1 Introduction

Climate change is manifested by change in mean climatic conditions (such as surface temperature and precipitation), and also by the increase in the intensity and frequency of related climate extremes (such as heat wave, frost, drought, and flood) (Lobell et al. 2013; Ding et al. 2010). Generally, crop growth and development are more sensitive and vulnerable to climate extremes than climate change (Schlenker and Roberts 2009). Thus, along with the important impact of changes in key climate variables (i.e., radiation, temperature, and precipitation) on crop phenology and yield (Xiao and Tao 2016; Tao and Zhang 2011), extreme climate events exerted more serious negative effects on some physiological processes and behaviors of crops, resulting in very large losses in crop yield (Chen et al. 2016; Zhang et al. 2014). For example, heat stress and drought triggered by supra-optimal temperatures and low precipitation detrimentally affect crop transpiration and photosynthesis processes and yield traits (Mahrookashani et al. 2017). Excessive precipitation or heavy rain is more likely to cause waterlogging over a long period that harms crop growth and reduces grain yield in cooler areas in conjunction with poorly drained soils (Li et al. 2019). Overall, the climate-related risks faced by crop production mainly depend on the degree of exposure to various extreme climate events (Butler and Huybers 2015), occurring with varying severity at different temporal and spacial scales and affecting various aspects of crop growth and development, and even farming practices (Lobell et al. 2011).

Maize (Zea mays L.), as a vital source of food, feed, and fuel in the world, plays an important role for agricultural and national economy. Compared to other grain crops, maize is currently the crop with the largest sown area and highest production in China (Xiao et al. 2020a). The North China Plain (NCP) is an important maize production region in China, accounting for nearly 40% of national maize production (Tian et al. 2019). Since maize planted in the NCP is usually grown in summer season (June to September), exposure to high temperature stress, drought, and/or heavy precipitation are relatively common risks (Xiao and Tao 2016). Especially, extreme heat events around flowering and grain filling periods (i.e., reproduction growth period) can cause serious damage to maize yield in the NCP (Wang et al. 2018). Moreover, related study in this region indicated that agricultural water conditions would reduce final maize yield when the accumulative precipitation during the maize growth season was less than 300 mm or greater than 500 mm (Wang et al. 2020). This conclusion notes that both drought and/or heavy precipitation will play an important role in maize production in the NCP.

There are increasing concerns on the impacts of climate extremes on maize growth and development (Lobell et al. 2012). Exceptionally large numbers of extreme climate events have been observed in recent years, and many studies highlight the threat that extreme climate pose on maize phenology and yield (Lobell et al. 2013; Van der Velde et al. 2010). Previous studies mainly focused on the impact of extreme weather/climate conditions on maize production over the past few decades due to the easy availability of historical climate records and crop observations (Zhang et al. 2014; Lobell et al. 2011, 2012, 2013; Ding et al. 2010). Along with the continuous warming of global climate, extreme climate events will occur more and more frequently, and the extreme degree will become more and more serious in the future (Bai et al. 2021; He et al. 2018). Understanding and predicting the spatiotemporal changes of extreme climate events under future climate scenarios is essential for formulating adaptation strategies to reduce the risks of climate to maize production (Bai et al. 2021). Therefore, it is particularly important to explore the future occurrence of extreme climate events during the maize growth period under the background of future climate change (Zhang and Yang 2019).

In general, the probability of occurrence of extreme climates in future climate scenarios can be projected by global climate models (GCMs) (Wang et al. 2016). According to GCMs data from the Coupled Model Inter-comparison Project phase 5 (CMIP5), several studies projected the future changes in extreme climate in the NCP (Bai et al. 2021; He et al. 2018). CMIP6 could better project future climate change arising from natural, unforced variability and in response to changes in radiative forcing in a multi-model context (Eyring et al. 2016). However, there are very few studies to project the change in future extreme climate events using GCMs from CMIP6 in the NCP (Xiao et al. 2020a). Additionally, in previous studies, exploring the change of extreme climate during the crop growth period under the future climate scenario often failed to consider that the crop phenology itself also will change in the future climate warming (Bai et al. 2021; He et al. 2018). Therefore, we first needed to assess the potential shift of crop phenology and growth stages under future climate scenarios to accurately quantify the future extreme climate change during the crop growth period (Zhang and Yang 2019). Process-based crop models mainly driven by climate data are widely accepted and applied as effective tool for assessing crop growth and development process under changing environment conditions (Xiao et al. 2016). Therefore, based on the crop model calibrated and validated using historical weather and phenological data, it is possible to quickly and acceptably simulate and predict changes in crop phenology under future climate scenarios.

