The field experiment was conducted in 2018 and 2019 at the experimental base (115°09’E, 28°31’N) in Shanggao County, Jiangxi Province, China. Double rice cropping systems consist of early-season rice and late-season rice followed by winter fallow period. Late-season rice is growing from June sowing to November harvesting. The daily average temperature, daily precipitation, and sunshine duration during the late rice seasons in 2018 and 2019 are presented in Figure S1 . The topsoil (0–15 cm) properties before the experiment in 2018 were: pH of 5.5, organic carbon concentration of 20.5 g kg^-1, total nitrogen concentration of 2.0 g kg^-1, alkaline hydrolyzable nitrogen concentration of 190.1 mg kg^-1, available phosphorus concentration of 20.0 mg kg^-1, and available potassium concentration of 65.1 mg kg^-1.

There were two temperature treatments: (1) ambient temperature treatment (ambient) and (2) experimental warming treatment (warming), with three replicates for each treatment with random arrangement. Each plot was 5 m wide and 10 m long. Details of the FATI system were reported in our previous study (Yang et al., 2020). Briefly, one infrared heater (1500 W, 180 cm in length, 20 cm in width) was suspended 75 cm above rice canopy (in the middle of last leaf) in each warming plot for a cultivar. A ‘dummy’ heater with the same size was suspended to imitate the shading effects of the heater. Each infrared heater formed a 1.5 m × 1.8 m sampling area with uniform and reliable warming effects ( Figure S2 ). Warming treatments carried out from transplanting to maturity. The rice canopy temperature of each cultivar was monitored by a digital temperature monitor (ZDR–41, Hangzhou Zeda Electronic Instrument, Hangzhou, China) in 1 h intervals. The daily average temperature of the rice canopy under two temperature treatments are shown in Figure 1 . Compared to ambient treatments, warming increased the daily average temperature of TY, JXN, and YY by 1.9°C, 2.1°C, and 2.1°C over the two years, respectively.

The tested late-season rice cultivars were Taiyou398 (TY), Jiuxiangnian (JXN), and Yongyou1538 (YY), which are popular rice cultivars in Jiangxi Province, China. Taiyou398 is an indica hybrid rice cultivar with a short growth duration of approximately 110 d; JXN is an indica inbred rice cultivar with a long growth duration of approximately 135 d; YY is an indica-japonica hybrid rice cultivar with a long growth duration of approximately 132 d. The sowing, transplanting, heading, and maturity dates of the three rice cultivars are showed in Table S1 . Rice seedlings were manually transplanted at a hill space of 25 cm × 13 cm with three seedlings per hill. Urea (46.0% N concentration), calcium magnesium phosphate (12.0% P₂O₅ concentration), and potassium chloride (60.0% K₂O concentration) were used as nitrogen, phosphorus, and potassium fertilizers, respectively. The application rates of the nitrogen, phosphorus, and potassium fertilizers were 210.0, 46.5, and 156.2 kg ha^-1 in late rice seasons, respectively. Nitrogen fertilizer was applied in three splits, 40% was applied as basal fertilizer (1 day before transplanting), 20% was top-dressed at early tillering (7 days after transplanting), and 40% was top-dressed at panicle initiation. All of the phosphorus fertilizer and 70% potassium fertilizer were applied as basal fertilizers with the remaining potassium fertilizer applied at panicle initiation. The field was kept flooded from transplanting to mid-season drainage, and was then intermittently irrigated until maturity. Other field management measures, including diseases, weeds, and insects were the same as double cropping rice production.

All statistical analyses were carried out using IBM SPSS ver. 24.0. We analyzed the grain yield, yield components, and grain quality by using three-way analyses of variance (ANOVA). The treatments for each cultivar in the same year were compared using Student’s t-test and the differences were determined to be statistically significant when 0.01< P ≤ 0.05 and P ≤ 0.01.

