Changes in seasonal patterns of temperature and ... - Springer Link

ADVANCES IN ATMOSPHERIC SCIENCES, VOL. 24, NO. 3, 2007, 459–473

Changes in Seasonal Patterns of Temperature and Precipitation in China During 1971–2000

Ý Ë), A. J. CANNON

SONG Lianchun∗1 ( 1

2,3

, and P. H. WHITFIELD3

Institue of Arid Meteorology, China Meteorological Administration, Lanzhou 730000 2

Department of Earth and Ocean Sciences, University of British Columbia, Vancouver, British Columbia V6T 1Z2, Canada

3

Meteorological Service of Canada, Environment Canada, #201–401 Burrard Street, Vancouver, British Columbia V6C 3S5, Canada (Received 4 August 2005; revised 27 September 2006) ABSTRACT

Many studies have shown evidence for significant changes in surface climate in different regions of the world and during different seasons over the past 100 years. Based on daily temperature and precipitation data from 720 climate stations in China, cluster analysis was used to identify regions in China that have experienced similar changes in the seasonal cycle of temperature and precipitation during the 1971–2000 climate normal period. Differences in 11-day averages of daily mean temperature and total precipitation between the first (1971–1985) and second (1986–2000) halves of the record were analyzed using the MannWhitney U test and the global k-means clustering algorithm. Results show that most parts of China experienced significant increases in temperature between the two periods, especially in winter, although some of this warming may be attributable to the urban heat island effect in large cities. Most of western China experienced more precipitation in 1986–2000, while precipitation decreased in the Yellow River valley. Changes in the summer monsoon were also evident, with decreases in precipitation during the onset and decay phases, and increases during the wettest period. Key words: k-means clustering, seasonality, trends, temperature, precipitation, China DOI: 10.1007/s00376-007-0459-1

1.

Introduction

Human activity, economic development, and ecological systems have been affected by changes in the global climate that have occurred in past centuries (Karl et al., 1995). Observed and projected global warming in the twentieth and twenty-first centuries has affected and will continue to affect agriculture, the hydrological cycle, environmental conditions, and the development of China’s society and economy (Tao et al., 2003; You, 2001; Liu et al., 2003). Smit and Cai (1996) describe possible implications for agriculture in China resulting from climate change. Current agricultural production is finely tuned to climatic features and is sensitive to competition for water resources from other sectors. While some crop yields ∗ E-mail:

[email protected]

may increase in a warmer climate, moisture deficits may threaten the stability of production. Based on GCM simulations of future climate, Xu and Yan (2001) forecast a northwards shift in the distribution of Pinus koraiensis, with a potential expansion in area of 3.4%. Based on results from a regional climate model, Chen et al. (2003) suggest the possibility of dramatic changes to “life zones” in China, with existing zones changing location and new zones appearing. In terms of hydrology, Guo et al. (2002) found that humid basins in China may be less sensitive to climate change than semi-arid basins, with runoff exhibiting greater sensitivity to changes in precipitation than temperature. Qian and Zhu (2001) identified recent changes in low-flow conditions in the Yellow River and northwest China and increased flood frequency in

460

SEASONAL PATTERNS OF TEMPERATURE AND PRECIPITATION IN CHINA

the Yangtze due to changes in the monsoon. The Chinese climate and its recent trends vary over time and space. The average temperature in the whole country has increased over the past 50 years, most notably since the 1980s (Wang et al., 2004). China experienced two warm periods during the 1940s and 1990s and a relatively cold period during the 1950s and 1960s (Chen et al., 2004). Significant warming started in 1986, with the largest effects observed in northeast China in winter (Qian and Zhu, 2001). Zhai et al. (1999) found that the annual temperature increased significantly throughout China, particularly at higher latitudes, during the past 40 years. In an analysis of temperature extremes for China, Zhai and Pan (2003) found decreasing trends in the number of frost days, cool days, and the frequency of cool nights, and increasing trends in the number of warm days and warm nights. Since the 1950s, a belt of cooling centered over the Sichuan Basin has been observed (Chen et al., 2004). Jones and Moberg (2003) identified a warming trend in annual temperatures over China of +0.07◦C per decade. With the exception of a cooling signal during fall in northern China, warming was found in all seasons. Tang and Ren (2005) investigated countryaveraged surface temperature anomalies for the period 1905–2001. Monthly mean temperatures increased by 0.08◦ C per decade, which was larger than the global average. More warming was found in winter and in spring. Urbanization effects on warming, which were not accounted for, needed to be incorporated to better detect regional changes due to global climate change. Zhou et al. (2004) identified a significant urbanization effect on surface temperatures (+0.05◦ C per decade) in southeastern China, the region where the most rapid changes in land use have occurred in recent decades. Zhang et al. (2005) showed that urbanization and other land use changes may contribute to the observed 0.12◦ C (10 yr)−1 increase for daily mean surface temperature in China. Similarly, Chu and Ren (2005) found that the average warming rate for the last 40 years near Beijing was higher in the urban stations (+0.22◦ C per decade) than in the rural stations (+0.06◦ C per decade). The contribution of urbanization effects on warming trends reached 71.1% over the past 40 years, but only 48.5% for the last 20 years, possibly due to greater influence of large scale global climate change. The lowest values of precipitation over China were observed during the 1920s, with a maximum reached in the 1950s, and a gradual decrease observed since (Chen et al., 2004). After the 1970s, rainfall amounts have shown only a weak oscillation, although since the 1980s, the pluvial region moved southward from northern China to the middle and lower reaches of

