Spatiotemporal Characteristics of Droughts and Floods in Shandong Province, China and Their Relationship with Food Loss

Drought is a condition in which precipitation is significantly lower than the multi-year average during a given period (Dracup et al., 1980). There is no universally accepted definition of flood, and it is strongly correlated with increased precipitation (Chen et al., 2020a). Deficient and excessive water resources, respectively, can lead to drought and flood events, which pose a significant threat to climate-sensitive economic sectors, particularly agriculture (Simelton et al., 2012; Liu et al., 2018; Shi et al., 2020b). According to the Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report, the frequency of extreme heat and precipitation in East Asia has increased due to climate change (Intergovernmental Panel on Climate Change, 2021). As the largest agricultural nation in East Asia, China’s droughts and floods pose a significant threat to regional food security (Su et al., 2018; Zhang et al., 2020b). Shandong Province is an important grain-producing region in China, ranking first in China for many years in terms of total agricultural output value (Yang et al., 2021; Zhang et al., 2022). The province is dominated by rainfed agriculture (Zhang, 2008), and precipitation is closely linked to this agricultural pattern and is a significant factor in determining the success or failure of food production (Zarei et al., 2020). Due to this, droughts and floods have become significant meteorological disasters affecting agricultural development in the region (Wang et al., 2019; Shi et al., 2020a). The quantitative study of the spatiotemporal characteristics of drought and flood is crucial for ensuring food security.

Drought indices are frequently employed in quantitative studies of drought in diverse global regions (Mukherjee et al., 2018; Ndlovu and Demlie, 2020). Based on different sources of calculated data, drought indices can be divided into those based on ground climate data (Zhao et al., 2019; Ahammed et al., 2020), and those based on remote sensing monitoring (Zhou et al., 2020; Nie et al., 2020). In Shandong Province of China, the drought indices based on ground climate data are commonly used to identify drought events (Zuo et al., 2018; Zhang et al., 2019a; Yao et al., 2021). The most prevalent drought indices based on ground climate data are the Palmer Drought Severity Index (PDSI) (Palmer, 1965), the Standardized Precipitation Index (SPI) (McKee et al., 1993), and the Standardized Precipitation Evapotranspiration Index (SPEI) (Vicente-Serrano et al., 2010). The PDSI, a drought index based on water supply and demand, represents a significant step in the evolution of drought indices (Palmer, 1965). However, the index’s fixed time scale and autoregressive characteristics restrict its application. The SPI is utilized to analyze the impact of drought on various types of water demand by measuring precipitation deficiencies on multiple time scales (Bhunia et al., 2020). However, the SPI does not address the impact of temperature on drought, despite the fact that temperature significantly impacts drought (Tirivarombo et al., 2018). Since the SPEI overcomes the limitations of the PDSI and SPI (Wang et al., 2017), it is now the principal index for identifying drought in Shandong Province (Zuo et al., 2018; Yao et al., 2021). Since the SPEI is derived from precipitation and potential evapotranspiration, it can also be used to identify flood events (Ayugi et al., 2020; Zeng et al., 2020; Tao et al., 2022).

On the basis of the drought index’s quantification, pertinent studies have been conducted to further examine the correlation between drought and grain yield loss (Madadgar et al., 2017; Chen et al., 2020b). For example, Zhang et al. (2019b) used the Drought Severity Index (DSI) to monitor agricultural drought, and the relationship between crop yield and drought-affected areas was further determined. Kim et al. (2019) analyzed the relationship between drought and crop yield through linear regression methods, which can be used to capture agricultural drought loss patterns. It can be concluded that the Pearson correlation coefficient method can be used to describe the relationship between drought and flood conditions and grain yield, and the linear method can be used to quantify this relationship. In addition, studies have been conducted to determine how the flood affected crop yields (Chen et al., 2018; Gao et al., 2019), e.g., Wang et al. (2021) used SPEI to compute drought and flood conditions in three northeastern provinces of China and further analyzed the correlation between drought/flood conditions reflected by SPEI and spring corn yield. From these, it is evident that the drought index can be used to describe flood events and the relationship between floods and grain yield can be further examined. In short, drought indices can be used to describe drought and flood events, and exploring the relationship between drought/flood events and food losses is helpful for food security work.