Averaged across cultivars and years, warming significantly reduced grain yield by 2.2% and the spikelet number by 3.1% but had no effect on head rice yield, panicle number, filled grain percentage, and grain weight ( Table 1 ). Grain yield, head rice yield, and yield components varied among the three cultivars. Temperature and cultivar had interactive effects on grain yield, head rice yield, and spikelet number ( Table 1 ). Warming had no effect on the grain yields and spikelet number of TY and JXN but significantly reduced those of YY by 4.9% and 6.5%, respectively ( Figures 2A, C ). The head rice yield of JXN was significantly increased by 5.3%, while that of TY and YY were not changed under warming conditions ( Figure 2B ). In addition, grain yield, head rice yield and spikelet number varied with experimental years. Grain yield and spikelet number were higher and head rice yield was lower in 2018 than those in 2019 ( Table 1 ).

Compared to ambient treatment, warming significantly increased head rice rate and the chalky grain rate by 4.3% and 23.8%, respectively ( Table 2 ). Head rice rate and chalky grain rate varied among cultivars and experimental years, and significant interactions between temperature and cultivar were observed to affect head rice rate and chalky grain rate ( Table 2 ). Warming significantly increased the head rice rate by 6.6% of JXN and by 7.8% of YY, whereas no significant effect was observed on TY ( Figure 3A ). The chalky grain rates increased significantly by 79.1%, 21.6%, and 7.6% of TY, JXN, and YY under warming conditions, respectively ( Figure 3B ). Furthermore, head rice rate was lower and chalky grain rate was higher in 2018 than in 2019 ( Table 2 ).

Amylose content and setback were decreased by 6.2% and 41.5%, respectively, under warming conditions relative to the ambient treatment, whereas peak viscosity and breakdown were not significantly affected by warming ( Table 2 ). Amylose content and RVA parameters varied with cultivars and experimental years. Significant interactive effect of temperature and cultivar on amylose content was observed, and on RVA parameters was not found. As shown in Figure 3C , warming had no significant effect on amylose content of TY but significantly decreased that by 7.5% and 8.8% of JXN and YY, respectively. Compared to those in 2019, amylose content and breakdown were higher and peak viscosity and setback were lower in 2018 ( Table 2 ).

The grain yields in response to warming varied strongly among the late-season rice cultivars ( Table 1 ). Differently, previous studies found that warming by using FATI facilities significantly reduced the grain yields for both indica rice (Rehmani et al., 2014) and japonica rice (Cai et al., 2016; Wang et al., 2016; Wang et al., 2018a; Wang et al., 2018b) in middle rice cropping systems. The negative effects of warming on grain yields were due to the decrease in panicle number (Wang et al., 2018b), spikelet number, filled grain percentage, and grain weight (Rehmani et al., 2014; Cai et al., 2016). In this study, the grain yields of indica rice (TY and JXN) did not change under warming conditions, which was attributed to the minor changes in yield components, while warming reduced the spikelet number of indica-japonica rice (YY), resulting in a decrease in its grain yield ( Figure 2A ).

The impact of warming on grain yield formation depends on ambient air temperature (Chen et al., 2017; Chen et al., 2020). The optimal temperature for rice tillering is approximately 28.4°C, ranging from 16.4°C to 35.3°C; for rice anthesis is approximately 26.3°C, ranging from 16.2°C to 37.0°C; and for grain-filling is approximately 24.2°C, ranging from 20.7°C to 31.3°C (Sánchez et al., 2014). In this study, the canopy average temperature of ambient treatments was 31.6°C to 32.1°C from transplanting to panicle initiation, 26.8°C to 29.4°C from initial heading to full heading, and 20.5°C to 26.4°C from full heading to maturity ( Table S2 ). Thus, experimental warming (1.9–2.0°C) did not exceed the upper temperatures range limit for the three late-season rice cultivars and had little effects on their panicle number, filled grain percentage, and grain weight.

