Climatic Causes of Maize Production Loss under Global Warming in Northeast China

: Maize ( Zea mays L.) is one of the most important staple crops in Northeast China, and yield losses are mainly induced by climate anomalies, plant diseases and pests. To understand how maize yield loss is a ﬀ ected by global warming, daily precipitation and temperatures, together with provincial agricultural data sets, were analyzed. The results showed that the accumulated temperature, an important factor in agricultural productivity, increased by 5% in 1991–2017, compared to 1961–1990, and that the frequency of low temperatures decreased by 14.8% over the same time period. An increase in drought by 21.6% was observed from 1961–1990 to 1991–2017, caused by decreased growing-season precipitation by − 4 mm / decade. In addition, days with heavy rain in August and September increased slightly in Northeast China. In general, maize growth responded positively to the increased thermal conditions; in 1961–1990, 22.7% of observed maize yield-loss cases were due to low temperatures, but only 10% in 1991–2017. However, during the same time, the number of drought-induced yield loss cases increased from 27.3% to 46.7%. Moreover, yield loss cases caused by heavy rainstorms increased from 4.5% to 13.3%, indicating that heavy rainstorms have become an increasing threat to agriculture in Northeast China over the last three decades. In total, at least 70% of cases of provincial yield losses in Northeast China over the last three decades could be attributed to climatic factors. The frequency of climate hazards has changed under global warming, resulting in new challenges for agriculture. While drought and low temperatures were the primary causes for climate-induced yield losses before the 1990s, negative impacts from extreme events, mainly drought but also heavy precipitation, have increased in the last three decades, associated with global change. Farmers, agricultural scientists, and government policy makers could use these results when planning for adaptation to climate change.


Introduction
Climate change has already had an observable impact on the global food production [1]. In this context, extreme climate and weather events can have significant negative impacts on agriculture, depending on their duration, intensity and distribution [2]. If several of the world's breadbaskets are exposed simultaneously to extreme climatic conditions, it could have severe negative global impacts [3].
forms ears in August. Thus, an additional climate hazard affecting crop yields in NC is waterlogging caused by rainstorms associated with the East Asian summer monsoon [44].
There have been large efforts to describe and predict spatiotemporal changes in key agricultural factors, such as precipitation and temperature extremes, but less focus has been on how changes in various climate hazards related to global warming have impacted maize production. This information is needed when guiding the actions of government agencies and crop breeders to provide a more resilient agriculture sector that is better adapted to changing climate conditions. Therefore, the objective of this study is to assess the influence of key climate hazards, drought, low temperature and rainstorms, on maize yield loss in NC since the 1960s.

Study Area and Data
Here, Northeast China (NC) is defined as consisting of the provinces Heilongjiang, Jilin, and Liaoning (38 • N-55 • N and 118 • E-135 • E). This part of China is cool due to its high-latitude location, and can be characterized as having a temperate monsoon climate. The annual mean surface air temperature of NC is only 5.4 • C (1981 to 2010), which is 7.8 • C lower than the average for China. The total annual precipitation ranges from 460 to 750 mm.
In order to investigate the influence of climate hazards on maize yield loss, daily precipitation and mean temperatures provided by the China Meteorological Administration, the period 1961-2017 were used. Only stations with data series covering more than 57 years with low rates of missing data (≤5%) were included in the analysis (Figure 1), resulting in a total of 169 stations for NC. Furthermore, relative soil moisture data from 10 cm depths, observed on the 8th, 18th, and 28th of every month from 1995 to 2010, were also used. Provincial-level maize yields were obtained from the Ministry of Agriculture of the People's Republic of China for the period 1958 to 2017. Sustainability 2020, 12, x FOR PEER REVIEW 3 of 13 typically forms ears in August. Thus, an additional climate hazard affecting crop yields in NC is waterlogging caused by rainstorms associated with the East Asian summer monsoon [44].
There have been large efforts to describe and predict spatiotemporal changes in key agricultural factors, such as precipitation and temperature extremes, but less focus has been on how changes in various climate hazards related to global warming have impacted maize production. This information is needed when guiding the actions of government agencies and crop breeders to provide a more resilient agriculture sector that is better adapted to changing climate conditions. Therefore, the objective of this study is to assess the influence of key climate hazards, drought, low temperature and rainstorms, on maize yield loss in NC since the 1960s.

