Extreme climate events over northern China during
the last about 50 years
Han Hui, Gong Dao-yi
(Key Laboratory of Environmental Change and Natural Disaster, Institute of Resources
Science, Beijing Normal University, Beijing 100875, China)
Abstract: Climate extremes for agriculture-pasture transition zone, northern China, are
analyzed on the basis of daily mean temperature and precipitation observations for 31 stations
in period 1956-2001. Analysis season for precipitation is May-September, i.e., the rainy season.
For temperature is the hottest three months, i.e., June through August. Heavy rain events,
defined as those with daily precipitation equal to or larger than 50mm, show no significant
secular trend. A jump-like change, however, is found occurring in about 1980. For the period
1980-1993, the frequency of heavy rain events is significantly lower than the previous periods.
Simultaneously, the occurring time of heavy rains expanded, commencing about one month
early and ending one month later. Long dry spell are defined as those with longer than 10 days
without rainfall. The frequency of long dry spells displays a significant (at the 99% confidence
level) trend at the value of +8.3%/10a. That may be one the major causes of the frequent
droughts emerging over northern China during the last decades. Extremely hot and low
temperature events are defined as the uppermost 10% daily temperatures and the lowest 10%
daily temperatures, respectively. There is a weak and non-significant upward trend in frequency
of extremely high temperatures from 1950s to the middle 1990s. But the number of hot events
increases as much as twice since 1997. That coincides well with the sudden rise in mean
summer temperature for the same period. Contrary to that, the frequency of low temperature
events have been decreasing steadily since the 1950s, with a significant linear trend of
–15%/10a.
Keywords: Agriculture-pasture transition zone, climate extremes, trends
CLC number: P461 Document code:
1 Introduction
The agriculture-pasture transition zone over the northern China is one of the most sensitive regions
to global climate changes. And the environment there as well as the regional eco-activity often
suffers from the climate extremes (Ye and Chen, 1992). There are numerous studies analyzing the
climate changes over the transition zone and the vicinities, most of them focused on the mean
condition of climate variables, e.g. the monthly-seasonal mean temperature and total precipitation.
Based on the analysis of meteorological observations, it was reported that the precipitation in
northern China experienced a decreasing trend and notable inter-decadal fluctuations (Huang et.,
1999; Gong and Shi, 2001; Ren et al., 2000). However, to investigate the change of climate extremes
might be of interest, particularly in such an arid-semiarid region. For instance, most of the
precipitations in northern China come forth in the form of strong convective precipitation. Typically,
several heavy rainfalls can contribute a vast amount of the total annual precipitation. Furthermore,
strong convective precipitation is well known for the non-uniformity in temporal and spatial
distribution. Therefore, the same rainy season precipitation amounts may show different significance
for their environmental impacts, if the rainfall’s temporal and spatial distribution is uneven. Thus,
the long-term variations of extreme climate events have significant meanings for environment and
ecological system. Some previous studies analyzed the variations of climate extremes in temperature
and precipitation (e.g., Zhai et al., 1999a; Zhai and Ren, 1997; Zhai et al., 1999b). Zhai et al (1999a)
Received date: Accepted date:
Foundation items: Huo Ying-Dong Education Foundation-81014
Author: Han Hui, Candidate for Master degree, specialized in physical geography. Email. hhandsome@irs.bnu.edu.cn
found that in association with the decreasing of the annual precipitation amount, some extreme
precipitation events also experienced considerable changes in the last 45 years in the northern China.
For example, the one-day maximum and three-day maximum precipitation have decreased, and the
concurrence of heavy rains (including those of daily precipitation in excess of 50mm and 100mm).
While, the mean intensity of those extreme precipitations is getting larger. However, most of these
studies on climate extremes focused on the entire China. Few of studies paid attention on the climate
extremes in the agriculture-pasture transition zone. Clearly, the large spatial extent may ignore
regional features. Here we aim to study the long-term variations of the precipitation and temperature
extremes in the transition zone over the northern China, on the basis of daily precipitation and daily
mean temperature observations during the last about 50 years.
