Effect of Indian Ocean–Pacific SST Pattern in Autumn on Winter Wheat Climatic Yield in the North China Plain in the Following Year and a Possible Mechanism

Ensuring stable crop yield increases to meet rising demand is an important issue globally, particularly when accounting for climate change. In this study, using observations, reanalysis datasets, and the Hodrick and Prescott filter method, we find that changes in a distinct pattern of Indian Ocean–Pacific five-pole (IPFP) SST (sea surface temperature) are strongly linked to the ensuing year’s winter wheat climatic yield (the part of yield that fluctuation caused by climatic factors change) in the North China Plain (NCP), which is the main production region of winter wheat in China. Here we define a normalized IPFP index (IPFPI) and demonstrate that the autumn IPFPI (1948–2014) is well correlated with the ensuing year’s winter wheat climatic yield (1949–2015), particularly for October (r = 0.69; n = 67; P < 0.001). A composite analysis shows that the October IPFP is correlated with sowing-period and emergence-period climate factors in the NCP. When the October IPFP is in a positive phase, the atmosphere geopotential height fields and water vapor flux are bebefitial to rainfall formation in NCP, and the precipitation and soil moisture are higher in NCP and benefit winter wheat growth, thus increasing the climatic yield. In addition, accumulated rainfall and soil water content might influence winter wheat growth from sowing and emergence (autumn) to the returning green stage (following spring).

. Time series of annual winter wheat yield (t/ha) in the NCP . (a) Original annual winter wheat yield (black solid line) and corresponding HP filter trend yield (red solid line); and (b) climatic yield of winter wheat determined using the HP filter.
Because NCP winter wheat is harvested each year in June, we computed the correlation coefficients of winter wheat climatic yield and global monthly SST one year prior to the winter wheat harvest. Thus, monthly SST data from July of the previous year to June of the harvest year were used in the correlation analysis with winter wheat climatic yield. We found that from July to December (because October has the strongest signal, we focus on results from October) of the year prior to the harvest, there were five highly correlated centers following the pattern "positive-negative-positive-negative-positive" from the southern Indian Ocean through the Indo-Pacific Warm Pool to the northwest Pacific (Fig. 2).
According to the domain and pattern of the highly correlated regions, the normalized SST sequence was used to define an Indian Ocean-Pacific SST five-pole index (IPFPI): where A1, A2, A3, B1, and B2 denote the regional mean SST in ( Fig. 3), indicating a significant positive correlation (P < 0.001). However, the relationship captured by the IPFPI between remote SST signals and local NCP atmospheric circulations, and the effects on local climate factors during key periods of winter wheat growth (the seeding and emergence stages) as well as the overall impact on winter wheat production, require further analysis.
Relationships between local NCP climatic factors, the IPFP of the same period, and winter wheat climatic yield. Temperature and precipitation are the main climatic factors influencing crop growth, development and yield fluctuations. Suitable temperatures and abundant rainfall are necessary conditions for normal crop grow and high yield. In addition, soil moisture affects crop yield, particularly in rain-fed agricultural areas. . Three positive-correlation areas (white boxes) and two negative-correlation areas (black boxes) were chosen to construct the IPFPI. The correlation coefficients of ±0.2, ±0.24, and ±0.31 are statistically significant at the 90%, 95%, and 99% levels, respectively (Student's t-test), (map produced by NCL Version 6.6.2, http://dx.doi.org/10.5065/D6WD3XH5) 46 .  shows that the local NCP temperature in October has a significant negative correlation with the October IPFPI during 1951-2014. This suggests that the IPFP negatively impacts the NCP local temperature in October, which is the sowing period for winter wheat in the NCP. Figure 4b shows that the temperature in October during 1951-2014 has a weak negative influence on the corresponding winter wheat climatic yield in NCP in the following year . This suggests that temperature fluctuations in the sowing period have a negative impact on the following year's winter wheat climatic yield. These findings are similar to those of others studies. Zhuang et al. 17 found that in northern China, higher temperatures during the seeding and emergence stages result in a reduced yield of winter wheat. The rate of yield decrease was about -40 kg ha −1 °C −1 . On a countrywide scale, rising temperatures might have contributed to a 4.5% reduction in wheat yield during 1979-2002 18 , and warmer daytime temperatures were likely to have decreased wheat yield by 6%-20% per 1 °C of temperature increase 16 . These results indicate that winter wheat yield are negatively affected by warming conditions. This may be because winter wheat is a chimonophilous crop (i.e., it prefers a cool climate). If temperatures increase, particularly during key growth stages such as the seeding and emergence periods, winter wheat growth might accelerate and advance the phenological period to before winter. This could expose winter wheat to low temperatures, wind, and drought during winter and the following spring.