Study Area and Data
Here, Northeast China (NC) is defined as consisting of the provinces Heilongjiang, Jilin, and Liaoning (38° N-55° N and 118° E-135° E). This part of China is cool due to its high-latitude location, and can be characterized as having a temperate monsoon climate. The annual mean surface air temperature of NC is only 5.4 °C (1981 to 2010), which is 7.8 °C lower than the average for China. The total annual precipitation ranges from 460 to 750 mm.
In order to investigate the influence of climate hazards on maize yield loss, daily precipitation and mean temperatures provided by the China Meteorological Administration, the period 1961-2017 were used. Only stations with data series covering more than 57 years with low rates of missing data (≤5%) were included in the analysis (Figure 1), resulting in a total of 169 stations for NC. Furthermore, relative soil moisture data from 10 cm depths, observed on the 8th, 18th, and 28th of every month from 1995 to 2010, were also used. Provincial-level maize yields were obtained from the Ministry of Agriculture of the People's Republic of China for the period 1958 to 2017.

Methods
Temperature and precipitation during the GS are the main limiting factors of climate induced maize yields in NC [45]. If the AT during the growing season does not reach a certain threshold, it may have negative impacts on the yield, so we selected AT as an index which would identify low temperatures that could be harmful for maize growth. Further, maize yields are vulnerable to the extent and duration of drought, so in our analyses we used the accumulated drought index (AD), which is based on an index of the weighted average of precipitation [37], and is more flexible and versatile than the SPI when monitoring drought on daily timescales. Finally, another climate hazard

Methods
Temperature and precipitation during the GS are the main limiting factors of climate induced maize yields in NC [45]. If the AT during the growing season does not reach a certain threshold, it may have negative impacts on the yield, so we selected AT as an index which would identify low temperatures that could be harmful for maize growth. Further, maize yields are vulnerable to the extent and duration of drought, so in our analyses we used the accumulated drought index (AD), which is based on an index of the weighted average of precipitation [37], and is more flexible and Sustainability 2020, 12, 7829 4 of 13 versatile than the SPI when monitoring drought on daily timescales. Finally, another climate hazard affecting maize yields in NC is heavy precipitation events, so rainstorm days were chosen as an index in this paper.

Low Temperature
We used the 10-degree threshold for calculating the AT according to Yan et al. [46], where the AT is obtained from the sum of average daily temperatures that are equal to or above 10 • C from May to September. Because we were interested in low temperatures, we used the anomaly of AT as a low temperature, which was defined as: where AT x represents the AT anomaly during May-Sep in year i, relative to the mean AT during the previous three years, and T represents daily temperatures equal to or greater than 10 • C. Low temperatures are defined when AT x is negative.

Droughts and Rainstorms
Drought conditions were characterized using the WAP (the Weighted Average of Precipitation) index. Drought variability can be described by a simplified physical model using only precipitation. The model is expressed as: where t is the time, and f (t) is the drought intensity at time t, P(t) is the precipitation at time t, bf (t) is the effect of the decay of drought extent from the demands of runoff, evapotranspiration, groundwater flow, and percolation [37], where b represents the strength of the decay. The WAP index is calculated as: WAP = N n=0 a n P n N n=0 a n where a = e −b∆t < 1 represents the contribution of P 1 of the previous day's precipitation to the drought extent f 0 of the present day. The SPI and precipitation of the present day can be considered as two extreme cases (in which a is changing from 0 and 1). When a approaches zero, the WAP will become P 0 . When a reaches 1, WAP will be equivalent to the simple average of precipitation. In this paper, WAP was standardized. Detailed information about the index can be found in Lu [37].
To relate maize yield loss to the duration of drought, we used accumulated drought index (AD) derived from the WAP: Rainstorms in NC they are mainly induced by mesoscale convective systems (MSC) in summer and autumn [47]. A MSC generally has a spatial scale of 10-1000 km, lasting from about one hour to one or two days. The large-scale weather system that mainly affects the occurrence and development of MSCs is the northeast cold vortex. When a blocking high exists, the position of the northeast cold vortex becomes stable, often producing rainstorms, gale winds, hailstorms, etc. Maize is sown in May and will become a tall crop after July. After heavy rains and/or strong winds, maize is prone to lodging, either by the strong winds or by the soils becoming waterlogged (saturated), or by flooding due to changes in the water balance [48][49][50], then resulting in crop failure. Thus, we focused on heavy rainstorm days in August and September. We defined a heavy rainstorm day as one with precipitation greater than 50 mm for one day.