2 Data
The extent and location of the agriculture-pasture transition zone varies among authors as their
studies aimed to different purposes. But, the critical region has no significant difference among them
(Wang et al., 1999). In the present manuscript, the agriculture-pasture transition zone refers to the
area between 105E and 121E in northern China, covering a rather broad extent and including
agriculture-pasture transition zone and the vicinities (See Figure 1). The daily precipitation and mean
temperature data set are derived from the China Meteorological Administration, with the time span
of 1951-2001. We chose 31 gauge stations, which have at least 46 years of data available and have
no more than 10 missing days. Among these 31 stations, there are 30 rainfall stations and 25
temperature stations. The geographical positions of these stations are plotted in Figure 1. There are
some missing records in the early period, thus we employed all data for the time period of 1956
through 2001, 46 years in total. The number of the missing data is small. The interpolation of these
missing data may inappropriate since the daily precipitation is of small spatial-temporal scale. The
interpolation might introduce larg errors. Therefore, we simply ignored these missing records in our
analysis.
Daily precipitation data include six types, namely, (1) no precipitation, (2) trace precipitation
(daily precipitation readings less than 0.1mm), (3) precipitation from frost, fog and dew, (4)
precipitation from pure snow, (5) precipitation from snow and rain, (6) precipitation from rainfall.
Since the precipitation amount of the third type is considerably low, all days with this type of
precipitation readings are regarded as non-precipitation days in our analysis. Analysis on
precipitation extremes focuses on the frequency of heavy rains and the dry spells in rainy season, i.e.,
May-September.
Analysis season for temperature is the warmest three months, i.e. June, July and August. Study
contents include the secular changes of extremely high and low temperatures.
3 Precipitation extremes
3.1 Changes in the frequency of heavy rainfall
Here we defined heavy rain events as those with daily precipitation equal to or larger than 50mm,
including the usually named heavy rainfall and the excessively heavy rainfall (>100mm/day). Since
the excessively heavy rainfall occurs rather rarely in the target region, so we combined it into the
class of heavy rainfall. The frequency of heavy rainfall decrease from south to north, it occurs about
2 times per year on average in North China Plain, and decrease to less than 0.5 times in Inner
Mongolia Plateau. All 30 precipitation stations averaged over the period 1956-2001, the frequency of
heavy rainfall is 0.544 times per year. Zhang and Liu (2003) estimated the frequency of heavy
rainfall over a much narrow region, got a value of 0.5 times per year, that is similar to our result. The
mean precipitation intensity of the heavy rainfall is 72.5mm/day. This suggests that most heavy
rainfall is close to the lower limit of the definition (i.e., the 50mm/day). Although the number of
heavy rainfall days consists of only 1.1% of the total number of rainy days in rainy season, the
precipitation contributes 12.8% of the total May-September precipitation amount.
During the last 46 years, the frequency of heavy rainfall shows a slight decreasing linear trend of
–2.3%/10a, but this trend is not statistically significant. However its temporal variation displays
evident decadal-scale change before and after 1980. In the 1980s and the early 1990s, the frequency
is apparently lower than in the previous period. Comparing of the mean frequency in 1980-1992 with
that in 1973-1979, we found there is notable difference. The t-test value is –3.6, significant at the
99% confidence level. Comparing with the whole period of 1956-1979, the t-value is –2.2, also in
excess of the 95% confidence level. We also tested these statistical values using Monte Carlo method,
2
and yielded the same confidence levels. Analysis shows that the change in the heavy rainfall
frequency may independent of the simultaneous changes in the precipitation amount. The
precipitation shows a slight decreasing trend of –2.6%/10a, but not statistically significant. In
addition, there is no evident decadal change around about 1980. The mean precipitation in
1980-1992 is a little less than that of 1973-1979, but the difference is small. The t-value is –1.8,
below the significant of 95% confidence level. Comparing with the period of 1956-1979, the t-value
is –1.1, not significant too. It shows that the probability of heavy rainfall has significantly changed in
about 1980, though the rainy seasonal precipitation amount does not experience evident change. To
clearly show this feature, we presented the relative frequency of heavy rainfall, defined as the
occurrence in each 1000mm precipitation, in Figure 2. Clearly, the reduction of relative frequency of
heavy rainfall in about 1980 shows up. The relative frequency in 1980-1992 is evidently lower than
in 1973-1979, with a t-value of –3.4. Compared to period of 1956-1979, the t-value is –2.4,
statistically significant too.
