Weather extremes like droughts and heat waves are becoming increasingly frequent worldwide, with severe consequences for agricultural production and food security. Although the effects of such events on the production of major crops is well-documented, the response of a larger pool of crops is unknown and the potential of crop diversity to buffer agricultural outputs against weather extremes remains untested. Here, we evaluate whether increasing the diversity of crop portfolios at the country level confers greater resistance to a country's overall yield and revenues against losses to droughts and high temperatures. To do this, we use 58 years of annual data on weather, crop yields and agricultural revenues for 109 crops in 127 countries. We use the spatial distribution of each crop and their cropping cycle to determine their exposure to weather events. We find that growing greater crop diversity within countries reduces the negative impacts of droughts and high temperatures on agricultural outputs. For drought, our results suggest that the effect is explained not only by crop diversity itself, but also by the sensitivity of the most abundant crops (in terms of harvested areas) to this extreme. Countries dedicating more land to minor, drought-tolerant crops reduce the average sensitivity of country-scale crop portfolios and show greater resistance of yield and revenues to drought. Our study highlights the unexploited potential for putting crop biodiversity to work for greater resilience to weather, specifically in poorer developing countries that are likely to suffer disproportionately from climate change impacts.
1.Introduction
Weather extremes, including heat waves and droughts, are major determinants of food production shocks at the global scale (Ray et al 2015, Lesk et al 2016, Anderson et al 2019, Cottrell et al 2019). Effects of these shocks are particularly severe in developing countries where a high proportion of people rely on local, mainly rain-fed agriculture for consumption and livelihoods, where political unrest and chronic poverty reign, and where financial insurance mechanisms are inexistent or inaccessible (Morton 2007, Maxwell and Fitzpatrick 2012, Harvey et al 2014, Carleton 2017). More intense and frequent extreme weather events are expected in the future (Pörtner et al 2022), fueling concerns about the vulnerability of food systems (Thiault et al 2019) and humanity's ability to combat hunger (Sustainable Development Goals 2, UN General Assembly 2015). In this context, international organizations are calling for bold actions to evaluate risk-adverse solutions that can help maintain food security under conditions of more unpredictable weather (FAO, IFAD, UNICEF 2021).
Agricultural adaptation strategies to cope with climate change have so far mainly focused on technological fixes. But on the ground, greater use of fertilizers and modern varieties have not shown clear success in reducing the sensitivity of crops to climate extremes (Schlenker and Lobell 2010, Zaveri and Lobell 2019). Moreover, increased reliance on fertilizers and irrigation has generated numerous negative consequences (Foley et al 2011, Poore and Nemecek 2018). These include greenhouse gas emissions (Crippa et al 2021), groundwater depletion and pollution (Dubois 2011), that compromise efforts towards climate change mitigation and sustainability (Sayre and Taraz 2019, Blakeslee et al 2020, Ryan and Sudarshan 2022). Finally, these solutions tend to be expensive and their adoption may be slow in the developing world, where farmers are poor and limited liquidity constrains agricultural decisions (Tittonell and Giller 2013, Karlan et al 2014).
Attention to technological solutions may mirror a tendency of research to focus on a few number of globally dominant crops, i.e. maize, wheat, rice and soybeans (Lobell and Field 2007, Schlenker and Lobell 2010, Lobell et al 2011, Osborne and Wheeler 2013, Iizumi et al 2014, Ray et al 2015, Iizumi and Ramankutty 2016, Lesk et al 2016, Anderson et al 2019, Li et al 2022), whose management facilitates the implementation of technology. These crop-specific approaches have led to a better evaluation of the fine-scale spatial variability of the effects of temperature and precipitation on the yield of major crops. However, this approach may hide the way a wider diversity of crops influences responses to weather variability. It remains unknown whether accounting for the diversity of crops and their responses changes the relative balance of the effects of temperature and rainfall variability on the aggregated response of a crop portfolio.
