Behavioural plasticity of a pest species may aggravate global wheat yield loss under climate change

Global food production has to grow by more than 70% by 2050 to meet the demand of world population growth¹. However, pests and pathogens have caused more than 40% of crop losses every year across the globe². Importantly, the crop losses by pests are predicted to increase under global climate change^3,4,5. Therefore, the crop yield loss caused by pests is of great significance to global food security in the context of ongoing climate change and world population increase^4,5,6,7.

Climate change leads to an increase in both mean temperature and frequency of extreme events⁸. All life activities of ectothermic pests, from metabolic rate to population growth, depend largely on ambient temperatures. Increase in mean temperature is predicted to aggravate the pest damage and yield loss^4,7,9,10,11, while extreme events are generally thought to inhibit pest populations¹². Insects can actively initiate mitigation approaches, such as behavioural, physiological and evolutionary mechanisms, to buffer stress and increase their survival probability under extreme events^12,13,14. However, the true roles of these processes in pest population dynamics and crop damage have rarely been identified and modelled. Behavioural plasticity is one of the most important processes of insects to cope with extreme high temperature^{13,15,16,17,18,19}, implying an important role in enabling pest population growth. Nevertheless, no studies have incorporated behavioural plasticity in pest damage and crop loss predictions so far.

When predicting pest damage under climate change, mean temperature increase is usually treated as the main driver in models for metabolic rate³ and feeding rate¹⁰, phenology and voltinism^8,9, population growth rate³ of pest insects, or disease infection risk⁶. The data used for establishing these models are commonly derived from constant temperature experiments. However, field insects experience temperature fluctuations through space and time^20,21,22 within their microhabitats^12,13,15,23. Constant temperatures failed to reflect the irreplaceable role of daily thermal extremes in driving insect performances^12,24,25. Moreover, in studies on the impacts of fluctuating temperatures and thermal extremes on pests, insects are often confined in limited spaces such as leaf clip cages^26,27 or containers²⁸. As a result, these studies overlooked the buffering role of pests’ behavioural thermoregulation to extreme conditions^{12,13,15,16,17,18} by exploiting microclimate heterogeneity in nature^{21,22,23,29,30,31}. Importantly, since crop yield loss is often a function of pest density³², neglecting insects’ buffering activity will likely hide the true effects of daily thermal fluctuations and thereby lead to inaccurate predictions of insect density due to the accumulation of daily errors during the whole crop growing season. Unfortunately, despite a few predictions incorporating behavioural thermoregulation of rather large ectotherms at the landscape scales or across kilometres^16,33, fine-scale behavioural thermoregulation (for instance, by moving across a single plant) has not been considered in predictions of pest population growth and the consequent crop yield losses. Thus, it remains unclear if and to what extent pest insects can use microclimate heterogeneity and behavioural thermoregulation to buffer climate impacts^13,23, and whether this buffering effect will cause more abundant pests and more serious crop yield losses¹³.

In the present study, by using wheat Triticum spp. and the grain aphid Sitobion avenae as a worldwide pest-crop model system³⁴, we quantify the effect of fine-scale thermoregulation on global abundance of a pest and a crop yield loss under climate change. First, we conducted a series of experiments (Fig. S1) to understand if and to what extent aphid fine-scale behavioural thermoregulation (FSBT) can buffer the impact of extreme high temperatures (EHTs). Then, we estimated and compared the global abundance of S. avenae and wheat yield loss across spatial-temporal scales with and without aphid thermoregulation in 697 locations during 1977–2017 (Fig. S2). Meanwhile, we illustrated how pest thermoregulation aggravated wheat yield loss (AYL, in %) in different countries/regions and we linked these results to global food security. Our study allows for an understanding of the importance of fine-scale thermoregulation of pest insects when assessing global food security and pest status at the macro-scales, although some other main factors, such as agronomic practices, crop genetics/varieties, other pest insects and diseases also need to be considered in future estimations.

Behavioural plasticity of aphids and its effects on survival and population growth

Behaviour experiment – First, we observed S. avenae behaviour under simulated daily fluctuating temperatures involving natural EHTs to determine if the aphids displayed FSBT at the individual plant level (surface of leaves and the soil surface below the plant). Overall, aphids can exploit spatial temperature heterogeneity within microhabitats and avoid overheating through thermoregulation (Fig. 1a, b). In the presence of daytime EHT, temperature was the lowest on the soil surface, intermediate on the leaf and the highest in the air (Fig. 1a). The temperature difference between soil surface and air increased with air temperature, and could be greater than 12 °C at midday. As a result, the aphids began to move from plant to soil surface as the temperature increased, resulting in a decreased proportion of the aphids residing on the host plant. When the temperature decreased after midday, aphids went back to leaves to a large extent. By contrast, in the absence of EHT, aphids did not leave the plant and the proportion of host plant residence remained high through time (Fig. 1b).

Then, we conducted two parallel experiments (clip cage and soil moisture experiments) to test if aphid FSBT can reduce the negative effects of EHT on key life-history traits and population growth of the aphids.