In contrast to the temperature effects, Fig. 5a shows that the local NCP precipitation in October has a significant positive correlation with the October IPFPI during 1951-2014, and the results similar with 0.5 × 0.5° grid data's results (we don't show the figures here). This suggests that the IPFP has a positive impact on local NCP precipitation in October. Figure 5b shows that precipitation in October during 1951-2014 had a significant positive influence on the ensuing year's winter wheat climatic yield in the NCP . This suggests that increased   www.nature.com/scientificreports www.nature.com/scientificreports/ precipitation during the sowing and emergence periods leads to higher winter wheat climatic yield the following year. Soil moisture is closely related to precipitation. Soil moisture at various depths in October is positively correlated with the IPFPI of the same period in the NCP during 1982-2010 (Fig. 6a,c,e). This suggests that the IPFP increases soil moisture at various depths in October. Similarly, soil moisture at various depths in October during 1982-2010 is significantly positively correlated with the corresponding winter wheat climatic yield in the NCP of the following year (1983-2011; Fig. 6b,d,f). This suggests that during the sowing and emergence periods, soil moisture has a significant positive stimulating effect on winter wheat growth, development, and climatic yield.
Drought results in reduced crop yield in semi-arid to semi-humid areas [33][34][35] , particularly for a rain-fed agricultural system 36-38, . Before and after October corresponds to the winter wheat seeding and emergence periods in the NCP 7,10 . Precipitation and soil moisture at these stages are important to winter wheat growth 4 . This applies not only to October, but also to the several months before the seeding period, as precipitation and soil moisture have continuous effects on winter wheat 39 . Because the NCP has a temperate monsoon climate, severe droughts occur, particularly in the winter half of the year (October to March). Thus, precipitation and soil moisture play an important role in winter wheat production from emergence to the turning-green and jointing stages. Hence, our results demonstrate that water supply conditions during the vegetative growth phase are closely linked to winter wheat growth and yield, in agreement with previous work 5,17 .
Possible mechanism of IPFP influence on local NCP climatic factors. In this study, composite analysis was used to investigate how the IPFP affects local NCP climatic factors. The five highest and five lowest October IPFPI values along with the corresponding winter wheat climatic yield of the following year were selected for the analysis (Table 1). A significant difference exists between the averages of the highest and lowest October IPFPI values, and a similar difference exists between the average climatic yield (Table 1). Figure 7a,b shows that the composite SST anomalies of the five highest (and lowest) October IPFPIs are related to a positive (and negative) phase of the five cells of the India Ocean-Pacific SST. The composite difference also reflects the www.nature.com/scientificreports www.nature.com/scientificreports/ significant difference between the IPFP positive and negative phases (Fig. 7c). Thus, different IPFP phases correspond to different SST anomaly patterns. This may lead to different atmospheric cycling patterns through air-sea interactions and teleconnections that impact local NCP climate and circulation.
The different geopotential height fields during the highest and lowest IPFPI years were selected for composite difference analysis. The results of composite difference showed that 1000 hPa (Fig. 8a) and 850 hPa (Fig. 8b) in NCP showed negative anomalies in geopotential height, while positive anomalies in geopotential height were observed at 500hPa (Fig. 8c) and 200hPa (Fig. 8d), and statistically significant at 90% level. This showed that when IPFPI appears positive abnormal high value year, there were low pressure anomalies near the ground and low altitude atmosphere in NCP, while high pressure anomalies appear in the middle and upper atmosphere. This kind of high and low pressure configuration at vertical height was benefit for the updraft in NCP, which made the temperature of NCP at the lower level and the precipitation increased, which was beneficial to the growth and yield increase of winter wheat, and vice versa.