In this study, along with predicting potential shift in maize phenology under future climate scenarios, we investigated six extreme climate indices (ECIs) closely related to maize production in the NCP based on daily climate data statistically downscaled from 20 GCMs in CMIP6. The objectives are to (1) determine the potential phenological changes of maize in the NCP under future climate scenarios, (2) identify the possible spatial and temporal changes of extreme climate events during the maize growth period in the future, and (3) explore the ability to adjust the sowing date of maize to reduce extreme climate events.

2 Materials and methods

2.1 Study area

The NCP study area is located between 32.9–40.5°N and 113.7–122.7°E (Fig. 1). During 1961–2014, the mean annual temperature in the region is 9.6–16.0 °C, and mean annual precipitation across the plain is 451–955 mm, of which more than 50–80% of the amount fell between July and September (Xiao et al. 2016). In the NCP, the rotation of summer maize and winter wheat in 1 year is the typical cropping systems over the past few decades and recently (Xiao and Tao 2016). Summer maize (hereinafter referred to as maize) is sown in early June and harvested in September (Xiao et al. 2020b). In terms of field management practices, irrigation is conducted 1–2 times every year, fertilizer is applied at sowing time (about 120 kg N ha−1) and jointing stage (about 60 kg N ha−1), and pesticides are used to control pests and diseases. Due to interannual weather variability and agronomic management, sowing dates are shifted and due to breeding progress cultivars are changed every 5–8 years (Xiao and Tao 2016).

Fig. 1
figure 1

Locations of the observed 52 meteorological stations and 37 agricultural meteorological stations in the North China Plain

Full size image

2.2 Climate data

We obtained historical daily climate data (1981–2014) for 52 meteorological stations (Fig. 1) from Chinese Meteorological Administration (CMA), which included sunshine hours (Sh), maximum (Tmax) and minimum temperature (Tmin), and precipitation (Prec). Daily solar radiation (Rad) was calculated from the Sh using the Angstom-Prescott equation (Xiao et al. 2016). Historical data of temperature (Temp) and Prec were used to calculate observed ECIs during the historical period and correct biases of GCM outputs for the statistical downscaling procedure (Bai et al. 2021). In addition, daily climate data including Rad, Tmax, Tmin, and Prec were used to drive the APSIM model (Brown et al. 2014) to simulate maize phenology under future climate scenarios.

Future climate data were acquired from 20 GCMs (Table 1), which was provided by the CMIP6, which integrated future Shared Socioeconomic Pathways (SSPs) and the Representative Concentration Pathway (RCP) from CMIP5 simulations (O’Neill et al. 2016). The SSPs represent alternative evolutions of future society under climate change and policy. In this study, we selected two representative scenarios, i.e., SSP245 (combining SSP2 with RCP4.5) and SSP585 (combining SSP5 with RCP8.5). In detailed, SSP2 assumes a central pathway in which trends continue their historical patterns without substantial deviations, while SSP5 envisions an energy intensive, fossil-based economy (O’Neill et al. 2016). RCP4.5 represents a stabilization scenario which is a radiative forcing pathway stabilized at 4.5 W m−2 by 2100, and RCP8.5 represents a high emissions scenario which is an increasing radiative forcing pathway to 8.5 W m−2 after 2100.

Table 1 List of the 20 global change models (GCMs) for future climate scenarios used in this study
Full size table

Generally, GCM output data are at spatial resolutions of hundreds of kilometers, and so are too coarse to be applied directly in small-scale or regional scale climate analysis (Xiao et al. 2020b). The spatiotemporal downscale processing can obtain the required high-resolution or site-specific climate data (Wang et al. 2016). In this study, we used a statistical downscaling method developed by Department of Primary Industries of New South Wale (DPINSW), Australia (Liu and Zuo 2012). This approach uses monthly gridded GCM climate data and relative parameters derived from GCM projections and climate observations to generate a realistic time series of daily climate data. Firstly, monthly GCM simulations were downscaled to desired sites using the inverse distance-weighted interpolation (IDW). A bias correction procedure was conducted to correct site-based monthly GCM values. Secondly, daily climate variables (Rad, Tmax, Tmin, and Prec) were temporally scaled for each site from the spatially downscaled projections by using the WGEN stochastic weather generator (Liu and Zuo 2012). More detail introduction for this downscaling method can refer to Liu and Zuo (2012). Future daily climate data generated by this approach have been widely employed in future climate studies (Bai et al. 2021; Xiao et al. 2020a, b; Wang et al. 2016).