Demographic experiment (clip cage) – We used clip cages to fix the aphids on the plant to limit their FSBT (Fig. 1c, d). We designed a full factorial experiment of FSBT and EHT to investigate the impact of S. avenae thermoregulation on survival and population growth (Fig. 1c, d) and other important life-history traits (Fig. S3, Table S1 & S2). The aphids in the two treatments without EHT experienced similar mild temperatures, survived similarly well (χ²=0.28, df = 1, p = 0.5987), and had a relatively high intrinsic rate of increase (r_m) (Fig. 1d) due to faster development, higher fecundity, etc (Fig. S3). However, in the two treatments with EHT, aphid survival decreased. Importantly, FSBT increased survival significantly (χ²=12.83, df = 1, p = 0.0003) and reversed the negative r_m observed in the treatment without FSBT to a positive index, by exploiting a wide range of temperatures within the microhabitats (~28–40 °C at midday).

Abundance experiment (soil moisture) – In this experiment, the aphids were allowed to move freely. We used moist soil to maximize the thermal heterogeneity across the plant and the soil surface, allowing FSBT of aphids. By contrast, we used dry soil to decrease thermal heterogeneity as much as possible to limit the efficiency of FSBT. Since these soil moisture conditions did not affect S. avenae performance (Figs. S4 & S5), this experiment allows us to understand if FSBT of aphids can contribute to their population growth under EHT. The aphids in the different treatments also experienced different fine-scale microhabitat temperatures (Fig. 1e). As a result, in the two treatments without EHT, aphid abundance increased rapidly (Fig. 1f). In the presence of EHT, the aphids with limited FSBT (dry soil) underwent high maximal microhabitat temperature (~35–40 °C), leading to a population decline. By contrast, the aphids with FSBT (moist soil) during EHT exploited a larger microhabitat temperature range (~29–40 °C), resulting in a population increase.

EHT-survival experiment and survival models – Finally, we conducted an experiment to determine daily maximum temperature-dependent survival under conditions with and without FSBT. We built two survival models with these experimental data to quantify S. avenae survival during EHT with and without FSBT (Fig. 1g, h). Aphid survival decreased with daily maximum air temperature, following an inverse sigmoid curve. The aphids survived better when they had the possibility to thermoregulate behaviourally. However, as daily maximum air temperature continues to increase, the potential for aphid FSBT may decline due to the simultaneous increase in temperatures at the soil and plant surfaces.

Global abundance of S. avenae and wheat yield loss with and without FSBT

FSBT is powerful to buffer aphids against exposure to deleterious temperatures during EHTs. Thus, FSBT will allow a continuous increase in S. avenae abundance despite EHT, potentially contributing to wheat yield loss. Here, we calculated the aggravated yield loss (AYL) in wheat production caused by FSBT globally by following these steps (Fig. S2). First, to estimate the global aphid abundance and wheat yield loss across spatial-temporal scales, we collected the global distribution data of the aphid and the global daily temperature data during the past 41 years (1977–2017). We used these data to estimate the following variables during wheat growing season under the conditions with and without aphid FSBT: (i) the number of days with EHT which can depress aphid survival based on the EHT-survival models (Fig. 1h), (ii) the aphid peak density which was determined by the EHT frequency-dependent survival and the temperature-dependent relative growth rate of the aphid population in field conditions according to Ciss et al (2014)‘s model³⁵ assuming a given initial aphid density (1 aphid/tiller at the decimal code = 45 of wheat phenology) and (iii) the relative wheat yield loss (%) which was derived from aphid peak density using a yield loss model (Fig. S6). To quantify the effects of FSBT, we calculated the differences between the above paired (with and without FSBT) estimations, respectively. Then, we quantified the role of aphid FSBT in assessing global food security. We first estimated the global yearly aggravated wheat production loss caused by aphid FSBT (Fig. S7). Then, we compared the speed (% per year) of this production loss with the speed of the observed increase for global wheat production.

The results showed that FSBT reduces EHTs depressing aphid survival during the wheat growing season globally, and thereby increases aphid peak density and wheat yield loss by aphids (Fig. 2a–c). Specifically, the increase in aphid peak density and wheat yield loss (AYL) could be 0.15 aphid/tiller and 0.04%, respectively, for each EHT day buffered by aphid FSBT (Fig. S8). Aphid thermoregulation caused an aggravated loss of global wheat production of 8.7 million tons/year, and this aggravated loss increased at a rate of 0.15 million tons/year during 1977–2017 (Fig. 3a). Importantly, this increased rate of aggravated global production loss (2.6% per year) was faster than the increase of global wheat production (2.0% per year) (Fig. 3b), indicating that aphid thermoregulation not only greatly aggravates wheat yield loss but also generates a hidden risk for global food security that was never anticipated before.

Furthermore, we calculated the mean AYL (1977-2017) for each country and compared the AYL between various geographical regions and between the countries with different income levels. The results showed that the AYL due to FSBT by S. avenae is greater for underdeveloped countries in Africa, Asia, and South America than other developed countries (Fig. 4a, b). Among them, Eastern Africa (1.91%), Southern Asia (1.84%), Western Asia (1.75%), Central Asia (1.62%) and South America (1.21%) had higher AYL than the global annual average (1.18%). Moreover, wheat yield loss due to aphid FSBT varied according to the countries’ income levels (Fig. 4c). The AYL increased as the income level decreased, with much larger losses in low-income countries (2.10%) and low-income and food-deficit countries (2.27%).