Yield Loss Rate
Maize yields are influenced by technological development and by climate. Maize yield losses are caused by weather events, as well as by plant diseases and insect pests. The increasing trend in yields has largely been driven by advances in agricultural technology, but the effects of these advances are quite short-for example a new variety usually increases yields for only around 3-5 years in China. The yield loss rate was defined using Equation (5), which expresses the yield for a given year relative to the yield averaged over the preceding three years according to Song et al. [10]: where Y l is the yield loss rate caused mainly by climate hazards; Yi is the maize yield in a given year, i; Yt is the maize yield during the three years previous to the given year.

Change of Climate Hazards
The growing season temperature in NC showed large inter-annual variability, but an overall increasing rate of 0.  [51], so the AT seems to have become increasingly more suitable for maize over the last three decades. Because our results show a distinct change in thermal conditions from 1991, in the subsequent analyses we analyzed climate events and yield losses during those two intervals: 1961-1990 and 1991-2017. The AT increased by 5% during 1991-2017, relative to 1961-1990 ( Figure 2). From the low temperature index (Equation (1)), it was found that the frequency of low temperature occurrences decreased by 14.8% in 1991-2017 relative to . For example, in 1969, which was a very cold year in NC, the AT was 2565 • C·d, lower by 199 • C·d than that in the previous three years. In 1976, another cold year, the AT was 2605 • C·d, lower by 163 • C·d than that in the previous three years in NC. The year with the coldest growing season temperature after 1991 was 2003, when the AT was 2845 • C·d, 132 • C·d lower than that in the previous three years.
Sustainability 2020, 12, x FOR PEER REVIEW 5 of 13 yields has largely been driven by advances in agricultural technology, but the effects of these advances are quite short-for example a new variety usually increases yields for only around 3-5 years in China. The yield loss rate was defined using Equation (5), which expresses the yield for a given year relative to the yield averaged over the preceding three years according to Song et al. [10]: where Yl is the yield loss rate caused mainly by climate hazards; Yi is the maize yield in a given year, i; Yt is the maize yield during the three years previous to the given year.