VOL. 24

the Yangtze River. In general, less precipitation and higher temperatures have been observed in the Yellow River valley. In the most recent 20 years, precipitation in northern China deceased, especially in summer, while precipitation in northwest China exhibited an increasing trend (Song and Zhang, 2003). Changes in the climate of China show complexity in both the spatial and temporal dimensions (Shen and Varis, 2001; Niu et al., 2004). In addition, changes are superimposed on a climate that is characterized by considerable spatial and temporal variability, reflecting, for example, monsoon, maritime, and continental influences. All areas in China are subject to large seasonal and interannual variations in temperature and precipitation. We examine these changes in more detail in this study. Statistical techniques to identify changes in seasonality and its spatial distribution were proposed by Leith and Whitfield (1998) and Whitfield and Taylor (1998). Whitfield and Cannon (2000) extended this methodology by applying a statistical clustering procedure to assess spatial patterns of changes in Canada. Whitfield et al. (2002) demonstrated that the spatial separation of the clusters could be improved by analyzing shorter time intervals, e.g., pentad or 11-day averages, rather than monthly or seasonal averages, which tended to mask significant changes. The present study adopts these techniques to identify regions of China that have experienced similar changes in seasonal cycles of temperature and precipitation during the 1971–2000 climate normal period, with a focus on the spatial and temporal distribution of recent variations in the climate system. Trend analysis and cluster analysis methods are applied to Chinese reference and base climate station data split into two periods: 1971–1985 and 1986–2000. Results from this study will help to resolve variations in climate between these two periods, detecting both recent spatial and temporal variations in temperature and precipitation across China. While these 30 years of records are insufficient to determine long-term trends, they do provide a large enough contemporary dataset to allow examination of recent spatial patterns of change. 2.

Data and methods

Daily mean temperature and total daily precipitation records were obtained from reference and base climate stations in China for the 1971–2000 climate normal period, which coincides with the period of greatest global warming during the 20th century (IPCC, 2001). We obtained daily climate data from the Climate Data Office of the National Meteorological Information Center in China. We used daily values for mean tempera-

NO. 3

SONG ET AL.

461

Fig. 1. Indices used to determine the number of clusters (a) for 11-day temperature and (b) for 11-day precipitation. Vertical axis labels indicate whether the optimum number of clusters is given by the local minimum or maximum of the second derivative of each index (min 2nd or max 2nd), or whether they are given by the local minimum or maximum (min or max) of each index. Note: calculation of the second derivative removes two points from the beginning of each series.

Fig. 2. Temperature stations clustered at the eight-cluster level in China.

ture and total precipitation that have been extensively quality-controlled at three levels: observation station, provincial data office, and national data office. In total, 143 reference stations and 577 base stations were selected. Here we treat both of these types of stations equally. Of the 720 available stations, 633 stations met completeness criteria. Calculations of temperature and precipitation series to be used in subsequent statistical analyses were

based on 11-day averages, which separate the year into 33 periods. This level of smoothing is slightly longer than the average lifespan of synoptic-scale systems. Whitfield et al. (2002) concluded that the use of 11day averages resulted in better spatial resolution of changes in seasonality than monthly or annual series. The technique used to detect significant differences in temperature and precipitation between the first (1971–1985) and second (1986–2000) halves of the

462


VOL. 24

Fig. 3. Temperature cluster 1 at the eight-cluster level. Shown are (a) boxplots summarizing the distribution of 1-p-value scores over stations, along with curves showing station median temperatures for 1971–1985 (solid black line) and 1986–2000 (dashed grey line); arrows indicate increases or decreases between the 15-year periods; and (b) maps showing distribution of cluster members.

