Long-term spatiotemporal variability of precipitation and its linkages with atmospheric teleconnections in the Yellow River Basin, China

Precipitation is an irreplaceable link in the local or global hydrological cycle process. Its variability regulates the plant reproduction and growth, the amount of river runoff, and phenology as well as the ecosystem stability in the arid and semi-arid region of the Yellow River Basin (YRB) (Wang et al. 2017; Xue et al. 2017). Moreover, part of the terrestrial freshwater mainly comes from precipitation, which is a limited and scarce resource of the region. It is not only owing to its unpredictable and unstable availability but also its severe scarcity in the YRB (Liu et al. 2019a). The precipitation variations have shown significant temporal and spatial variability at the basin due to the complexity of climate patterns and terrain of the basin (Gao & Wang 2017). Meanwhile, human-induced greenhouse gas emissions combined with climate changes have contributed to global warming, which is further intensified by the redistribution of global precipitation resources in space and time (Liu et al. 2019a). Consequently, the trends and variations of precipitation need to be continuously explored and studied, especially the precipitation spatiotemporal patterns and underlying causes.

Precipitation is a typical climate variable that varies over time and space, which would be more suitable for analyzing precipitation variability by linking temporal and spatial changes. Moreover, with the rapidly changing environment, the precipitation series have a significant nonstationary behavior because of the effects of multifarious driving factors (Guan et al. 2019). Hence, substantial studies have been made in exploring the spatiotemporal and trend changes of precipitation in various parts of the world (Qin & Xie 2016; Wu et al. 2016; Yang et al. 2020; Mathew et al. 2021). Zhang et al. (2014) pointed out that the precipitation was gradually decreasing and the rainstorm events were unevenly distributed in the YRB. The change in precipitation also attracts more attention in the southern Mongolian plateau, including the duration and frequency of the precipitation (Wang et al. 2021b). Moreover, Gherardi & Sala (2019) found that extreme precipitation events tended to increase over the global arid regions at different temporal scales. The extreme precipitation events also exhibited an increased tendency in spatial coverage and frequency in the the Yangtze River Basin since 1970 (Li et al. 2021). Global climate change has brought varying degrees of impact on precipitation in various regions. Consequently, the study of spatiotemporal patterns and trend variability of precipitation is imperative in the Yellow River Basin (YRB).

The spatiotemporal variability of precipitation is strongly correlated with global atmospheric teleconnections (Nalley et al. 2019; Wang et al. 2020). There was an intimate association between the precipitation dipole pattern and the summer North Atlantic Oscillation (NAO) in the Tibetan Plateau. However, the relationship had been weakened because of the regional variation of atmospheric circulation, which is caused by the asymmetrical wave propagation around Eurasia since the late 1990s (Liu et al. 2021). Mohammadi et al. (2020) investigated the temporal and spatial teleconnections among Peruvian precipitation and other oceanic oscillations. And they pointed out the El Niño/Southern Oscillation (ENSO) is the major climate control factor for extremely dry or wet conditions in Peru. Xu et al. (2015) mentioned the change in Pacific Decadal Oscillation (PDO) phase might affect the precipitation variation of southern China because of the atmospheric circulation characteristics over Eurasia. Besides, substantial research has indicated that the Arctic Oscillation (AO) was a major synoptic factor influencing the change of precipitation in the middle to high latitudes of the Northern Hemisphere (Givati & Rosenfeld 2013; Wang et al. 2020). Recently, the comprehensive effects of various atmospheric circulation elements on spatial consistency and time variation of precipitation have been a research hotspot (Jiang et al. 2013; Nalley et al. 2019; Li et al. 2021). For instance, the ENSO combined with PDO has a dominant integrated effect on the winter regional extreme rainfall at large scales on time in the Yangtze River Basin (Li et al. 2021). Nalley et al. (2019) emphasized that the collective influence of multiple teleconnection factors should be considered when explaining the precipitation variability. Additionally, the YRB has a large range of longitude and latitude, and therefore, the characteristics of the climate system and its causes are very complex (Liu et al. 2019b; Wang et al. 2022). The precipitation variability over time and space of the YRB is not only affected by internal dynamical causes and thermal conditions, but also by the external forcing factors such as the thermal forcing of the Tibetan Plateau and various atmospheric circulations (Liu et al. 2008, 2021; Zhang et al. 2014; Xu et al. 2015).

Nevertheless, there are fewer studies on long-term precipitation trend variations and spatiotemporal consistency in the YRB. And then, the individual or coupled effects of various teleconnection factors on the precipitation are especially far less known. The individual or coupled effects of various teleconnection factors on precipitation are especially far less known. And then, the bivariate wavelet coherence (WTC) provided us with effective methods to reveal the interactions of two nonlinear hydrometeorological time series at different time–frequency domains (Wang et al. 2021a). However, the WTC method would lose its usefulness when more than three variables are involved in the study. Hence, the novel approach of multiple wavelet coherence (MWC) is proposed based on the WTC method (Hu & Si 2016). Moreover, Hu & Si (2016) also compared the MWC with multivariate empirical mode decomposition and multiple spectral coherence methods and they pointed out that the MWC method had a significant advantage in revealing the scale independence relationships of multiple variables.

