HTML
-
The China Meteorological Data Service Center provided data on total daily rainfall and maximum temperature (Tmax). In this study, a day is considered a no-precipitation (NP) day if the quantity of precipitation is 0 mm, and a hot day is one where the Tmax exceeds the 85th percentile (35 ℃) of Tmax during the summer of 1970 to 2018 as suggested by Liu et al. [46], and a DH day is one when both conditions are true. Because of the difference in duration, DH days can be subdivided into three different groups: short-term events (1–2 days), moderately persistent events (3–5 days), and long-term events (more than five days).
NCEP/NCAR reanalysis data (Kalnay et al. [47]) from 1970 to 2018 were used to analyze atmospheric circulation anomalies for the difference in the number of DH days key climate factors' positive and negative phases. The Physical Sciences Laboratory of the NOAA provided the monthly Southern Oscillation (SO) Index (SOI). The EASM Index (EASMI), the SCSSM Index (SCSSMI), and the SWAC Index (SWACI) are the monsoon indices employed in this study (Li and Zeng [48–50]). The intensity index of the WPSH and Indian Ocean Basin warming (IOBW) in spring was from the National Climate Center.
-
Mann-Kendall (MK) trend test (Mann [51]; Kendall [52]) was used to calculate the long-term trend of DH days and test the significance of the trend, which is a non-parametric approach and one of the most frequently applied methods for detecting changes in hydrology and climatology.
-
We opted for the Butterworth high-pass filter (BHPF) due to its flat frequency response within the passband and minimal fluctuations in the curve. It ensures a smooth transition and avoids sudden discontinuities at the cutoff frequency D0. The system function of the BHPF, with the cutoff frequency positioned at a distance from the origin D0, can be defined as follows:
$$ H_{\mathrm{BHPF}(u, v)}=\frac{1}{1+\left(\frac{D_0}{D(u, v)}\right)^{2 n}} $$ where n represents the order of the Butterworth filter, D0 represents the center of the frequency domain, and D(u, v) represents the distance from the center of the frequency domain to the plane of the frequency domain, which is the cutoff frequency.
-
Considering the constraints posed by the available data timeframe, this study employs a recycling independent test using data from 1970 to 2008 as the training set. The objective is to assess the stability and robustness of the forecasting model. Specifically, the model is trained using observation data from 1970 to 2008 to predict the number of dry and hot days in the years 2009 to 2018. Subsequently, observation data from 1970 to 2009 is utilized to predict the number of dry and hot days from 2010 to 2018. Following this pattern, predictions are made for each year, with 10 predictions for 2018, 9 predictions for 2017, and a decreasing number of predictions from 2016 to 2009. This approach allows for a comprehensive evaluation of the model's performance over multiple forecast periods.
-
To examine the correlation between climate factors and DH days, we employed Pearson correlation analysis along with the two-tailed Student's t-test. Furthermore, we utilized multiple linear regression to explore the relationship between the interannual trend of climatic factors and DH days. Additionally, we conducted a relative importance test on the climate factors to determine their respective contributions in predicting DH days. Using a multiple linear regression model, we predicted the change characteristics of DH days based on these factors.
2.1. Data
2.2. Analysis methods
2.2.1. TREND ANALYSIS
2.2.2. HIGH-PASS FILTER
2.2.3. RECYCLING INDEPENDENT TEST
2.2.4. OTHERS
-
The variation characteristics of the number of NP days, hot days, and DH days in South China from 1970 to 2018 are shown in Fig. 1. The spatial distribution of the trend of the number of hot days and DH days from 1970 to 2018 is almost identical, while the spatial distribution of NP days is different. The number of NP days shows a decreasing trend at most stations in eastern South China but an increasing trend in most stations in the western and southern regions (Fig. 1a). The number of hot days at most stations has a significant upward trend and a few coastal stations show a downward trend (Fig. 1b). The spatial variation characteristics of DH days are similar to hot days (Fig. 1c). In terms of relative changes (relative to the average of 1970 to 2018), the variations among different stations are more consistent (see Fig. S1 in Appendix).
Figure 1. Spatial variation trends in the number of NP days (a), hot days (b), and DH days (c) in the summer of 1970 to 2018 (units: day year–1, stations marked with a cross ("×") inside the circle are at the 95% confidence level).
