Origin of the 30–60-day intraseasonal oscillation of streamflow in source region of Yellow River in China: a perspective of the atmospheric signals from mid-high latitude

Abstract

Streamflow in source region of Yellow River (SRYR) matters with regard to the adjacent and downstream water resources. Intraseasonal oscillation (ISO) in the streamflow in SRYR is of great significance to the sub-seasonal prediction of streamflow in SRYR, but is unknown. Here, we first report a 30–60-day ISO in the streamflow in SRYR, which is regulated by the atmospheric 30–60-day ISO at mid-high latitude over North Eurasia. The 30–60-day ISO in atmosphere is featured by a Rossby wavetrain, and the wave energy propagates southward onto the TP, which causes anomalous wind response over TP. The leading anomalous high (low) with anti-cyclonic (cyclonic) wind anomalies over the TP favors dry (wet) air in lower troposphere in SRYR, via enhancing the water vapor divergence (convergence). Dry (wet) air always results in strong (weak) evaporation from the Yellow River, which causes the later streamflow valley (peak) and thereby the 30–60-day ISO in the streamflow in SRYR.

Introduction

Yellow River is the second greatest river in China, it originates from the Tibetan Plateau (TP) and serves as the most important water source in North China. The source region of Yellow River (SRYR) is located in the northeastern TP, where the streamflow accounts for nearly 40% of the total annual amount in the Yellow River, and plays a very important role in water resource supply along the Yellow River basin (Hu et al. 2011; Wu et al. 2018; Jin et al. 2022; Lan et al. 2004). Multi-timescale variability of streamflow in SRYR greatly matters to the downstream industries, domestic households and agriculture. Therefore, fully understanding the streamflow variability in SRYR will benefit the social-economic development, water management, as well as its multi-timescale climate forecast.

The source of streamflow in SRYR mainly comes from the surrounding precipitation, melting ice, snow and permafrost, as well as the subsurface water (Jia et al. 2008; Lan et al. 2010a; 2010b). The climate change, and human activities including land use, vegetation destruction, water resource facilities, are important factors affecting the variability of streamflow in SRYR (Liang et al. 2015; Piao et al. 2010; Wang et al. 2009). Of which, the local precipitation, air temperature and evaporation is regarded as the three dominant meteorological variables influencing the streamflow in SRYR (Shi et al. 2007; Wang et al. 2012; Lv et al. 2015). However, responses of the streamflow to the precipitation and temperature are diverse at different timescales in distinct reaches of Yellow River (Labat 2010; Su et al. 2017; Lan et al. 2010a, b; Jiang and Li 2011; Chiew et al. 1995; Kamga 2001; Zhang et al. 2007).

The streamflow in SRYR shows a robust seasonal cycle, with the largest amount in summer and the smallest in winter (Bai et al. 2016; Jiang and Li 2011). There is a decreasing trend in the streamflow in SRYR since 1960s (Li et al. 2012; Lan et al. 2010a; Zhang et al. 2017; Chu et al. 2018; Hu et al. 2021; Shi et al. 2017; Yang et al. 2021, 2023), and the streamflow exhibits an increasing trend since 2000 (Shi et al. 2007; Bao et al. 2021; Zhang et al. 2020; Jiang and Li 2011). In addition, the streamflow in SRYR shows an irregular interannual and decadal oscillation, with the periods of 3–4, 5–8, 15–18, 22–25, and 42 years, depending on the study periods and methods used to detect the periods (Ma 1995; Hou et al. 2020; Jiang and Li 2011; Li et al. 2012; Chu et al. 2018).

As stated above, the linear trend, decadal, interannual, and seasonal variations in streamflow in SRYR have been widely discussed (see the preceding paragraph). But, few attentions have been paid to the variability of streamflow in SRYR at intraseasonal (or sub-seasonal) timescale. The intraseasonal oscillation (ISO) is referred to as the 10–90-day variability in surface air temperature, rainfall, and atmospheric variables. This timescale bridges the synoptic and seasonal climate variability, and is regarded as the important target of sub-seasonal prediction (Vitart et al. 2015). However, in contrast to the ISO in surface air temperature and rainfall, the ISO in streamflow in SRYR is less known. Accordingly, in the present work, we intend to explore the ISO in the streamflow. Here we find a 30–60-day ISO in streamflow in SRYR that cannot be explained by surrounding precipitation. Therefore, we will focus on the 30–60-day ISO in streamflow in SRYR and try to reveal the underlying mechanism.

