Spatio-temporal variation of annual precipitation in China and its response to ENSO

The El Niño/La Niña Southern Oscillation (ENSO) is not only the strongest inter-annual oscillation in the earth's climate system but also an important factor that indirectly affects precipitation through the teleconnection between ocean and atmosphere, which has a significant impact on the climate in most parts of the world, including China (Chen & Lian 2020). Therefore, discussing the relationship between annual precipitation and ENSO in China is one of the most important research tasks at present (Wang et al. 2022).

Based on observational or dynamic simulation data, many scholars have used wavelet coherence (which is a time-normalized and scale-resolved method to specify the correlation between two time series) (Araghi et al. 2017), singular value decomposition (Liu et al. 2019b), Empirical Orthogonal Function decomposition (Sapna et al. 2019), Mann–Whitney U hypothesis test (Lee & Julien 2016), Community Earth System Model (Liguori et al. 2022), and other methods to explore the relationship between precipitation and ENSO in the globe (Ropelewski & Halpert 1987; Dai & Wigley 2000), as well as South America (Souza et al. 2021), Asia (Chen et al. 2018; Wen et al. 2019), the tropical Pacific (Christine & Scott 2016), Canada (Nalley et al. 2019), China (Wang et al. 2022), northern Mexico (Omar et al. 2020), and other places under the scale of month, season and year. The results showed that ENSO was the link between SSTA and atmospheric anomalies in the tropical Pacific Ocean, which can not only affect the climate conditions in the tropical Pacific Ocean but also cause precipitation anomalies, even large-scale extreme precipitation events (Chai 2022) globally through teleconnection (Shimizu et al. 2017). Global land mean annual precipitation decreased by about 12 mm in El Niño years, and increased by about 11.7 mm in La Niña years on average (Li & Zhao 2019). ENSO was also one of the main factors modulating precipitation in South America (Souza et al. 2021), Africa (Emmanuel 2022), China (Liu et al. 2019a, 2019b), and other places. Precipitation in spring, autumn, and winter in Central Asia was highly correlated with ENSO (Chen et al. 2018). When ENSO was in the positive/negative phase, there was obviously less/more precipitation in Canada (Nalley et al. 2019), while there was more/less precipitation in the Andes (Jonaitis et al. 2021) and southern California (Du et al. 2020). When ENSO was in the positive phase, the summer precipitation anomaly in eastern China showed a ‘tripolar’ spatial distribution of drought in the north and south and flood in the central part. When ENSO was in the negative phase, the summer precipitation formed a ‘northwestward-southwestward’ rain-band with less rain along the ‘Hu-Huanyong Line’ (Wang et al. 2022). There was a strong correlation between the El Niño phase and precipitation in northern Mexico (Omar et al. 2020). Some scholars also proposed that there were temporal differences in teleconnection between precipitation and ENSO (Torbenson et al. 2019). For example, the correlation between precipitation and ENSO in the southeast of the United States from April to September was significantly lower than that from October to next March (Nam & Baigorria 2015). There was a strong response relationship between precipitation and ENSO in summer and autumn in Indonesia, but in spring and winter, such a relationship was quite the opposite (Rahman et al. 2014). The correlation between ENSO and precipitation in China had strong regional and seasonal differences, as did its impact on precipitation (Gong & Wang 1999). For example, there was a significant positive correlation between summer precipitation and ENSO in northeastern China after 1997 (Han et al. 2017). The correlation of winter precipitation with ENSO and mid-latitude north Atlantic sea surface temperature in northwest China was weak in the mid-1990s, and then became significantly stronger. When ENSO was in the positive phase and the mid-latitude north Atlantic sea surface temperature was in the negative phase, the precipitation increased significantly (Yin & Zhou 2020). There was a decrease in precipitation in the arid region in El Niño years but an increase in precipitation in the spring and summer of the following year (Li & Zhao 2019).

Existing relevant results demonstrated that precipitation had different sensitivities in response to different types of ENSO events (Jia & Zhang 2020). Yet the types of ENSO events were usually divided by the location of occurrence, and rarely used accumulated SSTA as an indicator, divided ENSO event with intensity to quantify the sensitivity of precipitation to ENSO events with different types and intensities. Most of the research was carried out on a seasonal scale (Zhang et al. 2022a), with fewer inter-annual scales.

Based on what has been mentioned above, to make the research results highly representative and comprehensive, this article chose China as the study area, for it covered a vast expanse of land with a large number of climate types. Based on precipitation and SSTA raster data from January 1901 to December 2020, the response of precipitation and ENSO with different intensities was comprehensively revealed, which can enrich the research results in this field and provide a reference for disaster prevention.

• (1)

  The model builder in Arcgis software was used for coordinate transformation and extraction to obtain monthly data of SSTA in the Niño 3.4 region and precipitation in China.

• (2)

  The band collection statistics tool in Arcgis software was used to calculate the regional mean value of precipitation in the study area.