Plant diversity is considered to be a promising 'nature-based' solution to meet multiple ends in agriculture (Beillouin et al 2019, Dainese et al 2019), including the reduction of climate-driven risks of crop failure (Renard and Tilman 2019). Decades of ecological research have shown that increasing plant taxonomic, genetic or functional diversity in natural ecosystems confers greater ecological stability (Donohue et al 2016) through increased resistance to climate stresses (Isbell et al 2015) and lower year-to-year variability of biomass production (Tilman and Downing 1994, Gross et al 2014, García-palacios et al 2018, Schnabel et al 2021). Comparatively, research on agricultural systems critically lags behind (Beillouin et al 2019, 2021). The few existing studies do support the notion that increasing crop diversity leads to lower year-to-year variability in both yield (Smith et al 2008, Gaudin et al 2015, Prieto et al 2015, Raseduzzaman and Jensen 2017, Renard and Tilman 2019) and incomes (Falco and Chavas 2008, Chavas and di Falco 2012, Auffhammer and Carleton 2018, Noack et al 2019) from fields to countries. However, the potential of crop diversity to buffer crop production and revenues against extreme weather events, and thus increase resistance, needs to be evaluated.
Hypotheses about the mechanisms governing the diversity–stability relationships involve mainly two processes (Thibaut and Connolly 2013, Xu et al 2021). The first one is based on a selection effect: the more species there are in an ecosystem, the more likely it is that at least one species is less sensitive to a given climate condition. The second process is based on a complementarity effect. Complementarity arises from species differences in their ecological niche (Barry et al 2019), a principle that underpins the concept of the insurance effect of biodiversity (Tilman et al 1998, Ives and Carpenter 2007, Loreau and Mazancourt 2008, Gross et al 2014). Accordingly, in a diversified system, one crop may fail, but multiple failures are less likely. Because different crop genotypes, varieties or species exhibit different responses to temperature and precipitation (i.e. contrasting ecological niches), fluctuations in productivity among plants can compensate each other and lead to greater resistance in a given year or greater stability through time. At the country level, Mahaut et al (2021) showed that fluctuations in the yield of crops that cover large proportions of harvested area (i.e. selection effect), and the asynchronic dynamics that occur between crops (i.e. insurance effect), both modulate climate-induced temporal variability of total yield of countries. However, only a few studies so far tested for processes underlying the crop diversity–stability relationship at large scale.
Here, our aim is to fill important gaps in the research linking biodiversity and stability in agricultural systems. We test whether crop diversity within countries increases the resistance of total caloric yield and economic revenues (all crops included) to weather extremes using 58 years (1961–2018) of annual data. We focused our study on droughts and high temperatures, two types of extremes that are recognized as major threats to food supply (Pörtner et al 2022). Agricultural data were matched with weather data based on the specific spatial extent of individual crops. We further test whether the selection and/or the complementarity (through the insurance) effects can explain the increased resistance to weather extremes conferred by greater crop diversity at the country level.