Data and method

TangNaiHai hydrological station (TNH) is located in the upstream of Yellow River (Fig. 1a), the streamflow at which is rarely affected by human activities and mainly determined by meteorological factors (Bai et al. 2016; Zhang et al. 2017). It has been suggested that the variation in streamflow at TNH performs well in representing the change in streamflow in SRYR (Hu et al. 2011; Wu et al. 2018; Jin et al. 2022; Lan et al. 2004). Here we use the daily observed streamflow at TNH to reveal the intraseasonal variability of the streamflow in SRYR. The daily observed streamflow covering May–August of 1997–2018 is provided by the Yellow River Conservancy Commission of the Ministry of Water Resources. The daily atmospheric components are calculated from the hourly ERA5 reanalysis of the European Centre for Medium-Range Weather Forecasts (ECMWF), with a resolution of 0.25° × 0.25° (Hersbach et al. 2020). The daily rain gauge observed precipitation at 13 stations along the Yellow River in SRYR is provided by the information center of China Meteorological Administration (CMA), and the total rainfall amount at the 13 stations is considered as the income of streamflow in SRYR.

Fig. 1

a Locations of the Yellow River (blue line) and TNH (red triangle). The shadings represent the elevation, and TP is the area with elevation higher than 3000 m. b Ratios of the power spectrum of the streamflow at TNH to the red noise spectrum at α = 0.05 significance level. c Ratios of the power spectrum of the precipitation in SRYR to the red noise spectrum at α = 0.05 significance level. In b and c, values greater than 1 indicate the periods passing 95% confidence level. Y-axis and x-axis denote the period (unit: day) and time (unit: year), respectively

We calculate the power spectrums of the streamflow at TNH and the rainfall amount in SRYR to detect the dominant ISO signals. In addition, we apply a band-pass filter method of Lanczos (Duchon 1979) to extract the ISO components. The mean square deviation (MSD) of the band-pass filtered variable is used to measure the ISO intensity. Furthermore, we calculate the wave activity flux (WAF) (Takaya and Nakamura 2001; Tam and Li 2006) to explore the wave activity at intraseasonal timescales. The WAF equation given in Takaya and Nakamura (2001) is shown as follows:

$$\mathbf{W}=\frac{p \; \text{cos}\phi }{2\left|\mathbf{U}\right|}\left(\begin{array}{c}\frac{U}{{a}^{2}\; {\text{cos}}^{2}\phi }\left[{\left(\frac{\partial {\psi }^{\prime}}{\partial \lambda }\right)}^{2}-{\psi }^{\prime}\frac{{\partial }^{2}{\psi }^{\prime}}{\partial {\lambda }^{2}}\right]+\frac{V}{{a}^{2}\; \text{cos}\phi }\left[\frac{\partial {\psi }^{\prime}}{\partial \lambda }\frac{\partial {\psi }^{\prime}}{\partial \phi }-{\psi }^{\prime}\frac{{\partial }^{2}{\psi }^{\prime}}{\partial \lambda \partial \phi }\right]\\ \frac{U}{{a}^{2}\; \text{cos}\phi }\left[\frac{\partial {\psi }{\prime}}{\partial \lambda }\frac{\partial {\psi }^{\prime}}{\partial \phi }-{\psi }^{\prime}\frac{{\partial }^{2}{\psi }^{\prime}}{\partial \lambda \partial \phi }\right]+\frac{V}{{a}^{2} }\left[{\left(\frac{\partial {\psi }^{\prime}}{\partial \phi }\right)}^{2}-{\psi }^{\prime}\frac{{\partial }^{2}{\psi }^{\prime}}{\partial {\phi }^{2}}\right]\\ \frac{{f}_{0}^{2}}{{N}^{2}}\left\{\frac{U}{a\; \text{cos}\phi }\left[\frac{\partial {\psi }^{\prime}}{\partial \lambda }\frac{\partial {\psi }^{\prime}}{\partial z}-{\psi }^{\prime}\frac{{\partial }^{2}{\psi }^{\prime}}{\partial \lambda \partial z}\right]+\frac{V}{a }\left[\frac{\partial {\psi }^{\prime}}{\partial \phi }\frac{\partial {\psi }^{\prime}}{\partial z}-{\psi }^{\prime}\frac{{\partial }^{2}{\psi }^{\prime}}{\partial \phi \partial z}\right]\right\}\end{array}\right)+{\text{C}}_{U}M.$$

