Impact of Droughts on Winter Wheat Yield in Different Growth Stages during 2001–2016 in Eastern China

Remote sensing data are often used to monitor vegetation growth due to their large observation area, long observation time, and near real-time monitoring performance (Peters et al. 2002; Abduwasit et al. 2007). The remote sensing data for the computation of the drought severity index (DSI) include MOD09A1 V006 and MOD16A2 V006 downloaded from the Land Processes Distributed Active Archive Center (LP DAAC)¹ with a temporal resolution of 8 days and a spatial resolution of 500 m. The MOD09A1 data set covers the period from 1 February 2000 to the present, of which data for the 166th and 177th days are missing in 2001; the MOD16A2 data set covers the period from 1 January 2001 to the present, of which data for the 177th day in 2001 and 10 days of data in 2008 are missing. Considering the continuity and consistency of the data, images from 2001 to 2016 were analyzed in this study, and a few missing data were interpolated using long-term mean values. After the preprocessing procedure on the MOD09A1 dataset, the NDVI for each grid can be obtained as:where R_nir is the near infrared reflectivity and R_red is the infrared reflectivity. The total winter wheat yield for the five provinces of the study region during 2001–2016 was retrieved from the China Statistical Yearbook 2002–2017 released by the National Bureau of Statistics.² The winter wheat yields of most prefecture-level cities in the study area were obtained from the statistical yearbooks of Hebei, Henan, Shandong, and Anhui Provinces, but the data for the prefecture-level cities of Jiangsu Province were collected from the statistical yearbooks of each individual city. Data of the exact growing and developing periods for winter wheat were obtained from the agricultural database of the Division of Plantation Management of the Ministry of Agriculture, China.³ Growing and developing of winter wheat is divided into different stages, such as seeding, sprouting, tillering, overwintering, green, jointing, heading, flowering, filling, and maturity, and the specific classification is shown in Table 1.

Mu et al. (2013) proposed the drought severity index (DSI) based on the NDVI and evapotranspiration computed from remote sensing data. It was corroborated that the DSI performed well in drought monitoring practice in southwestern China (Zhang and Yamaguchi 2014). By exploring the performance of the DSI in northern China, Zhang et al. (2016) found it can accurately reflect drought in this region as well.

Computation of potential evapotranspiration (PET) based on MOD16A2 was done via the Penman–Monteith equation and then ET/PET can be obtained. Computation of the DSI was done by the following procedure: standardization of the NDVI and ET/PET; summation of the standardized NDVI and ET/PET to obtain the Z value; and then standardization of the Z value to obtain the DSI. The algorithms are as follows: where \(Z_{\text{ET/PET}}\) denotes the standardized ET/PET; \(R_{\text{ET/PET}}\) denotes the ET/PET for a certain time interval in a single year during 2001–2016; \(\bar{R}_{\text{ET/PET}}\) denotes the mean ET/PET for the same time interval during the entire 2001–2016 period; \(\sigma_{\text{ET/PET}}\) denotes the standard deviation of ET/PET for the time interval during the 2001–2016 period; \(Z_{\text{NDVI}}\) denotes the standardized NDVI; NDVI denotes the NDVI for the same time interval in the same year during 2001–2016; \(\overline{\text{NDVI}}\) denotes the mean NDVI for the time interval during the entire 2001–2016 period; \(\sigma_{\text{NDVI}}\) denotes the standard deviation of the NDVI for the time interval during the 2001–2016 period; \(\bar{Z}\) denotes the mean Z value for the time interval of the 2001–2016 period; and \(\sigma_{Z}\) denotes the standard deviation of the Z value for the time interval during the 2001–2016 period. Mu et al. (2013) proposed the DSI and classified droughts into different drought intensities—the specific classification is shown in Table 2.

Crop yields are generally composed of trend, climatic fluctuation, and random components. Trend in yields reflects the contribution of factors that affect productivity development while the climatic fluctuation component of yield is the contribution of climate fluctuation, which is mainly associated with meteorological disasters. To investigate the relationship between drought intensity and winter wheat yield, a critical step is to separate the trend from the impact of climatic fluctuation. Therefore, four fitting methods—triple point moving average, exponential smoothing, Logistic, and HP filter—were compared in this study for the separation of trend in yield. The results indicate that the HP filter method has higher accuracy and can be recommended for this study. Therefore, the HP filter method was used for the detrending treatment of winter wheat yield, and the specific calculation algorithm can be found in the relevant literature (Harvey and Trimbur 2008).

The Pearson correlation coefficient is a measure of the linear correlation between two variables (Pearson 1895). According to the DSI values, crop areas affected by droughts of different intensities in each prefecture-level city were extracted, and the correlation coefficient between drought-affected crop area (by drought intensity) and winter wheat yield (detrended) was calculated. The correlation coefficient was used to show the impact of droughts on winter wheat yields in different growth stages.