Gong and Shi (2001) indicated that the position of the transition zone, defined as the mean
latitudes of 250mm isohyets in summer, experienced notable changes in 1978/1979. The position in
1980s and the early 1990s, is located in southern locations compared to the previous decades. That is
consistent with the changes in the frequency of heavy rainfall as shown in the present study. It is
interesting to note that the decadal changes of heavy rainfall frequency in 1980 is also synchronous
to the large-scale regime changes in summer rainfall over East Asia near 1979/1980. Thus, the
change in frequency of heavy rainfall in the agriculture-pasture transition zone over northern China
is likely related to the large-scale climate change.
3.2 The distribution of heavy rainfall events within rainy season
Figure 3 shows the frequencies of heavy precipitation events during 1 May to 30 September since
1956. Here the frequency is estimated on the basis of pentad, i.e., we counted the total number of
heavy rainfall events in all stations in each year for each 5 days. Each consecutive 5 days is taken as
a pentad starting at 1 May. Since there are 153 days in total, so we defined the latest 8 days as a
pentad. There are 30 pentads in total. The climatological frequency estimated from the entire data
period is also presented in the Figure 3. The highest frequency appears in the 15th-20th pentads, i.e.
from the mid-July to the early August. The maximum peaks in the 18th-20th pentads, i.e., lasting
about half month from the late July to the early August.
Generally speaking, most of the heavy rainfalls occurred in a relatively short time period in all
years. However, there is also a notable change occurred in the 1980s to the early 1990s. In this
period, the temporal distribution of heavy precipitation events shows a more scattered feature within
the rainy season. High frequency can be found in a broader time span, from the 10th to the 25th
pentad (i.e. from the mid-June to the early September). Comparing with other decades, the
commence date of high frequency heavy rainfalls is about one month early, and the ending date is
one month later. Heavy precipitation events for 10th-15th pentads and 20th-25th pentads have
markedly increased, while the frequencies for 15th-20th pentads decreased evidently. This suggests
that the heavy precipitation events may appear in more months during the period of 1980s to the
early 1990s, though the total frequency of heavy rainfall decreased simultaneously.
3.3 Maximum daily rainfall
We also investigated the maximum daily precipitation. From each station each year, the maximum
values are picked out in all 153 days. Then, we averaged these maxima, yielding the regional mean.
Figure 4 shows the corresponding time-series. There is no evident secular trend in the maximum
daily rainfall. This is similar to the precipitation amount and heavy rainfall frequency. However,
there is a notable step-like change around 1980. For the period of 1980s to the early 1990s, the
maximum rainfall is apparently lower than the previous period. In the mid-1990s, the values slightly
increase again.
3.4 Long dry spell duration
A drought usually caused by consecutive days without rainfall or very small rainfall. Consecutive
dry days lead to non-linear increase of the drought stress, and bring great impact on agriculture and
pasture. In order to investigate the precipitation aspect of droughts, we analyzed the secular changes
of the dry spells. Here we defined dry spell as consecutive days without rainfall, the number of days
3
measure its length. Long dry spell is defined as those with longer than 10 days without rainfall.
Precipitation from frost, fog and dew are regarded as no precipitation. We counted all dry spells from
1 May to 30 September station by station for each year. Then, we averaged all stations to produce the
regional mean time series. Figure 5 shows the frequency of the dry spells. Here both the long-dry
spells and the slight dry spells (defined as those shorter than 10days) are shown for comparison.
In the last 46 years, the total frequency of dry spells shows a significant decreasing trend of
–1.63%/10a, above the 95% confidence level. We found that remarkable reduction of the decreasing
trend is mostly related to the strong reduction in the frequency of slight dry spells. Decreasing trend
of the slight dry spells is –2.49%/10a (significant at the 95% confidence level). It is of importance to
note that the frequencies of long dry spells displays a strong increasing trend of +8.34%/10a in
excess of the 99% confidence level. This linear trend suggests that the long dry spell has increased
by about 38% in the last several decades. It can be inferred from the analyses that drought stress,
which caused by changes in precipitation character, is intensively increasing over the
agriculture-pasture transition zone and the vicinities. It might be one of the reasons for the frequent
drought there.