W is the phase-independent wave-activity flux, p = (pressure/1000 hPa), $\psi^ {\prime}$ is perturbation stream function. a is the earth’s radius, $\phi$ and $\lambda$ are latitude and longitude, respectively. U and V are zonal and meridional basic flow, respectively. U indicates the vector of a steady zonally inhomogeneous basic flow. N² is the buoyancy frequency squared. C_U is the vector that represents the phase propagation in the direction of U, and the calculation formula of quantity M can be found in Eqs. 26 and 27 in Takaya and Nakamura (2001). The other symbols are of conventional use in meteorology. The detailed introduction for the WAF equation can be found in Takaya and Nakamura (2001).

Eight phases of the 30–60-day ISO in streamflow in SRYR are identified by conducting the following steps. First, 30–60-day ISO component of the streamflow in SRYR is extracted by using the 30–60-day band-pass filter, and then the filtered series is standardized. Second, peaks and valleys of the filtered streamflow are identified, and the peak phases are recorded as phase 1 and valley phases as phase 5. Third, times between the peak and valley phases, at which the filtered streamflow equals zero (simply as “zero times”), are detected, and the zero times following the peak (valley) times are recorded as phase 3 (phase 7). Fourth, times in the middle between phases 1 and 3 are defined as phase 2, and those between phases 5 and 7 are defined as phase 6.

In order to reveal the causation of the ISO in streamflow in SRYR, the environmental conditions before the valley and peak phases of the filtered streamflow in SRYR are analyzed.

Original signal of 30–60-day ISO in streamflow in SRYR

Figure 1b and c shows the ratios of the power spectrum of streamflow at TNH and precipitation in SRYR to the red noise spectrum at α = 0.05 significance level. The values in Fig. 1b and c greater than 1.0 imply the periods exceeding 95% significance level. Significant 10–20-day and 30–60-day ISO are detected in streamflow at TNH (Fig. 1b). The 10–20-day ISO in streamflow can be explained by the counterpart of precipitation, however, the 30–60-day ISO in streamflow cannot be resembled by the counterpart of precipitation (Fig. 1c). We speculate this 30–60-day ISO in streamflow is mainly originated from the local land–atmospheric interaction.

To explore the origination of atmospheric 30–60-day ISO, we calculate the atmospheric ISO intensity and WAF. The average MSDs of the 30–60-day filtered geopotential height at the near surface level over the TP (550 hPa) during May–August of 1997–2018 is shown in Fig. 2a. Stronger 30–60-day ISO in geopotential height appears at mid-high latitude with the largest MSDs being over northwest of the TP. Relatively weak 30–60-day ISO is observed over the northern TP. Actually, atmospheric 30–60-day ISO at mid-high latitudes was firstly discovered in 1980s (Anderson and Rosen 1983; Krishnamurti and Gadgil 1985; Zhang 1987), and two primary circulation patterns were reported in Li (1990), i.e., Eurasian-Pacific teleconnection pattern and Pacific-North America pattern. The WAF indicates the propagation of the wave energy at 550 hPa (Fig. 2b). Because WAF is independent on ISO phases (Takaya and Nakamura 2001), the distributions of WAF are similar in each phase of the 30–60-day ISO in streamflow in SRYR. Thus, WAF in phases 2–4 (the 8 phases of ISO in streamflow are shown in Fig. 3) are illustrated in Fig. 2b for example. Wave energy of the strongest 30–60-day ISO at mid-high latitude northwest of the TP generally propagates eastward north of the TP, with a branch propagating southward onto the TP, implying that the 30–60-day ISO over the northern TP is attributed to the wave energy propagating from the mid-high latitude. Similar distribution of the ISO intensity and propagation of the wave energy are also observed in upper troposphere (Fig. 2c and d). Regressions of the 550-hPa geopotential height and wind against the annual streamflow at TNH averaged during May to August are provided in Fig. 2e. An anomalous high with anti-cyclonic wind is observed northwest of the TP at mid-high latitude, and anomalously low appears over the TP, accompanied by cyclonic wind. Therefore, large (small) streamflow in SRYR corresponds to anomalous high (low) northwest of the TP at mid-high latitude and anomalous low (high) over the TP, and the anomalous high (low) northwest of the TP fall in the area with the strongest 30–60-day ISO (Fig. 2a).