Change of Climate Hazards
The growing season temperature in NC showed large inter-annual variability, but an overall increasing rate of 0. 24 [51], so the AT seems to have become increasingly more suitable for maize over the last three decades. Because our results show a distinct change in thermal conditions from 1991, in the subsequent analyses we analyzed climate events and yield losses during those two intervals: 1961-1990 and 1991-2017. The AT increased by 5% during 1991-2017, relative to 1961-1990 ( Figure 2). From the low temperature index (Equation (1)), it was found that the frequency of low temperature occurrences decreased by 14.8% in 1991-2017 relative to . For example, in 1969, which was a very cold year in NC, the AT was 2565 °C•d, lower by 199 °C•d than that in the previous three years. In 1976, another cold year, the AT was 2605 °C•d, lower by 163 °C•d than that in the previous three years in NC. The year with the coldest growing season temperature after 1991 was 2003, when the AT was 2845 °C•d, 132 °C•d lower than that in the previous three years. Determining a drought index threshold is important when characterizing the impact of drought on maize production. Using the observed values of relative soil moisture, the relationship between the WAP and relative soil moisture was investigated. The results showed significant correlations between the WAP and relative soil moisture (r = 0.76, p < 0.0001, N = 367). It was found that the Determining a drought index threshold is important when characterizing the impact of drought on maize production. Using the observed values of relative soil moisture, the relationship between the WAP and relative soil moisture was investigated. The results showed significant correlations between the WAP and relative soil moisture (r = 0.76, p < 0.0001, N = 367). It was found that the relative soil moisture reached approximately 60% when the WAP 0 was −1.0 (Figure 3), so based on earlier studies (see above), we chose this value as the limit for defining drought events. Next, the accumulated drought index (AD) for the maize growing season was calculated. The annual precipitation decreased slightly by −1.2 mm/decade from 1961 to 2017 in NC, which is line with the precipitation change over the whole of China [15]. Maize growing season precipitation, on the other hand, decreased by hand, decreased by −4.0 mm/decade. Because of the decreasing precipitation, drought conditions during the maize growing season increased in NC, especially during the past two decades. Accumulated drought index (AD) values for the maize growing season showed decadal variability, being 12.0 in 1960s, 13.7 in 1970s, 11.4 in 1980s, 13.9 in 1990s, 17.4 in 2000s, and 13.4 from 2011 to 2017. In a multi-decadal perspective, the drought index increased by 21.6% in 1991 to 2017 relative to 1961 to 1990 (Figure 4).
The results of our analysis showed that the regional average number of rainstorm days in August-September increased slightly during the analyzed period by 7.8% during 1998-2017 compared to 1961-1980. Moreover, extreme rainstorm days also increased, especially in the Jilin and Liaoning provinces. For example, maximum rainstorm days in the Jilin province were in average 0.6 days in the 1970s, but increased to 1.0 days in the 1990s. In Liaoning province, the maximum rainstorm days was 1.9 days in the 1980s and increased to 2.4 and 2.3 days in the 1990s and 2000s, respectively (Table 1).

Climatic Reasons for Maize Production Loss
Climate variations from daily to inter-annual timescales affect a number of physical, chemical, and biological processes that drive the productivity of maize. The latitudinal distribution of maize is a function of several climate factors, such as temperature, precipitation, and daily solar radiation (photoperiod). Extreme weather events and hazards influence the production of maize. Depending on the development stage, maize growth may be negatively affected by short-duration events, but The results of our analysis showed that the regional average number of rainstorm days in August-September increased slightly during the analyzed period by 7.8% during 1998-2017 compared Sustainability 2020, 12, 7829 7 of 13 to 1961-1980. Moreover, extreme rainstorm days also increased, especially in the Jilin and Liaoning provinces. For example, maximum rainstorm days in the Jilin province were in average 0.6 days in the 1970s, but increased to 1.0 days in the 1990s. In Liaoning province, the maximum rainstorm days was 1.9 days in the 1980s and increased to 2.4 and 2.3 days in the 1990s and 2000s, respectively (Table 1).