Global trends of aggravated wheat yield loss with climate warming

To understand how past climate warming has altered the trends in AYL (annual change rate) caused by FSBT in S. avenae, we calculated the slopes (% per year) of the linear regressions between AYL and year for the globe and for each country. Overall, under the global warming scenarios for wheat planting areas in 1977-2017 (0.03 °C/year) (Fig. 5a), aphid FSBT led to an accelerated global AYL (0.08%/°C) (Fig. 5b). We also found that the change rate differed largely between countries (Fig. 5c). More than 4/5 of the countries (61/71) showed increased aggravation with warming, including most of the largest wheat production countries (15/19) and all the low-income and food-deficient countries (11/11) (Fig. 5d).

The increase of mean temperature under climate change is predicted to aggravate the damage of pests and diseases to crops^4,7,9,10,11. By contrast, the more frequent extreme temperatures (e.g., heat waves) are often supposed to have a negative effect on pest populations¹², thereby suggesting less insect damage to crops. Here we challenge this paradigm. We show that insect pests can alleviate the adverse effects of extremely high temperatures through FSBT, significantly improving their survival and thus promoting pest population growth and aggravating the crop yield loss. Furthermore, such aggravated yield loss is more prominent in the main wheat production regions and low-income and food-deficit countries. This effect brings forward a major yet unanticipated risk for global food security.

Thermoregulation across heterogeneous environments can reduce the need for physiological adjustment or adaptive evolution³⁶ and therefore is an economic and realistic way for organisms to avoid overheating^{13,16,17,18,19}, particularly in the case of limited potential to cope with climate warming through physiological thermal plasticity³⁷ or adaptive evolution³⁸. The mosaic of microclimates in nature can provide plenty of relatively cooler microsites for species to hide^15,23,30 and thus plays a critical role in survival probability under climate extremes^13,23,29,39. While few studies have highlighted the importance of FSBT in mitigating climate extremes^40,41, the influence of FSBT on the response of populations’ growth rate to EHTs has rarely been quantified. More importantly, the role of insect FSBT in buffering climate extremes has never been considered in the estimations of crop losses to pest insects. In this sense, here we provide a quantitative study illustrating to what extent pest insect FSBT will promote their population growth and cause aggravated crop loss.

Here, we focused on the direct damages of a single aphid species (S. avenae) on wheat, an economically important crop. Nevertheless, several aphid species commonly co-occur in the field²⁵, and these aphids usually infest wheat together with other pests^6,42. Multiple co-occurring pest species can cause more crop damage than the sum of damage of each species alone⁴³. More importantly, these aphid species are virus-borne vectors transmitting the yellow dwarf viruses (YDVs), a series of viruses infecting wheat globally and causing severe yield losses (ranging 5%-80%, 30% on average)⁴⁴, implying an even heavier aggravation of wheat losses associated with pests than reported here. Moreover, wheat aphids could also infest many other cereal crops such as barley, oat, maize and rice, which suggests that the aggravated yield losses reported here for wheat may apply more broadly to all major crops. Importantly, economic crop pests such as rice planthoppers¹³ and some potato pests⁴⁵, usually live within microhabitats with dense crop canopies, weakening direct solar radiation and thus preventing soil surface from overheating⁴⁶. Together with commonly used farmland irrigation reducing microclimate extremes⁴⁶, this suggests that the potential for herbivorous pest insects to buffer climate extremes through FSBT and consequently for causing severe crop loss is high and widespread.

Global and regional food security has been increasingly challenged by climate warming^47,48. Crop yield loss due to pests and diseases is one of the main problems^2,5,6. Here we find that thermoregulation can relieve S. avenae from being exposed to heat stress caused by EHT during the wheat growing season globally. In the case of moderate aphid density without any pest management, the resulting increase in aphid peak density causes an aggravated wheat yield loss of 8.7 million tons per year, which compares to the amount of food needed to feed about 133 million people every year⁴⁹. Besides the direct damage from pests, main abiotic stresses such as heat waves, extreme droughts, storms and floods may also reduce global crop production either independently⁵⁰ or interactively with biotic factors (pests and diseases)⁵¹, which implies an insufficient potential to meet future global food demand under climate change⁴⁷. Moreover, our results showed that the aggravated losses are higher in Africa, Asia, and South America, coinciding with the regions where the per capita wheat consumption is substantially increased during the past 60 years⁵². This indicates that food security in countries with low income and food deficits is more vulnerable to climate change. These countries are the most food insecure according to the 2022 Global Hunger Index⁴⁸ because of the lowest average crop yield⁵³, the largest negative impacts by warming⁵⁴, the least capability to adapt to climate change in sustainable agriculture and their important reliance on food imports⁵⁵.

Some potential limitations can be further considered to improve the estimations in future studies. On the one hand, our estimations in the global mitigation of insect FSBT against EHT assume a global homogeneity of both aphid adaptations and the temporal-spatial availability of cooler microclimates across the globe. First, the different genetic backgrounds of the aphid among geographic regions likely generate differences in heat tolerance and thermoregulatory behaviour. These differences may potentially alter the extent to which aphid FSBT would mitigate the impacts of EHT and thereby modulate our estimations of S. avenae population growth. Second, although commonly used irrigation can provide the aphid a moist soil surface for thermoregulation, the differences of irrigation regimes (such as sprinkler vs. flood irrigation) in their contributions to aphid thermoregulation need to be considered in future research. Irrigation regime may not only modify the soil moisture and the availability of cooler microclimates⁵⁶, but also modulate crop water stress with consequences for both pest survival and crop yield. On the other hand, we focused on temperature in our estimations of aphid population growth. As expected, greater changes of EHT can result in dramatic shifts in EHT-dependent aphid survival, thereby affecting aphid peak density and wheat yield loss (Fig. S9). Although temperature has been proven to be the most important factor affecting the abundance of cereal aphids⁵⁷ and other insects^58,59 under climate change, other factors such as greenhouse gas concentration and natural enemies certainly matter⁵. Furthermore, considering the complexity of actual yield loss caused by insect pests, some other main factors, such as the availability of aphid-resistant wheat varieties and their long-term effectiveness and local input levels of pest management, also should be considered in estimating actual yield loss caused by the aphids.

