exposure (AOT40/M7) for a 4-month period (120 days), which is typical of cultivars popular in the NW-IGP. We investigate the following five periods:

Period 1: 16 May (emergence) to 15 September (maturity);

We derive specific exposure–yield relationships for Indian wheat and rice cultivars using a two-pronged approach.

Firstly, we use our ozone measurements conducted at a suburban site in Punjab and a number of field studies conducted in the region that reported variations in the sowing date of crops , which lead to an unintentional change in ozone exposure, and one study that reported co-located yield and ozone measurements to derive an empirical exposure–yield relationship for rice and wheat. The empirical field data support the need to revise the exposure–yield relationship for Indian cultivars and demonstrate that for rice optimizing, the sowing date can be a suitable strategy to minimize ozone exposure and maximize crop yields.

Secondly, we derive India-specific exposure–yield relationships by plotting relative yields (RY) and ozone exposure for all OTC studies on Indian cultivars reported in the peer-reviewed literature and fitting the data to obtain an exposure–yield relationship . For maize, only one OTC study on two Indian cultivars has been conducted, and we use the fit of these data to obtain an exposure–yield relationship . We compare these exposure–yield relationships for rice and wheat with RY observed for cultivars commonly grown in Pakistan and Bangladesh to investigate to what extent the results can be extrapolated to all of southern Asia. We refrain from including cultivars popular in south-east Asia in our study, as they have been reported to show a very different sensitivity to ozone exposure . We provide an upper and lower limit for RY and crop yield losses for a set of five different sowing dates for rice and wheat, of three for cotton and of two for rabi and kharif maize, using both exposure-dose–response relationships established in several studies in the west (Table ) to provide a lower limit and our new India-specific functions to provide an upper limit to the possible loss.

We use both the old AOT40-based exposure–yield function and our revised AOT40-based relationship to calculate crop production losses and economic cost losses and contrast the two.

Table  summarizes the ozone exposure-dose–response relationships for relative yield loss (RYL) for wheat, rice, maize and cotton based on AOT40 and M7 values collected from the peer-reviewed literature.

All the ozone exposure-dose–response relationships previously reported in the literature are based on field studies conducted in the USA or in Europe. Relative yield loss is defined as the crop yield reduction from the theoretical yield that would have resulted without O3-induced damages , calculated using Eqs. (3) and (4): RYLi=1-RYi,CPLi=RYLi1-RYLi×CPi, where RYi stands for relative yield in the year i, CPLi stands for crop production loss in the year i and CPi stands for the crop production of the same year. The crop production per fiscal year was taken from the database of the .

Economic cost loss (ECL) for any crop is defined as the financial loss due to O3-induced damage in a given financial year. The minimum ECL is calculated for different crops based on corresponding minimum support prices (MSPs) of the same fiscal year using the following equation: ECLi=CPLi×MSPi. The MSPs are recommended by Commission for Agriculture Costs and Prices and are announced by the Government of India at the beginning of each season for each year. These prices are defined as the fixed price at which government purchases crops from the farmers. All our crops of interest come under the MSP valuation process. It should be noted, however, that the MSP is typically approximately 50 % less than the market value of the crop and often lower than the production costs. The upper limit for the ECL is calculated using the relationship between CPL due to deficient monsoon rains and the Indian GDP established by using the following equation: ECLi[%GDP]=RYLi[%]×0.36.

Figure  shows the seasonal box-and-whisker plot of the daytime (08:00–19:59 LT) 1 h average ozone mixing ratios for the period from October 2011 to January 2014. The highest ozone levels are observed in the summer season in April, May and June, with median ozone mixing ratios of 60–80 nmol mol-1 and peak ozone mixing ratios of approximately 130 nmol mol-1. This is expected, as conditions such as high temperature, low humidity and high solar radiation favour the photochemical production of O3 regionally.

After summer, the next highest ozone levels are observed during the post-monsoon season (October and November), with median ozone mixing ratios of 50–60 nmol mol-1. The post-monsoon season is characterized by lower levels of solar radiation (range of daytime maxima ∼ 480–720 W m-2) compared to the summer season (range of daytime maxima ∼ 600–920 W m-2), but the occurrence of large-scale agricultural burning emissions of ozone precursors and a lower boundary layer still result in comparably high ozone levels.

The lowest median daytime ozone mixing ratios of approximately 30 nmol mol-1 are observed in August, during the peak monsoon season, when cloudiness and wet scavenging of ozone precursors limits the photochemical ozone production, and during peak winter (December and January). During winter, a reduction in the solar radiation, low temperatures and fog result in less photochemical production of O3.