2.3 APSIM-maize model and phenology simulation

The APSIM is a cropping systems model developed by agricultural production systems research unit of Australia (McCown et al. 1996). APSIM contains an array of modules for simulating growth; development and yield of crops, pastures, and forests; and their interactions with the soil (Keating et al. 2003). APSIM-maize is a biophysical model developed to simulate maize phenology and yield at the field scale (Wang et al. 2018). Overall, APSIM-maize model simulates maize growth and development, biomass accumulation and partitioning, and LAI as well as root, stem, leaf, and grain growth in daily step from sowing to maturity (Holzworth et al. 2014, 2018). Daily solar radiation, precipitation, and maximum and minimum air temperatures are the minimum weather input requirements of the model.

In this study, APSIM-maize model was used to simulate the growth process of maize under the future climate scenarios. In simulating the whole growth and development process of maize, APSIM-maize defines 11 crop growth stages, and the start of each stage is determined by accumulated thermal time, except the sowing to germination stage that is driven by soil moisture. The phase between germination and emergence is determined by both thermal time and sowing depth. The duration between emergence and flag leaf appearance is determined by total number of leaves on the plant and the rate of leaf appearance that is in turn determined by temperature. The total number of leaves equals that at germination plus that subsequently initiate until floral initiation. Hence, the time of floral initiation determines total leaf number and time of flag leaf appearance. Moreover, thermal development rate is sensitive to photoperiod between end of juvenile phase and floral initiation (Xiao et al. 2016). The main parameters used in APSIM-maize model for simulating maize phenology of a given cultivar includes thermal time from emergence to end of juvenile stage (°C day), thermal time from flowering to maturity (°C day), thermal time from flowering to start of grain-filling (°C day), and required thermal time change for floral initiation per hour photoperiod increase (°C h−1) (Xiao and Tao 2016).

The observed records of maize growth stages (e.g., sowing, flowering and maturity dates) at the 37 agro-meteorological stations in the NCP (Fig. 1) were collected from CMA. Enlisted agro-technicians in each station are used to record the dates of main phenological phases of crops (Xiao et al. 2015), including maize. Therefore, the quality of the phenological data obtained is guaranteed. The APSIM-maize model was calibrated and validated based on observed phenological data (i.e., sowing, flowering, and maturity dates) for the selected stations. Firstly, the typical cultivar cultivated in 2001–2010 was identified for each station. Then the first 3 years of all phenological data from selected stations were used to calibrate the cultivar parameters in APSIM-maize model, and the remaining phenological data for each selected cultivar were used to validate the model for corresponding cultivar. The model performance was evaluated based on the slope of regression equation and coefficient of determination (R2), root-mean-square error (RMSE), and agreement index (d); d is a better indicator of model performance. A d value close to 1 indicates low model bias, while a d value of 0 indicates no relationship between observed and simulated variables (Willmott 1982).

$$\mathrm{RMSE}=\sum_{i=1}^N{\left({P}_i-{O}_i\right)}^1/N$$
(1)
$$d=1\left[\sum_{i=1}^N{\left({P}_i-{O}_i\right)}^1/\sum_{i=1}^N{\left(\left|{P}_i^{\prime}\right|+\left|{O}_i^{\prime}\right|\right)}^2\right]\cdots 0\le d\le 1$$
(2)

where Pi is simulated values, Oi is observed values, N is number of values, and \({P}_i^{\prime }={P}_i-\overline{O}\) and \({O}_i^{\prime }={O}_i-\overline{O}\); \(\overline{O}\) are average observed values.

Finally, the validated model was run by the future daily climate data for 2015–2100 to simulate the dates of flowering and maturity in the selected station. Moreover, we applied the well-validated APSIM-maize model to simulate the maize phenology at different sowing dates (SDs) under future climate scenarios to investigate the impacts of SD adjustments on ECIs during maize growth period. We set five simulation groups according to different SDs, which are observed SD during 1981–2014 (S0), SD advance 30 days (S_A30) and 15 days (S_A15), and SD delay 15 days (S_D15) and 30 days (S_D30).