Since grain yields are affected by multiple factors, including low-temperature frost and flooding (Shi et al., 2020b), we can not be certain that the decrease in food production is attributed to drought or flood. Therefore, the response mechanisms of grain yields to individual disasters are flawed. Given the availability of disaster loss data, the Percentage of Production Loss method is recommended for estimating the amount of food lost due to a specific disaster (Zhu et al., 2015; Xiao et al., 2017). We used the calculated drought-related food losses to analyze the relationship between them and droughts (drought duration, drought severity, drought peak, etc.). Similarly, the relationship between the amount of food loss due to floods and floods (flood duration, flood severity, flood peak, etc.) is determined. The output’s results can provide quantitative direction for food security work.

This study aims to: 1) estimate the amount of food lost as a result of droughts or floods, and identify the drought and flood factors significantly associated with grain yield reduction. 2) quantify the relationship between the amount of grain loss caused by drought/flood and drought/flood factors, and analyze the trend and spatial frequency distribution of drought/flood factors. The study’s findings can be used to provide quantitative guidance for food security work in Shandong Province and to mitigate food losses caused by droughts and floods. Furthermore, this study could provide a way to accurately quantify the relationship between drought/flood conditions and food losses.

Shandong Province (Fig. 1) is located in China’s east coa-st (114°47′30″‒122°42′18″E, 34°22′54″‒38°24′0.6″N). There is a temperate monsoon climate in the province, with annual mean temperatures of 11‒14°C and annual precipitation of 550‒950 mm. Shandong Province is becoming more and more vulnerable to drought as its population and water requirements increase (Zuo et al., 2018). Floods are also serious in Shandong Province, with 1.074 × 10⁷ ha of crops flooded from 1984 to 2010, resulting in a total economic loss of 5.473 × 10¹⁰ yuan RMB (Shi et al., 2020a). Because Shandong is a major agricultural province (Xu et al., 2020), the study of droughts and floods could help to save food.

Figure 1.  Location of Shandong Province in China and the distribution of meteorological stations

Except for the meteorological stations with missing data, there are 19 meteorological stations in Shandong Province (Fig. 1) with complete meteorological data from 1984 to 2020. Monthly meteorological data are derived from daily meteorological data of the China Meteorological Data Network (https://www.ncei.noaa.gov/). Food loss data from various disasters in Shandong Province are obtained from Statistics on the ‘China Agricultural Statistical Material’ (The Ministry of Agriculture of the People’s Republic of China, 1985‒2006) from 1984 to 2005. Furthermore, data on food loss from 2006 to 2020 are derived from the ‘China Water and Drought Disaster Bulletin’ (The Ministry of Water Resources of the People’s Republic of China, 2006‒2020). Food sown area data are derived from the ‘Shandong Statistical Yearbook’ (Shandong Bureau of Statistics, 2010‒2021) and ‘Statistical Material of 60 years of Agriculture in New China’ (The Ministry of Agriculture of the People’s Republic of China, 2010).

The Standardized Precipitation Evapotranspiration Index (SPEI) is calculated by normalizing the cumulative probability distribution values of precipitation and potential evapotranspiration difference value series. The SPEI was created in 2010 and can be used to display a region’s wet and dry conditions (Vicente-Serrano et al., 2010). The index’s calculation steps are as follows: 1) calculate potential evapotranspiration. We use the Thornthwaite method to calculate potential evapotranspiration. It is reasonable to calculate the potential evapotranspiration of Shandong Province using this method (Zuo et al., 2018; Ren et al., 2021; Yao et al., 2021). 2) Determine the difference in monthly precipitation and potential evapotranspiration. 3) Calculate the cumulative function by fitting the difference to the three-parameter Log-Logistic distribution. 4) Determine the SPEI value. Drought and flood levels (Table 1) are classified using SPEI values and previous literature (Tao et al., 2022).

The run theory (Fig. 2) can be used to identify drought/flood events (Yevjevich, 1967). Drought duration (Dd), drought severity (Ds), and drought peak (Dp) are all components of a drought event. Set the drought threshold to ‒0.5. If the SPEI value is less than ‒0.5, it means the month is in drought. The duration of a drought event, Dd, is the number of months with SPEI values less than ‒0.5. The severity of a drought event, Ds, is the sum of the absolute values of SPEI values over the course of the drought. The peak value of a drought event, Dp, is the maximum absolute value of SPEI values over the course of the drought.

Figure 2.  Schematic diagram showing the run theory. SPEI values are the Standardized Precipitation Evapotranspiration Index values; F1 and F2 represent two flood events; Fd represents flood duration; Fs represents flood severity; Fp represents flood peak; D1 and D2 represent two drought events; Dd represents drought duration; Ds represents drought severity; Dp represents drought peak

A flood event also includes flood duration (Fd), flood severity (Fs), and flood peak (Fp). Set the flood threshold to 0.5, indicating that the month is in flood when the SPEI value exceeds 0.5. The duration of a flood event, Fd, is the number of months with SPEI values greater than 0.5. The severity of a flood event, Fs, is the sum of the SPEI values over the course of the flood. The maximum SPEI value during the flood duration is Fp, the peak value of a flood event.