We checked the relationship between long dry spells and the precipitation amount in the
corresponding years, found that there is no significant, stable relationship between them. An increase
in the precipitation amount for a given year does not necessarily mean there is a consistent reduction
of long dry spells. The years with the highest frequencies of long dry spell are 2001, 1999, 1972,
1966 and 1982 respectively. Nevertheless, the five years with the lowest precipitation amount are
1965, 1972, 1980, 1973 and 1984. The years with the lowest frequencies of long dry spells are 1964,
1967, 1985, 1973 and 1984 respectively. But the five years with the highest precipitation amount are
1959, 1969, 1956, 1964 and 1973. Clearly, there is evident difference in the exact years. Of course,
the long dry spell usually gets more when the precipitation amount is lower than the normal. If we
calculate the frequency difference of the long dry spell between high and small precipitation years;
or calculate the difference of mean precipitation amount between the high and low dry spell
frequency years, the differences are statistically significant. However, it should be pointed out that
there is never a case-to-case connection. In some years such as 1966, the precipitation anomaly is
–6.5%, nearly the normal, but the frequency of long dry spell in this year is the fourth highest in the
past several decades. The main reason for this situation might be the enhancement of precipitation
intensity. Since the precipitation occurred in a shorter time, as a consequence, the number of days
without rainfall would increase. We found the frequency of heavy rainfall is higher than normal in
1966 (see Figure 2). That supports the above notion. Similarly, in 1984 the frequency of long dry
spells is the fifth highest since 1956, whereas the precipitation anomaly is only –0.15%.
Therefore, we can infer that the slight increase of precipitation amount does not necessarily
mean there is a decrease in the probability of drought occurrence or mitigation of the drought, and on
the other hand, the decrease of precipitation amount does not necessarily mean that there is an
increase in the probability of drought event. Similarly, the years with the same negative anomalies of
precipitation amount may cause heavy drought, may not. The variations of precipitation intensity
and rainfall distribution within rainy season have great influence on the occurrence of drought and
its intensity. The increase in frequency of long dry spell in the target region, are mainly accounted
for by the decrease in the number of light rain days. The number of light rain days (with daily
precipitation less than 10mm) shows a significant decreasing trend since the 1950s. This trend is
equal to –1.42 days per 10 years, significant at the 95% confidence level (Gong et al., 2003). In
addition, the changing number of light rain days is also probably related to the changes in the cloud
amount and humidity; Kaiser (1998, 2000) has found that the cloud amount is descending in most
region of China since 1950s. The decrease is particularly notable in the northern and northeastern
China. From 1950s to the mid-1990s, the decreasing trend of cloud amount in the northern China is
–3% per 10 years, significant at the 95% confidence level. At the same time, the ground relative
humidity shows a decreasing trend of –1.0% to –1.5% per 10 years. These all can give rise to the
reduction of the number of light rain days, resulting in more and longer dry spells, consequently,
yielding more droughts there.
4 Extreme temperature events
4.1 The definition of extreme temperature events
Following an often used method (e.g, Houghton et al., 2001), the highest 10 percent of mean daily
temperatures are defined as the extremely high temperatures, the lowest 10 percent are the extremely
4
lows. Different from the situation of precipitation, here the analysis months for temperature extremes
are confined to June, July and August, i.e., the hottest three months. If we considered the entire
May-September, most anomalously low temperatures will take place in May or September. There all
analysis for temperature is based on the 1 June to 31 August. Firstly, we sorted all daily temperatures
for a given station in the descending order. The first 10% temperatures are the extremely high values,
and the last 10% are the extremely low temperatures. Their corresponding dates then are determined
and summed to produce the frequency of the extremes each year. After averaging the time series
for all stations, we got the regional mean series of the frequency of temperature extremes. For each
station there is 423 days of extremely high temperature and the same number of extremely low
temperatures, 9.2 days per year (i.e., 10% of the number of summer days is 9.2, there are 92 days
from 1 June to 31 August).