2.Material and methods
2.1.Agricultural data
The FAOSTAT database provided data on annual cropland and harvested area (in ha), production (in tonnes) and agricultural revenues (in 2015 equivalent US dollars) for an initial set of 246 countries and 176 crop commodities. From the initial list of crops, we discarded aggregated crop categories (e.g. citrus or other cereals) and crops grown in fewer than ten countries. Because agricultural yields and revenues, and their response to weather extremes, are likely to be influenced by conditions linked to agricultural management, we used annual total application of nitrogen and total cropland equipped with irrigation at the country level from the FAOSTAT database as covariates. Quality of FAOSTAT data varies among countries, depending on whether they have been estimated by the Food and Agriculture Organization of the United Nations (FAO) or communicated directly by countries. We accounted for this difference by removing 33 countries for which at least 20% of data on area harvested or production over the entire time period had to be estimated (following Renard and Tilman 2019). We also discarded Ireland, New Zealand and Netherlands, all of which use much of their fertilizer on pastures rather than for croplands, and Egypt, which has 100% of its cropland equipped with irrigation for the entire time period. For each of the remaining countries (n = 127) and for each year, we calculated the total annual caloric yield (millions of kcal ha−1) by (1) multiplying the production of each of its crops by commodity-specific kcal conversion factors from the United States Department of Agriculture's Nutrient Database (United States Department of Agriculture 2013); (2) summing these kcal harvests across all crops; and (3) dividing this sum by the sum of all harvested land for all these crops. Similarly, for each country and each year, we calculated the total annual agricultural revenues per hectare by dividing the total revenues by the total harvested area (in constant 2015 USD $/ha). We quantified crop diversity using the Shannon–Wiener diversity index (H') based on the proportion of total area harvested represented by each crop grown annually by each country studied. We calculated the annual rate of application of nitrogen and irrigation per hectare by dividing their use by the total cropland area in each country and each year. Sources of all the data are summarized in supplementary table 1.
2.2.Weather data
We obtained global 0.5° gridded temperature (in degrees Celsius) data from the Climate Research Unit of the University of Anglia (CRU TS v.3.23, Harris et al 2014) for the 1961–2018 period. We used these data to identify years with high growing season temperatures (we provide details of the calculation below). In years with high growing season temperatures, crops are expected to experience heat stress that affect their metabolic processes and biomass production. Droughts are mostly related to a reduction in the amount of precipitation, but they are also influenced by other variables including temperature and evapotranspiration (Vicente-Serrano et al 2010a). To quantify wet and dry annual weather events, we used the global 0.5° gridded Standardized Precipitation-Evapotranspiration Index (SPEI) from Vicente-Serrano et al (2010b), an index that is based on both precipitation and potential evapotranspiration (PET) data. The SPEI12 index was calculated for a period of one year using the monthly climatic water balance (i.e. precipitation minus PET) from the CRU database. SPEI index values are centered around zero to allow comparison of drought severity through time and space (Vicente-Serrano et al 2010a). In order to simplify interpretation, we used the negative of the SPEI index so that water deficits, hereafter referred to as droughts, have positive values.
2.3.Combining weather and crop data
To match the weather data to the agricultural data, we first calculated temperature and SPEI for each country across all cropland areas (i.e. all crops included). We derived monthly temperature and SPEI values for the cropland area in each country by taking the monthly values of each grid cell in a country, weighted by the proportion of cropland in each grid cell (Monfreda et al 2008). Then, we calculated temperature and SPEI for each country across annual growing seasons by averaging monthly weather values over months of country-specific crop-growing-season calendars (Sacks et al 2010). Second, we derived annual temperature and SPEI conditions for the 154 individual crop commodities for which the global harvested area at 5 min resolution was mapped by Monfreda et al (2008). We derived monthly temperature and SPEI values for each crop in each country by taking the monthly values of each grid cell in a country, weighted by the proportion of harvested area each crop occupied. Then, we calculated crop-specific temperature and SPEI across countries and annual growing seasons by averaging monthly weather values over months of country-specific growing-season calendars. Finally, we tested for the role of extreme temperatures by identifying years with a growing season temperature one standard deviation above the 1961–2018 growing season temperature average. This was made for both the cropland and the crop-specific analysis.