(1) The crop disaster-covered rate, the crop disaster-affected rate, and the crop extinction rate

The crop disaster-covered area represents the sown area with grain yield reduced by more than 10% due to the disaster, which is used to calculate the crop disaster-covered rate (Eq. (1)). The crop disaster-affected area represents the sown area with grain yield reduced by more than 30% due to the disaster, which is used to calculate the crop disaster-affected rate (Eq. (2)). The crop extinction area represents the sown area with grain yield reduced by more than 80% due to the disaster, which is used to calculate the crop extinction rate (Eq. (3)). These three indicators show the relative magnitude of losses and the extent of damage caused by droughts/floods to food production, which are calculated as follows:

where $ {{A}}_{{i j}}{} $ is the crop disaster-covered rate in year j from the ith disaster; $ {{B}}_{{i j}} $ is the crop disaster-affected rate in year j from the ith disaster; $ {{C}}_{{i j}} $ is the crop extinction rate in year j from the ith disaster; $ {{E}}_{{i j}} $ is the crop disaster-covered areas in year j from the ith disaster (ha); $ {{F}}_{{i j}} $ is the crop disaster-affected areas in year j from the ith disaster (ha); $ {{G}}_{{i j}} $ is the crop extinction areas in year j from the ith disaster (ha); $ {{M }}_{{j}} $ is the grain sown area in year j (ha).

(2) The method of Percentage of Production Loss

After obtaining data on the food yield reduction for various disasters, we can calculate the amount of food loss caused by multiple disasters using the Percentage of Production Loss method (Zhu et al., 2015; Xiao et al., 2017).

where $ {{L}}_{{i j}} $ is the amount of food lost in year j from the ith disaster (kg); $ {{E}}_{{i j}} $, $ {{F}}_{{i j}} $, and $ {{G}}_{{i j}} $ are the same as in Eqs (1), (2), and (3), respectively; $ {{B }}_{{j}} $ is the grain yield per unit area in year j (kg/ha).

(1) Pearson correlation coefficient

The Pearson correlation coefficient method is a statistical method for accurately describing the closeness of two variables and the correlation coefficient value r is between ‒1 and 1. The closer r is to ‒1, the stronger the negative linear correlation between the two variables; the closer r is to 1, the stronger the positive linear correlation between the two variables. If r = 0, there is no linear relationship between the two variables. The literature describes the Pearson correlation coefficient methods and the significance tests (Zeng et al., 2022). Set the level of significance to 0.05. The calculation results are considered significant if the significance level is less than 0.05.

(2) Linear Regression Model

The linear regression model can be used to investigate the relationship between disaster-related factors and food losses. The literature describes the linear regression methods and the significance tests (Sein et al., 2021). The trend rate of change is ten times the linear regression coefficient.

(3) Mann-Kendall test

The Mann-Kendall test is a non-parametric test method recommended by World Meteorological Organization (WMO) for detecting sequence trends and mutations. Furthermore, in hydrology, the test is commonly used in trend analysis (Jia et al., 2018; Sein et al., 2021). The Mann-Kendall test procedure is described in detail in the literature by Huang et al. (2022).

The Mann-Kendall test is used to test the trend of the time series sample while avoiding outlier interference (Jia et al., 2018). We examine the trends in annual mean temperature (T), annual precipitation (P), annual potential evapotranspiration (PET), and annual drought/flood conditions in Shandong Province from 1984 to 2020 using the Mann-Kendall test and the Linear Regression model. Fig. 3 depicts the trends: 1) the temperature in Shandong Province has increased significantly since 1988, with a rate of increase of 0.4°C/10 yr. 2) Annual precipitation exceeded 800 mm in 1990, 2003, 2007, and 2020, and the changing trend of annual precipitation is insignificant, increasing at a rate of 15 mm/10 yr. 3) The annual trend in potential evapotranspiration follows the temperature trend, increasing at a rate of 25 mm/10 yr. 4) The SPEI value of the 12-month scale (SPEI-12) in December of each year shows the annual drought and flood conditions. There is a significant trend of aridity in Shandong Province from 1986 to 1993, 1998 to 2006, and 2014 to 2020. From 1984 to 2020, the linear trend in the province shows an aridification trend, with a reduced rate of 0.05 /10 yr in SPEI-12.