4.2 Extremely high temperatures
Figure 6 shows the regional mean frequency of extremely high temperature events. It shows an
strong increasing trend of +20.9% per 10 years, this trend is statistically significant above the 99%
confidence level. However, it should be pointed out that the trend is mainly caused by an abrupt
increase that occurred in the late 1990s. The average frequency of the extremely high daily
temperatures in the latest 10 years is 14.1 days per year; it is two times the value of 7 days per year
for period of mid-1950s to mid-1960s. When the period of 1997-2001 excluded, the trend in
1956-1996 is only +4.1% per 10 years, not significant. Therefore, we can infer that the frequency of
extremely high daily temperature events displays a slight and not significant increasing trend from
the mid-1950s to 1996, whereas the frequency increases sharply since 1997.
4.3 Extremely low temperatures
Figure 6 also shows the time series of the regional mean frequency of the extremely low daily
temperature events. Generally, the frequency displays a rather steady and consistent decreasing trend.
That is clearly different from that in the changes of high temperature extremes. On average the
frequency of low temperatures is about 10 days per year during the mid-1950s to the early 1960s. In
the recent decade the frequency decreases to about 6 days per year. The linear trend estimated from
the entire data period is –15.09%/10a, significant above the 99% confidence level.
Changes in climatic mean and variability both can lead to variations in the frequency of
temperature extremes (Houghton et al., 2001; Katze and Brown, 1992; Robeson, 2002). The
variations in summer mean temperature show a rather significant warming trend since the late 1970s;
the trend is 0.65C per 10 years during the period of 1976-2001. For the entire data period of
1956-2001 the linear trend is 0.24C per 10 years, both significant at the 99% confidence level. This
long-term trend is consistent with the trends of the temperature extremes. In addition, the
inter-annual fluctuations between the low temperature frequency and mean temperature also display
good agreement. The frequency of extremely high temperatures increases with the mean temperature;
while the frequency of low temperatures decreases as the mean temperature gets warmer. It can be
inferred that there are close relations between the increasing mean temperature and the frequency of
temperature extremes. Some studies (e.g., Zhai and Ren, 1997) indicated that the minimum
temperatures in northern China showed an increasing trend from 1950s to 1990s. That is consistent
with our analysis too.
It should be noted that the increase in mean temperature is most prominent since the late 1990s.
The warming trend in the previous period is not remarkable. That feature is rather similar to the
changes in the high temperature extremes. Thus, the recent abrupt increase in the frequencies of
extremely high temperatures might be largely related to the evident warming of mean temperature.
5 Discussion
The immediate causes for the anomalous daily precipitation and temperature are the variations of
daily synoptic conditions. To thoroughly analyze the relationship between the daily circulation
characteristics, patterns and the surface weather conditions is a very difficult work. In fact, many
synoptic analyses are based on the individual case study. Considering the variations in climate
extremes have close relations with the mean weather situations; furthermore, some indices such as
long dry spell itself and its long-term changes are related to the low frequency process of the
atmospheric circulation. Therefore, here we study the corresponding monthly-seasonal mean
circulation condition. That would provide useful information of circulation background for
understanding the climate extremes.
5
Figure 8 shows the composites of 500hPa geopotential height field for four cases. The 500hPa
difference between 5 highest long dry spell frequency years and the 5 lowest frequency years are
calculated and plotted, similarly, the 500hPa changes between largest and smallest precipitation
years, between highest and lowest hot temperature frequency years, and between warmest and
lowest mean temperature years all are estimated, for each case 5 highest-value and lowest-value
years are considered. For the temperature cases the concerning months are June-August, for
precipitation cases are from May-September. There are great similarity between Figures 8(a) and
8(b), both displays a center situated in 40-50N and 90-100E with the large height anomalies; at
the same time, a smaller and weaker center is located over the Korean peninsular. This kind of
circulation pattern suggests that in association with the dry conditions in the transition zone, there
tend to emerge an anomalous anticyclone circulation in the western vicinities, and a cyclonic
circulation in the Korean peninsular and the vicinities, and vice versa. The circulation anomaly can
exert direct influence on the precipitation in the transition zone. It is clearly that the anomaly center
in the landmass is very strong, implying the importance of the influence from the inland.
