Keywords

1 Introduction

Increasingly severe climate change and human activities lead to more and more frequent global extreme climate disasters (Shi et al. 2020). The drought-flood abrupt alternation (DFAA) is caused by the complex relationship between drought and flood, which has developed into one of the most destructive disasters (Shi et al. 2021). It can cause enormous socioeconomic losses and seriously threaten agriculture, ecological environment, and human health. Therefore, understanding the spatiotemporal pattern of DFAA and its driving mechanism is of great significance for water resources management and early warning and mitigation of DFAA.

DFAA is quantified and monitored by the standardized precipitation index, long-/short-cycle drought-flood abrupt alternation index (LDFAI/SDFAI), and runoff abrupt change index. Among these indices, LDFAI and SDFAI were proposed in 2006 (Wu et al. 2006) and have the advantages of single input, namely only precipitation data is needed, and a simple calculation process, which become the most popular. The government and the public have paid great attention to when and where DFAA occurs. Numerous scholars have carried out multiple regional DFAA assessments. For instance, Li et al. (2017) defined the DFAA index using streamflow to identify DFAA of the Poyang Lake catchment in the past 50 years and found that drought to flood (DTF) and flood to drought (FTD) were concentrated in March to April and July to August, respectively. Spatiotemporal evolution characteristics of DFAA in Shanxi Province in summer were fully revealed using LDFAI (Liu et al. 2017).

The above studies mainly focused on the DFAA index with a single time scale to evaluate the changes of DFAA in a specific period. There was a lack of evaluation for DFAA in different periods. Therefore, it is necessary to strengthen the quantitative study of DFAA in different periods based on multi-time scale indicators. For example, Shan et al. (2015) comprehensively evaluated the spatiotemporal distribution of DFAA in the middle and lower reaches of the Yangtze River (MLYR) based on LDFAI and SDFAI and found that long-/short-cycle DFAA in this region was dominated by FTD. In addition, The change trends of LDFAI, SDFAI, and monthly precipitation in the growth period of tobacco were analyzed (Zhang et al. 2019).

DFAA is often affected by a variety of factors, such as precipitation, temperature, and evaporation, which directly affect DFAA. Moreover, many studies have shown that large-scale climate factors are closely related to DFAA. For instance, DFAA in the MLYR is related to phenomena of continuously low SST in the Nino 3.4 region, especially the La Nina phenomenon, which can provide an early warning for DFAA (Shan et al. 2018). Precipitation anomaly and its dynamic factors and water vapor sources during the DFAA period in the MLYR from May to June in 2011 were also comprehensively analyzed (Wang et al. 2014). However, these studies only focus on the influence of a single atmospheric index, or the causes of DFAA change in a particular year, while the relationship between different atmospheric indexes with DFAA and the driving mechanism of DFAA in historical/recent years has not been fully revealed. Therefore, the main objectives of this study are to: (1) analyze the spatiotemporal variation characteristics of long-/short-cycle DFAA; (2) comprehensively evaluate the potential driving mechanisms of various atmospheric factors on DFAA. The results would help to understand the evolving pattern of DFAA under global warming and provide guidance for regional water resources management.

2 Study Area and Data

2.1 Study Area

The Yellow River basin (YRB), which is located between 96°– 119° E, 32° –42° N (Fig. 1), is surrounded by the Bohai Sea, Bayankara mountain, Qinling Mountains, and Yinshan Mountains in the east, west, south, and north, respectively (Ji et al. 2021). The YRB has a total drainage area of 795,000 km2 including 42,100 km2 of the inner basin, with a length of about 1900 km from east to west and a span of about 1100 km from north to south. The climatic characteristics of the YRB are complicated due to the monsoon and atmospheric circulation. The annual average precipitation and the annual average evaporation are concentrated in 200– 800 mm and 800 – 1800 mm, respectively, which is one of the typical climate-sensitive areas in China (Zhao et al. 2019). Under the influence of global warming and human activities, drought/flood disasters occur frequently in the YRB. Furthermore, the YRB is an important agricultural base in China, these frequent natural disasters seriously threaten the safety of agricultural products. According to the characteristics of the drainage system, the YRB is divided into the upper (UYRB) and middle and lower (MLYRB).

Fig. 1.
figure 1

The geographical position of the Yellow River basin.

