Phenology can intuitively indicate changes in seasons and climate (Zheng et al., 2012; He et al., 2015), thus being an important biological indicator for measuring regional climate change. According to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC, 2014), the average global surface temperature has increased by 0.89°C from 1901 to 2012. An increase in air temperature can promote enzyme activity and accelerate the phenological processes of plants, thus shortening plant growing seasons. In turn, changes in the length of plant growing seasons can modify vegetation productivity and structural composition, as well as water, heat, and carbon exchange rates in the soil-vegetation-atmosphere system, thereby affecting the climate system and exacerbating climate change (Wang et al., 2015; Cui et al., 2018). As agricultural production is directly affected by climate change (Guo, 2015), studying the patterns of phenological changes in crops is of great importance for guiding regional agricultural production and mitigating the negative impacts of climate change.

Under the scenario of climate change, maize phenology has been examined worldwide. Based on the Agro-IBIS model (a process-based ecosystem model adapted from the Integrated Biosphere Simulator), Sacks and Kucharik (2011) found that the sowing date and maturity date of maize advanced 0.40 day per year and delayed 0.10 day per year, respectively, between 1981 and 2005 in the US Corn Belt. Chinese scholars have found that, due to climate warming, the tasseling date and maturity date at more than 80% of the country’s maize production sites advanced by an average of 3.2 days and 6.0 days, respectively, over each 10-year period (Tao et al., 2014). In northeast China, the sowing date of maize advanced and the maturity date delayed by 4-21 days from 1981 to 2007 (Liu et al., 2013). In Inner Mongolia, the sowing date and maturity date of maize advanced by 1.0 day and delayed by 3.3 days per 10-year period, respectively, from 1981 to 2009 (Xiao, 2015). Fu et al. (2016) conducted temperature increment experiments in Tibet Autonomous Region and found that air temperature increments generally led to advancing trends in the phenological phases of maize to different degrees. Under the influence of increasing average air temperature, decreased precipitation with increased latitude, and substantial decline in sunshine duration, the seeding, jointing, and silking dates of summer maize in Jing-Jin-Ji (Beijing-Tianjin-Hebei) region and Shandong Province were significantly delayed, and the seeding, jointing, and silking dates of summer maize in Henan Province advanced considerably (Meng et al., 2015).

Based on the above results, previous studies mainly focused on the effect of air temperature increase on maize phenological phases in the context of climate change. In addition to temperature, moisture and sunshine duration can also affect crop growth (Liu et al., 2018a). For example, shortened sunshine duration can promote flowering of short-day plants and advance their flowering stage, while extended sunshine duration delays the flowering stage. Zheng et al. (2002) found that the main factor influencing flowering in winter and early spring plants is air temperature, whereas the flowering of late spring and early summer plants is simultaneously affected by air temperature and sunshine duration; the onset of pollen spreading in late summer plants is mainly affected by sunshine duration. Under certain sunshine conditions, changes in air relative humidity can also cause changes in plant phenology (Wang et al., 2010). In addition, limited by observation data and model simulation capabilities, studies conducted so far have mostly focused on individual key phenological phases, such as seeding, tasseling, and ripening dates, and lack a systematic analysis on the changing patterns of multiple phenological phases during crop growth. Moreover, due to the influence of local and regional climates, the same phenological stages of the same crop significantly differ among different regions, and most studies have been conducted at the regional scale or include only a few sites. For example, Zhao et al. (2015) explored the effects of sowing date adjustment and species change on maize growing season and yield in five maize production sites in northeast China (two in Heilongjiang, two in Jilin, and one in Liaoning). Wang et al. (2015) selected three to four representative sites in northeast, north, and southwest China to examine the uncertain factors in model simulation of maize phenology. The growth process of maize is influenced by multiple climate factors, but changes in the patterns of multiple consecutive phenological stages, especially changes in the trends of spatial differentiation patterns for each phenological stage during maize growth, hence further examination is still needed.

Therefore, based on long-term observation data (1981 to 2010) from 114 agro-meteorological stations in the four major maize growing zones in China, this study aims to: (1) quantify spatial and temporal changes occurring in the differentiation characteristics of maize regarding sowing, seedling, three-leaf, seven-leaf, jointing, tasseling, milk-ripe, and ripening dates; and (2) examine the changes in the patterns of the phenological phases and length of growing seasons for different planting types of maize in different regions in the context of changes in key climate factors (average air temperature, precipitation, and sunshine duration). The results obtained here could provide a theoretical reference for regional agricultural production practices.

This study examined spring, summer, and spring-summer maize grown in the northwest maize zone, northern spring maize zone, Huang-Huai spring-summer maize zone, and southwest maize zone, which are the four main maize production zones in China, based on the Chinese maize planting zonations (Tong, 1992). The sowing date of spring maize is from late April to early May, and that of summer maize is from early June to mid-July. Spring-summer maize is sowed at intervals between April and June or between May and June, and its sowing date is mainly affected by the planting time of the previous seasonal crops.