Climatic Reasons for Maize Production Loss
Climate variations from daily to inter-annual timescales affect a number of physical, chemical, and biological processes that drive the productivity of maize. The latitudinal distribution of maize is a function of several climate factors, such as temperature, precipitation, and daily solar radiation (photoperiod). Extreme weather events and hazards influence the production of maize. Depending on the development stage, maize growth may be negatively affected by short-duration events, but yields, especially on province level, may not decrease if there is a subsequent improvement in growth conditions. Weather events extending over longer time, such as drought, are more likely to have negative impacts on yields. Thus, it is important to investigate the limit of hazards that can affect the province-level yields.
Overall, the increasing temperatures in the study area had a positive effect on maize growth [10]. In addition, given the new conditions, new maize varieties have been bred that are adapted to warmer conditions in NC, which can grow over an extended growing season. It is clear that warmer GS temperatures alone have had little negative impact on yield loss in NC. However, our analysis showed significant correlations between the yield loss rate and low ATx (r = 0.50, p < 0.0001). The temperature was cold during the growing season in 1969, and the low temperature index was −249.3 • C·d in Heilongjiang and −205 • C·d in the Jilin province, corresponding to decreased maize yields of 32.0% and 15.3% for these two provinces, respectively (Table 2). In general, our results suggest that maize yields decreased when the AT was 135 • C·d lower than those observed during the previous three years ( Figure 5). Therefore, the low temperature limit for yield loss was set as −135 • C·d. However, it is clear that very low AT has decreased in recent years.  Associated with climate warming, the main reasons for maize yield losses changed during the analyses period. We found 53 cases of maize yield loss in the Heilongjiang, Jilin, and Liaoning provinces from 1961 to 2017 using equation 5. Between 1961 and1990, 22.7% of the maize loss cases were caused by low temperatures, compared to 10% in 1991-2017 ( Figure 6). Focusing on drought, 27.3% of maize yield loss cases could be attributed to drought in 1961-1990. This number increased to 46.7% in 1991-2017 ( Figure 6). Only 4.5% of the maize yield loss cases were induced by heavy rainstorms during the 1961-1990 period, but this impact increased by almost three times to 13.3% in 1991-2017 ( Figure 6). Our results showed that maize yield loss is associated with the extent and duration of drought. A significant correlation (r = 0.39, p < 0.0001) between yield loss rate and AD was found. As an example, drought was very serious in 2000 in NC. The maize growing season precipitation was only 347 mm, 35% less than in a normal year in the Liaoning province, and the AD was 40.9, leading to a decrease of 31.5% in the maize yield (Table 2). In the same year, precipitation was only 426 mm (−14%) in Jilin and 374 mm (−15%) in Heilongjiang provinces, the AD amounted to 24.7 and 27.3 for those two provinces, respectively, corresponding to decreases in maize yields by 32.9% for Jilin and 6.2% for Heilongjiang. In general, maize yields decreases when the AD exceeds 22.8 ( Figure 5). The correlation between yield loss rate and rainstorm days was −0.28, which is significant (p < 0.005). On a provincial level, maize yields were negatively affected when the average rainstorm days during August-September exceeded 1.5 ( Figure 5). For example, rainstorms which occurred in the western and northern regions of the Liaoning province on August 6 in 1994, had precipitation amounts of 349, 309, and 344 mm in Yixian, Xinmin and Kaiyuan stations, respectively. Ten days later, an additional rainstorm affected eastern Liaoning, accompanied by winds of 20-25 m/s [52], and that year, the maize production decreased by 29.5% in Liaoning.
Associated with climate warming, the main reasons for maize yield losses changed during the analyses period. We found 53 cases of maize yield loss in the Heilongjiang, Jilin, and Liaoning provinces from 1961 to 2017 using equation 5. Between 1961 and1990, 22.7% of the maize loss cases were caused by low temperatures, compared to 10% in 1991-2017 ( Figure 6). Focusing on drought, 27.3% of maize yield loss cases could be attributed to drought in 1961-1990. This number increased to 46.7% in 1991-2017 ( Figure 6). Only 4.5% of the maize yield loss cases were induced by heavy rainstorms during the 1961-1990 period, but this impact increased by almost three times to 13.3% in 1991-2017 ( Figure 6). Associated with climate warming, the main reasons for maize yield losses changed during the analyses period. We found 53 cases of maize yield loss in the Heilongjiang, Jilin, and Liaoning provinces from 1961 to 2017 using equation 5. Between 1961 and1990, 22.7% of the maize loss cases were caused by low temperatures, compared to 10% in 1991-2017 ( Figure 6). Focusing on drought, 27.3% of maize yield loss cases could be attributed to drought in 1961-1990. This number increased to 46.7% in 1991-2017 ( Figure 6). Only 4.5% of the maize yield loss cases were induced by heavy rainstorms during the 1961-1990 period, but this impact increased by almost three times to 13.3% in 1991-2017 ( Figure 6).