Fig. 4. As in Fig. 3, but for temperature cluster 2 at the 8-cluster level.

1971–2000 climate normal period are described by Leith and Whitfield (1998) and Whitfield et al. (2002). For each 11-day period, the two 15-year medians, the sign of the difference in the medians, and the level of statistical significance (p-values) for differences between the first and second halves of the record, were calculated. The statistical significance of differences between the 15-year median values was then determined using the Mann-Whitney U test—a nonparametric, rank-based test that is robust against nonnormality in distributions of two test samples. In all cases, values of one minus p-value were multiplied by the sign of the observed difference, and then normalized to eliminate seasonal differences in variability. To distinguish regions exhibiting similar changes in the seasonal cycle of temperature and precipitation, results from the significance test were clustered using

the global k-means algorithm (Likas et al., 2003). The global k-means algorithm was selected to improve clustering performance and to reduce the influence of initial seed selection on the cluster solutions. The number of clusters to be retained was determined by inspecting four separate measures of cluster separation: that of Calinski and Harabsz (1974), Scott and Symons (1971) (SS index), Milligan and Cooper (1985) (Trace W index), and Davies and Bouldin (1979) (DB index), all as adapted by Weingessel et al. (2002). Each separation measure was calculated for numbers of clusters ranging from 2 to 10. The most suitable cluster levels for temperature and precipitation were determined from a comparison of the top four results for each index and visual inspection of the spatial arrangement of clusters, similar to Whitfield et al. (2004).

NO. 3

SONG ET AL.

463

Fig. 5. As in Fig. 3, but for (a–b) temperature cluster 3 and (c–d) temperature cluster 3U (stations in large urbanized centers) at the eight-cluster level.

Fig. 6. As in Fig. 3, but for temperature cluster 4 at the eight-cluster level.


464 3.

Results

For both temperature and precipitation, results from the cluster separation indices suggested the eightcluster solution as being “best” (Fig. 1). It should be noted that the cluster analysis was run without any constraints that might force the clusters to be spatially contiguous. While contiguity-constrained clustering is possible, it was felt that the data should be the arbiter of spatial homogeneity. This resulted in some geographically separated stations being grouped together—a phenomenon called “aggregation error” by Gong and Richman (1995). While they found the kmeans algorithm to be more resistant to this type of error than other clustering algorithms that they tested, it is a statistical artifact of cluster analysis procedure. Since the clustering was carried out on the differences between the first and second halves of the climate normal period rather than on the climatologies, a greater likelihood of aggregation errors was expected. Discussion of the clustering results has been limited to the major contiguous regions; outliers are treated as statistical artifacts. While outliers are obvious in the figures, they do occur relatively infrequently (11 stations or 1.7% of the total for temperature, and 25 stations or 3.9% of the total for precipitation). 3.1

Temperature cluster results

The spatial distribution of the eight temperature clusters is shown in Fig. 2. The station locations for each cluster are plotted over the topography and province boundaries within China. These provide the reader with visual aids with respect to station locations and possible controls on spatial contiguity. The statistical results for the clusters are shown in Figs. 3–10. Each figure shows the station median temperatures during the first (1971–1985) and second (1986– 2000) halves of the climate normal period, along with boxplots summarizing the distribution of the index scores from the Mann-Whitney U test (one minus the p-value multiplied by the sign of the change between the first and second halves of the climate normal period). The legend in the upper left corner shows the scale for the index scores, ranging from −1 (strongly significant cooling) to +1 (strongly significant warming). The line in each box gives the median score over all stations in the cluster; the boxes extend from the 25th to 75th percentiles, with the whiskers extending to 1.5 times the interquartile range. The stations in cluster 1 are located in western and northern Xinjiang (Tarim Basin and Junggar Basin) and eastern Sichuan Province (Sichuan Basin) (Fig. 3). The climatology of both of these regions is characterized by cold winters and hot summers. The average