Our study has implications for pest management and food security. Thermoregulation can diminish the impacts of daily temperature extremes on pests and thereby enable population increase under EHT. These findings point to a potential strategy consisting of altering fine-scale microclimates by agricultural practices such as modifying crop canopy structures to interfere with thermoregulatory behaviours of pests⁶⁰ and even reduce crop diseases mediated by pests⁶¹. Meanwhile, since fine-scale microclimates and behavioural thermoregulation within microhabitats shape organismal responses to climate change¹², increasing heterogeneity and accessibility of microclimate, such as building climatic refugia, can contribute to population persistence and biodiversity conservation for non-crop ecosystems located around the crops in the context of increasing climate extremes^62,63. This may, in turn, enhance biodiversity-mediated natural pest control and facilitate crop production^64,65.

In conclusion, climate change increases not only the average temperature, but also the frequency, intensity, and duration of heat waves⁷. The rise of mean temperature is predicted to promote pest population growth and thus increase crop yield loss^4,7, while frequent extreme temperatures are expected to suppress pest abundance¹³ and should lead to reduced yield loss. By contrast, here we find that pests can mitigate the negative effects of extreme temperature through fine-scale thermoregulatory behaviour, favouring their population growth and aggravating crop damage and yield loss. Although our results were derived from wheat and aphid system, these findings may generally apply to other crop-pest systems due to the prevalence of insect behavioural thermoregulation within their microhabitats. In addition, our method can be a reference for predictions of other pests and crop loss by incorporating the extreme temperature and fine-scale thermoregulation behaviour in modelling the response of crop yield to climate change.

Thermoregulation experiments in the laboratory

Here, we designed four experiments to clarify if and to what extent S. avenae can behaviourally buffer the impact of extreme high temperatures (Fig. S1).

Insect preparation

The aphids of Sitobion avenae were collected from a winter wheat field near Beijing (39^o48′ N, 116^o28′ E). The aphids were reared in the laboratory on 5–20 cm high wheat seedlings in screen cages (30 × 30 × 25 cm) under standard rearing conditions at constant 20 °C, 50–70% relative humidity and 16:8 h (light: dark) photoperiod. The winter wheat seeds were sown evenly in plastic pots (diameter = 10 cm, height = 8 cm) containing nutrient soil to provide essential nourishment. The wheat seedlings were used to feed the aphids and were renewed once a week.

Generating EHTs, microclimate heterogeneity and aphid FSBT

EHTs – To simulate the ecologically relevant EHTs during growing season in nature, we used a “8h-fluctuating EHTs” thermal regime: during daytime, the temperature increased linearly from 20 °C at 08:00 and reached the maximum of 40 °C (EHT) at 12:00, then, decreased linearly to 20 °C at 16:00; in the remaining 16 hours, the temperature was set to constant 20 °C (Fig. 1a, c, e). Here, the ramping rate was 0.08 °C/min, which is very close to the rate of temperature change in nature^66,67. We chose 40 °C as the highest temperature because this temperature was often recorded as extremely high temperatures in the field when S. avenae occurs^24,25,68. The daily average temperature was kept at 24 °C, which is within the suitable temperature range of S. avenae^24,25. A constant 24 °C was used as the thermal regime without EHT (Fig. 1b, c, e) to eliminate the possible influence of the mean temperature difference between treatments. While we cannot test all natural thermal fluctuations, the constant temperature serves as a simplified representation of optimal conditions for insect performance. Previous research shows that mild temperatures (22–25 °C) with moderate daily fluctuations (±5 °C) are common in wheat fields⁶⁹. Under these moderate conditions, the effects of constant and variable temperatures are similar⁷⁰. Therefore, using a constant temperature as a control is reasonable.

Microclimate heterogeneity – We generated the microclimate heterogeneity by manipulating the soil surface moisture (see below). When the ambient temperature gradually increased from 20 °C to 40 °C, the moist soil surfaces were cooler than the ambient air and plant leaf surfaces, and the maximum temperature of the moist soil surface was <28 °C (Fig. 1a). Therefore, the soil surface can provide a cooler microsite for aphids to avoid the detrimental impacts of EHT. This manipulation simulated the thermal landscape in wheat fields, whereby the temperature at different heights of wheat plants differs greatly at midday, showing an obvious trend of higher temperature at the upper part and lower temperature at the lower part^71,72, which provides a temperature gradient for the aphids to regulate their body temperature by moving downward.

Fine-scale behavioural thermoregulation (FSBT) – In our study arena (see below), the aphids need to move from their host plant to the moist soil surface to avoid EHT. Here, we recorded the positions of the aphids and the moving behaviour of aphids between the seedlings and the soil surface to characterize their behavioural thermoregulation. Since aphids rarely transfer between host plants and soil surface in the absence of biotic and abiotic environmental disturbances³⁴, we can use the transfer behaviour as an indicator of the behavioural thermoregulation.