2.4 Extreme climate indices

Maize in the NCP is mainly planted in summer season (period from June to September), and high temperature stress, drought, and heavy precipitation tend to be the most common extreme climate disasters (Xiao and Tao 2016). Generally, the sensitivity of maize to high temperature stress varies with specific phenological stage. Based on previous studies, maize grown in the NCP is mainly damaged by high temperature (Tmax greater than 34 °C) during the flowering period and grain filling stage (i.e., reproductive growth period (RGP)) (Wang et al. 2018). In this study, we define RGP as the period from 10 days before flowering to 30 days after flowering to facilitate the calculation of extreme high temperature index of this stage. To comprehensively evaluate the occurrence and impact of high-temperature stress on maize in the NCP, we developed three extreme high temperature indices (e.g., HD (hot days), HSI (heat stress intensity), and HCD (consecutive hot days)) during maize RGP (Table 2). In order to evaluate the occurrence of heavy precipitation, continuous precipitation, and drought during the whole growth period of maize, we selected three extreme precipitation indices (e.g., R20 (heavy precipitation days), CWD (consecutive wet days), and CDD (consecutive dry days)) during the whole growth period of maize (Table 2), which were recommended by the CCI/WCRP/JCOMM Expert Team on Climate Change Detection and Indices (ETCCDI) (Bai et al. 2021). The detailed information for above six ECIs was shown in Table 2.

Table 2 Definitions of 6 extreme climate indices (ECIs) used in this study
Full size table

The six ECIs during maize growth period for the historical period (1961–2014) at each meteorological station were calculated based on the observed phenology dates. For the future period (2015–2100), the six ECIs were calculated based on the average phenological stages of maize in the historical period as the mode without considering future phenology change, and based on the maize phenology simulated by APSIM-maize model as the mode with considering future phenology change. For some meteorological stations without observed and simulated maize phenology, we used the phenological information from the nearest agro-meteorological stations to calculate ECIs.

2.5 Multi-model ensemble methods

In order to reduce the uncertainty caused by an individual GCM, the averaging result from multiple GCMs was widely used in climate projection, which based on an assumption that errors tend to cancel if the GCMs are independent (Tebaldi and Knutti 2007). However, the GCMs from CMIP6 cannot be regarded as completely independent. Bishop and Abramowitz (2013) proposed a new independence weighted mean method (IWM) that accounted for inter-model dependence defined using covariance of GCM errors. In this study, we employed IWM to calculate ECIs specific to maize. We calculated six ECIs at each station during the period of 1961–2014 based on the 20 GCMs. The purpose of IWM is to find the linear combination of an ensemble of model simulations that minimizes mean square difference with respect to the observations.

$$\sum_{j=1}^J{\left({\mu}_e^j-{y}^i\right)}^2\kern0.75em \mathrm{where}\kern0.5em {\mu}_e^j={\mathbf{w}}^T{\mathbf{x}}^j=\sum_{k=1}^K{w}_k{x}_k^i$$
(3)

where j = 1, 2, …, J are time steps of annual index values and k = 1, 2, …, K for GCMs (J = 54 and K = 20 in this study); \({\mu}_e^j\kern0.5em\)is the multi-model ensemble of ECIs for the jth time step and yj is the jth time step observed ECI; w = [w1, w2, …, wK]T;\({\mathbf{x}}^j={\left[{x}_1^j,{x}_2^j,\dots {x}_K^j\right]}^T\); wk is the kth model coefficient in the linear combination; and\(\kern0.50em {x}_k^j\) is the ECIs for the jth time step of the kth GCM.

To ensure that \(\sum \limits_{k=1}^KW_k=1\), this constraint term is solved using a Lagrange multiplier (λ).

$$F\left(\mathbf{w},\lambda\ \right)=\frac{1}{2}\left[\frac{1}{J-1}\sum_{j=1}^J{\left({\mu}_e^j-{y}^i\right)}^2\kern0.5em \right]-\lambda \left(\left(\sum_{k=1}^K{w}_k\right)-1\right)$$
(4)

The solution of Eq. 2 is expressed as

$$\mathbf{w}=\frac{{\mathbf{A}}^{-1}\mathbf{1}}{{\mathbf{1}}^T{\mathbf{A}}^{-1}\mathbf{1}}$$
(5)

where 1T = [1, 1, …, 1]; A is the sample-based estimate of the covariance of the bias-corrected errors between all of the ensemble members.

$$\mathbf{A}=\frac{\sum_{j=1}^J\left({\mathbf{x}}^j-{\mathbf{y}}^j\right){\left({\mathbf{x}}^j-{\mathbf{y}}^j\right)}^T}{J-1}$$
(6)

In order to evaluate the changing trend of extreme climate in the future, we mainly investigated extreme climate stress on maize in three 30-year periods base on IWM simulated ECIs. The baseline period was 1981–2010 and two future periods of 2031–2060 (2040s) and 2071–2100 (2080s).