• (3)
  The linear regression model was used to simulate the variation of precipitation. The formula is (Ma et al. 2016):

  

  (1)

  where is the annual change rate; is the precipitation in year i; here, n was 120.
• (4)
  The moving average method was used to conduct an auxiliary analysis of the variation in precipitation. The formula is (Shang et al. 2018):

  

  (2)

• Here, n was 11.

• (5)
  The coefficient of variation (CV) is a value to measure the fluctuation degree of a variable (here it was precipitation, which is indicated by x in the following formula), a higher CV value indicates a higher degree of fluctuation in the variable (Pan 2022). The formula is:

  

  (3)

In addition, according to the fluctuation degree of precipitation, it was divided into five levels, which were: CV > 0.4 (high fluctuation), 0.3 < CV ≤0.4 (relatively high fluctuation), 0.2 < CV ≤ 0.3 (moderate fluctuation), 0.1 < CV ≤ 0.2 (relatively low fluctuation), and CV < 0.1 (low fluctuation).

• (6)
  The Mann–Kendall trend test was used to identify the significance of precipitation change (Saleem et al. 2021). The formulas are as follows:

  

  (4)

  

  (5)

  

  (6)

  

  (7)

  where is the value of precipitation in year j (k); S is the Kendall statistic; is the variance of S; Z is the test statistics.
• (7)
  The Pearson correlation coefficient between precipitation and ENSO was calculated using following formula:

  

  (8)

  where is the precipitation in year i; is the ONI in year i.

According to the definition of ONI and ENSO given by NOAA: ENSO is a phenomenon in the equatorial Pacific Ocean characterized by a five consecutive 3 month running mean of SSTA in the Niño 3.4 region that is above (below) the threshold of +0.5 °C (−0.5 °C). This standard of measure is known as the ONI. On the basis of this standard, a total of 24 El Niño events and 41 La Niña events were identified this time.

In order to explore the response relationship between precipitation and ENSO more comprehensively, based on previous research results (Li & Zhai 2000), this article divided ENSO events into two categories and 10 subcategories with intensity: extremely strong El Niño/La Niña, strong El Niño/La Niña, moderate El Niño/La Niña, weak El Niño/La Niña and extremely weak El Niño/La Niña. See Table 1 for details.

The average annual precipitation in the study area was 547.3 mm, the maximum value of precipitation appeared in 2016 (642.4 mm), and the minimum value appeared in 2011 (508.6 mm). The precipitation decreased from 1901 to 1925 and then increased from 1926 to 1950. After entering the 1950s, the precipitation generally showed a downward trend again, and this trend continued until the early 21st century. By 2007, the overall trend in precipitation was upward. In terms of space, the maximum value of the multi-year average was distributed in the southeast of the study area, and the total annual precipitation exceeded 1,400 mm as a whole. The minimum values were concentrated in Xinjiang, western Tibet, and western Inner Mongolia, where the annual precipitation was mostly less than 200 mm, the overall distribution increased from northwest to southeast.

It can be seen from Figure 3(c) that during the T1 period, the overall precipitation in the study area showed a downward trend, and this part of the area accounted for 68.3% of the total. Among them, the area where precipitation decreased significantly accounted for about one-third of the total area (the area passed the significance test at a confidence level of 95%), mainly distributed in the Continental Basin, the Yellow River Basin, and the junction of the Huaihe River Basin and the Yangtze River Basin. Except for the Haihe River Basin and Huaihe River Basin, where the precipitation fluctuated greatly, the precipitation in other areas was relatively stable, and most of them belonged to the low fluctuation area or the relatively low fluctuation area. The area where the precipitation increased significantly was mainly concentrated at the junction of the Yangtze River Basin and the Pearl River Basin, accounting for 13.5% of the total study area. The rising rate was greater than 6 mm/a, and the variation over the years was relatively stable. The precipitation in other areas did not change significantly, which failed the significance test, and the variation rate was mostly between −3.6 and 2.4 mm/a. The overall precipitation in the study area during the T2 period was mainly on the upward trend, this part of the area accounted for 55.4% of the total, and the areas passing the significance test accounted for 38.7% of the total, mainly distributed in the Continental Basin, the Yellow River Basin, and the Southeast Basin. Among them, the Continental Basin had the fastest rising rate, followed by the Yellow River Basin and the Southeast Basin. Except for the Southeast Basin, which belonged to the moderate fluctuation area, the precipitation in other areas was relatively stable, and most of those areas belonged to the low fluctuation area. In addition, 44.6% of the study area showed a decreasing trend, mainly distributed in the northern part of the Continental Basin, the southern part of the Southwest Basin, and the Songhua and Liaohe River Basin. The variation rate was concentrated between 1.5 and 2.5 mm/a, which failed to pass the significance test, and the decreasing trend was not significant. In the T3 period, except for a weak downward trend in the west of the Southwest Basins and a weak upward trend at the junction of the Huaihe River Basin and the Yangtze River Basin, the overall precipitation in the study area increased in the west and decreased in the east. The two covered a nearly equivalent part of the study area and accounted for 52.1 and 47.9% of the total study area, respectively. The precipitation in the former area increased significantly in the Continental Basin and passed the significance test at a confidence level of 95%, yet the degree of fluctuation was relatively large (most of them were in the moderate fluctuation area or above). The precipitation in the latter area decreased significantly in the eastern Songhua and Liaohe River Basin, the eastern Yellow River Basin, the central Yangtze River Basin, and the Haihe River Basin. Among them, except for the central Yangtze River Basin, with less fluctuation, the rest of the areas all belonged to the moderate fluctuation areas. After entering the T4 period, except for the downward trend in the central Continental Basin and the Southwest Basin, the precipitation in the study area was mainly on the rise, this part of the area accounted for 81.2% of the total study area (the area that passed the 95% significance test accounted for 22.5% of the total), the rising rate mainly 1.2–10.8 mm/a, and the overall rising rate was mainly increased from the northwest to the southeast. Among them, except for a small part in the northeast and west of the Continental Basin, the rest of the areas belonged to the moderate fluctuation areas.