The research sites selected for the present study met the following four criteria: (1) typical site and representative of the agricultural production status of local maize; (2) with long-term (1981-2010) maize phenological observation data; (3) with meteorological observation station; and (4) with field management records, including information on maize variety, irrigation, and fertilization, among other features. According to data quality, 114 sites with complete observation records were selected (Figure 1). These sites were distributed in the northwest maize zone (17 sites), northern spring maize zone (54 sites), Huang-Huai spring-summer maize zone (32 sites), and southwest maize zone (11 sites). Maize growth season was divided into three key growth stages: vegetative growth stage (sowing to tasseling date), reproductive growth stage (tasseling to ripening date), and the entire growth stage (sowing to ripening date). The lengths of these three growth stages were calculated separately. Climate data for the 114 sites during the 1981-2010 period were obtained from the data sharing website (http://data.cma.cn/site/index.html).

2.2.1 Trend analysis

The trends observed for changes in the length of phenological stages and growth stages of maize were calculated using the following linear regression equation, considering year as the independent variable (1):

${\theta }_k=\frac{n\times \sum _{i=1}^n(k\times P_{i,k})-\sum _{i=1}^{n}k\times \sum _{i=1}^n{P}_{i,k}}{n\times \sum _{i=1}^n{k}^2-{\left(\sum _{i=1}^{n}k\right)}^2}$ (1)

where n is the number of years in the analyzed period, P_i,k is the length of phenological stage k in year i (DOY: Day of Year) or growth stage length (days) of the observation site in year i, i represents the ith year, and θ_k is the trend observed for changes in the length of phenological stage or growth stage (d/a). If θ_k is greater than 0, it means that the length of the phenological stage or growth stage is delayed or extended, respectively; for θ_k lower than 0, the length of the phenological stage or growth stage is advanced or shortened, respectively. A two-tailed t-test was used to evaluate the significance of the regression coefficients.

2.2.2 Calculation of growing degree-days

Growing degree-days (GDD) is a heat index for analyzing crop growth. In the present study, it refers to the effective GDD with a daily average air temperature ≥10°C. It was calculated according to McMaster and Wilhelm’s (1997) formula (2):

$GDD=\sum _{d_s}^{d_e}[\left(\frac{T_{max}+T_{min}}{2}\right)-T_{base}]$ (2)

where d_s represents the starting date of the research phase, d_e represents the ending date of the research phase, T_max represents the daily maximum air temperature, T_min represents the daily minimum air temperature, and T_base represents the minimum biological temperature in the maize growth stage.

Based on the observed dates for phenological phases, the starting date (i.e., the date of sowing) and the ending date (i.e. the date of ripening) of maize growth were determined. Daily meteorological data were used to calculate average air temperature, precipitation, sunshine duration, and GDD during each maize growth stage at each site. Trends in average air temperature, precipitation, sunshine duration, and GDD during maize growth stages in 1981-2010 were also calculated using equation (1).

The spatial distribution of the trends in average air temperature, precipitation, sunshine duration, and GDD changes during maize growing season at each selected site from 1981 to 2010 is shown in Figure 2. Throughout the country, average air temperature and GDD increased from 1981 to 2010 by 0.03°C/yr and 4.92°C d/yr, respectively. Precipitation and sunshine duration showed decreasing trends, and the decrease rates were -0.16 mm/yr and -2.07 h/yr, respectively (Table 1). Average air temperature and GDD showed increasing trends at 90% and 83% of the selected sites, respectively, and the increase in air temperature was greater at high than at low latitudes.

Precipitation showed obvious regional differences (Figure 2b). In most zones, precipitation showed a decreasing trend (55%), especially in northern spring maize zone, northwest maize zone, and southwest maize zone; however, precipitation in the Huang-Huai spring-summer maize zone showed an increasing trend. In terms of planting type, precipitation decreased by 1.30 mm/yr during the spring breeding growth stage of maize, while during the growth stage of summer maize and spring-summer maize it increased by 2.49 mm/yr and 1.44 mm/yr, respectively. Sunshine duration decreased by 63% of the selected sites, but it was most significant in the Huang-Huai spring-summer maize zone, showing a decrease of -7.30 h/yr (Figure 2c and Table 1); on the contrary, sunshine duration in the northern spring maize zone showed an increasing trend of 0.51 h/yr (Table 1). On average, sunshine duration during the growing season of spring maize was 0.31 h/yr, and -5.37 h/yr and -10.03 h/yr during the growing season of summer and spring-summer maize, respectively.