3 Results

3.1 Comparison between downscaled and observed extreme climate indices

The interannual variability of individual GCM simulated and observed values for each ECI across the NCP during the period 1961–2014 were shown in Fig. 2. Both simulated and observed ECIs showed large interannual fluctuations. However, the IWM ensemble results from 20 GCMs had less variability than individual GCM and observation (Fig. 2). The RMSEs and agreement index (d) calculated between individual GCM, multi-model arithmetic mean (AM), and observations for six ECIs were shown in Table 3. For each ECI, RMSE of individual GCM was larger than that of AM and IWM, and the RMSE of IWM was smallest. Moreover, the d values for IWM were significantly higher than that of individual GCM and AM, suggesting that the IWM ensemble results were well agreed with observations. Overall, the ensemble results of IWM could be used to analyze the changes of the six ECIs under future climate scenarios.

Fig. 2
figure 2

Observed (OBS) and simulated extreme climate indices (ECIs) including HD (a), HSI (b), HCD (c), R20 (d), CWD (e) and CDD (f) from individual models and multi-model ensemble (IWM) from 1961 to 2014 across the North China Plain. HD is hot days; HSI is heat stress intensity; HCD is consecutive hot days; R20 is heavy precipitation days; CWD is consecutive wet days; and CDD is consecutive dry days

Full size image
Table 3 The Root-mean-square errors (RMSEs) and agreement index (d, values in parentheses) between each global change model (GCM), the multi-model arithmetic mean (AM), independence weighted mean (IWM), and observed values for the 6 extreme climate indices (ECIs) during 1961–2014
Full size table

3.2 The effect of phenology shift on extreme climate indices

We first applied field trial data about maize phenology across the NCP to calibrate and validate the APSIM-maize model to predict crop phenology changes under future climate scenarios. Overall, the field-observed and model-simulated dates of flowering and maturity agreed well for all investigated stations (Fig. 3). Overall, the RMSEs between simulated and observed dates of flowering and maturity were 3.0 and 3.2 days, respectively; the R2 values for the simulated flowering and maturity dates exceeded 0.72; and the d values exceeded 0.93, suggesting good agreement between the simulated and observed values (Fig. 3). Thus, the simulation results from APSIM-maize model for maize phenology were reliable and therefore applicable.

Fig. 3
figure 3

The validation of the APSIM-Maize model simulated against observed flowering date (a) and maturity date (b) for the investigation stations in the North China Plain. DOY is day of year

Full size image

Future climate change has considerable impacts on the maize phenology in the NCP. Due to the climate warming in the future, the flowering and maturity dates of maize showed significant advance trends across all GCMs (Fig. 4). Moreover, the changes of maize phenology under SSP585_2080s were higher than that under other three future climate scenarios (i.e., SSP245_2040s, SSP245_2080s, and SSP585_2040s) (Fig. 4).

Fig. 4
figure 4

Simulated changes of flowering and maturity dates based on 52 investigated stations across the North China Plain in the 2040s (2031–2060) and the 2080s (2071–2100) under SSP245 and SSP585 compared to the baseline (1981–2010). Box boundaries indicate the 25th and 75th percentiles across 52 stations, whiskers below and above the box indicate the 10th and 90th percentiles. The black lines and crosshairs within each box indicate the multi-model median and mean, respectively

Full size image

Future changes of maize phenology had an important impact on the five ECIs (HD, HIS, HCD, R20, and CDD) except for CWD (Fig. 5). Along with maize phenology was significantly advanced under future climate condition, the increase in the three extreme high-temperature indices (HD, HIS, HCD) when considering future phenology changes was greater than that without considering phenology changes (Fig. 5a–c). However, when considering phenology changes, the increase in R20 under the four climate scenarios was lower than the value without considering phenology change (Fig. 5d). Overall, if the phenology changes of maize were not considered, the increase in extreme high temperature is often underestimated when analyzing the ECIs during the maize growth period in the NCP under future climate scenarios (Fig. 5a–c). However, the increase in extreme precipitation (R20) was overestimated (Fig. 5d). Moreover, the continuous drought (CDD) during the maize growth season was reduced under the future climate scenarios, but this reduction is greatly underestimated if phenology changes were not considered (Fig. 5f). Therefore, it is very important to consider the shift in crop phenology in evaluating the extreme climate during the crop growth period under future climate change scenarios.