Fig. 2

Average MSDs of the 30–60-day filtered geopotential height during May–August of 1997–2018 at a 550 hPa and c 300 hPa. WAF (vectors; unit: m² s⁻²) and stream function (shadings; unit: m² s⁻¹) in phases 2–4 of the 30–60-day ISO in streamflow at TNH at b 550 hPa and d 300 hPa. e Regressions of the 550-hPa geopotential height (shadings) and wind (vectors) against the annually averaged streamflow at TNH during May to August. In e, the regressed geopotential height with the statistical significance exceeding 95% confidence level is marked by green lattices, and only the regressed wind passing 95% confidence level is shown. In a–e, the thick purple lines represent the topographic contours of 3000 m

Fig. 3

Eight phases of the 30–60-day ISO in streamflow at TNH (x-axis) based on the standardized 30–60-day filtered streamflow at TNH (black; y-axis on the right), and the standardized 30–60-day filtered geopotential height (orange; y-axis on the left), mean evaporation rate (brown; y-axis on the left), specific humidity (green; y-axis on the left), and air temperature (purple; y-axis on the left) averaged in SRYR (96–103°E, 33–35°N) at each phase of the streamflow

Eight phases are defined based on the standardized 30–60-day filtered streamflow at TNH (black line in Fig. 3), among which phase 1 corresponds to the peak of 30–60-day ISO in streamflow at TNH and phase 5 the valley. Accordingly, regional averaged circulation regime in 96–103°E, 33–35°N, covering the SRYR, at each phase of 30–60-day ISO in streamflow at TNH is shown in Fig. 3. Note that negative value of the mean evaporation rate in ERA5 data represents evaporation and positive value represents condensation. It is observed that the strongest (weakest) of evaporation appears at phase 4 (phase 8), which leads the valley (peak) of the streamflow for 1 phase, implying that the 30–60-day ISO in the streamflow is possible a response to the variation in evaporation. Air temperature and humidity are two primary factors affecting the evaporation. The air temperature varies synchronously with the streamflow, but the specific humidity reaches the valley (peak) at phase 3 (phase 7), which leads the strongest (weakest) evaporation for 1 phase. This demonstrates that the leading variation in specific humidity is responsible for the late response in evaporation. Furthermore, the valley (peak) of specific humidity follows the peak (valley) of the geopotential height at phase 2 (phase 6), which implies the important role of atmospheric circulation in modulating the specific humidity. The above sequence implies a possible chain effect producing the 30–60-day ISO in streamflow at TNH, which initiates from the atmospheric circulation anomalies.

Mechanism of 30–60-day ISO in streamflow in SRYR

Here we select valley and peak phases of the 30–60-day filtered streamflow at TNH, which are simply referred as valley and peak phases of the streamflow, respectively, hereafter. In order to explore the alternation between the valleys and peaks of the streamflow, we analyzed the leading environmental conditions of the valley (peak) phase of the streamflow (Fig. 4). Accordingly, composites of variables in phases 2–4 and phases 6–8 are provided in Fig. 4 for investigating the causations of the valleys and peaks of the streamflow, respectively.