The phenological stages of maize were mainly delayed nationwide, at 68%, 54%, 53%, 54%, 55%, 74%, and 77% of the sites showing a delaying trend in the sowing, seedling, three-leaf, jointing, tasseling, milk-ripe, and ripening dates, respectively. However, 54% of the sites showed an advancing trend in the seven-leaf stage. Sites with delayed sowing, seedling, three-leaf, jointing, and tasseling dates were mainly distributed across the Huang-Huai spring-summer maize zone, although there were some sites in the eastern part of the northern spring maize zone and in the northwest maize zone. Sites where the seven-leaf stage tended to advance were mainly located in the southeast of the northern spring maize zone (Figure 3d).

There was a significant inter-regional difference in the changes observed in maize phenological stages (Figure 3 and Table 2). Within the same zone, the phenological phases were significantly different for different maize planting types. In the northwest maize zone, spring maize showed an advancing trend from the sowing to the jointing date (0.05<θ<0.24 d/yr), while the tasseling, milk-ripe, and ripening dates were delayed by 0.02, 0.12, and 0.06 d/yr; all phenological phases of summer maize, from sowing to ripening date, were delayed (0.21<θ<0.87 d/yr). All phenological phases of spring maize in the southwest maize zone were advanced (0.13<θ<0.53 d/yr). In contrast, the phenological phases of spring maize in northern China were mainly delayed (0.01<θ<0.30 d/yr), as only the seven-leaf stage was advanced (-0.14 d/yr). In the Huang-Huai spring-summer maize zone, the phenological phases of summer maize were delayed from the sowing to the ripening date (0.09<θ<0.35 d/yr). The phenological phases of spring-summer maize were also delayed, and sowing data had a significant delay (0.96 d/yr); the sowing date of spring-summer maize was significantly delayed, resulting in the shortening of the entire growth stage. Due to the implementation of the intermittent growth system comprising spring maize, winter wheat, and summer maize in this zone, the sowing date of maize was affected not only by climate factors but also by the planting and harvesting time of the previous seasonal crop.

Different growing zones also presented significant differences in the direction and magnitude of the trends observed for changes in the phenological phases of maize of the same planting type. For instance, the phenological phases of spring maize were mainly advanced in the southwest maize zone, but mainly delayed in the northern spring maize zone. The delay in the phenological phases in the northwest maize zone (0.21<θ<0.87 d/yr) was significantly greater than the delay of the phenological phases of summer maize in the Huang-Huai spring-summer maize zone (0.09<θ<0.35 d/yr).

Maize vegetative growth, reproductive growth, and entire growth stages were extended at 34%, 71%, and 57% of the selected sites, respectively, with extension ranges of 0.08-0.76, 0.61-0.97, and 0.03-1.07 d/yr, respectively. On average, the vegetative growth stage of spring maize was shortened by 0.06 d/yr. Reproductive growth and entire growth stages were extended by 0.18 d/yr and 0.16 d/yr, respectively. Spring maize in the northwest maize zone and southwest maize zone showed an extension trend at the three growth stages. The vegetative growth stage of spring maize in the northern spring maize zone was shortened, while reproductive growth and entire growth stages were mainly extended (Table 2). The vegetative growth and entire growth stages of summer maize were shortened by 0.20 d/yr and 0.05 d/yr, respectively. The reproductive growth stage was extended by 0.14 d/yr. The three growth stages of summer maize were shortened in the northwest maize zone, while vegetative growth stage was shortened in the Huang-Huai spring-summer maize zone; reproductive and entire growth stages of this maize type were extended in the Huang-Huai spring-summer maize zone (Table 2). The reproductive growth stage of spring-summer maize in the Huang-Huai spring-summer maize zone was extended (0.14 d/yr), but vegetative and entire growth stages were shortened by 0.83 d/yr and 0.60 d/yr, respectively (Figure 4).