Winter wheat yield in the five study provinces is in negative correlation with drought intensities. Lower drought intensities roughly link with higher winter wheat yield. However, decreasing drought intensities do not necessarily imply decreasing frequency of droughts across the study region—severe droughts with short durations can still be observed in smaller areas. Both 2001 and 2002, for example, were characterized by severe droughts in the study region. Based on an 8-day averaged DSI, annual dry spells can be observed during 2001 and 2002 and the winter wheat yield was also lower than the long-term annual average winter wheat yield. Severe drought conditions continued until January–March 2003, followed by shifts between dry and humid conditions thereafter. Winter wheat yield in 2003 was also lower than the long-term annual average. Related studies (Wang et al. 2016, 2018; Zheng et al. 2017) have shown that, although the overall regional drought strength gradually decreased, the frequency of regional and periodic drought significantly increased, and severe droughts in the region often result in crop failure. The period of 2004–2006 witnessed lower intensity and shorter duration droughts and the winter wheat production during that period persistently increased to the long-term annual mean level. From 2007 onwards, with the exception of the 2009 winter season to the summer of 2010 and 2011, a positive DSI dominated and the study area was relatively humid. Winter wheat yield also shifted from negative to positive anomalies. Yields were higher than the average and persistently increased from 1 year to another. The period from winter 2009 to summer 2010 was characterized by dry climate conditions and was also the growing season of winter wheat. However, the winter wheat yield in 2010 was slightly increased. Because irrigation was widely practiced in the drought-affected areas, the negative impacts of the droughts were reduced. Figure 2 shows the remarkable correlation or linkages between the DSI and winter wheat yield, and further investigation of the impact of droughts on winter wheat yield during specific critical growth stages will shed light on the mechanism of impact.

The spatiotemporal evolution of droughts with different drought intensities were investigated to evaluate the impact of droughts on crops during specific critical growth stages and the related impact on winter wheat production. In general, severe to extreme droughts occurred mainly during 2001–2009. Specifically, 2001 and 2003 were dominated by severe and extreme droughts. Droughts occurred mainly in two time periods—March–May and August–October. Consecutive droughts can sometimes be observed during summer and autumn in Anhui, Hebei, Henan, and Shandong Provinces. Few droughts were detected in Jiangsu Province. Higher drought frequency was still observed during 2003–2010. Droughts with lower intensity and lower frequency were observed during 2011–2016 and the study region has tended to be humid in recent years.

Figures 3 and 4 show the seasonal distribution of droughts of different intensities in the five provinces. In general, drought frequency was higher in the spring and autumn seasons, while the frequency of droughts was relatively lower during the summer (particularly in July and August) and winter seasons. Within Anhui Province, droughts occurred mostly during March–June and September–December (Fig. 4a). As many as nine drought events existed on the 57th day of the year during 2001 to 2016—in more than half of the years. Moderate and severe droughts occurred often in March to April and in September to November. Serious droughts occurred mostly during the spring and autumn seasons. Consecutive drought hazards occurred during both 2001 and 2002 (Fig. 3). In 2001, moderate and severe droughts occurred in April to May, and mid- and late August to mid-October and mid-November; and moderate and severe droughts occurred from mid-September to mid-November in 2002. Generally, autumn is the seeding period for winter wheat and spring is the critical period for the growth and development of winter wheat. Consecutive and long-lasting droughts in those periods can have serious impacts on winter wheat yield.

A different temporal distribution of droughts within 1 year is observed in Hebei Province compared to that in the other four provinces. Except for the days from November to December and from January to March, a relatively even temporal distribution of drought frequency can be observed (Fig. 4b). During typical drought years, severe consecutive droughts occurred in the spring–summer and autumn–winter seasons—in May–July of 2001 and September–October of 2002 (Fig. 3b). The temporal distribution of droughts within 1 year in Henan Province is similar to that in Anhui Province—droughts occurred mainly during spring and autumn, but relatively high drought frequency and intensity also can be observed in mid-June to July (Fig. 3a, c). The intensity and frequency of droughts in Jiangsu Province are relatively low, mild and incipient droughts are dominant. Droughts occurred mainly during spring—in March to May. The temporal distribution of droughts in Shandong Province is relatively uniform throughout the year and the mean occurrence frequency of incipient droughts ranges from 5 to 6. However, the differences between Shandong Province and Anhui and Henan Provinces lie in the fact that a lower frequency of moderate, severe, and extreme droughts can be detected in Shandong Province. A higher frequency of droughts with lower intensity can be observed in Shandong Province when compared to those in Anhui and Henan Provinces (Fig. 4e).