Figure 3.  Trends of temperature (a), precipitation (b), potential evapotranspiration (c), and drought/flood condition (d) in Shandong Province, China during 1984‒2020. T represents annual mean temperature; P represents annual precipitation; PET represents annual potential evapotranspiration; SPEI-12 represents Standardized Precipitation Evapotranspiration Index values on the 12-month time scale. The UF curve is a sequence of statistics for a positive-order time series for the Mann-Kendall test; the UB curve is a sequence of statistics for a reverse time series for the Mann-Kendall test

To characterize the multi-scale drought, we compute the SPEI for each time scale separately (Fig. 4). The SPEI values for 1984‒1985, 1993‒1996, and 2003‒2013 are primarily positive, indicating that there were more flood events in these years. The SPEI values for 1986‒1992, 1997‒2002, and 2014‒2020 are primarily negative, indicating that there were more drought events in these years.

Figure 4.  Standardized precipitation evapotranspiration index (SPEI) values at different time scale in Shandong Province, China during 1984‒2020. SPEI-1, SPEI-3, and SPEI-6 represent the Standardized Precipitation Evapotranspiration Index values of 1-, 3-, and 6-month time scales, respectively

The frequency of drought and flood events decreases as the time scale increases in Fig. 4. The SPEI values of the 6-month scale (SPEI-6) and 12-month scale (SPEI-12) are too coarse (Figs. 3 and 4). The SPEI value of the 1-month scale (SPEI-1) is not suitable for drought identification (Zhang et al., 2020a), so we calculate the SPEI value of the 3-month scale (SPEI-3) for the following study.

We classify the drought levels in Shandong Province over 37 yr based on the SPEI-3 values and Table 1. The frequency of slight, moderate, severe, and extreme drought is then calculated for each meteorological station. Finally, based on the inverse weight interpolation method, maps of the frequency distribution of different levels of drought in each region are produced. The Jenks natural breaks classification method is used to divide the occurrence frequencies into six ranges to maximize the distinction between the spatial distribution of different frequencies (Asbury and Aly, 2019).

The frequency maps of the spatial distribution of various drought levels are depicted in Fig. 5. The distribution characteristics are as follows: 1) the average frequency of slight drought in Shandong Province was 68 months, and the highest frequency areas were concentrated in Rizhao and Linyi. 2) The average frequency of moderate droughts was 45.21 months, with a higher frequency in all regions except the southwestern part of Shandong Province and parts of Rizhao, Dongying, and Binzhou. 3) The average frequency of severe droughts was 24 months, with high-frequency areas more concentrated in the eastern part of Weihai. 4) The average frequency of extreme drought was 4.11 months, with high-frequency areas more concentrated in Heze, Jining, Qingdao, and Dongying.

Figure 5.  Spatial distribution of the duration of different drought/flood levels in Shandong Province, China during 1984‒2020

Similarly, the SPEI-3 values can be used to categorize flood levels. The frequency distribution characteristics of various flood levels are as follows (Fig. 5): 1) there were approximately 64.47 months of slight flood events in Shandong Province from 1984‒2020, with high-frequency areas in all regions except parts of Jining, Dezhou, Yantai, and Rizhao. 2) The frequency of moderate floods was 44.42 months, with high-frequency areas more concentrated in Rizhao, Linyi, and Tai’an. 3) There were approximately 25 months of severe flood events, with high-frequency areas more concentrated in the coastal areas of Bohai Bay. 4) The frequency of extreme flood events was 7.05 months, with a lower frequency in the province’s central region.

SPEI-3 can be used to describe the conditions of drought and flood on the seasonal time scale, and the drought and flood conditions in spring, summer, autumn, and winter are quantified using SPEI-3 values in May, August, November, and the following February, respectively. The seasonal drought and flood characteristics of Shandong Province are shown in Fig. 6.

Figure 6.  Frequency spatial distribution of seasonal droughts/floods in Shandong Province, China during 1984‒2020

Droughts are classified according to their seasonal distribution (Fig. 6): 1) the frequency of spring drought in Shandong Province was 30.73%, with high-frequency areas concentrated in Shandong Province’s southwestern and northwestern regions. 2) The summer drought frequency was 32.29%, with high-frequency areas in Linyi, Yantai, and Weihai. 3) The frequency of autumn drought was 32.29%, and the frequency of autumn droughts in Linyi was high. 4) The frequency of winter drought was 30.58%, which was higher in all areas except Jining.