Table  shows the monthly increment in AOT40 and the monthly M7 for the period October 2011 to January 2014. The yearly maximum and minimum monthly values for all indices correspond to the same months, May and August, in both years. All indices show maxima during summer (May and June) and post-monsoon (October and November) and minima during the monsoon (July to September) and winter (December to February); however, the difference between the cumulative metric (AOT40), which gives higher weight to high values and low or no weight to low values, and the average-based metric (M7) comes out very clearly. For AOT40 the amplitude between peaks (∼ 14 000 nmol mol-1 h) and minima (∼ 500 nmol mol-1 h) is very high. The annual peak values are 30 times higher for AOT40 compared to the annual minima. For M7 peaks are only 2–3 times higher compared to the minima.

Few studies have so far reported ozone exposure indices over the IGP; however, a number of studies have reported average diel profiles for each month of the year or a time series of average daytime ozone for their site .

Table  shows the M7 or average daytime ozone calculated from the data in those studies. The seasonality and monthly average daytime ozone levels are similar for all urban and suburban sites in the IGP and the adjoining mountain valleys. However, sites located further to the east report lower M7 values during May and June, due to the higher frequency of summer rain, lower temperatures and the earlier onset of the monsoon in the eastern part of the IGP. The only site further to the west for which ozone measurements have been reported is located close to the centre of the summertime “heat low” over the NW IGP; it reports summertime and monsoon season M7 that are higher than those observed at our site and also a strong anticorrelation of the observed ozone during monsoon season with the intensity of the monsoon rainfall.

Given the fact that the most reliable crop-yield–exposure indices are based on AOT40 and not M7 values, there is urgent need to relate the available observations to AOT40 values. did so using a linear relationship. When applied to our data presented in Table , the relationship estimates reasonable AOT40 values (slope AOT40 predicted vs. AOT40 observed: 0.93; R2 = 0.87) but performs poorly, while reproducing peak AOT40 values. We find that at our site the actual data follow an exponential curve, AOT40=0.0201×M73.0765R2=0.94, and AOT40 values predicted using this curve match peak AOT40 observations better (slope AOT40 predicted vs. AOT40 observed: 1.03; R2 = 0.97).

Several studies attempted to model ozone levels and exposure metrics over the IGP. modelled AOT40 over the Indian region for the year 2003 using the model REMO-CTM (REgional MOdel chemistry transport model). For the north-western part of the IGP, close to the foothills, REMO-CTM models 5000–6000 nmol mol-1 h in May, 1500–2000 nmol mol-1 h in July and 6000–7000 nmol mol-1 h in October. We find that the model underestimates the observed AOT40 in the north-west IGP by a factor of 2 to 3 during May and July and reproduces the observations well during October. Consequently, the model would be able to predict crop production losses during rabi season better and would underestimate crop production losses during zayad and kharif seasons.

In a more recent study conducted using WRF-Chem (weather research and forecasting chemistry model), predicted ozone daytime concentrations of ∼ 50 nmol mol-1 for kharif season and ∼ 40 nmol mol-1 for rabi season for the Chandigarh UT. However, the authors considered only the time windows of 15 June to 15 September and of December to February for kharif and rabi seasons respectively. For these time windows, predicted ozone daytime concentrations agree well with the measured M12.

intercompared model-predicted ozone with surface observation for the HANK model. The model could not resolve the daytime ozone peak in Delhi and, hence, will perform poorly in predicting AOT40. Comparing the reported values for Chandigarh with our measurements, we find that the model has equal difficulty in resolving the seasonality, in particular the high ozone levels in summer.

compared MATCH (Model of Atmospheric Transport and Chemistry)-modelled M7 values with measured surface ozone for Varanarsi and Lahore and found good agreement between model and observations for both cropping seasons. For our site, too, there is excellent agreement between modelled and observed M7 values (model: 40–50 nmol mol-1 for rabi season and 50–70 nmol mol-1 for kharif season; observations: 40–52 nmol mol-1 for rabi season and 47–64 nmol mol-1 for kharif season).

used a global model (TM5 – TM stands for transport model) to predict surface ozone over India, and the model reproduces surface observations for our site equally well.

Crop yield losses and associated economic losses due to ozone are well constrained for the USA and Europe (Avnery et al., 2011a). The analyses of crop production losses made so far for India are based on model-derived O3 mixing ratios and apply O3-dose–plant-response metrics and formulae developed in the US or in Europe . Such predictions may underestimate crop yield losses. It has already been pointed out above that for some models, the model predicted daytime O3 mixing ratios or AOT40 values tend to be lower than the observed O3 mixing ratios or AOT40 in particular for zayad and kharif seasons. Hence, model predictions need to be validated and improved using in situ ozone measurements.