Fig. 4

Composites of 30–60-day filtered environmental conditions in a, c, e phases 2–4 (i.e., 1–3 phases ahead of the valley phase) and b, d, f phases 6–8 (i.e., 1–3 phases ahead of the peak phase) of the 30–60-day ISO in streamflow at TNH. a and b Geopotential height (shadings; unit: gpm) and wind (vectors; unit: m s⁻¹) at 550 hPa; the areas with the geopotential height passing the 95% confidence level are dotted, and only the wind vectors passing the 95% confidence level are plotted. c and d Vertically integrated water vapor flux (vectors; unit: kg m⁻¹ s⁻¹) and water vapor flux divergence (shadings; unit: 10^–5 kg m⁻² s⁻¹); only the water vapor flux exceeding 95% confidence level are shown, and the areas with water vapor flux divergence exceeding 95% confidence level are marked by green dots. e and f Specific humidity (contours; unit: 10^–5 kg kg⁻¹) at 550 hPa and the mean evaporation rate (shadings; unit: kg m⁻² s⁻¹); the areas with evaporation rate exceeding 95% confidence level are marked by green dots; the red triangles represent the location of TNH; the thick red and black lines represent the Yellow River, in which the red lines indicate the section upstream TNH and the black lines represent the section downstream TNH. The thick purple lines in a–d represent the topographic contours of 3000 m

Figure 4a and b shows the composites of 30–60-day filtered geopotential height and wind at 550 hPa. Ahead of the valley phase of the streamflow (Fig. 4a), there is an anomalous low located northwest of the TP and an anomalous high northeast of the TP, corresponding to the two wave activity centers at mid-high latitude (Fig. 2b). In addition, an anomalous high appears over the TP, which enhances the northerly wind anomalies over the TP and the anti-cyclonic wind divergence in northeastern TP. Similarly, ahead of the peak phase of the streamflow (Fig. 4b), the changes in geopotential height and wind anomalies lead to cyclonic wind convergence in northeastern TP. The composites of 30–60-day filtered vertically integrated water vapor flux and water vapor flux divergence are provided in Fig. 4c and d. Ahead of the valley phase of the streamflow (Fig. 4c), the water vapor is transported southward over the TP, and diverges in northeastern TP, especially in SRYR. In contrast, ahead of the peak phase of the streamflow (Fig. 4d), the water vapor is transported northward over the TP and converges in SRYR. Distribution of the water vapor flux over and around the TP coincides well with that of the wind fields at 550 hPa (Fig. 4a and b), implying that the water vapor conditions in SRYR mainly determined by the lower-level wind over the TP.

Because anomalous water vapor conditions can greatly affect the atmospheric humidity, and the latter is a crucial factor modulating the evaporation, the specific humidity and evaporation in SRYR are subsequently investigated. Composites of 30–60-day filtered specific humidity at 550 hPa and the surface mean evaporation rate are shown in Fig. 4e and f. Ahead of the valley phase of the streamflow (Fig. 4e), anomalously dry air in lower troposphere is observed in SRYR, which is favorable for the evaporation from the Yellow River. Correspondingly, evaporation anomalies (negative values) appear along the Yellow River upstream TNH, resulting in more water loss in this section of the Yellow River and thereby the later valley of the streamflow. The specific humidity and the evaporation ahead of the peak phase of the streamflow are the opposite of those ahead of the valley phase (Fig. 4f). That is, anomalously wet air and condensation (positive values) are found in SRYR ahead of the peak phase, reducing the water loss in upstream Yellow River and thereby contributing to the later peak of the streamflow. Eventually, the alternation between the valleys and peaks of the streamflow at intraseasonal timescale is produced.