VOL. 24

of 11-day daily mean temperature is below 0◦ C from late December to the beginning of February. The mean temperature during the summer is 25◦ C. In this cluster, there is moderate confidence for a warming trend in the winter half of the year and a cooling trend in the summer half of the year. The index of statistical change shows that significant increases occurred during winter (late December to early February) and significant decreases occurred during late March and early August. The maximum difference in temperature between 1971–1985 and 1986–2000 reached 2◦ C in winter. Stations in cluster 2 extend from western Guizhou and the northeast of Yunnan Province through to western Sichuan Province and into the eastern Tibet Autonomous Region (Eastern Tibetan-Qinghai Plateau) (Fig. 4). The climate of stations in this cluster is mild, lacking hot summers and cold winters, with no obvious seasonal boundary. Significant confidence is shown for a warming trend in November, the end of December to early January, and in early February. At the beginning of summer there is evidence of a warming trend. Moderate confidence for a cooling trend is seen in spring and fall. Cluster 3 is spatially discontinuous, with stations located in southwest Yunnan Province, northern Hebei, Beijing, Shanghai, central Inner Mongolia, Liaodong Peninsula, Laizhou Peninsula, Jiaodong Peninsula, and Hainan Island. There is strong evidence for a warming trend in all periods in the year, particularly in early February, late August to early September, and late December. The temperature in the latter half of the climate normal period is typically much higher than in the first half. The largest warming is observed in early February with a 3◦ C difference in temperature between 1971–1985 and 1986– 2000. It is possible that the warming trend in cluster 3 may be partially attributable to increasing urbanization in the Yangtze River Delta, Zhujiang River Delta, and the area around Bohai Gulf, such as Shanghai, Shantou, Shenzhen, Haikou, Beijing, Zhangjiakou, Qinhuangdao, Dalian, and Weihai, which are the important economic zones and urban centres in China. All of the stations in these areas were found in cluster 3. They have been separated into a separate cluster, 3U, with results plotted in Figs. 5c and d. These stations have experienced rapid urbanization since the 1980s, with evidence for warming due to the urban heat island effect (Zhou et al., 2004; Chu and Ren, 2005). Stations in cluster 4 are mainly located in the northwest of China in the Turfan Depression, including northern Shaanxi Province, western Gansu Province, Ningxia Autonomy, western inner Mongolia Auton-

NO. 3

SONG ET AL.

465



omy, eastern Xinjiang Autonomy and northern Shanxi Province (Fig. 6). The climate of stations in this cluster is characterized by strong seasonality, with significant intra-annual temperature differences, a short summer, and a long winter. The average temperature reaches 22◦ C in summer and −10◦ C in winter. There is evidence for a significant warming trend throughout the year, except for the middle of January, early April, early and middle August, and the end of October. Stations in cluster 5 extend from the middle and lower reaches of the Yellow River to the Yangtze River and Huaihe River regions and the Zhujiang River, including southern Hebei Province, most of Shandong and Shaanxi Provinces, eastern Gansu, southern Shaanxi, Henan Province, Anhui Province, Jiangsu Province, and Hubei Province (Fig. 7). Stations in this cluster exhibit hot summers and mild winters. There is strong evidence for a warming trend throughout the year, with only a few 11-day periods show-

ing neutral or cooling trends. Moreover, the warming magnitude in winter is quite significant. Cluster 6 is the only cluster that is spatially contiguous and has no overlap with other clusters. Its stations are located in the south of the Yangtze River, including Zhejiang, Fujian, Hunan, and Guangdong Provinces, the Guangxi Autonomy, and most of Guizhou Province (Fig. 8). This cluster represents the hottest area of China. Twenty-two of the 33 11-day periods show evidence of a warming trend, with significant warming at the beginning of January, the first period of February, and the last periods of December. Eleven periods show no change or a cooling trend, with the most significant cooling occurring in the last period of March. Cluster 7 is located in the Tibetan-Qinghai Plateau and the coastal region along the South Sea (Fig. 9). This cluster has the smallest intra-annual temperature variability, with a very mild summer season. Cluster 7

466


VOL. 24



exhibits significant confidence in warming in most periods of the year, especially in early winter. Temperature differences between the first and second halves of the climate normal period exceed 1◦ C in the first 11-day period of January, the first period of March, most of November, and the last period of December. Stations in cluster 8 are located in the northeast of China, including Jilin, Liaoning, and Heilongjiang Provinces, along with eastern Inner Mongolia and eastern Hebei Province (Fig. 10). These stations are marked by the coldest climate in China, with mean daily temperatures in the cluster reaching −15◦ C in winter. In comparison with the other clusters, there is very strong confidence for a warming trend throughout the year, particularly in late January and early February, the end of August, the beginning of October, and the end of December; only three of the 33 11-day periods show no evidence for a warming trend. The magnitude of warming in five of the 33 periods

exceeds 2◦ C. 3.2

Precipitation cluster results

The spatial distribution of the eight precipitation clusters is shown in Fig. 11. As in Figs. 3–10 for temperature, results from the Mann-Whitney U test and the cluster analysis for precipitation are shown in Figs. 12–19. In general, the high level of intra-seasonal variability in precipitation means that clear patterns in the seasonal trends are less evident than for temperature. Stations in cluster 1 are located in the eastern portion of southwest China in the Zhujiang Basin, with members in Guangxi Autonomy, Guizhou Province, the south of Yunnan Province, the west of Hunan Province, the west of Guangdong Province, the south of Hainan Province, and the Sichun Basin (Fig. 12). The climatology is characterized by large precipitation amounts in summer. Precipitation from June to July

NO. 3

SONG ET AL.