Full size image

2.2 Data Source

In this study, monthly precipitation data with the period for 1961–2020 was available from the Climate Research Unit (CRU) (https://www.cru.uea.ac. uk/data), which has a spatial grid of 0.5° × 0.5°. The nearest neighbor interpolation method was used to interpolate the precipitation data to a 0.25° grid. The monthly atmospheric circulation factors over the same data, including Arctic Oscillation (AO), Pacific Decadal Oscillation (PDO), El Niño Southern Oscillation (ENSO), and sunspot, download from the National Oceanic and Atmospheric Administration (NOAA) (https://psl.noaa.gov/arctic/data/).

3 Methods

3.1 Long-Cycle Drought-Flood Abrupt Alternation Index

The precipitation in the YRB has obvious seasonal characteristics, and mainly concentrates from June to September (Xu and Zhang 2006). Therefore, the precipitation during this period was selected as summer precipitation to quantitatively reveal the evolution of the long-/short-cycle drought-flood abrupt alternation in this basin. The time scale of the long-cycle drought-flood abrupt alternation index (LDFAI) is 2 months, namely, June-July is drought, August-September is flood, which is defined as drought to flood (DTF) (Shi et al. 2021), and vice versa.

$$LDFAI = (R_{89} - R_{{{67}}} ) \times (\left| {R_{67} } \right| + \left| {R_{89} } \right|) \times 1.8^{{{ - }\left| {R_{67} { + }R_{89} } \right|}}$$
(1)

where, \(R_{{{67}}}\) and \(R_{89}\) are standardized precipitation from June-July and August-September, respectively; \((R_{89} - R_{{{67}}} )\) and \(\left| {R_{67} } \right| + \left| {R_{89} } \right|\) represent intensity term of DFAA and drought/flood, respectively; \(1.8^{{{ - }\left| {R_{67} { + }R_{89} } \right|}}\) is the weighting coefficient. If LDFAI is greater than 1, it means DTF; if LDFAI is less than −1, it means flood to drought (FTD), and the rest are normal (Yuan et al., 2021).

3.2 Short-Cycle Drought-Flood Abrupt Alternation Index

The short-cycle drought-flood abrupt alternation index (SDFAI), with a time scale of 1 month, is similar to the definition of LDFAI (Zhang et al. 2019). The specific formula is as follows.

$$SDFAI = (R_{{\text{j}}} - R_{{\text{i}}} ) \times (\left| {R_{i} } \right| + \left| {R_{j} } \right|) \times 3.2^{{{ - }\left| {R_{i} { + }R_{j} } \right|}}$$
(2)

where, \(j = i + 1\) \((i = 6,7,8)\); the evaluation criteria of SDFAI are consistent with LDFAI.

4 Results and Discussion

4.1 Temporal Characteristics of Drought-Flood Abrupt Alternation

4.1.1 Temporal Pattern of Long-Cycle Drought-Flood Abrupt Alternation

LDFAI was calculated using the monthly precipitation of the YRB during 1961–2020. Temporal variation of LDFAI is of great significance for understanding the long period of DFAA events in the basin. Figure 2 presents the temporal variation of LDFAI in the UYRB and MLYRB, respectively. The red dotted line indicates the changing trend of LDFAI.

Fig. 2.
figure 2

Temporal variation of LDFAI in the UYRB (a) and MLYRB (b).

Full size image

As can be seen from Fig. 2(a), the inter-annual variation of LDFAI was large, reaching a maximum value of 2.69 and a minimum value of −3.87 in 2001 and 1999, respectively. Overall, the slope of LDFAI is −0.117/10a in the UYRB, indicating that LDFAI presents a downward trend in the UYRB. This further illustrates that FTD tends to increase while DTF tends to decrease, which is consistent with the recent study by Shi et al. (2021). The temporal variation of LDFAI in the MLYRB is the same as that in the UYRB, with a maximum value of 4.44 in 1985 and a minimum value of −2.96 in 2016 (Fig. 2(b)). The LDFAI in the MLYRB decreased at a rate of 0.186/10a, indicating that the region was also prone to FTD. In addition, the decreasing rate of LDFAI in the MLYRB is significantly higher than that in the UYRB, which illustrates that FTD of the MLYRB is more frequent in the flood season. Liu et al. (2012) found that the increase and decrease of precipitation in the flood season of the YRB during 1961–2010 mainly occurred in June and July-September, respectively, and the increasing trend of precipitation in June was higher than the decreasing trend in July. This conclusion confirms that FTD is prone to occur in the YRB during flood season.