The study region is dominated by a higher frequency of incipient droughts, and 40–50 droughts on average can occur in all seasons (Fig. 5d, h, l, p), accounting for about 25% of the total monitored droughts over the 16 years considered in this study. Discernable differences in spatial pattern of moderate and severe droughts can be identified across the study region. A higher occurrence frequency of moderate droughts can be observed in Hebei, Henan, Anhui, central Jiangsu, and eastern Shandong Provinces, and this is particularly true for the occurrence frequency of droughts in northern Hebei, southern Henan, and northern Anhui. The highest frequency of droughts can be detected mainly during spring, with about 30–40 drought events. Severe droughts occurred during the spring, summer, and autumn seasons, mainly concentrated in northern Hebei, southeastern Henan, and northern Anhui Provinces, especially in spring. In Hebei Province, droughts occurred in the southern is more serious than the northern part. Regional extreme droughts occurred only about 5–10 times during the study period. Generally, 15–20 extreme drought events can be observed during spring in southern Henan and during winter in northern Hebei.

Except for droughts in winter in Hebei Province, higher frequency of droughts can be detected mainly in the northern part of the study region rather than in the south. In Shandong Province, droughts occurred mainly in the eastern and the central areas. Generally, higher frequency of droughts can be identified across the entire Henan Province. Frequent droughts can be detected in southern Henan, the province’s region for winter wheat production, and frequent droughts pose serious challenges for stable production. Overall, droughts in Jiangsu Province are less severe than in the other provinces and occurred mainly in central Jiangsu. In spring and summer, the frequency of droughts in northern Anhui was higher and relatively low in southern Anhui.

Previous research mainly focused on accurate and real-time monitoring of the occurrence and development of droughts (Aghakouchak et al. 2015), but very few reports addressed the impacts of droughts on growing crops and related crop production. However, droughts with different intensities that occur during different growth stages of crops can have distinctly different impacts on crop yield (Çakir 2004). Therefore, it is critical to investigate the impacts of droughts on winter wheat yield by studying the occurrence timing of droughts relative to the growth stage of winter wheat.

Winter wheat is generally sown in October and harvested in June of the subsequent year. Yield in 1 year is usually affected by the drought conditions in October–December of the previous year and January–June of the current year. In this study, the correlation between the crop areas affected by droughts of different intensities during January to June of 2002–2016 and the detrended yield of winter wheat in each year from 2002 to 2016 was quantified. The correlation between the crop areas affected by droughts of different intensities during October to December of 2001–2015 and the detrended yield of winter wheat in each year from 2002 to 2016 was also quantified. The correlations were evaluated between crop areas affected by incipient, mild, moderate, severe, and extreme droughts and winter wheat production in order to explore the types of droughts that have the largest impacts on winter wheat production in terms of occurrence timing and intensities (Figs. 6, 7, 8, 9, 10).

Planting of winter wheat across the five provinces generally occurs in early October, and then winter wheat goes through sprouting, tillering, and overwintering. It starts to winter in Hebei and Shandong after mid-December, 10 days earlier than in Henan. But wintering in Jiangsu and Anhui is about 20 days later than in Shandong. This difference in timing of wintering is mainly due to the differences in solar radiation and thermal conditions that vary with latitude. The processes of greening, jointing, heading, flowering, filling, and maturity of winter wheat follow the ending of the winter season. The flowering stage of winter wheat can occur during early April in Jiangsu and Anhui Provinces. But the timing of the flowering stage of winter wheat in Shandong, Henan, and Hebei Provinces is 10–20 days later than in Jiangsu Province.

As shown in Fig. 6, from January to mid-March winter wheat is in the stage of overwintering and green. A weak negative correlation and positive correlation can be identified between incipient drought-affected crop areas and crop yields, implying that incipient droughts during the wintering and greening periods of winter wheat have no significant impacts on winter wheat yields and may even increase the yields of winter wheat slightly. Precipitation and air temperature are the main climatic factors that influence crop growth but they have different impacts on the growing crops in different growth stages. Regardless of the stage of crop growth, changes in cumulative precipitation during the growing season, especially during the earlier growing season, have the greatest impact on winter wheat. A persistent precipitation deficit will potentially have serious cumulative effects on the growth and development of winter wheat, and yield can be greatly reduced. Precipitation in the study region is mainly concentrated in July–September, accounting for about 70% of the total annual precipitation. Seeding of winter wheat occurs mainly in October. The good soil moisture conditions of the study area in early autumn provide good water conditions for the seeding and overwintering of winter wheat (Li et al. 2000). After a long wintering period, winter wheat in the study region enters the green stage. The increase of solar radiation and the warming temperature promote the growth of crops. During this period, under incipient drought conditions, and although winter wheat may suffer some slight water stress, the promotion effect of increased temperature and radiation is dominant in its growth, and these conditions may even increase winter wheat yield. When drought conditions aggravate from moderate to higher intensity (Figs. 8, 9, 10), however, a more serious water deficit will potentially result in a decrease in winter wheat production.

