3.1 Variation characteristics of various extreme precipitation
Table 1
Climatic inclinations of the regional average annual and seasonal extreme precipitations in GBA from 1961 to 2018
|
TM
|
PRE
|
R95P
|
R99P
|
RX1day
|
RX5day
|
R10day
|
R20day
|
SDII
|
CWDday
|
CDDday
|
unit
|
℃a− 1
|
mma− 1
|
mma− 1
|
mma− 1
|
mma− 1
|
mma− 1
|
daya− 1
|
daya− 1
|
mma− 1
|
daya− 1
|
daya− 1
|
annual
|
0.021*
|
5.517*
|
1.716
|
0.366
|
0.066
|
0.388
|
0.154*
|
0.073*
|
0.003
|
0.050*
|
-0.083*
|
DJF
|
0.027*
|
0.972
|
0.517
|
0.246
|
0.196
|
0.442
|
0.020
|
0.015
|
0.007
|
0.025
|
-0.078
|
MAM
|
0.016*
|
1.928
|
0.349
|
-0.145
|
0.008
|
-0.128
|
0.045
|
0.021
|
-0.005
|
0.072*
|
-0.055*
|
JJA
|
0.015*
|
2.515*
|
1.018*
|
0.632*
|
0.120
|
0.589*
|
0.086*
|
0.039*
|
0.023*
|
0.009
|
-0.032
|
SON
|
0.026*
|
0.126
|
-0.169
|
-0.367
|
-0.025
|
0.012
|
0.003
|
-0.002
|
-0.005
|
0.024
|
-0.037
|
* represents passing the statistical significance test with the confidence level of α = 0.05. |
Climatic inclinations of the regional average annual and seasonal extreme precipitations in GBA from 1961 to 2018 are investigated in Table 1. Trends and variations of these regional extreme precipitation anomalies are shows in Figs. 2, 3 and 4. We find that the trends of R95P and RX1day are negative in SON, positive in DJF, MAM and JJA. Differently, the trend of RX5day is negative in MAM and positive in DJF, JJA and SON. The trend of R99P is negative in MAM and SON, and positive in DJF and JJA. The confidence levels for R95P, R99P and RX5day reach a significant level of α = 0.05 only in JJA. The trends of R10day and R20day are positive except in SON. The confidence levels for R10day and R20day reach a significant level of α = 0.05 in JJA. The trend of SDII is negative in MAM and SON, and positive in DJF and JJA. The confidence level for SDII reaches a significant level of α = 0.05 only in JJA. The trends of CWDday (CDDday) are positive (negative) in all seasons, and the confidence levels reach a significant level of α = 0.05 in spring. In general, R95P, R99P, RX5day, R10day, R20day and SDII have a significant increasing trend in summer, while RX1day, CWDday and CDDday have no obvious trends. In spring, CWDday increases significantly, but CDDday decreases significantly, and there is no significant trend for other extreme precipitation. In winter and autumn, the nine indices all have no significant trends. Annually, R10day, R20day and CWDday have significant increasing trends, CDDday has a significant decreasing trend, but R95P, R99P, RX1day, RX5day and SDII have no significant trends.
Annually, the trend and inter-annual variation of R95P are basically consistent with that of R99P, RX1day and RX5day (Fig. 2a, b, c, d). The characteristics for these 4 kinds of extreme precipitation are also basically consistent in DJF (Fig. 2e, f, g, h), MAM (Fig. 2i, j, k, l), JJA (Fig. 2m, n, o, p), and SON (Fig. 2q, r, s, t), respectively. The correlation coefficients of R95P and R99P in the annual, DJF, MAM, JJA and SON are 0.95, 0.99, 0.93, 0.95 and 0.92, respectively. And then, the correlation coefficients of R95P and RX1day in the annual, DJF, MAM, JJA and SON are 0.64, 0.90, 0.81, 0.82 and 0.87, respectively. Also, the correlation coefficients of R95P and RX5day in the annual, DJF, MAM, JJA and SON are 0.65, 0.91, 0.82, 0.82 and 0.90, respectively. All of these coefficients pass the statistical test with a significant level of α = 0. 001. These 4 kinds of extreme precipitation are classified into the same category and represented by R95P which will be detected abrupt changes and inter-annual oscillations.