2.4.Statistical analyses
After the data cleaning process and hom*ogenization between agricultural and weather data, the database comprised 127 countries (see supplementary table 2 for the complete list) and 109 crop commodities. We first tested whether the response of total crop caloric yield and revenues to weather conditions depended on the diversity of crops grown within countries. We built models that account for management conditions (irrigation and nitrogen use) and other confounding factors. Using ordinary least squares (OLSs) regression we estimated:
where was the log transformed total output measured either in caloric yield or in revenues, aggregated across all crop species, in country j in year t. Weather is the weather during the growing season and cropland area of country j in year t. We ran separate regression models with high growing season temperatures and SPEI. We did not include both weather variables simultaneously since SPEI uses temperatures to calculate the PET. The three remaining variables of interest, crop diversity (Diversity), nitrogen fertilizer use (Nitrogen) and irrigation (Irrigation) have been calculated for each country j and year t. Here, we paid particular attention to the coefficients of the interaction terms. These measure the impact of crop diversity, fertilizer and irrigation on the responses of total caloric yield and revenues (
, respectively, to weather extremes. We expected
to be negative, reflecting a detrimental impact of droughts or high temperatures on
, and the interaction term
to be positive. A positive interaction would reflect the potential for increases in crop diversity to reduce or buffer the impact of droughts or high temperatures on
. Because Diversity, Fertilizer and Irrigation could respond to weather extremes and capture a direct effect of these events on yield and revenues, we ran separate models where these variables were lagged by one year. We expected fertilizer use, irrigation and crop diversity in the previous year to be good predictors of the same variables in the current year and served as tests of robustness for our results.
The last three terms were country fixed effects (country dummies), year fixed effects (year dummies) and the error term. These fixed effects capture the cross-country differences (including pedoclimatic conditions) and general changes in our variables over time due to technological progress (for example). We therefore only used deviations of diversity and output from country means and global trends to estimate our parameters. Given that variables in our models were measured in different units and showed heterogeneous value ranges, we normalized each predictor across all countries and over the full 58 year time period to enable the quantitative comparisons of the model coefficients (Schielzeth 2010). To evaluate whether crop diversity increased the resistance to weather extremes differently in different parts of the world, we ran separate models for each of seven world regions (as classified by the World Bank), excluding two regions with fewer than ten countries (i.e. North America and South Asia).
To disentangle how the selection of more resistant crops and complementarity between them underlined the effects of crop diversity on the resistance of agricultural production and revenues, we first evaluated the individual response of each crop to SPEI and high temperature. To do this, we used OLS regressions for each individual crop with the following form:
where was the total caloric yield of the crop i, in country j in year t. For weather, we ran separate models with high growing season temperatures and SPEI over cropland area specific to the crop i. The resulting coefficient
provided information about the direction and strength of the response of crop i to either droughts or high temperatures. Because we did not have crop-specific data on irrigation and fertilizer use, we did not include these variables in our model (2). Using simple regression models, we tested the relationship between individual crop responses to weather and total harvested areas or average revenues, respectively. Total global area and revenues were calculated across the 127 countries in our dataset. We hypothesized that main crops worldwide were also the most sensitive to weather extremes. Second, for each country, we computed the weighted mean (Wmean) and the weighted variance (Wvar) of individual crop responses to droughts and high temperatures to approximate the roles of selection and complementarity effects, respectively. To do this, we weighted individual crop responses by crop-specific area harvested measured at the country level. High Wmean values of the crop portfolio correspond to countries where the most abundant crops (in terms of harvested areas) showed low sensitivity to weather extremes, either high temperatures or droughts. High Wvar values indicate that the most abundant crops grown in a country display a great diversity of responses to weather extremes. We replaced crop diversity in our model (1) by values of Wmean and Wvar to test which of these two facets of the sensitivity of crop portfolio of countries drove the responses of caloric yield and revenues to weather extremes. All interactions among variables were included.
Statistical analyses were conducted using the R software v.1.4.1717, packages lfe (Gaure 2013) and Weighted. Desc. Stat (Parchami 2019).
3.Results
Total caloric yield significantly declined with droughts (figure 1(a)) and high temperatures (figure 1(b)) while total economic revenues significantly declined only with droughts. Consistently between models (using either drought or high temperatures), we found a stronger impact of weather on yield than on revenues. We also found that total caloric yield significantly increased with nitrogen and irrigation use and that total revenues strongly increased with crop diversity and nitrogen use at the country level (figure 1) (model results are detailed in the supplementary table 3).