4.1.2 Temporal Pattern of Short-Cycle Drought-Flood Abrupt Alternation

This study quantified the change of SDFAI to further reveal the evolution characteristics of DFAA in the flood season of the YRB. Figures 3, 4 and 5 show the temporal variation of SDFAI in the YRB from June to July, July to August, and August to September, respectively. The SDFAI of the UYRB and MLYRB showed a slight downward trend from June to July, and the changing rates were 0.104/10a (Fig. 3(a)) and 0.0748/10a (Fig. 3(b)), respectively. The variation trend of SDFAI from July to August (Fig. 4) is the same as that from May to June, which indicates that FTD occurs frequently in the YRB from June to July and July to August. It is worth noting that the variation trend of SDFAI in the UYRB is greater than that in the MLYRB, indicating that FTD is more likely to occur in the UYRB during this period. Interestingly, the change rate of SDFAI in the UYRB from August to September was 0.107/10a (Fig. 5(a)), indicating that SDFAI showed an upward trend, namely DTF occurred frequently in this region. However, SDFAI in the MLYRB decreased at a rate of 0.138/10a in this period, indicating that FTD in this region increased while DTF decreased. Meanwhile, DTF and FTD occurred before and after 1985 in the MLYRB, respectively. Xu and Zhang (2006) showed that precipitation in the lower of the YRB (LYRB) decreased significantly in September, which verified the reliability of this study. That is, the decline rate of SDFAI from August to September was significantly higher than that of other periods.

Fig. 3.
figure 3

Temporal variation of SDFAI67 in the UYRB (a) and MLYRB (b), SDFAI67 stands for SDFAI from June to July.

Full size image
Fig. 4.
figure 4

Temporal variation of SDFAI78 in the UYRB (a) and MLYRB (b), SDFAI78 stands for SDFAI in July-August.

Full size image
Fig. 5.
figure 5

Temporal variation of SDFAI89 in the UYRB (a) and MLYRB (b), SDFAI89 stands for SDFAI in August-September.

Full size image

4.2 Spatial Characteristics of Drought-Flood Abrupt Alternation

4.2.1 Frequency of Drought-Flood Abrupt Alternation

The frequency of DTF and FTD in the YRB are shown in Fig. 6 and Fig. 7, respectively, and there are significant spatial differences in DFAA frequency in different periods. The spatial distribution of precipitation in the YRB is uneven and affected by topography and geomorphology (Li et al. 2016), which results in regional and complex distribution of DFAA. In general, the frequency of long-period DFAA in the YRB is more than that of short-period DFAA.

Figure 6(a) shows that DTF frequency is high in most areas of the MLYRB, especially the below Huayuankou, while it is slightly lower in the UYRB. For short DFAA, DTF is mainly concentrated in June to July (Fig. 6(b)), followed by August to September (Fig. 6(c)), and finally in July to August (Fig. 6(d)). The source region and the central region of the YRB were vulnerable to DTF impact from June to July, while sporadic DTF damage areas appeared on the southern edge of the MLYRB from July to August. Furthermore, from August to September, DTF with high frequency is mainly distributed in the inner flow region of the UYRB.

For the long period (Fig. 7(a)), the spatial distribution pattern of FTD in the Yellow River is similar to that of DTF on the whole, but there was a slight difference locally. As can be seen from Fig. 7(b)-(d), the spatial distribution of FTD frequency is quite different in three periods of flood season. Among them, the high-frequency FTD in June-July and July-August was concentrated in the central region of the YRB, while the FTD in August-September was mainly distributed in the eastern edge of the MLYRB. Comparing Fig. 6(d) and Fig. 7(d), we can find DFAA types in the UYRB and MLYRB are opposite from August to September, namely DTF is dominant in the UYRB, while FTD is dominant in the MLYRB. Moreover, the frequency of FTD was highest from June to July, followed by July to August, and lowest from August to September. To sum up, short-cycle DFAA is mainly concentrated in June and July. FTD is more frequent than DTF, which is consistent with the results in Sect. 4.1, namely LDFAI/SDFAI shows a declining trend overall.