The following describes the seasonal distribution of floods (Fig. 6): 1) the frequency of spring floods in Shandong Province was 32.01%, and the frequency of spring floods in Dongying was low. 2) The summer flood frequency was 31.01%, with high-frequency areas concentrated in parts of Binzhou, Yantai, Linyi, Jining, and Zaozhuang. 3) The frequency of autumn floods was 31.01%, with a higher frequency in the southern part of Shandong Province. 4) The winter flood frequency was 28.88%, with high-frequency areas in Rizhao, Yantai, and Dezhou.

Using run theory and the SPEI-3, we obtain three drought characteristics at each meteorological station: Dd, Ds, and Dp. With this as a foundation, Dd′ is set as the sum of the Dd of all meteorological stations in the same year, called the cumulative drought duration. The cumulative drought severity Ds′ is the sum of the Ds of all meteorological stations in the same year. Dp′ is set as the average of the Dp of all meteorological stations in the same year, which is called the average drought peak.

The yearly patterns of Dd′, Ds′, and Dp′ are (Fig. 7): 1) since 1986, the cumulative duration of drought in Shandong Province has increased, with an increasing rate of 6.00 months/10 yr. Furthermore, the longest cumulative drought durations (139 months) occurred in 1999 and 2002. 2) The trend of Ds′ is consistent with Dd′, and the rate of increase of Ds′ was 8.00 units/10 yr. 3) Dp′ has shown an upward trend since 1986, with a growth rate of 0.08 /10 yr.

Figure 7.  Trends of drought/flood characteristic variables in Shandong Province, China during 1984‒2020. Dd′ represents cumulative drought duration; Ds′ represents cumulative drought severity; Dp′ represents average drought peak; Fd′ represents cumulative flood duration; Fs′ represents cumulative flood severity; Fp′ represents average flood peak. UF and UB curves are the same as in Fig. 3

Similarly, Fd′ and Fs′ are set as the sum of the Fd and Fs of all meteorological stations in the same year, respectively, and Fp′ is set as the average of the Fp of all meteorological stations in the same year. Fd′, Fs′, and Fp′ are called the cumulative flood duration, cumulative flood severity, and average flood peak, respectively. The year-to-year patterns of Fd′, Fs′, and Fp′ (Fig. 7) are as follows: 1) the trend of Fd′ showed cyclical changes, and there was an overall decreasing trend with a rate of 3.00 months/10 yr. 2) The trend of Fs′ is consistent with Fd′, and the decreasing rate of Fs′ was 4.00 /10 yr. Furthermore, severe floods occurred in 1990, 1998, and 2003. 3) Fp′ had more years of growth, but the overall growth trend is insignificant, with a growth rate of 0.01 /10 yr.

The cumulative spatial distribution of drought/flood characteristics over 37 yr is shown in Fig. 8. The spatial distribution of drought duration and severity is similar; areas with long drought duration and high drought severity were concentrated in southeastern regions of Shandong Province. Drought peaks were lower in the northwestern Shandong Province and higher in the southwestern. The areas with the most extended flood duration and highest flood severity were found in the southeastern part of Shandong Province. Flood peaks were higher in Weifang, Jining, Qingdao, and Yantai.

Figure 8.  Spatial distribution of drought/flood characteristic variables in Shandong Province, China during 1984‒2020

The trends of various drought/flood characteristic variables for each meteorological station are shown in Fig. 9, and the following are the trends in drought-related variables: 1) the annual drought duration and severity around Jinan and Tai’an are decreasing, i.e., the duration and severity of drought in this area are decreasing yearly. Except for these two areas, the duration and severity of drought in other regions are increasing each year. The northwestern part of Yantai has the highest increase rates in drought duration and severity, with 0.79 months/10 yr and 1.09 /10 yr, respectively. 2) There is an increasing trend of drought peaks in all regions of Shandong Province, which will certainly result in additional food losses. The following are the trends in flood characteristic variables: 1) The duration of annual flood events is increasing in Jinan, the western part of Rizhao, the northern part of Yantai, and the eastern part of Weihai, and there is a downward trend in most other regions. 2) The annual flood severity in Jinan and the northern part of Yantai is increasing, and the severity of flood in other areas are decreasing each year. 3) Flood events are becoming worse in Shandong Province’s central regions.