Experiment 1: Fine-scale behavioural thermoregulation under EHT

Experimental design – We quantified thermoregulation behaviour (the movements and the microsite choice of aphids) under the conditions with EHT (“8h-fluctuating EHT”, Fig. 1a) and without EHT, i.e., constant 24 °C (Fig. 1b). We tested the thermoregulatory behaviour for each nymphal stage (1st, 2nd, 3rd and 4th nymphal instars) and adults of the aphid separately. For each development stage, we tested 30 aphids placed individually on different wheat seedlings.

Aphid preparation for different development stages – To obtain the 30 test aphids of a certain development stage, and to ensure that the tested aphids were not related directly to each other, we randomly selected ~60 adult aphids from the stock rearing and put them individually into tubes (diameter = 1.5 cm, length = 7.0 cm). The number of aphids prepared was twice that of aphids tested to ensure enough test individuals. Inside each tube, a hydrated wheat seedling was provided for the adult aphid to feed and produce progeny for 12 hours. After that, only one newborn aphid from each tube was kept for testing the behaviour of 1st nymphal instars or for continuing to rear with a new tube containing a new wheat seedling until the aphid developed into 2nd, 3rd, 4th nymphal instars or adult stage. Different development stages of the aphids were obtained based on their respective development time at a constant temperature of 20 °C^24,25.

Experimental manipulation – Each test aphid was placed individually in an experimental arena. For this, we planted five wheat seedlings in a plastic pot (diameter = 10 cm, height = 8 cm), one in the centre and the other four around the centre at a distance of 3~4 cm, and then placed them in a rearing room with a constant 24 °C and 16 L:8D photoperiod. When the seedlings grew to about 10 cm height, they were used to identify the thermoregulation process of the aphids, i.e., leaving their original feeding site to search for other microsites. An aphid was transferred to the central seedling and allowed to settle down with a clip cage 12 hours before the experiment. The clip cage was removed gently when the experiment started. Each pot was covered by a transparent plastic cylinder (diameter = 9 cm, height = 21 cm), and the top of the cylinder was strapped using a piece of nylon gauze for ventilation. We recorded the aphid position every 10 minutes by eye from 08:00 to 16:00 hours – a procedure that was eased by the fact that only one aphid was present in the arena at a time. The temperature of different microhabitats (air, leaf surface of the central seedling, and soil surface) was recorded using thermocouples (type T) and data loggers during the experiment.

Response variables – A preliminary test indicated that aphids moved only rarely when feeding on their host plant under mild temperatures (20~30 °C). By contrast, the proportion of moving individuals increased at high temperatures. Therefore, the proportion of aphids displaying dispersal behaviour in response to the EHT can reflect the frequency of thermoregulation. Here, we took the changes in the proportion of aphids residing on their host plant as the response variables for thermoregulatory behaviour of the aphid across microsites (air, leaf and soil surface).

Experiment 2 (clip cage): Effects of FSBT on demographic responses to EHTs

Experimental design – To determine how EHTs affect key life-history traits and population parameters of the aphids and how FSBT can mitigate these negative effects, we designed a full factorial experiment of EHT and FSBT. The treatments were (1) EHT(+)FSBT(−), with EHT but without FSBT; (2) EHT(+)FSBT(+), with both EHT and FSBT; (3) EHT(−)FSBT(−), without EHT and FSBT; (4) EHT(−)FSBT(+), without EHT but allowing FSBT. The EHT was controlled as “8h-fluctuating EHTs” and without EHT was the constant 24 °C (Fig. 1c). We used clip cages to restrict the movement of aphids to create the treatments without FSBT.

Experimental manipulation – Here, we used the same experimental arena (a pot with 5 plants) as in experiment 1. When the wheat seedlings grew to about 10 cm height, one wingless adult aphid was gently transferred from the stock rearing to the central wheat seedling with a fine wool brush and then covered with a clip cage. After 12 hours, only one newborn nymphal aphid from the adult on the central wheat seedling was kept as the tested aphid by removing the adult aphid and other newborn nymphs. In the two treatments without FSBT, the test aphid was constrained on the central wheat seedlings with a clip cage (diameter = 5 cm, with ventilated windows on both sides) so that they could not move to the cooler soil surface during the whole experiment. In the two treatments with FSBT, each pot of wheat seedlings was covered by a transparent plastic cylinder (diameter = 9 cm, height = 21 cm), and the top of the cylinder was strapped using a piece of nylon gauze for ventilation. In this case, the aphid could move freely inside the cylinder from the host plant to the cooler soil surface, while it could not escape from the experiment arena.

We used four climate chambers (RG-400B, Hefei Yoke Instrument Co., Ltd., China, accuracy: ± 1.0 °C) to conduct the four treatments of EHT and FSBT (temperatures were recorded in the four chambers with hobo loggers to validate conditions). The pots with wheat seedlings and aphids were placed in the climate chambers with the corresponding treatment. The survival, development (moulting or not) and reproduction of the aphids were recorded twice a day at 8:00 am and 20:00 pm, and the deaths and offspring of the aphid were removed. When the aphids developed into adults, their body mass, body length and body width were measured by an analytical balance (MS105DU, METTLER TOLEDO International Trade (Shanghai) Co., Ltd., USA, accuracy: 0.01 mg) and a micro-object measurement and analysis system (V 1.01, Beijing Dongfang Nongren Biotechnology Co., Ltd., China, resolution: 768 × 576). Then, the aphids were returned to the experiment arena and their survival and reproduction were recorded twice a day, and the offspring of the aphid were then removed. For each treatment, 50–60 aphids were tested individually from birth to death. During the experiment, data loggers (U23-002, Onset HOBO, USA, accuracy: ± 0.1 °C) were used to measure the temperature of various microhabitats (i.e., ambient air temperature inside the cylinder, the leaf temperature within clip cage and soil surface temperature) every 5 minutes.