Because of the existence of different IPFP phases, different related circulation patterns might variously influence the NCP. Figure 9a shows that when the October IPFP is in a positive phase, it can cause a cyclone anomaly in the NCP at 700 hPa, which may decrease the air temperature and increase the precipitation in the winter wheat sowing and emergence periods. The composite differences of October water vapor flux and water vapor flux convergence at 700 hPa between highest and lowest IPFPI years have been showed (Fig. 10). We can find that the moisture come from the western Pacific Ocean transfer into and convergence in NCP, and may benefit for increase precipitation in this area. Figure 11a suggests a significant rainfall increase in the NCP when the IPFP is in a positive phase, consistent with the cyclone anomaly at 700 hPa in the NCP. However, when the October IPFP is in a negative phase, an eastern wind anomaly in the NCP exists at 700 hPa (Fig. 9b). Thus, the airflow does not converge in this area, which suppresses precipitation formation. Figure 11b shows a NCP precipitation decrease when the IPFP is in a negative phase in October. The October soil moisture is similar to the precipitation in the NCP. When the IPFP is in a positive phase, the NCP soil moisture anomaly shows a marked increase (Fig. 12a), www.nature.com/scientificreports www.nature.com/scientificreports/ and the NCP soil moisture anomaly decreases during negative IPFP phases (Fig. 12b), as also revealed in the composite difference (Fig. 12c).
These results indicate that the October IPFP could, through ocean-atmosphere interactions and atmospheric bridge effects 40 , impact the same-period local circulation in the NCP and further influence the temperature, precipitation, and soil moisture. Because of the high correlation between precipitation and soil moisture, when the IPFP is in a positive (negative) phase, precipitation and soil moisture anomalies increase (decrease) in the NCP during the winter wheat seeding and emergence stages, with important effects on subsequent winter wheat growth, leading to increased (decreased) yield. However, how autumn IPFP anomalies impact winter wheat growth remains to be determined.
To investigate the continuous effects of autumn IPFP, precipitation, and soil moisture, we examined the NCP soil moisture from after the sowing stage to the re-greening stage in the highest-IPFP years. Figure 13 shows that in years with strong precipitation anomalies before and during the sowing stage, the NCP soil retains a high water content from November (after sowing and emergence) to March (re-greening stage). This indicates that with high precipitation, the soil stores more water to support winter wheat growth and development for a long period. This is important for winter wheat development because this period corresponds to the dry season in the NCP, when severe cold and dry winter monsoons can lead to low temperatures and little rainfall. However, higher water content in the soil could play a role in retaining heat and moisture, thus helping winter wheat successfully over-winter and increasing yield in summer.   Fig. 7(a,b), but for October 700-hPa circulation anomalies (m/s), (maps produced by NCL Version 6.6.2, http://dx.doi.org/10.5065/D6WD3XH5) 46 .

Discussion and Conclusion
We used a HP filter to calculate the North China Plain (NCP) winter wheat climatic yield during 1949−2015. The yield is significantly positively correlated with a pattern of Indian Ocean-Pacific five-pole (IPFP) SST of the previous autumn during 1948−2014. Over the past 67 years, their correlation coefficient is 0.69, significant at the 99.9% confidence level (two-tailed t-test). Therefore, we conclude that a close link exists between the IPFP in autumn and winter wheat climatic yield in the NCP in the following year.
We also found that the autumn IPFP has a significant relationship with the atmosphere signals (geopotential height fields), and local NCP climate factors in the same period (negative correlation with temperature, positive correlation with precipitation and soil moisture). We further explain how the autumn IPFP could impact local NCP circulation through ocean-atmosphere interactions and atmospheric bridge effects 40 . We found that the atmospheric geopotential fields at different heights from 1000 hPa to 500 hPa, exhibit a five-pole distribution similar to the SST field. This result showed that when there was an abnormal change in the IPFP, the atmosphere also showed the corresponding changes and signals. At the same time, the correlation between the climatic yield of winter wheat in NCP and the different geopotential height fields in October of the previous year showed that there was a negative correlation between the low altitude atmosphere (925 hPa and 850 hPa) in NCP and a significant positive correlation in the middle atmosphere (700 hPa and 500 hPa) (Fig. R5), which indicated that the configuration of this atmospheric geopotential height was benefit for the updraft in NCP. Therefore, the temperature decreased, the precipitation and soil moisture increased during the sowing period of winter wheat, which was contribute to the growth and yield of winter wheat (we don't show the figures here). Thus, the IPFP's anomaly change could through impact atmospheric signals' variation and further influence the temperature, precipitation and soil moisture in the winter wheat sowing and emergence periods.
