Abstract

Chinese radiosonde data from 1970 to 1990 are relatively homogeneous in time and are used to examine the climatology, trends, and variability of China’s atmospheric water vapor content. The climatological distribution of precipitable water (PW) is primarily dependent on surface temperature. Influenced by the east Asia monsoon, China’s precipitable water exhibits very large seasonal variations. Station elevation is also a dominant factor affecting water vapor distribution in China.

An increase (decrease) in precipitable water over China is associated with an increase (decrease) of precipitation in most regions. Increases in the percentage of PW relative to climatology are greater in winter and spring than in summer and autumn.

Interannual variation and trends in precipitable water and surface temperature are closely correlated in China, confirming a positive “greenhouse” feedback. Interannual variations between precipitable water and precipitation are also significantly correlated.

1. Introduction

Atmospheric water vapor plays a part in all water and heat processes of the climate system. It is the most abundant greenhouse gas and makes the largest contribution to the “natural” greenhouse effect (IPCC 1996). It also plays an important part in the hydrologic cycle, but scientists were unable to study its spatial and temporal distribution until radiosonde data became available (Bannon and Steele 1960; Starr et al. 1965). Recently, the Global Energy and Water Cycle Experiment has emphasized the need to understand its role in global warming (Starr and Melfi 1990). IPCC (1996) presented a summary of the trends of water vapor in various regions over the last two decades and described the difficulties in monitoring long-term changes because of inhomogeneities in radiosonde data. Numerous efforts have been made to study water vapor distribution in different regions (see, e.g., Angell et al. 1984; Hense et al. 1988; Gaffen et al. 1992; Elliott et al. 1995).

The National Climatic Data Center has produced a worldwide upper-air dataset, called the Comprehensive Aerological Reference Data Set (Eskridge et al. 1995), that contains data from 1946 to 1994 and station histories. The difficulty in creating a reliable, long-term global water vapor time series is an issue of concern. China has a large national dataset of radiosonde observations taken since the late 1940s and adequate station histories, and these data have not been studied extensively.

Elliott and Gaffen (1991) found that the U.S. radiosonde humidity archives are not homogeneous because of changes in recording precision, instruments, and reporting practices. Chinese data also include temporal inhomogeneities, but they are relatively homogeneous after 1970 (Zhai and Eskridge 1996). Moreover, Chinese upper-air humidity data are free from low-humidity censoring, so that, with these data, the climatology and distribution of atmospheric water vapor content can be studied over China with some confidence.

China’s atmospheric water vapor research dates back to the 1950s (Xu 1958). Xu used data from 33 stations in eastern China for January and July 1957. Lu and Gao (1984) used 1960–70 monthly mean radiosonde summaries to calculate the climatological distribution of water vapor over China. Recent research has shown, however, that Chinese radiosonde temperatures and humidities include systematic errors due mainly to a change in instruments in the 1960s (Zhai and Eskridge 1996). Also, Elliott and Gaffen (1991) have shown that using monthly averaged temperature and dewpoint depression, as done by Lu and Gao (1984), can lead to an underestimation of precipitable water (PW).

In this paper, twice-daily observations are used to calculate the monthly, seasonal, and annual mean precipitable water. From these data, the distribution, interannual variation, and trends of atmospheric water vapor are analyzed. Possible connections between water vapor and precipitation, and between water vapor and surface temperature during 1970–90, are examined.

To place the climatology and trend of atmospheric water vapor over China in perspective, the results are compared with those from other regions, especially the United States, which, like China, occupies a large landmass in Northern Hemisphere midlatitudes.

2. Data sources and methods

Radiosonde and surface data were provided by China’s National Meteorological Center. The radiosonde data after 1980 come from the Chinese national telecommunication system, and prior to that date they were digitized on punched cards or via computer keyboard depending on the source. China has used the American Diamond Hinman, Finnish Vaisala RS 12, Russian RZ-049, and Chinese-designed GZZ model radiosondes (Gaffen 1993), with the observation time changing from 0300 and 1500 UTC to 0000 and 1200 UTC in April 1957 (Wang and Wang 1987). Temperatures have been corrected for radiation-induced errors since 1959. In this study, data from 63 radiosonde stations are used, all of which have been quality controlled (Alduchov and Eskridge 1996b), together with a monthly surface dataset from 378 Chinese stations. Inhomogeneity tests have been made for both radiosonde and surface data, but no effort was made to correct systematic errors or adjust the biases.