We found significant positive interaction effects affecting total caloric yield; between droughts and crop diversity and between high temperatures and crop diversity. The interactions mean that the negative impacts of the two weather extremes on total caloric yield were significantly reduced as crop species diversity within countries increased (figure 1, supplementary table 3). Standardized estimates of our models suggested that increasing crop diversity by one standard deviation at the country level canceled out the global impact of high temperatures (EstimateHigh temp = −0.027, Estimate High temp:Diversity = 0.028), and almost canceled out the impact of droughts on caloric yield (EstimateDrought = −0.024, Estimate Drought:Diversity = 0.020).
We found that nitrogen or irrigation use have comparably buffered the impacts of high temperatures on total caloric yield. Finally, model results suggested that increasing irrigation use significantly reduced the negative impact of droughts and high temperatures on total revenues of countries (supplementary table 3). Our results were robust to changes in the specifications of our statistical models (supplementary table 4).
At the regional level, total caloric yield in countries of East Asia & Pacific (n = 24), Middle East and North Africa (n = 20) and Sub-Saharan Africa (n = 35) were the most negatively impacted by droughts. In countries of Sub-Saharan Africa, crop diversity had the largest positive impact on total caloric yield. Increased crop diversity also significantly buffered the negative impact of droughts on yield in this region, as shown by the significant positive interaction terms. A comparable result was found for countries of East Asia & Pacific. As for the general models, regional agricultural revenues showed less impacts from droughts in comparison with yield. Only revenues of Sub-Saharan Africa were significantly affected and again, we found that crop diversity had a buffering effect. Countries of Sub-Saharan Africa were also the ones where total caloric yield and total revenues declined the most with high temperatures. Consistent with the models including drought, crop diversity had the largest positive impact on both outputs and buffered the effects of high temperatures in this region. Results from the regional analyses are detailed in supplementary table 5.
Caloric yield responses of individual crops to weather extremes showed substantial variation (figure 2). With few exceptions, caloric yield of main cereals (wheat, maize, barley) and oil crops (oil palm fruits, olive, sunflower, soybean) showed a negative response to droughts (figure 2(a)) and high temperatures (figure 2(b)) that was greater or equal to the average aggregated response across all crops. These results were supported by a significant and negative relationship between the global harvested area of crops and their yield response to droughts (F(1,107) = 6.63, P= 0.011, figure 3(a)). This result showed that crops occupying the largest harvested areas worldwide were also the most sensitive to droughts. Because these dominant crops were also the ones with the least average global economic value (F(1,107) = 39.62, P< 0.0001) high-value crops were also less sensitive to drought (F(1,107) = 34.62, P< 0.0001, figure 3(b)). Only a few crops, mostly tropical horticultural ones (e.g. banana, mango, coconuts, plantains, watermelons), experienced increasing yield related to location-specific drought events (figure 2(a)). However, we found no significant relationship between the global harvested area of crops and their crop yield response to high temperatures (F(1,107) = 2.33, P = 0.13).
The response of revenues from individual crops to weather extremes also showed variation (supplementary figure 1). We found no significant relationship of the individual response of revenues to droughts (F(1,107) = 1.66, P = 0.20) and to high temperatures (F(1,107) = 0.14, P = 0.71) with global harvested area.