Figure 9.  Trend maps of drought/flood variables at each meteorological station in Shandong Province during 1984‒2020. The numbers in the figure represent the interannual rate of change

The four variables of food loss due to droughts are calculated by Eq. (1) to Eq. (4), i.e., crop disaster-covered rate due to drought ($ {{A}}_{{1}} $), crop disaster-affected rate due to drought ($ {{B}}_{{1}} $), crop extinction rate due to drought ($ {{C}}_{{1}} $), and drought-related food losses ($ {{L}}_{{1}} $). Similarly, the four variables of food loss due to floods are crop disaster-covered rate due to flood ($ {{A}}_{{2}} $), crop disaster-affected rate due to flood ($ {{B}}_{{2}} $), crop extinction rate due to flood ($ {{C}}_{{2}} $), and flood-related food losses ($ {{L}}_{{2}} $). The trends of food losses due to droughts and floods are depicted in Fig. 10.

Figure 10.  Trends of food loss variables due to droughts/floods in Shandong Province, China during 1984‒2020. $ {{A}}_{\text{1}} $ represents crop disaster-covered rate due to drought; $ {{B}}_{\text{1}} $ represents crop disaster-affected rate due to drought; $ {{C}}_{\text{1}} $ represents crop extinction rate due to drought; $ {{L}}_{\text{1}} $ represents drought-related food losses; $ {{A}}_{\text{2}} $ represents crop disaster-covered rate due to flood; $ {{B}}_{\text{2}} $ represents crop disaster-affected rate due to flood; $ {{C}}_{\text{2}} $ represents crop extinction rate due to flood; $ {{L}}_{\text{2}} $ represents flood-related food losses. UF and UB curves are the same as in Fig. 3

The characteristics of the drought-related food loss variable are shown in Fig. 10: 1) There is a significant downward trend of $ {{A}}_{{1}} $ after 2007. Furthermore, the rate of decline of $ {{A}}_{{1}} $ was 6.9% /10 yr. 2) The abrupt decline in $ {{B}}_{{1}} $ occurred from 2004 to 2006, and the downward trend has been significant since 2010, with a rate of decline of 3.1% /10 yr for $ {{B}}_{{1}} $. 3) $ {{C}}_{{1}} $ fell precipitously in 2003, with a significant downward trend beginning in 2007 at a rate of 0.6% /10 yr. 4) The abrupt decline in $ {{L}}_{{1}} $ occurred from 2004 to 2006, and the downward trend has been significant since 2012, with a rate of decline of 6.3 × 10⁵ t/10 yr for $ {{L}}_{{1}} $. Drought-related food losses were highest in 2002 and 1992, both exceeding 5.0 × 10⁶ t.

The characteristics of the flood-related food loss variable are shown in Fig. 10: 1) $ {{A}}_{{2}} $ showed an insignificant increasing trend from 1990 to 2015, with an overall trend of 0.3% /10 yr since 1984. 2) The trend of $ {{B}}_{{2}} $ changed abruptly in 2013, 2017, and 2019, and there was an overall decreasing trend with a rate of 0.3% /10 yr. 3) The trend of $ {{C}}_{{2}} $ is insignificant, declining at a rate of 0.1% /10 yr. 4) From 1990 to 2016, the amount of food lost due to floods increased, with an overall growth rate of 1.0 × 10⁴ t/10 yr over 37 yr. Flood-related food losses exceeded 2.0 × 10⁶ t in 1993, 2003, 2012, and 2013.

The correlation and linear regression analyses between the bivariate variables are synchronous in our result, with no ahead or behind. Using the Pearson correlation coefficient, we can calculate the correlation between food loss due to droughts and drought-related elements (Table 2). The crop disaster-covered rate due to drought ($ {{A}}_{{1}} $) is highly correlated with T, P, PET, Dd′, and Ds′. Furthermore, $ {{A}}_{{1}} $ is the most strongly correlated with P, followed by PET, both have a strong negative correlation with $ {{A}}_{{1}} $. The crop disaster-affected rate due to drought ($ {{B}}_{{1}} $) is highly correlated with P, Dd′, and Ds′. The crop extinction rate due to drought ($ {{C}}_{{1}} $) is significantly correlated with P, Dd′, and Ds′, which is higher correlated with P. P, Dd′, Ds′, and Dp′ are significant impact factors affecting the amount of food lost $ {{L}}_{{1}} $.