The O3-dose–plant-response metrics used in the modelling studies conducted so far also underestimate crop production losses due to the fact that southern Asian wheat and rice cultivars are more sensitive to ozone . reviewed a large number of Asian OTC and plant chamber studies but refrained from deriving Asia-specific dose response curves for wheat and rice due to the large spread in the observational data. suggested that the spread could be due to the large variety of different cultivars studied or due to the diversity of experimental conditions. In the same year compared 20 different rice cultivars under identical conditions in a plant chamber and showed that most Oryza sativa L. Japonica cultivars were resistant to ozone damage (11 out of 12), while most Oryza sativa L. Indica cultivars showed significant yield losses (5 out of 8). A follow-up metabolomic analysis of selected cultivars by the same authors, , showed that the only japonica cultivar with high yield losses, Kirara 397, down-regulated proteins associated with photosynthetic electron transport as a response to ROS induced by ozone. One of the indica cultivars with high yield losses, Takanari, showed no noteworthy changes in the metabolic pathway of photosynthesis resulting from ozone exposure, but its yields were equally sensitive to ozone, and most down-regulated proteins were associated with protein destination and storage and unknown functions. In one of the japonica cultivars (Koshihikari), which did not suffer yield losses, ozone stress up-regulated the expression of certain proteins in the Calvin cycle of the energy metabolism. reported the expression of the RuBisCO, and several energy-metabolism-related proteins were adversely affected by ozone exposure in the two indica cultivars Malviyadhan 36 and Shivani. These results seem to indicate that the responses to ozone are indeed cultivar-specific. More studies are required to understand the damage mechanisms in different cultivars at a fundamental level and identify high-yielding cultivars that are resistant to ozone stress, which can be promoted by the relevant government agencies in affected areas.

Table  summarizes the relative yield loss calculated according to different exposure indices. In general, crop production losses calculated using the M7 index exposure–response relationships based on American studies conducted in the 1970s and 1980s tend to underestimate the actual yield losses of Indian cultivars, as the M7 index fails to capture the effect of extreme events on plant physiology and yields . The old AOT40 exposure–response relationship by does not capture the sensitivity of most southern Asian cultivars. Only Bangladeshi rice cultivars and a few select wheat cultivars follow this relationship, while most Indian wheat and rice cultivars are far more sensitive to elevated ozone levels. We propose a revised relationship (Table , Figs.  and ) based on a literature review of OTC studies conducted on Indian cultivars and demonstrate that this relationship adequately describes the empirical relationship between crop yield and AOT40 in field trials that were not aimed at studying the effect of ozone on crops. The revised equation (Table ) predicts that RYL for Indian cultivars are 1.5–2 times the RYL predicted based on the equation by .

A recent modelling study for the year 2005 predicted RYLs of 1 and 1.2 % for Punjab and Haryana respectively for wheat and 8.1 % for Punjab for rice . These relative yield losses are a factor of 15–30 lower compared to the RYL calculated using the same equation but employing in situ measurements for calculating AOT40 for wheat and a factor of 1.5 to 1.8 lower for rice (Table  Column RYLAOT40, ).

estimated the crop production loss of winter wheat based on a review of measured ozone mixing ratios published in the peer-reviewed literature for the years 2000–2007. The calculated relative yield losses, based both on the M7 exposure–response relationship for winter wheat proposed by of 10.8 % and on the AOT40-based exposure–response relationship by of 29.8 % RYL for Punjab and Haryana, agree well with crop yield losses calculated by applying the same equations to our in situ observations (Table ) for the years 2011–2014 (Table  Column RYLAOT40, ). This indicates that the underestimation of RYL by is due to an underestimation of the AOT40 values during the wheat growing season in the north-west IGP caused by the fact that Ghude and colleagues only considered December to February as the ozone-sensitive growth periods and excluded the months of March and April, which show the highest AOT40 values in the growing season of wheat. However, in the NW-IGP the grain filling stage of the crop is only reached in March, and wheat has been shown to be extremely sensitive to high ozone during the grain filling stage . used the MOZART-2 (Model for OZone and Related chemical Tracers, version 2) to predict a national average RYL of 25–30 % for wheat using the AOT40-based equation, which agrees well with our observations. , using the TM5 model, predicted RYL ranging from 20–30 % for wheat, 10–15 % for rice and 1–3 % for maize for the year 2000, which agrees well with the observations.