Fig. 6.
figure 6

Spatial distribution of DTF frequency in the YRB: (a) June-September, (b) June-July, (c) July-August, and (d) August-September.

Full size image
Fig. 7.
figure 7

Spatial distribution of FTD frequency in the YRB: (a) June-September, (b) June-July, (c) July-August, and (d) August-September.

Full size image

4.2.2 The Trend of Drought-Flood Abrupt Alternation

The Mann-Kendall (MK) test was used to detect the existence of possible trends in DFAA. Figure 8 presented the spatial distribution of test statistic Z values at different time scales during the study period. Z > 0 indicates that LDFAI/SDFAI has an increasing trend, namely DTF is dominant in this region, on the contrary, FTD is dominant. The Z values of LDFAI (Fig. 8(a)) and SDFAI (Fig. 8(b)) from June to July are negative and distributed in a patchy pattern in the YRB, indicating that FTD occurred frequently in these periods. From July to August, negative Z was mainly found in the UYRB, showing speckled distribution, while a few positive Z appeared in the eastern part of the MLYRB (Fig. 8(c)). On the contrary, the positive and negative areas of SDFAI trend from August to September were concentrated in the UYRB and MLYRB, respectively (Fig. 8(d)), and there was a clear dividing line, indicating that the UYRB and MLYRB are DTF and FTD disturbance areas, respectively. This result is consistent with Sect. 4.2.1.

4.3 Impacts of Atmospheric Circulation on Drought-Flood Abrupt Alternation

After understanding the evolution of DFAA, it is necessary to reveal its possible driving mechanism. Regional precipitation anomalies are closely related to the large-scale circulation background (Liu et al. 2016). DFAA is no exception, mainly influenced by solar radiation and atmospheric circulation (Zhang et al. 2021). Solar radiation, as reflected by the sunspot index, is a key factor in climate formation (Hendon and Glick 1995). In addition, DFAA is not only affected by the atmospheric circulation of the same period but also affected by the earlier period, namely there is a lag correlation (Sun et al. 2006). Therefore, this study calculated the correlation coefficients between LDFAI/SDFAI and atmospheric circulation factors (such as AO, PDO, ENSO, and sunspot) based on the Pearson correlation method, considering simultaneity and lag correlation. The maximum correlation coefficients are shown in Table 1.

Fig. 8.
figure 8

Spatial distribution of DFAA trend in the YRB: (a) June-September, (b) June-July, (c) July-August, and (d) August-September.

Full size image
Table 1. The correlation coefficient between LDFAI/SDFAI of YRB and atmospheric circulation factor.
Full size table

There was a significant positive correlation between LDFAI/SDFAI and AO in the YRB (Table 1). Previous study found that AO affects the middle and high latitudes of Eurasia mainly by influencing the Siberian high, which in turn changes the winter wind disturbance (Peng et al. 2021). In addition, a negative correlation was found for PDO between LDFAI/SDFAI. When PDO is in the cold phase, it tends to lead to more summer precipitation and shows a negative correlation (Setiawan et al. 2017). Compared with the UYRB, ENSO has a stronger correlation in the MLYRB. This is consistent with previous study (Dan et al. 2015), namely ENSO is weakly correlated with precipitation in the UYRB, while it is significantly negatively correlated with precipitation in the MLYRB, indicating that precipitation weakens when ENSO intensity increases. On the contrary, sunspot showed a weak negative correlation overall.

For long-cycle DFAA, LDFAI in the UYRB and MLYRB have the largest correlation coefficient with PDO and AO, respectively. For short-cycle DFAA, SDFAI67 and SDFAI78 of the YRB had the strongest correlation with PDO and AO, respectively. AO and PDO had the highest correlation with SDFAI89 in the UYRB and MLYRB, respectively. However, the interdecadal effect of atmospheric circulation on DFAA could not be derived using the correlation analysis, thus the cross-wavelet transform method was used to further investigate the driving process between them based on the maximum correlation coefficient.