Unlike the circulation situations related to the precipitation anomalies, the 500hPa height gets
higher over the region of 30-50N and east of 90E in association with the extremely high frequency
of hot extremes and mean temperatures. The center of the anomalous circulation is about 2-3 degrees
latitude north of the transition zone. This kind of pattern implies that under the control of anomalous
anticyclone circulation, surface temperature would increase and the number of extremely high
temperatures would get more. And can be seen from Figures 8(c) and 8(d), the anomaly heights is
east-west oriented, that suggests the subtropical high also locates more north than its normal position.
Therefore, in addition to the local circulations in the northern China some other factors in the tropics
may also be important for the climate and extremes in the northern China.
Besides the atmospheric circulation, other two important factors have also often been
emphasized. One is the global warming. Some studies (Hulme, 1996; Jones and Reid, 2001)
analyzed the relationship between the global warming and the precipitation or temperature in various
arid-semiarid regions, based on observations. But no overall relations were found. A probably
important reason is that the mechanism of global warming impacting regional climate is different
among regions. The studies of its influence on climate extremes are even more difficult. Recently,
the regional simulation results show that during a warmer climate the frequency of heavy rainfall in
northern China will increase (Gao et al., 2002), at the same time the maximum and minimum
temperatures will get higher. Clearly, as a consequence, the frequency of extremely high
temperatures will increase and the frequency of extremely low temperatures will decrease.
Simulations forced by the greenhouse gases also show that there are increasing trends in the
frequency of heavy rainfall and its contribution to the total precipitation amount in middle latitude
North America (Wilby and Wigley, 2002). The other factor is the land cover change. It was noticed
that the surface vegetation deterioration could influence temperature notably (Balling, 1991). Kalnay
and Cai (2003) recently indicated that the influence of land cover change and urbanization on
temperature is much larger than the estimation by previous studies. Based on the climate model
experiments, Zhao and Pittman (2002) reported that the influence of land cover change on extreme
temperature and precipitation is as large as the influence of the CO2-induced global warming. Of
course, these studies are preliminary. How do the global warming and land cover change influence
the climate and the extremes in agriculture-pasture transition zone need further study.
6 Conclusion
Our analyses show that there is no significant increasing linear trend in the frequency of heavy
rainfall in the agriculture-pasture transition zone and the vicinities. However, an evident
inter-decadal change occurs around 1980. The frequencies of heavy rainfall decrease significantly
during the 1980s to the early 1990s. At the same time the heavy precipitation events may appear in
more months, the commence date of high frequency heavy rainfalls being about one month early,
and the ending date being one month later. The climatological frequency estimated from the entire
data period shows that the highest frequency emerges in the mid-July to early August.
From the mid-1950s to 1996, the frequency of extremely high temperatures shows a slight
increasing trend, but not significant. However, the frequency of extremely high temperatures
increases abruptly since 1997. The frequency of low temperatures decreases steadily since the 1950s,
with a significant linear trend of –15.09% per 10 years.
There are no strict case-to-case relations between the long dry spells and the precipitation
6
anomalies for a given year. Although the precipitation amounts experience no notable trend, the
frequency of the long dry spells displays a quite significant increasing trend of +8.34%/10a. Thus,
the changing temporal distribution of rainfall within the rainy season is probably a very important
cause for the increasing drought stress in the agriculture-pasture transition zone.
References
Balling RC Jr. Impact of desertification on regional and global warming. Bulletin of the America Meteorological
Society, 1991, 72:232-234.
Gao XJ, Zhao ZC, Giorgi F. Changes of extreme events in regional climate simulations over East Asia. Advances in
Atmospheric Sciences, 2002, 19(5): 927-942.
Gong D Y, Shi P J, Wang J A. Daily precipitation changes in semiarid region over northern China. Submitted, 2003
Gong DY, Ho C.-H. Shift in the summer rainfall over Yangtze River valley in the late 1970s. Geophysical Research
Letters, 2002, 29:10.1029/2001GL014523.
Gong DY, Shi PJ. Variability of summer rainfall over northern China and its association with thermal condition at
early stage underlying surface. Journal of Natural Resources, 2001, 16(3): 211-215.
Houghton J T, Ding Y H, Griggs D J, Noguer M, van der Linden P J, Dai X, Maskell K, and Johson C A. (eds.).
Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third Assessment Report of
the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, 2001.