The underlying mechanism of the ISO in streamflow in SRYR is summarized in the schematic diagram (Fig. 5). In Fig. 5, only the physical mechanism for the valley of the streamflow is illustrated, and that for the peak is just the opposite but not displayed. The atmospheric 30–60-day ISO at mid-high latitude propagates southward onto the TP, leading to distinct wind fields over the TP ahead of the valley (peak) phase of the streamflow. The anomalous high (low) over the TP favor the dry (wet) air in lower troposphere via enhancing the water vapor divergence (convergence) in SRYR. The dry (wet) air induces local strong (weak) evaporation, which results in the later valley (peak) of the streamflow at TNH, and thereby the 30–60-day ISO in the streamflow in SRYR.

Fig. 5

Schematic diagram of the physical mechanism for the valley of the streamflow at TNH, and that for the peak is just the opposite

Conclusions and discussion

Conclusion

ISO is the important basis of sub-seasonal prediction, but the ISO in streamflow in SRYR is unknown. Here we find a 30–60-day ISO in streamflow in SRYR, which is not regulated by the contribution of precipitation. Our results show that the 30–60-day ISO signal of streamflow is originated from the atmospheric 30–60-day ISO at mid-high latitude over North Eurasia.

The energy of atmospheric 30–60-day ISO at mid-high latitude propagates southward onto the TP, leading to weak atmospheric 30–60-day ISO over the northern TP. The maximum (minimum) of the geopotential height, minimum (maximum) of the specific humidity, and the strongest (weakest) evaporation appear 3, 2, and 1 phase ahead of the valley (peak) phase of the streamflow at TNH, respectively. This kind of chronological sequence implies the causal relationships between the elements above, i.e., the changes in atmospheric circulation are responsible for the variations in atmospheric humidity, then the latter affects the evaporation from the Yellow River, and finally the streamflow is modulated. Furthermore, the specific underlying mechanism of the alternation between the valleys and peaks of the streamflow is investigated, by analyzing the leading environmental conditions of the valley and peak phases of the streamflow. Anomalous high appears over the TP in lower troposphere over the TP, accompanied by anti-cyclonic wind divergence. The wind divergence leads to anomalous water vapor divergence in SRYR, which induces local anomalously dry air. Dry air enhances the evaporation from the river, and thereby results in more water loss of the river. The physical process above consequently contributes to the subsequent valley of the streamflow at TNH. The mechanism for the peak of the streamflow at TNH is just the opposite. In this fashion, the 30–60-day ISO in the streamflow at TNH is produced.

It has been shown the atmospheric nonlinear interaction (Li 1990; Yang and Cao 1995) and topographic forcing (Luo and Li 2000; Yan 1998) are thought greatly contributing the atmospheric 30–60-day ISO at mid-high latitude.

Discussion

Given the streamflow is a temporal accumulation and spatial integral of runoff over the river source region, the 30–60-day ISO in streamflow is possibly coming from the accumulated effect of 10–20-day and seasonal variation in precipitation. On 10–20-day time scale, the lag-lead correlation analyses indicates that the streamflow in SRYR is closely related to the precipitation 5–10 days earlier. Actually, the peak of precipitation leads the peak of streamflow for about 3 phases on 10–20-day time scale; the interannual variance of 10–20-day precipitation during May–August is significantly correlated with that in streamflow. These imply the dominant contribution of accumulated effect of precipitation to the streamflow on 10–20-day time scale. However, the correlation coefficient of 10–20-day precipitation with 30–60-day streamflow in SRYR is merely 0.13, implying the weak contribution of 10–20-day precipitation to 30–60-day streamflow. Furthermore, we examined the 10–20-day filtered precipitation at 8 phases of the 30–60-day ISO in streamflow. Our results show there is no robust lag/lead relationship between the 10–20-day ISO in precipitation and the 30–60-day ISO in streamflow (figures are not shown).

The 30–60-day ISO in streamflow mainly occurs in rainy season, and its peak phase lags the maximum precipitation in climatology. On interannual variability, the peak phase of the 30–60-day streamflow is significantly correlated with that of the maximum precipitation. This implies that the peak of the 30–60-day streamflow comes late (early) when the precipitation peak happens late (early). However, no significant correlation is observed between the amplitude of maximum precipitation and that of the maximum 30–60-day filtered streamflow, implying that the seasonal precipitation has little contribution to the amplitude of the 30–60-day ISO in streamflow.