Response variables – Aphid survival (days) was selected as the key variable to identify the effects of EHT and FSBT on life-history traits of the aphids. Survival curves were estimated with the Kaplan-Meier method and compared using the log-rank test; pairwise log-rank tests were further conducted with Holm-Bonferroni adjusted P-values. The intrinsic rate of increase (r_m) was calculated as the indicator to reflect how EHT and FSBT affected the aphid population.

For other important life-history traits (Supplementary Data 1), We identified error distributions of different traits, i.e., development time, nymphal growth rate, body size at maturity (length, width, and mass), reproductive time, adult fecundity, and longevity, by using ‘fitdistr’ and ‘AIC’ functions from the ‘car’ and ‘MASS’ packages in R (version 4.0.0). We used generalized linear models (GLMs) with gamma distributed errors (for development time, growth rate, body mass, body length, body width, longevity) or normal distributed errors (for reproductive time, fecundity) to analyse the effect of 2 fixed factors (EHT and FSBT) and their interactions on these traits, with the ‘glm’ function and ‘car’ package. The multiple comparisons based on the GLMs between different treatments were conducted by using the ‘Tukey’ method, in the ‘emmeans’ function from the ‘emmeans’ package. The results are provided in the Supplementary Information (Fig. S3, Table S1 & S2).

Experiment 3 (soil moisture): Effects of FSBT on aphid abundance under EHTs

Experimental design – To test if aphid FSBT can contribute to promote their population, we tested the effects of EHTs and FSBT on aphid abundance across time. We used the same experimental design that in experiment 2: (EHT(+)FSBT(−), EHT(+)FSBT(+), EHT(−)FSBT(−) and EHT(−)FSBT(+)). To make our experiment more ecologically relevant, we allowed the aphids to move freely within the microhabitats, but we manipulated soil moisture to adjust the microclimatic gradient: a dry soil surface reflected a situation without FSBT (small temperature gradient), while a moist soil surface generated a situation with FSBT (large temperature gradient). Although this manipulation can potentially lead to changes in the relative humidity (RH) within microhabitats (on the leaf surface and on soil surface) under both dry and moist soil surface conditions (ranging from 40% to 90%), according to a preliminary test (see Supplementary Methods: Effects of EHT and RH on S. avenae performance), these RH conditions may not necessarily affect aphid performance (Fig. S4 & S5).

Experimental manipulation – We transferred 10 adult aphids from our stock rearing into a pot with 25 plants 12 h prior to the beginning of the experiment, allowing them to reproduce. Here, we increased the number of plants to provide sufficient host plants for the aphids, especially in the treatments without EHT (high growth rate). The pot was covered by a transparent plastic cylinder to avoid the aphids’ escape. After 12 h, we recorded the number of aphids in each pot as the initial aphid abundance, and then moved the pots together with aphids to the four climate chambers for the designed treatments, respectively. We recorded the number of aphids in each pot at 08:00 on days 0, 1, 3, 6, 9, 12, 15 and 18. We replicated 10 times each treatment. We used data loggers to measure the air temperature inside the cylinder and used thermocouples to measure various microhabitat temperatures (each of the random five leaves, soil surface and ambient air) every 5 minutes during the 8h-fluctuating EHTs (Fig. 1e).

Response variables – We calculated and used the number of aphids on each observation day for each pot as the aphid abundance to show the population trends in each treatment. We used a linear regression model to estimate the relationship between aphid abundance and experimental time for each treatment. The numbers of aphids on each observation day were log-transformed to avoid large dimensional differences between treatments. Meanwhile, we used the General Linear Model to test the differences among treatments and used Tukey’s method for multiple pairwise comparisons of different treatments.

Experiment 4: Aphid survival at different EHTs with and without FSBT

To quantify to what extent aphid FSBT can buffer the impact of different EHTs on their survival, we tested the 24-h aphid survival with and without FSBT (by restricting the movement of aphids using clip cages) under a constant 20 °C (control) and seven “8h-fluctuating EHT” regimes with maximum daytime temperatures of 34, 36, 38, 40, 42, 44 and 46 °C (nighttime temperature was constant 20 °C, Fig. 1g). We tested thirty individuals for each stage (1st, 2nd, 3rd, 4th nymph and adult) at each temperature regime. The aphids were individually settled on a wheat leaf in a pot for 1 hour prior to the experiment, allowing them to feed. We then used the transparent plastic cylinder to cover the pot to prevent the aphids to escape. Then, the survival of these aphids was checked after 24 hours. Since the survival data for all stages displayed a similar pattern across treatments, these data were pooled as source data to build the two EHT-dependent survival models with and without FSBT, respectively. We used the Gompertz model (reverse ‘S’ shape) to describe the aphid survival at day t caused by extremely high temperatures under the conditions with or without FSBT:

where EHTt is the daily maximum air temperature at any given day t. EHT0 is the EHT causing aphid survival equal to 0, and a and b are parameters required to fit the models. The parameters EHT0, a, and b differ between the conditions with or without aphid FSBT. The EHT-survival difference between the conditions with and without FSBT can quantify the extent to which the FSBT of the aphid can increase its survival probability at any given EHT.