467

Fig. 11. Precipitation stations clustered at the eight-cluster level in China.

Fig. 12. Precipitation cluster 1 at the eight-cluster level. Shown are (a) boxplots summarizing the distribution of 1-p-value scores over stations, along with curves showing station median precipitation for 1971–1985 (solid black line) and 1986–2000 (dashed grey line); arrows indicate increases or decreases between the 15-year periods; and (b) maps showing distribution of cluster members.

in the second half of the climate normal period (1986– 2000) was greater than in the first period (1971–1985), whereas less precipitation fell in late spring and early fall of the second period than in the first. In particular, June and July display moderate confidence for increased precipitation, whereas early April to the end of May, and from August to the end of September, display moderate confidence for decreased precipitation. The majority of stations in cluster 2 are distributed in the western portion of northwest China (Tarim and Junggar Basins), including Xinjiang Autonomy and the northwest area of Qinghai Province (Fig. 13). This

area is the driest region in China, with all stations exhibiting an arid or semi-arid climate. More precipitation occurred in the second half of the climate normal period than the first throughout most of year. There is strong confidence that increases in precipitation have occurred from the end of April to the end of August. Winter generally displays moderate confidence in increasing precipitation. This result is consistent with the results of Song and Zhang (2003). Cluster 3 stations extend from the northern portion of the middle and lower reaches of the Yangtze River to the southern portion of the middle and lower

468


reaches of the Yellow River (Fig. 14). The area includes Shanghai, Jiangsu Province, Hubei Province, Henan Province, and the east of Sichuan Province. The flooding season in these stations occurs in the middle of June to the middle of July due to the mei-yu Front. In this cluster, there are 17 periods showing evidence of decreases in precipitation, mainly from early September to the end of October, and 12 periods showing evidence of increases, mainly in early to mid January and from the beginning of February to the beginning of March. The stations of cluster 4 are mainly located in the eastern portion of northwestern China in the Turfan Depression, including Gansu Province, Shaanxi Province, Ningxia Autonomy, western inner Mongolia Autonomy, the east of Qinghai Province, and the north of Sichuan Province (Fig. 15). Most stations in the cluster have a semi-arid climate. Annual precipitation is between 200-400 mm and is concentrated in summer. In this cluster, the number of periods with evidence for a decreasing trend in precipitation (in January to the first part of February, the middle of April to the beginning of May, the middle of June to early July, and the end of August to early October) is greater than those showing evidence for an increasing trend (the middle of February to the first part of April). Cluster 5 stations are located in the western portion of southwest China, consisting of the western Sichun Province, western Yunnan Province, and eastern Tibet (Tibetan-Qinghai Plateau) (Fig. 16). Annual precipitation is mainly concentrated in summer and autumn, with the climatology of the region being famous for fall rains. More precipitation fell in the second half of the climate normal period than the first half during most of the summer and fall seasons, with moderate to strong confidence for an increasing trend in the early summer and from late summer to early fall. Cluster 6 stations are mainly located in northern China in the Yellow (Huanghe) River Delta, with the majority of stations in Shandong Province, Hebei Province, eastern Inner Mongolia Autonomy, and eastern Shanxi Province (Fig. 17). Two separate bands of stations are also found on Liaodong Peninsula and Jiuquan City in Gansu Province. Flooding season in these stations is from the end of June to the middle of August. Decreases in precipitation have occurred during most of the year. The index of statistical change indicates that there is moderate to significant confidence in a decreasing trend from the second period of January to the first period of March, from the second period of April to the end of April, in the second pe-