Thus, the main objectives of this study are as follows: (1) to assess the long-term spatiotemporal trends and distribution variations of precipitation in the YRB; (2) to evaluate the precipitation gradient against the elevation, longitude, and latitude; (3) to analyze the dominant spatiotemporal modes of annual precipitation; and (4) to explore the individual or comprehensive impacts of diversified atmospheric circulation factors (ENSO, NAO, PDO, and AO) on the precipitation variability in the YRB.

The database of precipitation covering every month from 1950 to 2019 is used to explore and analyze the spatial and temporal variations of precipitation in the YRB. The precipitation data of 82 national gauging stations are obtained from the National Meteorological Center of China (http://data.cma.cn/), and the distribution of the weather stations can be seen in Figure 1. It is an official and authoritative meteorological unit, and the data are of high enough quality after being reviewed by many parties. Besides, the missing precipitation data are extended based on the interpolation method in the study. The annual and monthly total precipitation of the YRB are aggregated from all stations by the Thiessen polygon method (Ye et al. 2020).

Global climate changes have significant indirect impacts on precipitation variations. In this study, four dominant circulation pattern indices are used to discuss and analyze the potential mechanism of spatial and temporal pattern changes of precipitation in the YRB, including the AO, the NAO, the PDO, and the ENSO. The empirical orthogonal function (EOF) method is employed to explore the winter sea level pressure (SLP). It found there is a zonal symmetric structure because of the leading mode of SLP variations poleward of 20N° in the Northern Hemisphere. This global scale circulation mode is regarded as the AO and the corresponding principal component (PC) time series is named the Arctic Oscillation Index (AOI) (Thompson & Wallace 1998). Similarly, the index of NAO is used to express the divergence of zonal mean monthly normalized SLP anomalies between low-pressure centers in Southwest Iceland and high-pressure centers in Gibraltar (Jones et al. 1997). The index of PDO can describe the major interdecadal changes in the large-scale climate in the Pacific Ocean. It is the outcome of the EOF analysis on the monthly mean sea surface temperature (SST) anomaly north of 20N° in the North Pacific Ocean (Mantua et al. 1997). The ENSO is the most crucial interannual ocean–atmosphere interaction mode in the equatorial Pacific, and it has an irreplaceable position in the change of global climate (Trenberth 1997). The monthly index of Niño3.4, AO, and NAO, PDO come from the NOAA Physical Sciences Laboratory (https://www.psl.noaa.gov/).

Generally, the eigenvalues are listed in the order ⁠. Each of them matches a column of the eigenvector called the EOF.

Evidently, the larger the eigenvalue ⁠, the greater contribution of the corresponding EOF to the total variance. Moreover, the North criterion is employed to test the significance level (95%) of the EOF patterns in this study (North et al. 1982), only the modes that pass the significance test are the signals with physical significance and can accurately reflect the variations of the original meteorological element field.

The Monte Carlo method is selected to examine the significance level (95%) of WTC and MWC in this study (Hu & Si 2016). The percentage area of significant agreement (PASC) could be calculated from the percentage of average wavelet coherence (AWC) or MWC values, which is passing the 95% significance test at all time scales (Hu & Si 2016; Nalley et al. 2019). The WTC/MWC and PASC are effective methods to assess the scale dependence between precipitation and large-scale circulation factors.

The location, including longitude and latitude, is also a significant driving factor affecting annual precipitation. Thus, the relationships between latitude and longitude and annual precipitation in the YRB were further investigated in this study (Figure 3(b) and 3(c)). Evidently, the annual precipitation gradient along with the longitude has increased in the YRB (10.73 mm/°E, R² = 0.075). The areas with high longitude in the YRB, close to the Western Pacific Ocean, are easily exposed to more water transported from the sea, so the precipitation in these areas is more compared with other regions.

Besides, the annual precipitation gradient along with the latitude exhibited a significant decreasing trend (57.24 mm/°N, R² = 0.44, p < 0.01), which indicates that annual precipitation is more easily affected by latitude in the basin. In addition, the East Asian monsoon also has a non-negligible impact on the precipitation in the YRB, especially in the region with low latitudes. This leads to spatial differentiation of precipitation in the YRB, and the south has more rainfall than the north.