Global estimations of S. avenae abundance and wheat yield loss

Modelling peak density of S. avenae and wheat yield loss

Wheat yield loss due to aphid infestation is a function of aphid peak density, which depends mainly on the host plant phenology and the ambient temperatures the aphids have experienced^32,35. Thus, first, we estimated aphid peak density using the data for wheat phenology and temperature. The aphid number at any given day t + 1 was calculated as:

where Aphid_Nt is the aphid number at day t, RGR_t is the relative growth rate of the aphid population at day t. The initial aphid number (Aphid_N0) was set to 1 aphid/tiller at the boot swollen of the wheat phenology (in decimal code (DC) = 45 at day 0). Survival_EHTt is the aphid survival at day t caused by extremely high temperatures under the conditions with or without FSBT. The maximum daily aphid number was chosen as the aphid peak density.

RGR preparation

We estimated the relative growth rate (RGR) of the aphid population according to Ciss et al (2014)‘s model³⁵ based on daily mean temperature (T) and the DC of wheat growth stage:

where T_Max is the lethal temperature for the aphid, DC_Max is the latest wheat growth stage allowing aphid feeding, DC_m is the position of the left inflexion point for the RGR response to wheat growth stage. a₁, a₂, b and k are parameters required to fit the model. Here we set the parameters as T_Max = 30, DC_Max = 92, DC_m = 32.91, a₁ = −0.53, a₂ = 0.53, b = −0.05, k = 1.11 according to Ciss et al.³⁵. We prepared the data for decimal code based on the growing degree-day (GDD) for the growth stages of cereal crops⁷³:

where a, b and GDD₀ are parameters required to fit the model. Here we used the GDD model (USDA: Metadata for Winter/Spring Wheat Growth Stage Models)⁷⁴ to calculate the degree-days of any growth stage of wheat:

where T_base is the base temperature, which was set as 0 °C⁷⁴ for spring wheat. For winter wheat, T_base was set as 0 °C before double ridge stage (0 < DC < 30), 3.3 °C from double ridge to heading (30 ≤ DC < 50), and 5.1 °C from heading to maturity (50 ≤ DC ≤ 100), respectively⁷⁵. Here we used a different T_base for winter wheat to make sure the estimations of GDD were in line with the field-recorded GDD of wheat growth⁷⁶. i is the day beginning at growth stage s1 and incrementing daily until the beginning of growth stage s2, T_meani is the average daily temperature, which we computed as (DT_max+DT_min)/2. For any given location, the planting dates of winter wheat and spring wheat were derived from the literature⁷⁶. The total number of days in a growing season was determined when GDD = 1825 degree-days because the aphid cannot continue feeding on the wheat plant at this mature growth stage³⁵.

Survival with and without FSBT

As ectothermic insect species with largely overlapping generations during the growing season, the aphid density at any given location and year depends mainly on how local climates (especially temperature extremes) constrain their survival and consequent population growth^12,24,25. Importantly, both the intensity and frequency of temperature extremes can matter during this process^12,25. Thus, here we first determined the total number of days at which the daily maximum air temperature (DT_max) reduces aphid survival during a given wheat growing season. Then we calculated the total survival of aphids with and without FSBT by multiplying the aphid survival for each day based on the data for DT_max of each site and on the EHT-survival models (Eq. 1). The estimated parameters of the two models are displayed in Fig. 1h. Note that this approach assumes that those EHT-survival models remain the same across locations.

Estimation of wheat yield loss caused by aphids

Since wheat yield loss is a linear function of the peak density of S. avenae^32,77,78, we estimated the wheat relative yield loss (%) using the following model:

where Aphid is the peak density (aphids/tiller) during the wheat growing season, a and b are parameters of the model. Here, the parameters were set as: a = 0.29 and b = 1.07 based on the extracted data derived from independent field experiments^77,78 (Fig. S6, Supplementary Data 1). The use of a model for relative yield loss (%), rather than some other models for quantitative production loss (e.g., kg/acre), allows us to integrate the yield loss across the globe, regardless of wheat production variations across time and space scales.

Estimating global impacts of FSBT during 1977–2017

To understand the unique impacts of aphid FSBT on the aphid abundance and wheat yield loss globally, we calculated the difference in key traits between the two conditions (with minus without aphid FSBT) during wheat growing seasons for all locations where the aphid is present in each year throughout the period 1977–2017: (i) the difference in EHT frequency (i.e., the extreme temperature actually experienced by aphids in the two FSBT scenarios), to quantify the buffering amplitude of FSBT, (ii) the difference in aphid peak density that results from the exposure to temperature, and (iii) the difference in relative yield loss to quantify the impact of FSBT on crop damage (see Fig. 2). In the process of model estimations, we set the maximum yield loss as 100% when the rare unreasonable estimated values (0.154% of the total numbers of data, 44/28577) exceeded 100% and excluded the extremely rare aberrant data (0.014% of the total numbers of data, 4/28577) to avoid the negative AYL estimations.