We conducted a linear regression analysis after obtaining the relevant factors affecting the drought-related food loss variables to quantify the degree of impact of those factors (Table 3). The explained variables are $ {{A}}_{{1}} $, $ {{B}}_{{1}} $, $ {{C}}_{{1}} $, and $ {{L}}_{{1}} $, while the explanatory variables are T, P, PET, Dd′, Ds′, and Dp′. The degree of grain yield reduction caused by drought-related elements is as follows: 1) when the annual mean temperature rises by 1°C, $ {{A}}_{{1}} $ decreases by 7.764%. 2) When annual precipitation increases by 1 mm, $ {{A}}_{{1}} $ decreases by 0.060%, $ {{B}}_{{1}} $ by 0.033%, $ {{C}}_{{1}} $ by 0.007%, and $ {{L}}_{{1}} $ by 7.930 × 10³ t. 3) When the annual potential evapotranspiration increases by 1 mm, $ {{A}}_{{1}} $ decreases by 0.145%. 4) When the cumulative drought duration increases by one month, i.e., the average drought duration per meteorological station increases by 1/19th of a month, $ {{A}}_{{1}} $ increases by 0.127%, $ {{B}}_{{1}} $ by 0.073%, $ {{C}}_{{1}} $ by 0.016%, and $ {{L}}_{{1}} $ by 1.939 × 10⁴ t. 5) When the cumulative drought severity increases by 1 unit, $ {{A}}_{{1}} $ increases by 0.097%, $ {{B}}_{{1}} $ by 0.057%, $ {{C}}_{{1}} $ by 0.013%, and $ {{L}}_{{1}} $ by 1.562 × 10⁴ t. 6) When the average drought peak increases by 1 unit, $ {{L}}_{{1}} $ increases by approximately 1.511 × 10⁶ t.

The Pearson correlation coefficient is used to calculate the relationship between flood-related food loss and flood-related elements (Table 4). The results of the analysis are as follows: 1) The crop disaster-covered rate due to flood ($ {{A}}_{{2}} $) is highly correlated with P, Fd′, Fs′, and Fp′. 2) The crop disaster-affected rate due to flood ($ {{B}}_{{2}} $) is highly correlated with T, P, Fd′, Fs′, and Fp′. 3) The crop extinction rate due to flood ($ {{C}}_{{2}} $) is significantly correlated with P, Fd′, Fs′, and Fp′. 4) P, Fd′, Fs′, and Fp′ are significant impact factors affecting the amount of food lost $ {{L}}_{{2}} $.

The linear regression analysis between variables with high correlation is shown in Table 5. The explained variables are: $ {{A}}_{{2}} $, $ {{B}}_{{2}} $, $ {{C}}_{{2}} $, and $ {{L}}_{{2}} $, while the explanatory variables are T, P, Fd′, Fs′, and Fp′. The degree of grain yield reduction caused by flood-related elements is as follows: 1) When the annual mean temperature rises by 1°C, $ {{B}}_{{2}} $ decreases by 1.319%. 2) When annual precipitation increases by 1 mm, $ {{A}}_{{2}} $ increases by 0.019%, $ {{B}}_{{2}} $ by 0.010%, $ {{C}}_{{2}} $ by 0.003%, and $ {{L}}_{{2}} $ by 3.000 × 10³ t. 3) When the cumulative flood duration increases by one month, the average flood duration per meteorological station increases by 1/19th of a month, $ {{A}}_{{2}} $ increases by 0.076%, $ {{B}}_{{2}} $ by 0.040%, $ {{C}}_{{2}} $ by 0.013%, and $ {{L}}_{{2}} $ by 1.134 × 10⁴ t. 4) When the cumulative flood severity increases by one unit, $ {{A}}_{{2}} $ increases by 0.058%, $ {{B}}_{{2}} $ by 0.030%, $ {{C}}_{{2}} $ by 0.009%, and $ {{L}}_{{2}} $ by 8.470 × 10³ t. 5) When the average flood peak increases by 1 unit, $ {{A}}_{{2}} $ increases by 6.647%, $ {{B}}_{{2}} $ by 3.468%, $ {{C}}_{{2}} $ by 1.176%, and $ {{L}}_{{2}} $ by 1.034 × 10⁶ t.