The trend and inter-annual variation of annual R20day are basically consistent with that of annual R10day and SDII (Fig. 3a, b, c). Also, the characteristics for these 3 kinds of extreme precipitation are also basically consistent in DJF (Fig. 3d, e, f), MAM (Fig. 3g, h, i), JJA (Fig. 3j, k, l), and SON (Fig. 3m, n, o), respectively. The correlation coefficients of R20day and R10day in the annual, DJF, MAM, JJA and SON are 0.93, 0.94, 0.95, 0.92 and 0.93, respectively. And the correlation coefficients of R20day and SDII in the annual, DJF, MAM, JJA and SON are 0.73, 0.91, 0.83, 0.91 and 0.77, respectively. All of these coefficients also pass the statistical test with a significant level of α = 0.001. These 3 kinds of extreme precipitation are classified into the same category and represented by R20day which will be detected abrupt changes and inter-annual oscillations.
The variations of CWDday and CDDday (Fig. 4) show their own characteristics and may not be classified into the above two categories. Their abrupt changes and inter-annual oscillations will be detected respectively.
Figure 5 shows the results of the Mann-Kendall abrupt change tests for the regional average annual and seasonal R95P, R20day, CWDday and CDDday in GBA. UF and UB are the forward and backward sequential statistics, respectively. The intersection of UF and UB curves denotes the abrupt change point. If the UF and UB curves are entangled and there is no obvious distance, the intersection points are not considered as abrupt change points. The null hypothesis is rejected when any of the points exceeds the confidence interval ± 1.96 that denotes a α = 0.05 confidence level. In JJA, the abrupt changes from less to more values can be seen in 1991 for R95P and in 1993 for R20day; R95P (R20day) has clearly been higher since 1991 (1993). In MAM, R95P shows three abrupt change points, from less to more in 1972 and 2009, and from more to less in 1994; R20day shows three abrupt change points, from less to more in 1965 and 2010, and from more to less in 1990. In DJF and SON, R95P and R20day show no obvious abrupt change points. Annual R95P and R20day show also no obvious abrupt change points. Annually and seasonally, the abrupt change can be seen in early 2010s for CWDday which has clearly been more and for CDDday which has clearly been less; in addition, CWDday in MAM occurs abrupt change points from less to more in 1966 and from more to less in1983.
Figure 6 shows the Morlet wavelet analysis of the regional average annual and seasonal R95P, R20day, CWDday and CDDday in GBA. Annually and seasonally, the variations of R95P have a significant periodic oscillation of 3–5 years, which is relatively significant in DJF and MAM during 1980-1990s, in JJA during 2000s, and in SON during 1960s, 1980-1990s and 2010s; the variations of R95P also have a significant periodic oscillation of 8–14 years, which is relatively significant in JJA and SON. The variations of R20day have a significant periodic oscillation of 3–5 years, which is relatively significant in DJF during 1980-1990s, in MAM during 1960-1070s, in JJA during 1960s, 1990-2000s, and in SON during 1960s, 1980-1990s and 2010s. The variations of CWDday have a significant periodic oscillation of 3–5 years, which is relatively significant in DJF during 1980-1990s, in MAM during 1070s, in JJA during 1960s, 1990-2000s, and in SON during 1970s. The variations of CDDday have a significant periodic oscillation of 3–5 years, which is relatively significant in all period in DJF, in MAM during 1070s, in JJA durng 1960s, 1980-1990s, and in SON during 1960s. The variations of CWDday and CDDday also have both a significant periodic oscillation of quasi-8 years.
3.2 Response of extreme precipitation to abrupt climate change
The annual and seasonal average temperatures show increasing trends in GBA (Fig. 7a, c, e, g, i). Climatic inclinations of the annual, DJF, MAM, JJA and SON average temperatures are 0.021 ℃a− 1, 0.027 ℃ a− 1, 0.016 ℃a− 1, 0.015 ℃a− 1 and 0.026 ℃a− 1, respectively (Table 1). Their confidence levels all reach a significant level of α = 0.05. The warming trend is the most obvious in winter and the weakest in summer. In GBA, the abrupt warming change occurs in 1995, 1989, 1998, 1995 and 1997 for the annual, DJF, MAM, JJA and SON, respectively (Fig. 7b, d, f, h, j). The mutation is the earliest in DJF and the latest in MAM.