Figure 2a shows the long-term trend of the number of NP days, hot days, and DH days in South China from 1970 to 2018. The three trends all show a significant upward trend from 1970 to 1990 and a downward trend from 1990 to 1997, and the trend is not important after 1997. After high-pass filtering (Fig. 2b), they all show significant interannual variation characteristics, with abnormally high values in 1996, 1998, and 2000 but abnormally low in 1997 and 1999. Except for Hainan Province, the long-term trend of NP days, hot days, and DH days in the other three provinces are basically consistent with the regional changing trends (see Figs. S2-S5 in Appendix).
Figure 2. The observed time series (a) and the time series after high-pass filtering (b) of the annual average of the number of NP days, hot days, and DH days in summer from 1970 to 2018 (units: days).
Continuous occurrences of DH days will aggravate the adverse effects; therefore, we further investigate the changing characteristics of DH events (Fig. 3). Since 1970, there has been an increasing trend in the frequency of all DH events. The frequency of short-term and moderate-duration events was relatively infrequent from 1971 to 1999 but increased substantially after 1999 (Figs. 3a and 3b). There is an obvious interannual variation for long-term events (Fig. 3c). It is evident from Fig. 3 that the change trends of short-term events and moderate-duration events align closely with the total number of hot and dry days. However, the long-term events exhibit distinct patterns. This suggests that short-term events and moderate-duration events have a more substantial influence on the annual variation of hot and dry days. Moreover, short-term events prove to be effective in capturing both the high and low values of the total number of hot and dry days.
-
Previous studies have highlighted the influence of various factors, such as EASM, WPSH, IOBW, SO, SCSSM, and SWAC, on DH days. In Table 1, we have summarized the correlation coefficients between these factors and the number of DH days. Upon examining Table 1, it is evident that SO and SWAC demonstrate a weak correlation with observed DH days. This suggests that their impact on the interannual variation of DH days is minimal. Consequently, we have narrowed our focus to EASM, SHI, IOBW, and SCSSM, as they exhibit stronger correlations. To observe their interannual variation trends over a 1–2 year period, we applied a high-pass filter to these four climate factors. Following the filtering process, the R between EASM, SHI, IOBW, SCSSM, and DH days are –0.56, 0.48, 0.58, and –0.53, respectively.
EASM SHI IOBW SO SCSSM SWAC Observed –0.39* 0.57** 0.63** 0.15* –0.46** –0.0001 High-pass filtered –0.56** 0.48** 0.58** –0.53** Note: *Significant at the 95% level based on a two-tailed Student's t test. **Significant at the 99% level. Table 1. Correlation coefficients between DH days and climatic factors.
Furthermore, we conducted a relative importance test on the climate factors to assess their respective contributions in predicting DH days. The results revealed that IOBW and EASM displayed relative importance exceeding 25%. This suggests that these two factors hold significant relevance and play a crucial role in predicting DH days.
These screened key climate factors were used to establish a multiple linear regression model to fit the time series of the number of DH days, and according to the coefficient of the multiple linear regression equation, the influence of the key climatic factors on the trend of DH days was quantified. As shown in Fig. 5, the fitted number of DH days and the observed number of DH days have consistent interannual variations and trends, with a correlation coefficient of 0.65. To eliminate the impact of the magnitude of each climate factor, a standardization was applied to these factors. The multiple linear regression model based on standardized climate factors is as follows:
Figure 5. (a) Standardized and (b) high-pass filtered observed and fitted time series of dry and hot days from 1970–2018.
The number of DH days = 0.56×IOBW–0.2×EASMI +3.5×10–16.
According to the coefficients of the fitting equation, IOBW contributed more to the number of DH days.
-
We selected the maximum and the minimum ten years after the standardization of DH days and key climate factors as their positive and negative phases to further analyze atmospheric circulation anomalies associated with DH days and to explore the process and mechanism of climate factors' influence on the interannual variation and trend of DH days. The positive and negative phase years of DH days and climate factors are listed in Table 2.
DH days and climatic factors Positive phase years Negative phase years DH days 2003, 1998, 2016, 2007, 1990, 2011, 2009, 2015, 2005, 2017 1995, 1977, 1982, 1976, 1994, 1999, 1974, 1975, 1973, 1997 IOBW 2002, 1991, 2014, 1988, 2003, 2005, 2015, 2010, 1998, 2016 1974, 1976, 1975, 1971, 1972, 2008, 1984, 1989, 1977, 1978 EASM 1974, 1976, 1975, 1971, 1972, 2008, 1984, 1989, 1977, 1978 1980, 1998, 1996, 2010, 1988, 1983, 2017, 2013, 2008, 2007 Table 2. Positive and negative phase years of DH days and climatic factors.