The spikelet number in response to warming were different between indica rice (TY and JXN) and indica-japonica rice (YY) in this study ( Table 1 ). Compared to indica rice cultivars, the growth of indica-japonica rice cultivars is obviously sensitive to temperature changes of in late rice seasons (Yin et al., 2021). In addition, we found that warming was not conducive to the dry matter accumulation at the heading stage of indica-japonica rice, especially the dry matter accumulation of the panicle ( Figure S3 ). Therefore, compared to indica rice cultivars (TY and JXN), indica-japonica hybrid rice (YY) might have poor heat resistance in this study. Warming might adversely affect the panicle differentiation or increase the ratio of spikelet degeneration and led to the decrease of spikelet number of YY ( Figure 2C ). In addition, the canopy average temperature from panicle initiation to initial heading was 30.5°C in 2018 and 31.1°C in 2019 ( Table S2 ). The higher temperature might be may be detrimental to panicle differentiation (Sánchez et al., 2014), resulting in lower spikelet number and grain yield in 2019. Consequently, climate warming will make temperatures unfavorable for indica-japonica rice (YY) production in the future. It is important to study the direct effects of warming on spikelet number and find countermeasures to prevent the yield reduction of late-season indica-japonica rice under future climate warming conditions.

High temperature on rice production focus on the grain yield and fail to account for the detrimental impact of high temperature on milling quality, which ultimately determine the head (edible) rice yield and market value. Lyman et al. (2013) found that warming (1.0°C) during growing season significantly reduced grain yield and further reduced head rice yield due to the increased percentages of chalkiness and broken kernels. Different with grain yields, the head rice yields of TY and YY were not changed under warming conditions, while the head rice yield of JXN was increased ( Figure 2B ). As discussed below, under warming conditions, the increase in the head rice rates of JXN and YY resulted in an increase in head rice yield of JXN and compensated for the reduction in head rice yield of YY ( Figure 3A ). In total, our results highlighted that the importance of planting long growth duration late-season rice cultivars (e.g., JXN and YY) in double rice cropping systems to maintain or increase the head rice yield and bolster food security under future climate warming conditions.

Generally, no matter what kinds of temperature increase facilities are used, experimental warming or high temperature during grain-filling stage always increases chalky grain rate and chalkiness and worsens rice appearance quality (Lanning et al., 2011; Xiong et al., 2017; Wang et al., 2021). In this study, this phenomenon was also confirmed by using three late-season rice cultivars under field warming conditions ( Table 2 ). There were several reasons for the cause of warming-induced grain chalkiness in this study. First, in the initial stage of grain-filling, high temperature accelerates the grain-filling rate, resulting in the loose packing of amyloplast (Dou et al., 2017). Second, high temperature-induced α-amylase is involved in the degradation of endosperm starch granules, forming hydrolysis traces, such as small pits, leading to an increase in chalkiness (Hakata et al., 2012). Third, warming tends to increase protein and amino acid contents and disturbs the accumulation balance of storage protein, amino acid, and starch, which coordinately controls the formation of grain chalkiness (Dou et al., 2018; Tang et al., 2019; Wang et al., 2021). Compared to short growth duration rice cultivar (TY), warming had less effects on the chalky grain rates of long growth duration rice cultivars (JXN and YY), might be due to their lower ripening temperature from full heading to maturity ( Table S2 ) and longer grain-filling stage (53–58 d, Table S1 ), which were beneficial to alleviate the adverse effects of warming on grain-filling. Furthermore, the higher chalky grain rate in 2018 might be attributed to the increased daily precipitation and decreased sunshine duration from heading to maturity ( Figure S1 ).