Conclusions and Discussion
Using daily climate data and provincial agriculture data, the climatic factors related to provincial maize yield loss under global warming were investigated. The results showed that the GS temperature increased by 0.9 • C in 2008-2017 compared to 1961 to 1970 in Northeast China. This resulted in a reduced frequency of low temperature occurrences, leading to a decrease in maize yield loss caused by low temperatures by more than half from 1961-1990 (22.7%) to 1991-2017 (10%). However, associated with increasing temperatures and slight changes in precipitation, there was a 20% increase in GS drought, resulting in an almost doubling of maize loss cases attributed to this extreme (27.3% in 1961-1990 vs. 46.7% in 1991-2017). Despite a moderate decrease in annual as well as growing season precipitation, heavy rainstorm days during a sensitive time in the maize growth cycle (August and September) increased slightly. While lodging caused by rainstorms did not cause too much loss in the early period (4.5%), it increased by almost three times in the latter period (13.3%), clearly showing that the impact of this particular extreme has increased in the last few decades. It has been pointed out that mean climate changes have overall benefitted crop yields and increased consumable food calories in major crops in China [1], and Meng et al. [23] showed that increased temperatures represented 35% of the overall yield gains in Heilongjiang province of China through farmers' adaptation. However, Song et al. [10] found that maize yield losses increased by 30% when drought increased by 100% in Northeast China. Global warming of 1.5 • C can lead to significant reductions in climatically suitable areas for summer maize cultivation in China [53].
Of the observed 53 cases of maize yield loss on provincial level from 1961 to 2017, at least 70% could be fully or partly attributed to climate hazards in NC, while the other 30% were due to plant diseases, insect pests, etc. Considering the three analyzed climate factors in this study, it is clear that their frequencies have changed as the climate has warmed. This means that the reasons for yield losses have also been changing. In the earlier three decades, drought and low temperature were the primary reasons for maize yield loss in NC, together accounting for almost 50% of the cases with equal contributions. However, in conjunction with global warming, the importance of drought increased so that it became the major maize-loss reason over the past three decades. Because irrigation infrastructure in NC is, in general, poor, and access to water was limited, maize growth is especially sensitive to drought in this region. Although maize is sensitive to drought throughout its development, yields are more severely affected during the filling stage (August-September), than in earlier development stages [54]. Moreover, due to global warming, there is a shift in the GS, so that the maturation occurs earlier [22]. The increased frequency of droughts in NC may be related to the weakening of the East Asian summer monsoon, but it is very likely that other factors, such as the Pacific Decadal Oscillation (PDO) also are of importance [55].
In addition to the impact of drought, it is evident that other extremes have to be taken into account when assessing the impact of global change on agricultural production. The almost threefold increase in yield loss to heavy rainstorms shows that this is an emerging risk. Regardless if wind felling or over-saturated soils cause the actual loss, this particular hazard needs more attention. The indirect impacts of climate change also need to be taken into consideration. Plant disease and/or insect pest outbreaks, both having severe negative impacts on yields, may also be related to adverse weather condition. For example, the plant disease (northern) leaf blight, which is caused by a fungus affecting the foliage, can have severe impacts on maize yields. Maize leaf blight usually appears when summers are wet [56]. Maize yield loss can be induced by the changes in weather events or the exogenous covariates (e.g., influence of insect pests). In this paper, we focused on those cases of maize yield loss when drought or low temperature or rainstorms happened at the same period, which indicated the yield loss was mainly or partly induced by climate events. For example, when maize yield loss happened, while no climate events happened at the same time, this indicated that the yield loss resulted from the exogenous covariates, and this case of yield loss was not studied in this paper. Of course, climate events could happen with the exogenous covariates (e.g., influence of insect pests) at the same time, and then the maize yield loss was induced by climate events and the exogenous covariates together.
In the future, if temperatures continue to increase, drought will likely become more serious in China [57], but there will also be more extreme precipitation events, such as heavy rainstorms [58]. It seems wise to breed maize varieties that are resistant to drought, rather than varieties resistant to low temperature, which has been chosen by farmers and governments in the past decades in NC. Moreover, it is important to make sure that the crops are planted in ways so that they are less sensitive to extreme winds and temperatures. Consequently, the results of this study are very important for farmers, plant breeders, and government policy makers in regard to developing strategies to adapt to climate change, producing more drought-resistant maize varieties, and taking measures to mitigate the negative impacts of drought on maize production in Northeast China under global warming.
Author Contributions: The contributions of Y.S., H.W.L. and G.Z. involved writing the manuscript; Y.L. and J.X. carried out data calculation. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.