The temperature and precipitation in autumn are important for winter wheat sowing and emergence in the NCP. Lower temperatures and higher rainfall benefit winter wheat production, in agreement with our results. Finally, soil moisture also plays a key role in this process. We found that autumn precipitation continuity effects Figure 10. As in Fig. 7(c), but for October water vapor flux (g·cm −1 ·hPa −1 ·s −1 ) and water vapor flux convergence at 700 hPa, (maps produced by NCL Version 6.6.2, http://dx.doi.org/10.5065/D6WD3XH5) 46 . Figure 11. As in Fig. 7(a,b), but for October GPCP precipitation anomalies (mm), (maps produced by NCL Version 6.6.2, http://dx.doi.org/10.5065/D6WD3XH5) 46  www.nature.com/scientificreports www.nature.com/scientificreports/ can, through soil moisture storage, aid winter wheat growth and development from autumn to the following spring, leading to fluctuations in the yield of winter wheat.
We have identified a connection between winter wheat yield and prior remote SSTs, and we have explained how remote SST impacts local climate in the NCP and influences the winter wheat yield. These results will help us to better understand how prior remote SSTs impact regional agricultural crop yield, which is important for decision makers in planning and deploying agricultural policies, and for farmers in improving agricultural activities  www.nature.com/scientificreports www.nature.com/scientificreports/ and optimizing planting structures. In addition, we have described a way to use remote SST signals from a specific period to predict crop yield in an agricultural region. Future work will include how to utilize the October IPFP to predict the ensuing year's winter wheat climatic yield in the NCP.

Data and Methods
Observational dataset and statistical methods. This study analyses the NCP (including Shanxi, Hebei, Shandong, Henan and Anhui provinces) winter wheat yield from 1949 to 2015. Meteorological data include surface air temperature (SAT), ER-SST, precipitation, geopotential height (GPH), u/v wind and soil moisture from NOAA (https://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.pressure.html). To validate these data, monthly temperature and precipitation data from 160 Chinese stations from 1951 to 2015 are also used. The five provinces of NCP and 160 stations' location showed in Fig. 14. Details of the datasets are provided in Table 2.
We used two-tailed t-test to calculate statistical significance. This study also used other statistical methods, including a Hodrick and Prescott (HP) filter, correlation analysis, linear regression, and composite analysis, which are not described in detail here, except for the HP filter.
Climatic yield. Actual crop yield is made up of three parts: the trend yield (that part due to economic and technical development, agricultural management, utilization of pesticides and fertilization, and other human factors), the climatic yield (that part due to changing climatic factors), and a random error term 17,41 . We quantify the yield as follows: where Y is the actual crop yield, Yt is the trend yield, Yc is the climatic yield, and ε is the random error term (for a Gaussian distribution with an average value of zero   42 . It is a decomposition method that is applied to one-time series in state space. It can be applied to, for example, crop production that comprises a long-term trend and a short-term fluctuation 43 . The following three points describe a HP filter: (i) a HP filter is an algorithm for extracting a smooth curve from time series data; (ii) a HP filter can be regarded as a special projection designed to extract a signal gt from time series data Yn (here, Yn is a superposition of gt and orthogonal noise ct); and (iii) a HP filter can be regarded as a type of high-pass filter. The high-frequency part passes through the filter and the low-frequency part is removed. It can be used to isolate high-frequency components whose period is <8 years 44,45 .
It is generally believed that there are two components of grain yield: a low-frequency component and a high-frequency component. These components can be separated using a HP filter. Let the crop yield sequence be Yt (t = 1, 2, 3, …, n, where n is the sample size), which is time series data that includes a long-term trend (gt) and a short-term fluctuation (ct). The HP filter can be used to separate Yt as follows: where gt is defined as a solution to the following minimization problem: When solving Eq. (4) as part of a HP filter, the results depend on the choice of the parameter λ. When λ = 0, the trend sequence satisfying the minimization problem, gt, is the actual grain yield sequence Yt. However, gradually increasing λ tends to smooth the trend sequence. As λ → + ∞, the trend sequence will approach one line infinitely. The most controversial part of this method relates to the choice of λ. Using other researches' experience 44 , we select the following values for λ: 100 (annual data), 1600 (seasonal data), and 14,400 (monthly data).
This study utilized the Eviews8 software for the HP filter treatment. Because crop yield are annual data, we chose λ = 100 for the analysis. We input the actual NCP winter wheat yield data to Eviews8, selected the Quick-Series Statistics-HP filter option under the data sequence menu, and used the output as the trend yield, gt. The actual yield, Yt, minus gt was used to obtain the winter wheat climatic yield, ct.

Data availability
The SAT (global), GPCP, CMAP, ER-SST V5, CPC Soil Moisture, GPH, u wind and v wind data all provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their Web site at https://www.esrl.noaa.gov/psd/ are all publicly and freely available.