According to Fig. 9(a), 1–3a antiphase resonance and 1-4a positive phase resonance existed in LDFAI and PDO during 1981–1986 and 1997–2005, respectively. Figure 9(b) shows a significant correlation between PDO and SDFAI from June to July, with periods of 14-18a and 0-2a during 1975–2004 and 2000–2006, respectively. AO had three significant resonance periods with SDFAI78, which were periodicities of 5a, 8-11a, and 1-4a during 1966–1975, 1997–1973, and 2012–2018, respectively (Fig. 9(c)). Moreover, there were three significant resonance cycles between AO and SDFAI from August to September, namely short-term oscillation of 1 and 1-5a (1968–1972 and 2012–2017) and long-term oscillation of 10-14a (2000–2012) (Fig. 9(d)). Previous research found that AO tended to increase and decrease during 1990–2000 and 2000–2014, respectively (Sun et al., 2018). Overall, AO showed a significant positive correlation with SDFAI78 and SDFAI89 after 2000, that is, AO can drive the decrease of SDFAI in the UYRB, indicating that AO is the key factor affecting the occurrence of FTD in this period.

In the MLYRB, LDFAI and AO showed significant oscillation cycles of about 8a, 5a, and 2-5a during 1969–1998, 1980–1986, and 2010–2017, respectively (Fig. 10 (a)). For short-scale DFAA, PDO had significant periods of 2-4a, 9a, and 3a during 1981–1991, 1990–2006, and 2008–2011, respectively, which passed the red noise test with a confidence level of 95% (Fig. 10(b)). There was a significant correlation between AO and SDFAI78 during 1984–1991 and 2008–2017, with a periodicity of about 17a and 4a, respectively (Fig. 10 (c)). Figure 10 (d) shows a significant positive correlation between PDO and SDFAI89 from 1980 to 1988, with a period of 3a. Moreover, a significant negative correlation with a periodicity of 12a was found during 1990–2014. Wang et al. (2016) found that PDO had a cold and warm phase exchange around 2000, which may be the main factor inducing short-cycle DFAA in the MLYRB. In addition, high index phase AO can weaken the summer monsoon and reduce the heat exchange between the north and the south, leading to regional precipitation anomaly and inducing DFAA of the YRB. Therefore, AO and PDO can be regarded as the dominant factor driving DFAA during the flood period in the YRB.

Fig. 9.
figure 9

Cross-wavelet transform between circulation factors with various indices in the UYRB, such as LDFAI (a), SDFAI67 (b), SDFAI78 (c), and SDFAI89 (d). The 95% confidence interval is represented by a thick black line; the relative phase relationship is represented by arrows (left anti-phase, right in-phase); the up and down arrows indicate drivers before or after DFAA, respectively.

Full size image
Fig. 10.
figure 10

Cross-wavelet transform between circulation factors with various indices in the MLYRB, such as LDFAI (a), SDFAI67 (b), SDFAI78 (c), and SDFAI89 (d).

Full size image

5 Conclusions

This study evaluated the spatiotemporal pattern of DFAA at multiple time scales for flood season from 1961 to 2020 of the YRB. Furthermore, the relationship between large-scale climate drivers, including AO, PDO, ENSO, and sunspot, with DFAA was fully revealed using the Pearson correlation method and cross wavelet method. Results showed that the spatiotemporal distribution of DFAA was significantly different. First, LDFAI/SDFAI almost tended to decline, that is, FTD occurred frequently in the YRB from 1961 to 2020. The frequency of long-period DFAA is significantly higher than that of short-period DFAA, which is distributed in patches throughout the YRB. Second, based on the MK trend test, LDFAI and SDFAI67 tend to decrease in the YRB, indicating that FTD is prone to occur in this period. In addition, SDFAI78 tends to increase in the UYRB and decrease in the MLYRB, while SDFAI89 has the opposite trend. Third, results indicated a strong influence of AO and PDO in the occurrence of the DFAA in the YRB. Meanwhile, the cross-wavelet results showed that the occurrence of DFAA is closely related to the enhancement of AO and the phase change of PDO in the YRB since 2000.

As a complex emerging disaster, DFAA is induced by many factors and its actual physical driving mechanism is complex, such as topography, vegetation coverage, and human activities, which are not considered in this study. Therefore, the influence of these factors on DFAA should be analyzed from the perspective of the mechanism. Nonetheless, we provide new insights into how large-scale climate factors force DFAA and should help promote research in this area.