We then investigated how two facets of the sensitivity of crop portfolios to weather extremes can confer greater agricultural outputs and greater resistance to these events at the country level. First, regression models showed that countries growing crop portfolios with high Wmean (abundant crops were also less drought-sensitive) and Wvar (abundant crops displayed various responses to drought) values had significantly greater total caloric yield and agricultural revenues. Although the effect of Wmean on caloric yield was smaller than that of Wvar, crop portfolios with high Wmean values contributed to buffer drought impact on caloric yield (Wmean response × drought, supplementary table 6). We further found that crop portfolios with greater crop diversity also had a greater diversity of responses to droughts (Wvar) (t = 3.67, P= 0.0004). However, we found no significant relationship between crop diversity and Wmean values (t = 1.18, P = 0.24). The weak effect of droughts on revenues found in our general model was no longer significant in the regressions including Wmean and Wvar (supplementary table 6). Finally, models considering high temperatures revealed no significant impact of either Wmean or Wvar on total caloric yield and no significant buffering effect (supplementary table 6). We only found a significant, positive effect of Wmean on total revenues. Only nitrogen and irrigation use significantly impacted total caloric yield and its response to high temperatures. Irrigation also enhanced total revenues in the model considering high temperatures.
4.Discussion
Our research shows that increases in crop diversity within countries have the potential to buffer yield losses caused by droughts and high temperatures without compromising overall total yield (i.e. from all crops), and to generate higher revenues. Our findings indicate that encouraging greater crop diversity can be an adaptive strategy and warrants more attention, specifically in poorer developing countries that are likely to suffer disproportionately from climate change impacts (Smith et al 2001, Mertz et al 2009). Finally, our work contributes to scale-up the research on the benefits of crop diversity (Gonzalez et al 2020), and provides a better understanding of the characteristics of crop portfolios within countries that can support higher yield and yield stability in the face of climate change.
Consistent with previous research, we showed that droughts and high temperatures have a strong negative impact on agricultural production (Cottrell et al 2019). We found the impact of droughts on total caloric yield to be about equal in magnitude compared to the effect of high temperatures at the country level. This result differs from previous research on climate change impacts showing a stronger impact of high temperatures compared to that of precipitation (Lobell and Field 2007, Lobell et al 2011, Iizumi and Ramankutty 2016). This can be partly explained by the fact that weather and the relative importance of temperature and precipitation may differ across different crop growing areas of large countries. Although we could measure the spatial heterogeneity of extreme weather events, our analyses were limited by the coarse resolution of our crop output data. Moreover, differences between our results and those of previous research may come from differences in approach. Our results suggest that the aggregated response of crop portfolio to weather extremes at the country level might not reflect the response of a subset of crops assessed individually. Within the wide variation in responses of individual crop to weather extremes, major cereals and oil crops showed among the greatest sensitivity to droughts and temperatures. Although global major crops constitute the majority of calories produced on a global basis, focusing research on these crops may provide a partial picture of the effect of extreme weather events on food production and the future of food security.
Accounting for both caloric yield and economic revenues allowed us to compare the effect of weather extremes on these two measures of agricultural output. We found that revenues were consistently less impacted by droughts and high temperatures than was yield in all of our models. A possible explanation is that increased prices in response to reduced supply partially offset drought-related losses in caloric yield. If so, this compensatory mechanism benefits producers, it hurts consumers, who may face higher food prices.
We found large negative impacts of past drought events on total caloric yield in countries of Sub-Saharan Africa, the Middle East & North Africa region and in the East Asia & Pacific region. Although floods, hailstorms and crop pests/diseases are sources of risk in these regions, drought is predominant (Miyan 2015, Ayanlade et al 2018, Tramblay et al 2020, Rahut et al 2021). These three regions contain a large share of current arid and semi-arid zones that are expected to expand in size and to experience more frequent and intense droughts in the future (Pörtner et al 2022). Additional research efforts will be needed to build climate change resilience in these 'hot spots' (de Souza et al 2015), especially in Sub-Saharan Africa where we found that agriculture was also severely impacted by high temperatures.