Table  shows the crop production loss and MSP for the fiscal years of 2012–2013 and 2013–2014. Data on crop production were obtained from the following sources: the and . Procurement data were obtained from the . For the fiscal year of 2013–2014, data for Punjab are based on estimates, while final data for Haryana were obtained from the . The table also presents economic cost losses calculated for wheat, rice, maize and cotton using the old and revised exposure–yield relationship. The losses are present for Haryana and Punjab, both separately and cumulatively.

The highest crop production loss is seen for wheat: 20.8 ± 10.4 million t in the fiscal year of 2012–2013 and 10.3 ± 4.7 million t in the fiscal year of 2013–2014 for Punjab and Haryana taken together. predicted crop production losses of only 0.25 million t for the year 2005 for both states. The discrepancy is mostly due to the fact that this study assumed that the ozone-sensitive growth period of wheat lasts only from December to February, and, hence, this study did not capture the effect of the high AOT40 during the grain filling stage of the crop in March (factor ∼ 15–30). Thus, the discrepancy is also partially due to the revision of the exposure–response relationship (Table ; factor ∼ 2). estimated crop production losses of 10.9 million t yr-1 on average for both states combined. The estimate falls within the same order of magnitude as our estimate. estimated a CPL of 26 million t for all of India but did not resolve losses for individual states. Economic cost losses amount to INR 244 ± 121 billion and INR 133 ± 60 billion in the fiscal years of 2012–2013 and 2013–2014 respectively. At an exchange rate of 60 INR/USD, this amounts to USD 4.1 ± 2.0 and 2.2 ± 1.0 billion respectively.

Rice shows crop production losses of 5.4 ± 1.2 million t in the fiscal year of 2012–2013 and 3.2 ± 0.8 million t in the fiscal year of 2013–2014 for Punjab and Haryana taken together. predicted crop production losses of only 0.85 million t for the year 2005 for both states. The discrepancy is caused both by an underestimation of the AOT40 due to the fact that the author considered a shorter ozone-sensitive growth period (factor 1.5–1.8) and by the revision of the exposure–yield relationship (Table ) to account for the sensitivity of Indian rice cultivars (factor 1.9). Economic losses amount to INR 67 ± 15 billion and INR 42 ± 11 billion for the fiscal years of 2012–2013 and 2013–2014 respectively. At an exchange rate of 60 INR/USD, this amounts to USD 1.1 ± 0.2 and 0.7 ± 0.2 billion respectively.

The Indian National Food Security Ordinance entitles ∼ 820 million of India's poor to purchase about 60 kg of rice or wheat per person annually at subsidized rates. The scheme requires 27.6 Mt of wheat and 33.6 Mt of rice per year. Cutting down ozone-related crop production losses in Punjab and Haryana alone could provide > 50 % of the wheat and 10 % of the rice required for the scheme.

Economic losses amount to INR 79.15 billion and 47.50 billion (USD 1.3 and 0.8 billion) for cotton and INR 0.18 billion and INR 0.24 billion (USD 3 and 4 million) for maize in the fiscal years of 2012–2013 and 2013–2014 respectively.

The total economic losses for the agricultural sector in Punjab and Haryana amount to INR 391 ± 136 billion (USD 6.5 ± 2.2 billion) in the fiscal year of 2012–2013 and INR 223 ± 71 billion (USD 3.7 ± 1.2 billion) in the fiscal year of 2013–2014. The loss estimates presented above underestimate the real economic losses due to ozone on several accounts.

Firstly, the crop is valued only at the MSP for common grade crops. The MSP is often even lower than the actual production cost and the economic value of the crop is typically much higher. This is particularly true for high-quality rice varieties such as basmati.

Secondly, we do not account for the losses in the food processing sector and other allied industries. The value gain from MSP to the final end consumer product ranges from a factor of 2 to 20 for food crops to a factor of > 100 for cotton.

Thirdly, this calculation does not consider the relationship between the rural demand for consumer products and rural income. Rural income is affected strongly by crop yields, 78 % of the rural population depends on agriculture as primary source of income.

Previous studies investigating the relationship between monsoon rainfall, food grain production and the nation's GDP for the years 1951–2003 found that a 1 % decrease in food grain production due to a deficient monsoon led to a 0.36 % decrease in India's GDP. Ozone-related crop production losses are likely subject to the same multiplication factor. With relative yields losses currently ranging from 10 to 58 % for the different crops (, van Dingenen et al., 2009), the real economic burden of current ozone levels in terms of India's GDP is likely to fall into the range from 3.6 to 20 % (Eq. 8).