VOL. 24

riod of September, and from the middle of October to the end of October. This pattern is consistent with frequent and severe spring and fall droughts that have occurred in northern China since the 1990s, which may be attributable to the evolution of the monsoon system in East Asia (Qian and Zhu, 2001). Cluster 7 is found in the coastal areas of northeast of China between the Zhujiang and Yangtze Rivers, including Jilin, Liaoning, and Heilongjiang Provinces, which are located in the highest latitudes in China (Fig. 18). Precipitation in this region is mainly concentrated in summer. Nineteen of the 33 11-day periods show small increases in precipitation over the climate normal period, with most periods in spring exhibiting small decreases. There is low to moderate confidence in increased precipitation during the flooding season from the end of June to the middle of August, and moderate to significant confidence in decreased precipitation from the end of January to the middle of February, from the end of March to the beginning of April, and at the end of April. Cluster 8 is located in the southeast of China, including Zhejiang, Fujian, Hunan, Jiangxi, and eastern Guangdong Provinces (Fig. 19). These stations are characterized by heavy precipitation, with large rains in May and June bringing what is referred to as the “floods ahead of flooding season”. More rainfall was observed in June and less rainfall in fall of the second half of the climate normal period relative to the first, with moderate to significant confidence in precipitation increases in January, the end of March, and in June, and moderate to significant confidence in decreases in precipitation in some 11-day periods in winter, spring, and fall. 4.

Discussion

Within different regions of China, complex patterns of temperature and precipitation differences between the first and second halves of the 1971–2000 climate normal period were observed. The goal of the cluster analysis is to achieve a purely statistical, objective clustering of the seasonality of changes in temperature and precipitation at the input stations. This may not always lead to results that are easily interpreted from a geographical standpoint. While the spatial cohesiveness of clusters facilitates discussions on regional trends, it is the statistical cohesiveness of clusters that represents groupings of stations with similar patterns of change. The main purpose of this work was the identification of regions exhibiting similar shifts in the seasonality of temperature and precipitation in China. The resulting temperature and precipitation clusters may provide a useful baseline

NO. 3

SONG ET AL.

Fig. 13. As in Fig. 12, but for precipitation cluster 2 at the eight-cluster level.



469

470


for future examination of climate changes in China and a definition of regions displaying similar patterns of inter-decadal change. The regions could also form the basis for examination of additional properties of the Chinese climate, such as the spatial distribution of longer-term trends. Many authors have examined long-term trends in time series of temperature observations, usually for annual, seasonal, or monthly datasets, whereas we have examined the spatial distribution of variations at a finer, 11-day, timescale for the most recent climate normal period. Yet, the results for clusters identified in this study share similarities with the results of studies presented by other authors in studies of longer-term trends. For example, Qian and Zhu (2001) found that the temperature of China as a whole has been rising continuously in the last two decades of last century, with the largest signal found in northeast China. These patterns are apparent in Cluster 8, which is located in a similar region; however, by examining changes at the 11-day timescale, we are able to provide a more detailed analysis of intra-seasonal changes. For example, we can show that significant warming in the year occurs from the middle of January to the beginning of April and in early summer. Chen et al. (2004) found that there is a cooling center in China located in Sichuan, Yunnan, and Guizhou Provinces, which is similar to what was found for cluster 2. In terms of the effects of the urban heat island, Zhou et al. (2004) showed that rapid urbanization in southeastern China in the past 25 years is responsible for a much lager warming rate than in other regions. We found that the three main economic zones and urban centers in China, the Yangtze River Delta, Zhujiang River Delta, and the area around the Bohai Gulf, all belong to Cluster 3, which shows the largest and most significant warming signal of all the temperature clusters. Our observation of much greater warming in winter, which is consistent with the results of Zhai et al. (1999), may be explained by the strengthening of the Arctic Oscillation, which has resulted in a weakening of the Siberian high and less strong cold air invading China (Gong and Wang, 2003). For 20th century precipitation, Chen et al. (2004) observed that the precipitation decreased in northern China and increased in northwest China in recent 20 years. Cluster 6 reveals more detailed patterns within the year in north China, while clusters 2 and 4 reveal different patterns between the western and the eastern regions of northwest China respectively. The primary reason for these shift patterns may be linked with the evolution of the East Asian monsoon system in the last 30 years of the 20th century (Qian and Zhu,

VOL. 24

2001), as the eastern monsoon region occupies almost half of China’s area. We also found that most of western China has experienced more precipitation during the second half of the 1971–2000 climate normal period than in the first half. 5.