Fig. 5
figure 5

Projected changes of multi-model ensemble (IWM) simulated six extreme climate indices (ECIs) including HD (a), HSI (b), HCD (c), R20 (d), CWD (e) and CDD (f) based on 52 investigated stations across the North China Plain in the 2040s (2031–2060) and the 2080s (2071–2100) under SSP245 and SSP585 compared to the baseline (1981–2010). Phe_N and Phe_C indicate that phenological changes are not considered and considered during the calculation of future extreme climate index changes. HD is hot days; HSI is heat stress intensity; HCD is consecutive hot days; R20 is heavy precipitation days; CWD is consecutive wet days; and CDD is consecutive dry days. Box boundaries indicate the 25th and 75th percentiles across 33 Global Change Models (GCMs), whiskers below and above the box indicate the 10th and 90th percentiles. The black lines and crosshairs within each box indicate the multi-model median and mean, respectively

Full size image

3.3 Multi-model ensemble projections of extreme climate indices

Spatial changes in six ECIs simulated by IWM ensemble between baseline period (1981–2010) and two future periods (2040s and 2080s) for two climate scenarios (SSP245 and SSP585) were shown in Fig. 6. We found that the three extreme high temperature indices (HD, HIS, HCD) would mainly increase in the 2040s and 2080s under both SSP245 and SSP585. Moreover, the increase in HD, HIS, and HCD under SSP585 scenario was much larger than that under the SSP245, with the largest increase under the SSP585_2080s (Fig. 6a, b, c). In terms of spatial characteristics, the increases of HD and HCD in the central part of the NCP were greater than that in the south and north parts (Fig. 6a, c), while HIS had no obvious spatial trend (Fig. 6b).

Fig. 6
figure 6

Spatial distribution of changes in multi-model ensemble (IWM) simulated extreme climate indices (ECIs) compared to 1981–2010 under SSP245_2040s (a1–f1), SSP245_2080s (a2–f2), SSP585_2040s (a3–f3), and SSP585_2080s (a4–f4). HD is hot days; HSI is heat stress intensity; HCD is consecutive hot days; R20 is heavy precipitation days; CWD is consecutive wet days; and CDD is consecutive dry days

Full size image

Both R20 and CWD showed significant increase trends under future climate scenarios (Fig. 6d, e). For R20, the increase in the 2080s period under the two scenarios (SSP245 and SSP585) was greater than that in the 2040s (Fig. 6d), while CWD had no significant difference under the two future periods (Fig. 6e). However, CDD showed significant decreasing trends across the NCP under all future climate scenarios (Fig. 6f). The decreasing trends of CDD under the SSP585 scenarios were greater than that of SSP245, and which in the 2080s period were greater than that in the 2040s (Fig. 6f). In addition, the changes of the three indexes (R20, CWD, and CDD) under future climate scenarios have no significant spatial characteristics (Fig. 6d, e, f).

Heat stress showed a consistent increase trend for both scenarios (SSP245 and SSP585) over the twenty-first century (Fig. 7a, b, c). There were no distinguishable differences between the two scenarios (SSP245 and SSP585) for three extreme high temperature indices (HD, HIS, HCD) before 2060. However, after the year of 2060, the gaps between SSP245 and SPP585 for HD, HIS, and HCD were gradually increasing due to the increase rates under SSP585 were larger than that under SSP245 (Fig. 7a, b, c).

Fig. 7
figure 7

Time series of multi-model ensemble (IWM) simulated extreme climate indices (ECIs) including HD (a), HSI (b), HCD (c), R20 (d), CWD (e) and CDD (f) under SSP245 and SSP585 across the NCP. The top and bottom bounds of the shaded area are the maximum and minimum values of 20 Global Change Models (GCMs) simulations. HD is hot days; HSI is heat stress intensity; HCD is consecutive hot days; R20 is heavy precipitation days; CWD is consecutive wet days; and CDD is consecutive dry days

Full size image

R20 and CWD showed slightly increasing trends under both SSP245 and SSP585 scenarios over the twenty-first century (Fig. 7d, e), but CDD (Fig. 7f) showed a slight decreasing trend. Moreover, there was no significant difference for R20, CWD, and CDD between the two scenarios (Fig. 7d, e, f).