The second dominant mode (EOF2; Figure 5(b)) showed the remarkable decreasing trends variability of average precipitation per year in the YRB. And the EOF2 and PC2 (Figure 6(b)) could explain approximately 14.3% of annual average precipitation variations. For PC2, its variation in the basin was significant and implied a significantly decreased trend during the entire research period. Moreover, from the EOF2, the average precipitation is decreasing in the lower of Tongguan, and the same conclusion could be obtained through the spatial variation of precipitation in the YRB (Figure 2(d)). The precipitation variances explained by EOF3 and PC3 account for 10.6% (Figures 5(c) and 6(c)), which is slightly lower than the percent variances explained by EOF2. The spatial pattern of precipitation reflected by EOF3 is significantly different from that of EOF1 and EOF2, while EOF3 mainly represents the distribution pattern of seasonal precipitation in the basin. The low center of EOF3 is located near Tongguan, which implies the decreasing trend of precipitation in spring or autumn in this area (Figure 5(a) and 5(c)). Besides, an opposite signal could be seen outside the region near Tongguan, and the precipitation in spring or autumn in this region showed an increasing trend during 1950–2019. PC3 presented a slightly increasing trend, which indicated an increasing wet pattern of precipitation in spring or autumn in the basin. And the EOF4 and PC4 modes (Figures 5(d) and 6(d)) explained about 5.6% of the precipitation variability. The EOF4 shows that there was a more humid phase than others throughout the study and the mean annual precipitation showed a remarkable rising tendency in the headstream of the Yellow River. The temporal mode of PC4 (Figure 6(d)) also showed a significant upward trend, and it showed that the annual average precipitation was relatively high basin headwater area of the basin from 1950 to 2019. The spatial pattern is also confirmed by the annual precipitation spatial trend analysis (Figure 2(d)).

We found that the annual average precipitation is easily affected by PDO in the YRB, the WTC and PASC are 0.387 and 7.40%, respectively. The EASM is itself influenced by the PDO, that is the weakened ocean–land thermal contrast occurred when the phase shifts from a negative to a positive PDO in north China (Qian & Zhou 2014). Therefore, the PDO indirectly regulates the precipitation variability through the EASM over the monsoon region of the YRB (Liu et al. 2019a). And the ENSO also has a remarkable effect on the change of precipitation in the basin, the WTC and PASC between ENSO and precipitation are 0.354 and 7.59%, respectively. The interannual periodic difference is not obvious between ENSO and the annual average precipitation from 1960 to 1990. In fact, many previous studies had been verified that ENSO had an important influence on climate change in Asia (Ouyang et al. 2014; Wang et al. 2021a). The warm ENSO, called El Nino, usually causes a notable reduction in precipitation in the middle and lower Yellow River (Ouyang et al. 2014). And the influences of them on precipitation were different from month to month and unevenly spatially distributed over the basin (Ouyang et al. 2014). The reason for more precipitation is most likely that the east of the basin is close to the ocean and affected by the East Asian monsoon (Zhou et al. 2012).

There is much evidence to indicate that precipitation is influenced in a complex manner by multiple driving forces simultaneously (Ouyang et al. 2014). And the coupled effects of multiple factors on the precipitation change are analyzed by the MWC method. Table 1 is the MWC and PASC (%) from the average annual precipitation and various coupled factors. Moreover, an additional explanatory factor should be considered to have a statistical significance only when the PASC was increased by at least 5% (Hu & Si 2016). Obviously, the WTC/MWC has a positive relationship with the change in the number of explanatory factors, while PASC has no prominent change relationship. The mode of AO–NAO could explain the precipitation variations best in all the two-factor combinations, the MWC and PASC are 0.626 and 10.35%, respectively.

In the study, the long-term spatiotemporal variations of precipitation are explored in the YRB based on the observation data. The precipitation gradient against the elevation, longitude, and latitude was evaluated using the SMR model. Besides, we explored the dominant spatiotemporal modes of annual precipitation by the EOF method. Furthermore, the WTC and MWC methods were selected to explore the individual or integrated effects of various atmospheric circulation factors (ENSO, NAO, PDO, and AO) on the precipitation variability in the YRB. The summary is as follows:

• (1)

  Diagnosis of the long-term tendency of precipitation reveals that the multi-year average of precipitation presents a slightly declined trend of about 0.88 mm/10a, while the precipitation in the upstream area showed a significant increasing trend. The dependence relationships of monthly precipitation gradient against elevation, longitude, and latitude by the SMR model indicated that the precipitation was more sensitive to longitude.

• (2)

  Due to about 71.9% of the total variance, four dominant patterns of annual precipitation could sufficiently reflect the spatiotemporal patterns of precipitation in the YRB. EOF1 represented the dominant precipitation distribution, which is highly consistent and homogeneous owing to the values of EOF1 showing a consistently positive feature. EOF2 elucidated that the precipitation downstream of the Tongguan showed a significant decreasing tendency, which was demonstrated by the spatial trend analysis of the precipitation. Moreover, the EOF3 and EOF4 showed the seasonal and regional spatiotemporal modes of precipitation in the YRB, respectively.

• (3)

  The wavelet coherence analysis indicated that the linkages between precipitation and individual circulation factor varied at different time scales. The MWC showed that the precipitation variations were controlled by multiple influencing factors simultaneously. The MWC and PASC (%) between precipitation and the combination of AO–NAO–PDO are 0.783 and 12.34%. Hence, it is vital to consider the comprehensive effects of multiple influencing factors when interpreting precipitation variations.

This study was funded by the National Natural Science Foundation of China-Shandong Joint Fund (U2006227, U1906234).

All relevant data are available from an online repository or repositories. (http://data.cma.cn/).

The authors declare there is no conflict.