Conclusions

Climate change, and its potential impacts on human activities, has been the subject of considerable discussion within the academic and policy communities, as well as among the general public. Implications of climate change for agriculture and food security are concerns worldwide, and they are very important for China. China’s population of more than one billion people depends directly upon agriculture for subsistence. The prospect of global climate changes makes the issue particularly urgent Smit and Cai (1996). Freshwater resources are also greatly affected by climate change. Under the monsoon climatic regime prevailing over much of China, water availability is highly variable over time, displaying seasonal or year to year variations. China’s water resources are seriously threatened and are clearly vulnerable to climate change (Ye, 1992). Variations in the seasonality of temperature and precipitation across China were identified between the first and second halves of the 1971–2000 climate normal period. We found eight clusters for both temperature and precipitation. These clusters are not distributed randomly in space; instead, different parts of China exhibit different, relatively coherent patterns. The result was the identification of regions with stations exhibiting similar patterns of change from 1971– 1985 to 1986–2000. In this study, our purpose was to identify regions displaying similar patterns of change from the first to the second halves of the record, but not to associate causal factors with the observed shifts. In the last 15 years, temperatures in most of China have become warmer, especially in winter. Cluster 3, including central northern China, peninsulas and islands, and the southwest of Yunnan Province displays warming throughout the year; the largest magnitude of warming was found in northeast China in cluster 8. Cluster 6 shows that while warming has been observed in the southern region of the Yangtze River, there is little statistical confidence that this warming is significant. We also found evidence for a region with cooling trends located in the east of southwest of China (cluster 2). Changes in precipitation are more complicated than changes in temperature. The western portion of northwestern China (cluster 2) and most of southwestern China (clusters 1 and 5) experienced more precipi-

NO. 3

SONG ET AL.




471

472


VOL. 24


tation during 1986–2000 than in 1971–1985, while less precipitation fell in the Yellow River valley (Cluster 4 and Cluster 6). Changes in the summer monsoon period were also evident, with decreases in precipitation during the onset and withdrawal periods, and increases during the wettest period. These changes may be linked with the evolution of the East Asian monsoon system since the 1980s. There is a need for an examination of the long-term changes within these clusters to determine whether the signals have persisted over longer periods of time or whether they are a more recent development. Temperature and precipitation clusters defined in this work provide a useful framework for such analyses of longterm trends. The questions remain why recent warming has been greater than in the past, why the climate in northern China is getting drier, and how these patterns have affected and will affect water resource supplies and ecosystems. More detailed studies on the causes and effects of the observed changes in seasonal patterns of temperature and precipitation are needed. Acknowledgements. We are grateful to the China Meteorological Administration (CMA) and the Meteorological Service of Canada (MSC) who sponsored the opportunity for this collaboration. The Climate Data Office of the CMA provided the climatic observations reported here. This work was supported by the National Natural Science Foundation of China (Grant No. 40475031). REFERENCES Calinski, R. B., and J. Harabsz, 1974: A dendrite method for cluster analysis. Commmunication in Statistics Stat., 3, 1–27. Chen Longxun, Zhou Xiuji, and Li Weiliang, 2004: Char-

acteristics of the climate change and its formation mechanism in China in last 80 years. Acta Meteorologica Sinica, 62(5), 634–646. Chen Xiongwen, Zhang Xinshi, and Li Bailian, 2003: The possible response of life zones in China under global climate change. Global and Planetary Change, 38, 327–337. Chu Ziying, and Ren Guoyu, 2005: Effect of enhanced urban heat island magnitude on average surface air temperature series in Beijing region. Acta Meteorologica Sinica, 63(4), 534–540. Davies, D. L., and D. W. Bouldin, 1979: A cluster separation measure. IEEE Trans. Pattern Anal. Mach. Intell., 1, 224–227. Gong, X., and M. B. Richman, 1995: On the application of cluster analysis to growing season precipitation data in North America east of the Rockies. J. Climate, 8, 897–931. Gong Daoyi, and Wang Shaowu, 2003: Influnce of Arctic Oscillation on Winter Climate over China. Acta Geographica Sinica, 58(4), 559–568. (in Chinese) Guo, S., J. Wang, L. Xiong, A. Ying, and D. Li, 2002: A macro-scale and semi-distributed monthly water balance model to predict climate change impacts in China. J. Hydrol., 268, 1–15. IPCC, 2001: Climate Change 2001: The Scientific Basis, Summary for Policymakers and Technical Summary of Working Group 1 Report. Houghton et al., Eds., Cambridge University Press, 94pp. Jones, P. D., and A. Moberg, 2003: Hemispheric and large-scale surface air temperature variations: An extensive revision and an update to 2001. J. Climate, 16, 206–223. Karl, T. R., R. W. Knight, and N. Plummer, 1995: Trends in the hight-frequency climate variability in the twentieth century. Nature, 377, 217–220. Leith, R. M., and P. H. Whitfield, 1998: Evidence of climate change effects on the hydrology of streams in south-central British Columbia. Canadian Water Resources Journal, 23, 219–230.