Huang RH, Xu YH, Zhou LT. Interdecadal variability of summer precipitation in China and drying tendency over
northern China. Plateau Meteorology, 1999, 18(4): 465-476.
Huang RH. Decadal variability of the summer monsoon rainfall in east Asia and its association with the SST in the
tropical Pacific. CLIVAR-Exchanges, 2001, 6:7-8.
Hulme M. Recent climatic change in the world's drylands. Geophysical Research Letters, 1996, 23:61-64.
Jones PD, Reid P. Temperature trends in regions affected by increasing aridity/humidity. Geophysical Research
Letters, 2001, 28:3919-3922.
Kaiser DP. Analysis of total cloud amount over China, 1951-1994. Geophysical Research Letters, 1998,
25:3599-3602.
Kaiser DP. Decreasing cloudiness over China: An updated analysis examining additional variables. Geophysical
Research Letters, 2000, 27:2193-2196.
Kalnay E, Cai M. Impact of urbanization and land use change on climate. Nature, 2003, 423:528-531.
Katze RW, Brown BG. Extreme events in a changing climate: Variability is more important than averages. Climatic
Change,1992, 21:289-302
Ren GY, Wu H, Chen ZB. Spatial patterns of change trend in rainfall of China. Quarterly Journal of Applied
Meteorology, 2000,11(3): 322-330.
Robeson S M. Relationship between mean and standard deviation of air temperature: implications for global warming.
Climate Research, 2002, 22:205-213.
Wang JA, Xu X, Liu PF. Landuse and land carrying capacity in ecotone between agriculture and animal husbandry in
northern China. Resource Sciences, 1999, 21(5): 19-24.
Wilby RL, Wigley TML. Future changes in the distribution of daily precipitation totals across North America.
Geophysical Research Letters, 2002, 29:doi:10.1029/2001GL013048.
Ye DZ, Chen PQ. China’s global change study: A primary perspective. Meteorological Press, Beijing, 1992.
Zhai PM, Ren FM，Zhang Q. Detection of trends in China's precipitation extremes. Acta Meteorologica Sinica, 1999a,
57(2): 208-216.
Zhai PM, Ren FM. On changes of China’s maximum and minimum temperatures in the recent 40 years. Acta
Meteorologica Sinica, 1997, 54:418-429.
Zhai PM, Sun AJ, Ren FM, Liu XN, Gao B, Zhang Q. Changes of climate extremes in China. Climatic Change,
1999b, 42:203-218.
Zhang WB, Liu BY. Spatial distribution of precipitation extremes in the ecotone between agriculture and animal
husbandry in northern China. Journal of Natural Resources, 2003, 19(3): 274-280.
Zhao M, Pitman AJ. The impact of land cover change and increasing carbon dioxide on the extreme and frequency of
maximum temperature and convective precipitation. Geophysical Research Letters, 2002, 29:
doi:10.1029/2001GL013476.
7
Figure 1 Study region and stations
Figure 2. (a) Time series for heavy rainfall events (daily rainfall exceeding 50 mm), (b)
precipitation amount in May-September, mm, (c) relative frequency of heavy rainfall occurs in
association with 1000mm of precipitation.
8
Figure 3 Heavy precipitation events counted on basis of pentads. The first pentad is 1-5 May;
the 30th pentad is for 23-30 September. (b) The climatological frequency for each pentad
estimated from the whole period.
Figure 4. Mean maximum daily precipitation (mm/day) averaged over all stations each year.
9
Figure 5. (a) Frequency of short dry spells (length is less than 10 days). (b) Frequency of long
dry spells (length is equal to or longer than 10days). (c) The total number of the dry spells.
Dashed lines are linear trends.
Figure 6. Frequency of extremely hight (a) and low (b) daily temperature events. Linear trends are
shown as dashed lines.
10
Figure 7. Regional mean temperature for June-August. Shown are anomalies with respect to
1961-1990. Dashed lines are linear trends.
Figure 8. Composites of 500hPa heights. (a) Five high dry-spell frequency years minus the
lowest five years. (b), (c) and (d) are same as in (a) but for precipitation amount, extremely hot
days, and mean temperature, respectively. Contour interval is 10gpm. Negative contours are
shown as dashed lines. (a) and (b) are for May-September, (c) and (d) for June-August.
11