3.4 The role of sowing date adjustment on extreme climate indices

Adjustment of maize SDs has important impact on ECIs during maize growth period (Fig. 8). Compared to the current SD (S0), the delay of SD could greatly reduce the extreme high temperature indices (HD, HIS, HCD) under all future climate scenarios. In contrast, the early SD increased the occurrence of extreme high temperature (Fig. 8a, b, c). In addition, the delayed SD (especially S_D30) significantly decreased R20 under all future climate scenarios (SSP245_2040s, SSP245_2080s, SSP585_2040s, and SSP585_2080s) (Fig. 8d). SD adjustment had little effect on CWD (Fig. 8e), but the adjustment of SD had a significant impact on CDD, and the advance or delay of SD would significantly increase CDD (Fig. 8f).

Fig. 8
figure 8

Multi-model ensemble (IWM) simulated six extreme climate indices (ECIs) including HD (a), HSI (b), HCD (c), R20 (d), CWD (e) and CDD (f) for different sowing date based on 52 investigated stations across the North China Plain in the 2040s (2031–2060) and the 2080s (2071–2100) under SSP245 and SSP585. S0 denotes the sowing date did not change under future climate scenarios; S_A30 and S_A15 denote that the sowing date was advanced by 30 days and 15 days, respectively; and S_D15 and S_D30 denotes that the sowing date was delayed by 15 days and 30 days, respectively. HD is hot days; HSI is heat stress intensity; HCD is consecutive hot days; R20 is heavy precipitation days; CWD is consecutive wet days; and CDD is consecutive dry days. Box boundaries indicate the 25th and 75th percentiles across 52 stations, whiskers below and above the box indicate the 10th and 90th percentiles. The black lines and crosshairs within each box indicate the multi-model median and mean, respectively

Full size image

4 Discussion

Generally, crop growth and development are more sensitive to extreme climates than climate change (Lobell et al. 2013). Therefore, it is vital to evaluate the extents of exposure to extreme climate condition for crop under the background of future climate change (Deryng et al. 2014). To some extent, climate warming could accelerate crop development and advance the phenology (Xiao et al. 2016). Our simulation results noted that the maize phenology across the NCP was significantly advanced under all future climate scenarios. However, previous studies often did not consider the changes in crop phenology when calculating the ECIs during the different crop growth period under future climate scenarios (Bai et al. 2021). Therefore, their projected results would overestimate or underestimate the future extreme climate conditions which limited crop growth and development (Bai et al. 2021; He et al. 2018). In this study, we aimed to accurately evaluate the occurrence of extreme climate event during maize growth period in the NCP under the future climate scenario. So, we first used the well-validated crop model (i.e., APSIM-maize model) to simulate the potential changes in maize phenology under future climate change. Furthermore, we compared the changes of ECIs during maize growth period with and without considering maize phenology shift, and we found that the extreme high temperature indices under considering the changes of maize phenology was much larger than that without considering the future phenological changes. With considering phenology changes, the increase in R20 was lower than that without considering phenology change, while the decrease in CDD under phenological changes was much greater than that without phenological changes. Therefore, our results indicated that the consideration of future phenological changes of crops had an important influence on the assessment of future extreme climate conditions during the crop growth period.

In the NCP, the advancement of maize phenology shifted the RGP to a period with warmer temperature (Xiao et al. 2016). Along with the future warming climate (Xiao et al. 2020a), three extreme high temperature indices (HD, HIS, and HCD) during maize RGP significantly increase under both SSP245 and SSP585 in the future. Additionally, the increase rates of extreme high temperature indices under SSP585 were higher than that under RCP245. Generally, the flowering and grain-filling stages of maize are more vulnerable to damage by high temperature stress. For example, a short episode (less than 1 h) of extreme high temperature during the flowering period can lead to pollination failure, thereby reducing kernel number and final yield (Edreira et al. 2014). In general, the occurrence of extreme high temperature during the sensitive stages of maize growth and development (flowering or grain filling period) could directly cause a significant reduction in maize production (Wang et al. 2018; Zhang and Yang 2019). Therefore, heat stress would be a more serious threat to maize production over the NCP under the future climate change scenarios.