NO. 3

SONG ET AL.

Likas, A., N. Vlassis, and J. J. Verbeek, 2003: The global k-means clustering algorithm. Pattern Recognition., 36, 451–461. Liu, J., N. Hayakawa, M. Lu, S. Dong, and J. Yuan, 2003: Hydrological and geocryological response of winter streamflow to climate warming in Northeast China. Cold Regions Science and Technology, 37, 15–24. Milligan, G. W., and M. C. Cooper, 1985: An examination of procedures for determining the number of clusters in a data set. Psychometrika, 50(2), 159–179. Niu Tao, Chen Longxun, and Zhou Zijiang, 2004: The characteristics of climate change over the Tibetan Plateau in the last 40 years and detection of climatic jumps. Adv. Atmos. Sci., 21(2), 193–203. Qian, W., and Y. Zhu, 2001: Climate Change In China From 1880 to 1998 and Its Impact On The Environmental Condition. Climatic Change, 50, 419–444. Scott, A. J., and M. J. Symons, 1971: Clustering methods based on likelihood ratio criteria. Biometrics, 27, 387–397. Shen, D, and O. Varis, 2001: Climate change in China. Ambio, 30(6), 381–383. Smit, B., and Y. Cai. 1996: Climate change and agriculture in China. Global Environ. Change Part A, 6, 205–214. Song Lianchun, and Zhang Chunjie, 2003: Changing features of precipitation over Northwest China during the 20th century. Journal of Glaciology and Geocryology, 25(2), 143–148. (in Chinese) Tang Guoli, and Ren Guoyu, 2005: Reanalysis of surface air temperature change of the past 100 years over China. Climatic and Environmental Research, 10(4), 791–798. Tao, F., M. Yokozawa, Y. Hayashi, and E. Lin, 2003. Future climate change, the agricultural water cycle, and agricultural production in China. Agriculture Ecosystems and Environment, 95, 203–215. Wang Zunya, Ding Yihui, and He Jinhai, 2004: An updating analysis of the climate change in China in recent 50 years. Acta Meteorologica Sinica, 62(2), 228–235. (in Chinese) Weingessel, A., E. Dimitriadou, and S. Dolnicar, 2002:

473

An examination of indexes for determining the number of clusters in binary data sets. Psychometrika, 67, 137–160. Whitfield, P. H., and A. J. Cannon, 2000: Recent variations in climate and hydrology in Canada. Canadian Water Resources Journal, 25, 19–65. Whitfield, P. H., and E. Taylor, 1998: Apparent recent changes in hydrology and climate of coastal British Columbia. Mountains to Sea: Human Interaction with the Hydrologic Cycle, Y. Alila, Ed., Proceedings of the 51st Canadian Water Resources Association Conference, 22–29. Whitfield, P. H., K. Bodtker, and A. J. Cannon, 2002: Recent Variations in Seasonality of Temperature and Precipitation in Canada 1976–95. International Journal of Climatology, 22, 1617–1644. Whitfield, P. H., A. W. Hall, and A. J. Cannon, 2004: Changes in the Seasonal Cycle in the Circumpolar Artic, 1976–95: Temperature and Precipitation. Arctic, 57(1), 80–93. Xu, D., and H. Yan, 2001: A study of the impacts of climate change on the geographic distribution of Pinus koraiensis in China. Environment International, 27, 201–205. Ye Duzheng, 1992: The Pilot Study of Global Change in China. China Meteorological Press, Beijing, 185–192. You, S. C., 2001: Agricultural adaptation of climate change in China. Chinese Journal of Environmental Science, 13(2), 192–197. (in Chinese) Zhai, P. M., and X. H. Pan, 2003: Trends in temperature extremes during 1951–1999 in China. Geophys. Res. Lett., 30(17), 1913. Zhai, P. M., A. J. Sun, X. L. Ren, B. Gao, and Q. Zhang, 1999: Changes of Climate Extremes in China. Climatic Change, 42, 203–218. Zhang Jingyong, Dong Wenjie, and Wu Lingyun, 2005: Impact of land use changes on surface warming in China. Adv. Atmos. Sci., 22(3), 343–348. Zhou, L. M., E. D. Robert, and Y. H. Tian, 2004: Evidence for a significant urbanization effect on climate in China. Proc. National Academy Science, 101(26), 9540–9544.