The SPEI based on the Thornthwaite method is appropriate for quantitative drought analysis in Shandong (Zuo et al., 2018; Ren et al., 2021; Yao et al., 2021); additionally, Zuo et al. (2018) recommended the use of SPEI-3 to assess agricultural drought conditions in Shandong. As a result, the calculations in this paper are reasonable. Over the next 20 yr, the global temperature rise is expected to reach or exceed 1.5°C (Intergovernmental Panel on Climate Change, 2021), significantly increasing future crop evapotranspiration. However, the change in precipitation is minor, resulting in an imbalance in the supply and demand for water resources and a significant aridity trend in the area. According to Zuo et al. (2018), the overall temperature trend in Shandong Province is increasing, and there is a trend of aridity. This is consistent with the findings of this paper, and both indicate that the trend of precipitation changes in Shandong Province is not significant. Ren et al. (2021) found a significant increase in temperature and a non-significant increase in precipitation in Shandong Province from 1980 to 2020, which is consistent with the findings of this study. Both this paper and Liu et al. (2015) found an increasing trend in drought duration in Shandong Province from year to year. Still, the year-to-year variation in drought severity described in this paper differed from theirs due to inconsistency in the study years. The study concluded that Linyi experiences frequent droughts and floods, which is supported by previous research (Lei et al., 2013; Yao et al., 2021). The root cause of this phenomenon is that January has the least amount of precipitation, accounting for 1.2% of the year, while July has the most precipitation, accounting for 28.6% of the year (Sun and Pei, 2018). As a result, Linyi is vulnerable to droughts in the winter and floods in the summer.

Since three drought characteristic variables are positively correlated with grain loss, there is the highest rate of increase in drought duration and severity in the northwestern part of Yantai (Fig. 9), which indicates that grain loss will be high in this area. Three flood characteristic variables are positively correlated with food losses. The duration of annual flood events is increasing in Jinan, the western part of Rizhao, the northern part of Yantai, and the eastern part of Weihai (Fig. 9), and the severity of annual floods is increasing in the northern parts of Jinan and Yantai (Fig. 9), so it is important to focus on flood control in these areas to reduce food losses. Furthermore, based on the amount of yearly increase in drought/flood characteristic variables for each region, we can predict the yearly increase in food loss. It is the product of the annual increase of drought/flood characteristic variables (Fig. 9) and the quantitative values of drought/flood and food loss (Table 3 and Table 5), which can provide decision support for disaster managers.

Because drought is spatially heterogeneous, correlation analysis on a small area scale is more accurate. Because small-scale disaster loss data are difficult to obtain, this paper quantifies the mechanism of droughts and floods on food losses at the provincial levels.

Droughts and floods in Shandong Province have severely harmed agricultural production. We investigated the relationship between drought/flood-related factors and food losses caused by drought/flood using methods such as the Mann-Kendall test, Pearson correlation coefficient, and linear analysis. Furthermore, the spa­tiotemporal characteristics of droughts and floods in Shandong Province were investigated. The study’s findings are as follows.

(1) On the time scale, the temperature and potential evapotranspiration in Shandong Province of China are increasing, while precipitation is not changing significantly. This has resulted in an aridity trend in Shandong Province, with year-to-year increases in drought duration, severity, and peak. On the spatial scale, drought severity is positively related to food losses, so extreme drought causes the most losses to food production. Extreme drought events occur mostly in the southwestern part of Shandong Province, mainly in Heze and Jining, so attention must be paid to drought mitigation in these areas. Slight floods are the most common flood events in Shandong Province, and extreme flood events that caused severe food losses occurred mostly in the western and northeastern parts of Shandong Province. At different seasonal scales, the spatial distribution of the frequency of drought and flood events varies significantly. Overall, spring and winter droughts are common in the northwest of Shandong Province, while summer and autumn floods are common in the southeast of Shandong Province.

(2) Based on historical disaster data, we accurately estimated the amount of food loss due to drought and flood. The food losses caused by droughts in Shandong Province are more than those caused by floods, and the food loss caused by drought shows a decreasing trend yearly, with a decrease rate of 6.3 × 10⁵ t/10yr. The amount of food lost due to floods increased with an overall growth rate of 1.0 × 10⁴ t/10 yr over 37 yr.

(3) The impact of precipitation on food loss due to drought/flood is significant, followed by potential evapotranspiration and temperature. Furthermore, the relationship between drought and flood conditions and food losses can be precisely quantified. The amount of food lost due to drought is positively correlated with the drought characteristics; a one-month increase in cumulative drought duration obtained by counting all meteorological stations data will result in 1.939 × 10⁴ t of grain loss. If the cumulative drought severity and average drought peak increase by one unit, food loss due to drought will increase by 1.562 × 10⁴ t and 1.511 × 10⁶ t, respectively. The amount of food lost due to flood is significantly positively correlated with flood characteristics, with an increase in cumulative flood duration of one month resulting in 1.134 × 10⁴ t of grain loss. If the flood severity and average flood peak increase by one unit, food loss will increase by 8.470 × 10³ t and 1.034 × 10⁶ t, respectively.