Rice is consumed primarily as intact kernels, and the head rice rate primarily determines market prices and producer revenues. Our results showed that head rice rate was significantly increased by warming but varied among the three late-season rice cultivars ( Table 2 ; Figure 3A ), which were different with previous reports (Rehmani et al., 2014; Jing et al., 2016; Dou et al., 2018; Tang et al., 2019). In general, high ripening temperature decreases head rice rate due to the increase in chalky kernel percentage (Wang et al., 2015; Siddik et al., 2019). The different observations in this study might be caused by the type and location of chalk formation, which is crucial to the degree of kernel breakage during milling (Lyman et al., 2013). According to Lyman et al. (2013), high temperature corresponding to the formation of white back or basal white chalky kernels did not result in an equivalent reduction in head rice rate. The optimal temperature for grain-filling is approximately 24.2°C (Sánchez et al., 2014). In the present study, the ambient temperature in late grain-filling stages were very low ( Figure 1 ). Therefore, warming (1.9–2.0°C) might be beneficial to the grain-filling and promote the maturity of inferior grains, especially for JXN and YY, reducing the broken kernel percentage during processing (Dou et al., 2017; Dou et al., 2018). In this study, we concluded that the chalk formed under warming conditions did not directly lead to the kernel breakage but would reduce the market value of head rice due to an increase in chalky grain rate. Consequently, planting long growth duration indica inbred rice cultivar (JXN) and indica-japonica hybrid rice cultivar (YY) in double rice cropping systems could improve the milling quality and reduce the deterioration of appearance quality under warming conditions.

It is generally acknowledged that the lower the amylose content, the better the taste of cooked rice (Chen et al., 2021). Our results indicated that warming decreased the amylose contents in head rice ( Table 2 ), which were in agreement with previous studies (Chun et al., 2015; Dou et al., 2018; Tang et al., 2019). Amylose in rice endosperm is synthesized by granule-bound starch synthase (GBSS), encoded by the Waxy gene (Huang et al., 2021). High temperature-induced the decrease in amylose content is mainly attributed to the reduction of GBSS activity (Ahmed et al., 2015; Cao et al., 2015). Besides, we noted that the amylose content of TY in response to experimental warming was relatively smaller, which might due to a high heat resistance of short growth duration indica hybrid rice ( Figure 3C ).

The RVA parameters of rice flour are useful tools to assess the eating quality of cooked rice (Chen et al., 2021). Setback from RVA, also known as short-term retrogradation of rice flour, is positively correlated with the hardness of cooked rice (Li et al., 2020). In general, cooked rice with better taste quality always has higher values of peak viscosity and breakdown and lower setback value (Chen et al., 2021). Therefore, warming was expected to improve the rice eating quality, as it did not affect the peak viscosity and breakdown but significantly decreased the setback, especially for long growth duration rice cultivars (JXN and YY) ( Table 2 ). Amylose content is an important factor in determining retrogradation characteristic because of the fast recrystallization rate of amylose molecules (Li et al., 2020). Thus, the lower setback values of rice flour under warming conditions might attribute to the decrease in amylose contents in the present study. Besides amylose content, other chemical compositions such as protein also negatively affect rice eating quality (Chen et al., 2021). Previous studies have reported that protein content in milled rice was enhanced under warming conditions (Dou et al., 2017; Dou et al., 2018; Yang et al., 2022b). Although the amylose content and RVA parameters indicated that warming improved the eating quality of late-season rice, the increase in protein content may reduce the eating quality. Thus, the actual effect of warming on the eating quality of late-season rice in the future should be evaluated by human tasting (Kim et al., 2017). Moreover, the fine structure of rice starch also affects the pasting property and eating quality (Tao et al., 2018; Chen et al., 2021), and its mechanism needs to be further studied under warming conditions.

In short, experimental warming has adverse effects on the grain yield, milling and appearance quality of late-season rice. Selecting suitable rice cultivars can compensate the adverse effects of warming on grain yield and quality of late-season rice. Other climate-smart strategies including innovative crop management and breeding should be developed to alleviate the negative impact of warming on rice production (Xu et al., 2021). For example, previous studies have shown that optimizing the rate and timing of mineral nitrogen fertilizer application, can also mitigate the adverse effects of warming on grain yield and quality (Tang et al., 2019; Yang et al., 2022a).