Preparation of temperature data

To estimate the density of S. avenae at the global scale during 1977–2017, we downloaded global daily temperature data (mean, maximum and minimum) for the past 41 years (1977–2017) from Berkeley Earth (1° × 1° grid, http://berkeleyearth.org/data/). Since microclimate models cannot calculate the temperatures at the microhabitat (cm) scale for small insects such as aphids^14,23, here we used a simple linear model (see Fig. S2, in Zhao et al.⁷⁹) to transform macroclimate data (gridded information) into microclimate (microhabitat for aphids). This model reflects the relationship between the temperature records at the local weather station and the temperature data collected using thermosensors inserted into the wheat spikelet, the favourite microhabitat for aphids to feed. As a result, this statistical model estimated ca. 5 °C higher daily maximum temperature on average in spikelet than the local weather station records. This temperature difference is similar to that experimentally observed in other wheat fields⁸⁰ and is frequently reported across various habitat types and geographic regions^30,81,82.

Determination of aphid distribution

To map the S. avenae infestation globally, first, we collected the distribution data of the aphid by using its scientific name “Sitobion avenae” as a keyword in the databases of GBIF and CABI. We also collected the distribution data of wheat by using the keyword “Triticum aestivum” in these two databases due to the prevalence of this aphid species to infest wheat globally. The GBIF and CABI databases integrate information from publications written in English. However, most information for China is provided in Chinese. Therefore, for the distribution records in China, we also searched for the literature in the database of China National Knowledge Infrastructure (https://www.cnki.net/) to reduce the publication biases^83,84 due to incomplete published records in Chinese. Then, we kept the aphid distribution data only for the locations that were included within the distribution maps of main wheat-producing areas for each country (according to USDA: Global Crop Production Maps by Region, https://ipad.fas.usda.gov/rssiws/al/global_cropprod.aspx). This selection process ensured that the aphid infestations on wheat are representative for each country and produced more reasonable global estimates of wheat yield loss induced by this aphid species. Then, the locations were combined according to the scales of the temperature data (1° × 1° grid) to avoid any repeated calculation. Finally, 697 locations were selected to estimate global aphid occurrence.

Furthermore, to understand the extent to which aphid FSBT will affect aphid peak density and aggravated wheat yield loss (AYL, in %) through buffering the impact of EHTs, we used linear regression models to describe the relationships between (i) the differences in EHT days and the differences in aphid peak density and (ii) between the differences in EHT days and the differences in wheat yield loss, both under the conditions with and without aphid FSBT (see Fig. S8). For example, when FSTB induces a decrease in the number of EHT days by 40 days, producing an increase of +6 in aphid peak density (because they survive better), then the yield loss worsens by +1.6% (Fig. S8). Then, to quantify how global aggravated wheat production loss caused by aphid FSBT changes with time, based on the estimated global yearly mean differences of relative yield loss (%) between the two scenarios with and without aphid BT, together with the data for global yearly wheat production (million tons) from FAO⁴⁹, we calculated the yearly global aggravated wheat production loss (million tons) due to aphid thermoregulation using the following function (see Fig. 3a):

where PL_FSBT is the aggravated global wheat production loss (million tons) caused by aphid FSBT, P is the global wheat production (million tons) obtained from FAOSTAT, YL_AP is a recently estimated wheat yield loss (%) to all wheat pests⁶, YL_FSBT is the global mean difference of wheat relative yield (%) between with and without aphid FSBT (see Table S4 for more details).

Then, we determined to what extent this trend of the aggravated wheat loss by FSBT of S. avenae impacted global food security. We compared the slope for the linear regression of yearly aggravated loss with the slope for the linear regression of the global wheat production (FAOSTAT) during 1977–2017, using the respective values in 1977 as the baselines. Here, the food insecurity was inferred when the global aggravated wheat loss to aphid FSBT increased faster than global wheat production, i.e., the former had a greater slope than the latter (see Fig. 3b).

Moreover, to know the differences of the aggravated yield loss between countries, we calculated the mean differences (1977–2017) of the aggravated yield loss for each country as well as for various geographical regions and for the countries with different income levels (see Fig. 4). The data for global wheat production by country were obtained from FAOSTAT. The data for development levels (income and food-deficit levels) were derived from The World Bank (https://data.worldbank.org/indicator/).

Sensitivity analysis of the aggravated wheat yield loss

Considering the complexity of the multiple models used for estimating the aggravated wheat yield loss, here we conducted sensitivity analyses for the potential factors that determine the global AYL estimations. The variables selected and their changes were shown in Table S5. Here, we determined the estimated values of AYL to artificial changes of variables in a set of models for the globe and for each geographical region (Fig. S9).

Trends in aggravated wheat yield loss under climate warming

To understand how climate warming will change the global trend in the estimations of aggravated wheat yield loss, we calculated the global mean temperatures (http://berkeleyearth.org/data/) and the global mean aggravated wheat yield loss in each year during 1977–2017. Then, to know which countries will be affected the most in the estimations of aggravated yield loss caused by S. avenae FSBT under warming, especially from the perspective of global food security, first we estimated the change rate (the slope of linear regression) of yearly estimated aggravated yield loss for each country. We then showed the slopes of global mean value and the values for the top 20 countries with the largest wheat production (FAOSTAT) (Australia was not included because the aphid species S. avenae was not widespread there, according to CABI), as well as the values for countries with both low-income and food-deficit (The World Bank) (see Fig. 5).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.