From 1981 to 2010, the average air temperature increased during the growth stage of spring, summer, and spring-summer maize in China (0.03, 0.02, and 0.05 °C/yr, respectively). In general, elevated air temperatures affect all stages of crop growth and development, including seedling, flowering, and ripening (Liu et al., 2010; Hou et al., 2014; Liu et al., 2017). For progressing from one growth stage to the next, crops need to accumulate enough effective accumulated temperature. Thus, an increase in air temperature can accelerate the accumulation of effective accumulated temperature and shorten each growth stage (Zhang et al., 2013; Zhang et al., 2018), leading to earlier phenological phases or shorter growth stages (Vitasse et al., 2011; Liu et al., 2018b). Precipitation and sunshine duration can also affect crop growth. According to our results, except for the slight decrease in average air temperature (-0.001 °C/yr) during the maize growth stage in the southwest maize zone, average air temperature and GDD in the four maize zones showed an increasing trend, whereas precipitation and sunshine duration showed mainly increasing and decreasing trends, respectively. The phenological phases of spring maize in China mainly advanced in the past 30 years. For example, phenological phases from the sowing date to the jointing date were advanced (0.05<θ<0.24 d/yr) in the northwest maize zone. The vegetative, reproductive, and entire growth stages in this zone were extended by 0.02, 0.03, and 0.10 d/yr, respectively, which is consistent with the findings of Xiao et al. (2015). The phenological stages of spring maize from seeding to maturation were advanced in the southwest maize zone (0.13<θ<0.53 d/yr), however, in the northern spring maize zone, the seven-leaf date showed an advancing trend (-0.14 d/yr); sowing, seedling, and ripening dates were delayed by 0.18, 0.04, and 0.30 d/yr, respectively. This is consistent with the findings of Li et al. (2009). From 1961 to 2010, the air temperature during the maize breeding growth stage in northeast farming areas increased significantly (Yin et al., 2015). Since 1991, the precipitation in the maize growing season (April-September) continued to decrease in northeast China. Since 1971, the accumulated temperature ≥10°C in this area has increased by 262.8°C, and the accumulated temperature ≥10°C (at 2700°C for example) has shifted about 200-300 km to the north and 50-150 km to the east (Ji et al., 2012). The northeast region is dominated by the planting of spring maize, and changes in climate factors during the maize growth stage are similar to the spatial variations of climatic factors observed for northeast spring maize in the northern spring maize zone. Accordingly, the three provinces in northeast China showed an advanced seedling date (0.02<θ<0.15 d/yr) and a delayed ripening date (0.18<θ<0.38 d/yr) for spring maize from 1990 to 2009 (Li et al., 2013). Liu et al. (2013) also reported that the sowing date in northeast China was advanced from 1981 to 2007, while the ripening date was delayed. Nationwide, the vegetative growth stage was shortened (-0.06 d/yr), while the reproductive and entire growth stages of spring maize were extended (by 0.18 and 0.16 d/yr, respectively). Summer maize and spring-summer maize showed a delaying trend in the phenological stages until the ripening stage. Whereas, due to the influence of local regional climate, the differences among regions were significant (Figure 3 and Table 2). The growth stages of summer maize in the northwest maize zone and spring-summer maize in Huang-Huai spring-summer maize zone were shortened (-0.65 and -0.60 d/yr, respectively), while the growth stage of summer maize in Huang-Huai spring-summer maize zone was extended (0.14 d/yr). Previous studies have found that during the growth stage of the summer maize in the North China Plain, air temperature decreased with latitude, sunshine duration declined significantly, and the number of days in the entire growth stage increased significantly from 1981 to 2009 (Meng et al., 2015). Our results for the changes in sunshine duration in the growth stage in Huang-Huai spring-summer maize zone are consistent with these findings. Therefore, the decrease in sunshine duration might be an important factor for extending maize growth season.

Many studies have examined spatio-temporal differences in changes in maize phenology and in the duration of each growth stage in China under different climate change scenarios. However, the effects of human activity such as cultivar shifts, use of fertilizers, and irrigation on crop phenology still need further investigation. Thus, combined crop models including the effect of climate changes, management measures, and other variables that might affect crop growth, should be considered in future studies. Changes in the length of each growth stage of maize are caused by changes in the corresponding phenological phases; therefore, discussing changes in maize phenological stages and growth stages at different scales, such as site, regional, and inter-regional levels, and including different planting types could help gain insight into the effects of climate change on crop phenology.

(1) During the past three decades (1981-2010), there were significant changes in maize phenological phases of China. While the phenological phases of spring maize were mainly advanced, those of summer maize showed a delaying trend in different regions. Specifically, sowing to jointing dates of spring maize in the northwest maize zone were advanced (0.05<θ<0.24 d/yr), as well as sowing to ripening dates in the southwest maize zone (0.13<θ<0.53 d/yr). Spring maize in the northern spring maize zone only advanced at the seven-leaf stage (0.14 d/yr). The delayed range of summer maize phenological phases in the northwest maize zone (0.21<θ<0.87 d/yr) was generally greater than in the Huang-Huai spring-summer maize zone (0.09<θ<0.35 d/yr).

(2) All phenological stages of spring and summer maize in Huang-Huai spring-summer maize zone showed a delaying trend (0.12<θ<0.96 d/yr). The changes in the phenological phases resulted in changes in the length of the corresponding growth stage. The vegetative growth stage of spring maize showed a shortening trend (-0.06 d/yr), and the reproductive growth and entire growth stages showed an extending trend (0.18 and 0.16 d/yr, respectively).

(3) The vegetative growth (-0.20/-0.83 d/yr) and entire growth (-0.05/-0.60 d/yr) stages of summer and spring-summer maize were shortened, while their reproductive growth stages were extended (0.14/0.14 d/yr). In general, the phenological phases of spring maize in China were advanced, while the phenological phases of summer and spring-summer were delayed; the length of each growth stage was mainly extended in different regions.