An analysis of meteorological parameters was performed to examine the characteristics during the positive and negative phases of DH days. Notably, there is a substantial increase in the geopotential height at 500 hPa (H500) over South China, reaching a maximum amplitude of 20 gpm, suggesting the dominance of a high-pressure ridge controlling the region (Fig. 6a). Similar to the pattern at 500 hPa, the geopotential height anomaly at 850 hPa also exhibits a positive anomaly. The abnormally high pressure in South China correlates with the anomalous intensification and westward extension of the WNPSH, as indicated by the 5880-gpm contour (Fig. 6c). In the negative phase of DH days, the western edge of WNPSH is situated around 125°E. However, during the positive phase of DH days, WNPSH extends further eastward, reaching approximately 145°E. Comparatively, during the positive phase of DH days, abnormal anticyclones appear at both 500 hPa (UV500) and 850 hPa (UV850) along the eastern coast of South China, providing further evidence of the abnormally strong and westward extension of WNPSH.
Figure 6. Geopotential height anomalies at (a) 500 hPa and (b) 850 hPa, and wind anomalies at (c) 500 hPa and (d) 850 hPa between DH Days positive and negative phases. The Orange dashed line in (c) denotes the WNPSH position for DH Days positive phase, and the black dashed line denotes the negative phase. The dotted area indicates a 95% significance test.
Furthermore, during the positive phase of DH days, the westerly jet stream in South China experiences a weakening trend (Fig. 7a). This weakening of the jet stream can potentially contribute to high temperatures and dry conditions in the region. Concurrently, there is a decrease in total cloud cover (TCC) over South China (Fig. 7b). Reduced cloud cover facilitates the penetration of shortwave solar radiation (DSSR), which, in turn, leads to an increase in surface temperatures (Fig. 7c). The warming of the surface intensifies the outward long-wave radiation (USLR) emitted from the surface (Fig. 7d), subsequently warming the atmosphere (SAT) in South China (Fig. 7e). Consequently, the frequency of hot and dry days in South China rises.
-
Anticyclonic anomalies are responsible for extreme hot events in various regions (Maheras et al. [53]). In eastern China, the anticyclonic abnormalities leading to extreme high temperatures are associated with the enhancement and westward or northwestward extension of the WPSH (Ding et al. [17]; Chen et al. [54]; Ren et al. [55]). The strong circulation anomaly at 500 hPa is associated with extreme heat events, and the strength of circulation patterns strength influences high-temperature days (Meehl et al. [56]).
Figure 8 shows the difference in the number of DH days in key climate factors' positive and negative phases. From the difference in IOBW (Fig. 8a), compared with the negative phase of IOBW, the number of DH days in the positive phase of IOBW increased dramatically. The difference between the positive and negative phases of EASM shows obvious differences in most stations in the northeast of South China and Hainan Province but not in other regions (Fig. 8b). In the positive phase of EASM, the number of DH days is less.
Figure 8. Anomalies of DH days in the years corresponding to positive and negative phases after standardization of (a) IOBW and (b) EASM.
To investigate the impact of IOBW on DH days in South China during the summer period from 1970 to 2018, we conducted a detailed analysis of meteorological parameters during the positive and negative phases of IOBW (Fig. 9). Our findings reveal a significant rise in geopotential height at 500 hPa (H500) and 850 hPa (H850) over South China, reaching a maximum amplitude of 25 gpm and 20 gpm, respectively, indicating the dominance of a high-pressure ridge in the region (Fig. 9a–9b). The presence of abnormally high pressure intensifies subsidence, reduces humidity, and diminishes rainfall, resulting in an increase in solar radiation reaching the surface (Fig. 9c) and enhanced vertical adiabatic heating (Black et al. [57]). During the positive phase of IOBW, the western boundary of WNPSH extends to approximately 125°E (Fig. 9e). In contrast, it is situated on the US west coast near the Pacific Ocean during the negative phase (Fig. S6). A noteworthy observation is the occurrence of abnormal anticyclones at both 500 hPa (UV500) and 850 hPa (UV850) along the eastern coast of South China in the positive phase of IOBW (Fig. 9e–9f), accompanied by an increase in surface outgoing longwave radiation (Fig. 9d). These atmospheric patterns contribute to elevated temperatures and reduced precipitation, consequently leading to an upsurge in the number of DH days.