However, negative correlations were detected between incipient droughts and winter wheat yield during early and mid-April (Fig. 6), although these correlations are not statistically significant except in southern Henan, Anhui, and northern Jiangsu Provinces. Further analysis indicates that regions dominated by negative correlations are also those dominated by winter wheat entering the stage of heading and flowering, and regions dominated by insignificant correlations are those where winter wheat is entering the jointing and heading periods. Innes and Blackwell (1981) indicated that water stress before flowering of the winter wheat can reduce the number of grains per spike, which can definitely affect the yield of winter wheat. Late May to late June (Day145–Day169 in Fig. 6) witnessed negative correlations between incipient droughts and yield across a larger area of the study region, implying that the incipient droughts that occurred during this period can cause apparent decrease of crop yield—late May to late June is also the flowering, filling, and maturity stage for winter wheat. Moreover, filling will follow flowering and the duration of filling will impact the fullness of the winter wheat grains (Zhang et al. 1998; Yang and Zhang 2006). The water requirement of winter wheat at this stage increased significantly when compared to that of the previous period. Long drought duration will result in poorer grains of wheat and reduce yield (Fang et al. 2017). A comparison between incipient drought-affected and mild drought-affected crop areas reveals no significant reduction of yield in northeastern Henan, southern Hebei, and central and western Anhui during late May to late June (Day145–Day169 in Figs. 6, 7), which can be attributed to the alleviation of droughts as a result of agricultural irrigation in these regions.

Figure 7 illustrates the correlation between mild drought-affected crop areas and crop yield. No significant correlations can be found during January and February (Day001–Day057 in Fig. 7). When compared with the impacts of incipient droughts on yield, a negative correlation dominates the mild drought-affected crop areas and winter wheat production. Expansion of the mild drought-affected crop areas links to reduced winter wheat production. Increased intensity of the mild drought largely offsets the positive impacts of increased temperature and solar radiation on the growth of winter wheat. Therefore, water deficit plays a critical role in the growth of winter wheat crops. In early and mid-April (Day097–Day105 in Fig. 7), a wide range of areas in northern Anhui, northwestern Jiangsu, and southern Henan were characterized by negative correlation coefficients and this time period corresponds to jointing and flowering. From late May to mid-June (Day137–Day161 in Fig. 7), a negative correlation between yield and mild drought-affected crop areas dominates large areas, and significant negative correlation was found at 0.01 significance level in Henan, Shandong, southern Hebei, and parts of Anhui Provinces. A closer look at these significant correlations indicates that mild drought had a profound impact on the yield of winter wheat during the filling and ripening stage. In the tillering stage of winter wheat growth (Day297), mild droughts still can potentially cause reduction of yield, especially in Anhui, Henan, and southern Hebei Provinces. When compared to the intensity of incipient droughts, the intensity of mild droughts is higher (Figs. 6, 7) and a stronger negative correlation can be observed between yield and mild drought-affected crop areas, implying that a higher intensity of mild drought will have more effects on winter wheat production, and this is particularly true for winter wheat yield during the greening to maturity periods.

Impacts of moderate droughts on winter wheat yield are similar to those of mild droughts (Figs. 7, 8). The difference is that mild drought during the wintering, greening, and jointing stages of winter wheat growth may have a promoting effect on yield. Moderate droughts have negative impacts on winter wheat yield during this same period and reduce winter wheat production (Zhang et al. 2008). From incipient drought to mild drought and moderate drought (Figs. 6, 7, 8), droughts of higher intensity have more significant impacts on winter wheat at its earlier growing season. For example, incipient drought can have discernable impacts on winter wheat yield during the filling period, while moderate droughts may have remarkable impacts on yield during the flowering, filling, and maturity period. In this sense, the impact of water stress on winter wheat yield is mainly reflected by the long-term cumulative effects (Huang et al. 2016). For severe and extreme droughts, the correlations between drought-affected crop areas and winter wheat production are less significant than those for mild and moderate droughts (Figs. 9, 10). What is particularly noteworthy is that the spatial pattern of severe and extreme droughts is significantly correlated with crop yield in October (Day281–Day297 in Figs. 9, 10), and this is also the time period for seeding and the sprouting stage of the winter wheat. Severe drought events can trigger lower sprouting rates of winter wheat and have negative impacts on yield.

These results show that droughts of different intensities occurring in different growth stages have distinct effects. It is unnecessary to irrigate when incipient droughts occur during the wintering and greening periods, but irrigation is indispensable to alleviate droughts when their intensity reaches the moderate level. Flowering and filling stage is the most sensitive period for winter wheat to water, and mild drought will lead to a sharp decline in yield. With the intensity of drought aggravating, the negative effect on winter wheat yield becomes more significant. Therefore, particular attention should be paid to this growing period for drought monitoring and irrigation.