Furthermore, increasing extreme precipitation (R20) and consecutive wet days (CWD) in the future would be another threat to maize production in the NCP. Generally, adverse effects of heavy precipitation are often related to waterlogging, which could severely limit crop growth (Wang et al. 2020). In addition, CWD could lead to insufficient light, which negatively affects the photosynthesis of maize and ultimately leads to a reduction in maize production (Xiao and Tao 2016). Moreover, if this CWD occurred in the late stage of maize growth, it would often lead to the lodging of maize, which will affect the grain filling rate and lead to a greater reduction in yield (Li et al. 2019). In the other hand, although the maize planted in the NCP mainly grows in the rainy season, it also could be subjected to drought if there is no or less precipitation for several continuous days under the high temperature in summer (Ming et al. 2015). Therefore, the drought caused by water deficiency and concurrent heat stress may also be a climatic disaster that caused the reduction of maize production in the NCP (Hu et al. 2014). In this study, our results indicate that the number of consecutive dry days (CDD) during the maize growth period show a decreasing trend under future climate scenarios. This result means that the drought stress on maize in the NCP would reduce under future climate scenarios, which will be beneficial to maize production in the future.

The higher risk of exposure to heat stress for maize in the NCP require us to propose special mitigation and adaptation strategies (Lv et al. 2020). Sowing date (SD) adjustment is considered to be an effective adaptive management measure to deal with extreme climate events (Zhang et al. 2019). To a certain extent, due to adjustment of SD can avoid extreme climate stress during the critical reproductive stage, the damage to crops can be mitigated (Zhou et al. 2016). In the NCP, related studies indicated that SD manipulation was a useful approach to alleviate heat stress for maize production (Tian et al. 2019). In this study, we found that the postponing SD could greatly decrease the extreme high temperature indices under future climate scenarios. The main reason is that the delay of maize SD may delay grain filling to the cooler season of late summer or early autumn (Xiao et al. 2020a). Therefore, delaying SD in the case of future climate warming will allow maize to avoid heat stress during RGP (Lv et al. 2020). Furthermore, more effort in breeding technology is crucial. Previous study indicated maize ideotypes under future climate should have a longer reproductive growing period, faster potential grain filling rate, large maximum grain numbers and larger radiation use efficiency (Xiao et al. 2020a). It is important to develop new cultivar that can cope with the increasing frequency and intensity of heat stress around flowering and grain filling. More importantly, heat-resistant cultivars should be widely planted in the study region under future climate change.

Attribute to GCMs could project future extreme climate condition (Wang et al. 2016), we use ensemble results of multi-GCMs from CMIP6 to assess the possible changes of six ECIs that can reflect heat and water stress on maize in the NCP. Despite the certain differences between various GCMs, there was still a well agreement between the GCM simulated data and observed data due to the empirical relationship that statistical downscaling method relied on (Bai et al. 2021). Moreover, since individual GCM cannot represent ECIs well, the reliability of extreme climate prediction can be improved by combining extreme data from multiple models using a weighting approach (Bai et al. 2021; Wang et al. 2016). In this study, we used independence-weighted mean (IWM) to estimate ECIs specific to maize in the NCP. We found that there is a high temporal consistency between the multi-model ensemble (i.e., IWM) and observed ECIs, and the RMSE is small. In addition, due to uncertainty of different GCMs, there may be certain uncertainties in the projection of future extreme climate stress based on individual GCM. Therefore, we minimized uncertainty by applying a diversity of GCMs (20 GCMs from CMIP6) and selecting two emissions scenarios (SSP245 and SSP585).

5 Conclusion

It is expected that climate change has already and will further increase the frequency and severity of extreme climate events. Our study offers a preliminary understanding of projected changes in extreme climate stress on maize production across the NCP based on the statistical downscaling data of 20 GCMs from CMIP6. Our finding indicates that maize production in the NCP would suffer higher frequency of exposure to heat stress under the future climate change scenarios. In addition, heavy precipitation (R20) and consecutive wet days (CWD) showed increasing trends, but consecutive dry days (CDD) showed a slight decreasing trend under both SSP245 and SSP585 over the twenty-first century. The reduction of drought stress under future climate scenarios would be beneficial to maize production in the NCP. Moreover, we found that the delaying of sowing dates could greatly reduce the extreme high temperature and even avoid heat stress under future climate scenarios. Thus, sowing date adjustment should be adopted to deal with the effects of future extreme climate on maize in the NCP.