According to our findings, Sub-Saharan Africa is the region that could benefit the most from promoting crop diversity to buffer the strong impacts of weather extremes on caloric yield and economic revenues. This finding resonates with ecological theories stating that biodiversity has its greatest moderating effect on ecosystem functioning in contexts of limited resources, including poor soil fertility (Lal 2009). We believe that our results have important implications considering not only the higher risks of climate shocks, but also the high level of vulnerability of African populations to climate change (Funk et al 2008, Thornton et al 2014). Agriculture supports the livelihoods of over 60% of the African population (FAO 2021a) and the continent is home to 37% of all malnourished people (FAO 2021b). In this context, studies assessing the benefit of biodiversification of agriculture still remain underrepresented in this region (Beillouin et al 2019). Our findings add urgency to the need to address this issue. The lack of emphasis on crop diversity as a form of climate insurance seems surprising considering that crop diversity is already a pillar of smallholder farming systems (Ricciardi et al 2018) and a lever of adaptation to climate risks (Koffi et al 2017, Mango et al 2018, Bellon et al 2020, Makate et al 2022). Evidence from local ethnobiological research suggests that societies farming in areas prone to high precipitation variability and/or to extreme weather events tend to make use of a wider range of crop varieties to secure their livelihoods (Kassie et al 2013, Matsuda 2013, Coomes et al 2015). Similarly, Baumgärtner and Quaas (2010) found that farmers diversified their crop portfolio in the face of uncertain climatic conditions and when financial insurance was not available or was too expensive for farmers to afford. A promising line of research consists of finding how to leverage the crop diversity that smallholding farmers in Sub-Saharan Africa make use of in order to secure better yields, limit yield stability and/or enhance revenues (though not all crops grown by smallholders are grown for sale). It is critical that research and policy efforts in this area catch up with the pace and scale of climate change.
Beyond trends, processes involved in the relationship between crop diversity and stability have mainly been evaluated at the local scale, within or between cultivated fields, where crops interact. However, mechanisms governing the diversity–stability relationship can be scale-dependent and the role of biotic interactions is expected to decline with spatial grain size (Gonzalez et al 2020). At the country level, our results suggest that processes close to the selection and complementarity effects promote total caloric yield, with the latter having the strongest effect. But to buffer drought impact on yield, crops occupying large shares of cropland within countries have to be drought-resistant, conferring high Wmean values to the crop portfolio. However, we found no relationship between Wmean and higher crop diversity and no effect of our metrics in models including high temperatures, meaning that there is still an additional way by which crop diversity buffered weather extreme impacts that we could not account for, and that merits further research.
Our results, however, could indicate how to appropriately constitute, or diversify, crop porfolios within countries. More specifically, we showed that relatively minor, high-value crops, were more stress-tolerant than major commodity crops. Although these minor crops represent a still untapped source of diversity to meet multiple challenges in agriculture (Zhang et al 2018), planting more rare crops may offer little stabilizing benefits. Instead, our results suggest that a diversification strategy based on promoting drought-resistant, high-value crops could help increase the Wmean value of the portfolio while increasing the diversity of responses to weather (Wvar) and thus leading to greater annual caloric yield and revenues.
Finally, we found that increasing irrigation and nitrogen use contributed to reducing the impact of high temperatures but not of droughts on caloric yield. Here, only irrigation use was found to increase resistance of revenues to the impact of both weather extremes. The lack of crop-specific data on management practices prevents us from going any further into the understanding of the buffering role of inputs. It is important to note, however, that there is, in general terms, a very limited degree to which increased irrigation can be used as a buffer against climate shocks in agriculture. Increasing yield and revenues through a more sustainable use of inputs could contribute to enhancing food security in other ways, for example by enabling farmers to constitute grain reserves and economic savings to compensate for crop and income losses in bad years.
Acknowledgments
Research leading to this contribution was supported by the French National Research Agency (ANR) under the 'Programme d'Investissem*nts d'Avenir' (reference 17- MPGA-0004). We thank DB McKey and S Prudham for reading and improving the English of this manuscript and finally we thank to the two reviewers for their constructive comments.
Data availability statements
The data that support the findings of this study are all publicly available following the links listed in the supplementary table 1 available from the corresponding author on reasonable request.
All data that support the findings of this study are included within the article (and any supplementary files). Data will be available from 28 February 2023.