Figure 9. Geopotential height at (a) 500 hPa and (b) 850 hPa, (c) net surface shortwave radiation flux, (d) upward surface longwave radiation flux, wind anomalies at (c) 500 hPa and (d) 850 hPa anomalies between IOBW positive and negative phases. The orange dashed line in (c) denotes the WNPSH position for the IOBW positive phase. The dotted area indicates a 95% significance test.
The difference in geopotential height and wind field between the positive and negative phases of EASM is shown in Fig. 10. When the EASM is in a positive phase, H500 in South China appears as a negative anomaly (Fig. 10a), and the East Asian trough deepens, resulting in static and stable weather. The negative anomaly of H850 is still obvious (Fig. 10b). Meanwhile, the western edge of WNPSH is located at about 127°E when the EASM is negative but retreats to about 145°E when the EASM is positive (Fig. 10c). Compared with the negative phase of EASM, abnormal cyclones appeared in South China at both 500 hPa and 850 hPa during the positive phase of EASM (Fig. 10c–10d), which would make the high-temperature atmospheric circulation easily meet with the warm and humid airflow to produce rainfall.
Figure 10. The same as Fig. 6, but for EASM.
Moreover, during the positive phase of EASM, the westerly jet (U200) intensifies (Fig. 11a), accompanied by a 2–4% increase in total cloud cover (TCC) (Fig. 11b). Additionally, there is an 8–12 W m–2 reduction in downward surface solar radiation flux (DSSR) (Fig. 11c) and approximately 4 W m–2 decrease in upward surface longwave radiation flux (USLR) (Fig. 11d). This combination of factors leads to a decrease in surface air temperature (SAT) by around 0–2 ℃ (Fig. 11e). Collectively, these conditions are unfavorable for the occurrence of DH days. Consequently, the positive phase of EASM is associated with a lower frequency of DH days.
Figure 11. The same as Fig. 7, but for EASM.
-
Climate factors can be used to predict precipitation and extreme heat in South China (Lü et al. [58]). For example, spring rainfall in eastern Australia and summer rainfall in northeastern Australia can be forecast months in advance using the ENSO indicator (Chiew et al. [59]). The SO can influence the estimation of global surface temperature anomalies (Halpert and Ropelewski [60]). However, there are few studies on predicting the number of DH days by using climatic factors in South China, and the key climate factors affecting DH days are different in different regions. Therefore, it is essential to identify accurate key climate factors and use them to predict DH days in South China.
In this study, we use the key climate factors, EASM and IOBW, finally identified in the previous section, as the predictor's input to the multiple linear regression model to predict the interannual trend of DH days. By using the adjusted optimal subset model (AOSM, Zhao et al. [61]), the prediction model of the number of DH days in summer in South China is established as follows:
Fitted number of DH days=6.5×IOBW–2.5×EASM +14.6
Figure 12a shows that the variation of the number of DH days fitted by multiple linear regression matches the observed variation of the number of DH days well. The correlation coefficient between the fitted and the observed number of DH days can be up to 0.7. In addition, the R2, MAE, and RMSE of the prediction model established by IOBW and EASM are 0.45, 2.8 days, and 2.2 days, respectively, indicating that the interannual variation of the number of DH days can be well reproduced by using this model. We input the 2019 predictors into a multiple linear regression model based on key climate factors to predict the number of DH days in 2019 (Fig. 12a). The predicted value of the number of DH days in 2019 is 13.1 days, and the observed value is 14.1 days (bias=1 day). The deviation between the observed and predicted values is slight, indicating that the model's predictive ability is acceptable. Besides the deviation, the stability of the prediction model is also an important aspect of testing the model′s performance. Therefore, a recycling independent test to assess the stability of the prediction was performed. As shown in Fig. 12b, the histograms of the number of DH days for each annual cycle prediction value are highly consistent, indicating that the model has good stability. The ten models' average R2, MAE, and RMSE were 0.43, 2.2 days, and 2.9 days, respectively, showing that the established model can predict DH days in South China stably and efficiently.