Precipitation showed a downward trend before the mid-1930s, especially in the Songhua and Liaohe River Basin and the Continental Basin. Subsequently, the precipitation in the Southeast Basin and the Haihe River Basin all began to show a significant upward trend, and such a trend gradually expanded westward. In the 1950s, precipitation rose in the west and fell in the east in general, and this trend continued until the beginning of the 21st century. By 2007, except for the central part of the Continental Basin and other areas, which showed a downward trend, the precipitation in the study area mainly showed an upward trend. The precipitation tendency rate in the Haihe River Basin has changed significantly since 1991, and the rate of precipitation variation in the eastern monsoon region has accelerated (Liu et al. 2021). The precipitation in North China showed a downward trend from 1970 to 2000 (Zheng et al. 2020), which was basically consistent with the conclusion of this study. However, some scholars also pointed out that from 1979 to 2018, the precipitation in the Yarlung Zangbo River Basin showed an upward trend at all time scales (Zhang et al. 2022a, 2022b), and the precipitation in southwestern Tibet from 1901 to 2017 showed a fluctuating and decreasing trend (Li et al. 2022), which was somewhat different from this study, possibly due to different time scales and areas of the study.

There were some similarities and differences in the correlation of precipitation with various ENSO events. Specifically, although precipitation was mainly negatively correlated with El Niño and La Niña at the same time and the strength of the correlation between precipitation and ENSO events was directly proportional to the intensity (that was, the correlation between precipitation and extremely strong/strong ENSO events was stronger than the correlation between precipitation and moderate/weak/extremely weak ENSO events), the correlation between precipitation and El Niño events was significantly stronger than the correlation between precipitation and La Niña events. The conclusion that different types of ENSO events had different effects on precipitation, and the intensity of ENSO events had a certain negative correlation with the level of drought and flood (Wang et al. 2016). When the intensity of EI Niño events increased, droughts in southwest China increased (Liu et al. 2019a, 2019b). All these conclusions were basically consistent with this study.

ENSO may also indirectly influence precipitation by affecting tropical cyclones (Du et al. 2020), the Asian summer monsoon (Wu et al. 2021), and others. Some scholars have also suggested that both the rearrangement of convection centers in the Walker circulation and the changes in the monsoon system during ENSO events may induce large precipitation anomalies in most parts of the world (Dai & Wigley 2000). The impact mechanism of ENSO on precipitation needs to be further studied in the future.

The study made a detailed analysis of the temporal and spatial variation features of precipitation in China and its relationship with ENSO, which had certain significance not only for revealing the temporal and spatial variations of precipitation in China and the correlation between such features and ENSO, but also for the improvement of the ecological environment and disaster prevention. However, there were several shortcomings. For example, the research method and data set used were not comprehensive, the study showed that wavelet coherence analysis is one of the best choices to achieve correlations in a time–frequency space (Araghi et al. 2017). Gridded precipitation products can be applied as an alternative, when precipitation records are not available, and multi-source gridded precipitation products had higher accuracy (Alireza et al. 2021). In the future, the response relationship between precipitation and ENSO should be compared and analyzed based on multiple data sets by using a combination of linear and non-linear methods at the same time.

The results showed that the precipitation in the study area was mainly on a downward trend before the mid-1930s, and then turned upward. In the 1950s, the precipitation generally showed a distribution of rising in the west and falling in the east, and this trend continued until the early 21st century. By 2007, except for the central part of the Continental Basin, the overall trend was mainly upward. In this study, 65 ENSO events were identified, including 24 El Niño events and 41 La Niña events. The precipitation was generally less when ENSO occurred. However, there were differences in the type and intensity of the events. For example, El Niño had a greater impact on precipitation than La Niña did, and extremely strong or strong El Niño/La Niña events had a more significant impact than moderate, or extremely weak ones. The correlation between precipitation and El Niño or La Niña events had some similarities and differences. For example, precipitation was mainly negatively correlated with El Niño and La Niña at the same time, and both correlations were proportional to intensity, but the correlation between precipitation and El Niño was significantly stronger than that of La Niña.

This research was funded by the Technology-Planning Project of Inner Mongolia (2020GG0074) and the Inner Mongolia Science Foundation (2020MS05054).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.