The average annual and seasonal precipitations show increasing trends in GBA (Fig. 8a, c, e, g, i). Climatic inclinations of the annual, DJF, MAM, JJA and SON precipitations are 5.517mma− 1, 0.972mma− 1, 1.928mma− 1, 2.515mma− 1 and 0.126mma− 1, respectively (Table 1). The confidence levels reach a significant level of α = 0.05 only in MAM and in annual. For the average precipitation in GBA, the abrupt changes from less to more values can be seen in 2012 in annual, DJF and MAM, and in 1993 in JJA; in addition, the abrupt changes from less to more in 1968 and from more to less in 1991 can be seen in MAM (Fig. 8b, d, f, h, j).
In summary, annually and seasonally, the abrupt warming changes of the average temperature in GBA occur around 1995. The abrupt increasing change of the average precipitation occur in 1993 for JJA, in 2012 for DJF, MAM and annual. The analysis in the 3.1 section shows that most of the abrupt changes of extreme precipitations occur in the early 1990s and around 2010. Based on the main abrupt change characteristics of temperature, precipitation and extreme precipitations, we divide the extreme precipitation change into three stages: 1961–1994, 1995–2009 and 2010–2018, to compare and analyze the difference of average values of extreme precipitations, and to determine the response degree of extreme precipitations to climate change.
Figure 9 shows changes of climate and extreme precipitations in GBA during three different stages. During 1961–1994, the annual, DJF, MAM, JJA and SON average temperatures are 20.4 ℃, 12.4 ℃, 20.5 ℃, 26.9 ℃ and 21.8 ℃, respectively. During 1995–2009, these values are 20.4 ℃, 12.4 ℃, 20.2 ℃, 27.0 ℃ and 21.7 ℃, respectively. There are little changes during 1961–1994 and 1995–2009 stages. But, during 2010–2018, these values are 21.1 ℃, 13.4 ℃, 20.9 ℃, 27.4 ℃ and 22.6 ℃, respectively, which are 0.7 ℃, 1.0 ℃, 0.7 ℃, 0.4 ℃ and 0.9 ℃ higher than those during 1961–1994. In recent decades, the warming of temperature in GBA is mainly manifested after 2010, with the most obvious in DJF and SON.
The annual, DJF, MAM, JJA and SON average precipitations are 1811.2mm, 149.6mm, 581.2mm, 791.0mm and 297.7mm, respectively during 1961–1994; 1874.1mm, 164.3mm, 700.0mm, 729.9mm, and 288.5mm during 1995–2009, respectively; 1966.2mm, 184.9mm, 635.2mm, 865.6mm and 289.4mm during 2010–2018, respectively. The annual and DJF precipitations in the latter stage are increasing compared with those in the former stage and have increased 155.0 mm and 25.3 mm during 2010–2018 than 1961–1994, respectively. The SON precipitation has little changes in the three stages and 2010–2018 is only 8.3mm less than 1961–1994. The MAM precipitation increases significantly during 1995–2009, and decreases during 2010–2018, but the general trend is increasing. The MAM precipitation during 2010–2018 is 54.0 mm more than 1961–1994. The JJA precipitation decreases during 1995–2009, but increases significantly during 2010–2018. The JJA precipitation during 2010–2018 is 74.6 mm more than 1961–1994.
The average annual, DJF, MAM, JJA and SON values account for 28.3%, 31.2%, 28.1%, 25.4% and 34.3% of the total precipitation for R95P, respectively; 15.8%, 15.7%, 12.8%, 11.4% and 17.0% of the total precipitation for R99P, respectively. Changes of annual and most seasonal R95P and R99P are consistent with those of precipitation, but the annual R99P is the most during 1995–2009. The average annual, DJF, MAM, JJA and SON values during 2010–2018 are 58.8mm, 16.4mm, 11.2mm, 33.4mm and − 2.4mm more than 1961–1994 for R95P, respectively; 10.5mm, 8.3mm, -6.0mm, 14.7mm and − 6.4mm more during 2010–2018 than 1961–1994 for R99P, respectively. The annual and seasonal R99P and R95P are increasing except for R95P in MAM and SON and R99P in SON.
The changes of the annual and seasonal R10day, R20day and SDII are consistent with those of precipitation, except for MAM SDII. The values in the latter stage are increasing compared with those in the former stage. The average R10day, R20day and SDII during 2010–2018 are 5.0day, 2.3day and 0.1mm more than 1961–1994 in annual, respectively; 1.1day, 2010–2018 are 0.6day and 0.6mm more than 1961–1994 in DJF, respectively. In SON, the average values in three stages are basically unchanged for R10day, R20day and SDII. In MAM, R10day, R20day and SDII all increase significantly during 1995–2009, and decreases during 2010–2018; the general trend is increasing for R10day and R20day, and decreasing slightly for SDII; the average R10day and R20day during 2010–2018 are 1.5day and 0.7day more than 1961–1994, respectively; but the average SDII during 2010–2018 is 0.2mm less than 1961–1994. In JJA, the average values are decreasing slightly during 1995–2009 and increasing significantly during 2010–2018 for R10day, R20day and SDII; the average R10day, R20day and SDII during 2010–2018 are 2.8day, 1.3day and 0.8mm more than 1961–1994, respectively.
The changes of RX1day and RX5day have their own characteristics, which are affected by both temperature and precipitation. The annual RX1day shows a slight decrease that is only 1.0 mm less during 2010–2018 than 1961–1994. But, the annual RX5day shows a significant increase that is 6.6mm more during 2010–2018 than 1961–1994. In DJF, RX1day and RX5day have same characteristic with precipitation and both of them increase in the latter stage compared with the former stage; the average RX1day and RX5day during 2010–2018 are 4.9mm and 13.3mm more than 1961–1994, respectively. In MAM, RX1day and RX5day in the latter stage are decreasing compared with those in the former stage; the average RX1day and RX5day during 2010–2018 are 3.6mm and 26.5mm less than 1961–1994, respectively. In JJA, the RX1day and RX5day have same characteristic with precipitation and are slightly decreasing during 1995–2009 and significantly increasing during 2010–2018; the average RX1day and RX5day during 2010–2018 are 1.2mm and 16.5mm more than 1961–1994, respectively. In SON, the average values are also basically unchanged in three stages for RX1day, but slightly increasing in the latter stage compared with the former stage for RX5day; the average RX5day during 2010–2018 is 4.5mm more than 1961–1994.
In DJF, MAM and SON, the changes of the CWDday are consistent with those of precipitation and the values are increasing in the latter stage compared with the former stage. The average CWDday during 2010–2018 are 1.1day, 1.6day and 0.4day more than 1961–1994 in DJF, MAM and SON, respectively. In JJA, the average CWDday is decreasing significantly during 1995–2009 and increasing significantly during 2010–2018; the general trend is to decrease; the average CWDday during 2010–2018 is 0.6day less than 1961–1994. The average annual CWDday is also decreasing significantly during 1995–2009 and increasing significantly during 2010–2018; but the general trend is upward; the average annual CWDday during 2010–2018 is 0.5day more than 1961–1994.
Annually and seasonally, the average values of CDDday are the highest during 1995–2009, except MAM. The average annual CDDday in 2010–2018 is 1.3day less than 1961–1994. In DJF and JJA, the average CDDday is increasing during 1995–2009 and decreasing during 2010–2018; the general trend is decreasing. The average CDDday during 2010–2018 is 0.8day and 0.7day less than 1961–1994 in DJF and JJA, respectively. In MAM, the value of the CDDday is decreasing in the latter stage compared with the former stage; the average CDDday during 2010–2018 is 2.3day less than 1961–1994. In SON, the average CDDday is increasing during 1995–2009 and decreasing during 2010–2018; the general trend is increasing; the average CDDday is 0.2day more in 2010–2018 than 1961–1994.