Because of the inhomogeneities found in Chinese radiosonde data in both temperature and dewpoint depression prior to 1970, only the data from 1970 to 1990 were selected. To ensure the reliability of the results, these data are further tested for inhomogeneities by the following procedure. Those stations that include obvious inhomogeneities are excluded. First, the method used by Zhai and Eskridge (1996) was applied to 0000 UTC and 1200 UTC annual mean temperature and dewpoint depression time series at 850, 700, and 500 hPa. Second, the Easterling–Peterson technique (Easterling and Peterson 1995) was applied to annual mean PW time series relative to a reference series of PW generated from surrounding stations. Third, annual mean surface pressure is checked before and after a station move. At least 5 yr of data before and after a potential move were used. The pressure difference is tested using Student’s t-test at the 99% significance level (results show that using the 99% instead of the 95% level more accurately detects a station move signal).

When inhomogeneities were found in the data, the station was rejected. The inhomogeneity tests show there are no universal inconsistencies in China’s upper-air annual mean times series. However, inhomogeneities exist in the data from some stations, mainly related to station moves.

Gaffen et al. (1991) obtained mean monthly value by using at least three observations a month that fall within the 10% confidence bands. In this paper, at least three observations at an observation time (0000 UTC and 1200 UTC, respectively) were needed to generate a monthly mean. That means that at least six observations were used to obtain a monthly mean. Otherwise, the monthly mean was considered missing. Stations with more than 3 yr of data missing at the same level were rejected. Upper levels were frequently missing some data. The missing data were estimated by linearly interpolation using anomalies from adjacent months and levels.

The original 63 evenly distributed radiosonde stations in China (except for southwest China) were reduced to a final 44 stations, mainly because of inhomogeneities and missing data (Fig. 1).

Fig. 1.

Selected and rejected stations and areas in China: 1—northeast China, 2—north China, 3—northwest China, 4—Tibet Plateau, 5—southwest China, Sichnan Basin in the north part of this region, 6—Yangtze River Valley, 7—south China.

Fig. 1.

Selected and rejected stations and areas in China: 1—northeast China, 2—north China, 3—northwest China, 4—Tibet Plateau, 5—southwest China, Sichnan Basin in the north part of this region, 6—Yangtze River Valley, 7—south China.

For monthly surface data, stations that moved vertically more than 50 m or horizontally more than 20 km were excluded. Song (1995) and Liu and Sun (1995) have performed homogeneity tests of China’s temperature and precipitation data. Stations that they identified as containing inhomogeneities were also excluded.

Dewpoint depression D was calculated for the digitized Chinese radiosonde data from relative humidity U and temperature T by (1). Equation (1) is from Gaffen (1993) and is valid for air temperature T greater than −60°C and pressure greater than 10 hPa:

 
formula

and

 
formula

Equations (2) and (3) (Alduchov and Eskridge 1996a) were used to calculate the saturation vapor pressure es, specific humidity q, T, U, and atmospheric pressure p:

 
formula

where

 
formula

and

 
formula

where

 
formula

The PW between the pressure levels pi and pi+1 can be calculated as

 
formula

Chinese radiosonde humidity data are cut off (censored) when the temperature is below −60°C or the pressure is below 10 hPa. Since the major portion of precipitable water is in the lower troposphere, the PW in the layer from the surface to 200 hPa (levels are the surface and 850, 700, 500, 400, 300, and 200 hPa) is considered to be the total amount of water vapor. To examine its profile, water vapor in the layers from the surface to 700 hPa, 700 to 400 hPa, and 400 to 200 hPa was calculated. In plateau areas, the surface pressure is below 700 hPa and the surface–400-hPa layer was used.

To calculate trends, the following linear regression method was used, where Y is the time series of the variable studied (precipitable water, temperature, or precipitation) and t is time:

 
formula

where a is

 
formula

and b represents the rate of increase or decrease:

 
formula

The summation is from i, equal to 1, to n (n is total number of years in study).

For precipitation and precipitable water, the percentage trend Bp is estimated by

 
formula

where b is from (5) and y is the yearly average.

Statistic significant parameter t = [(n − 2)/(1 − r2)]1/2 (Sx/Sy was used to perform Student’s t-test to measure the robustness of the trend estimates. Here, n − 2 is the degrees of freedom; Sx and Sy are the standard deviations of the meteorological variable and time, respectively; and r is the calculated regression coefficient.

The geographic regions that are referred to in this paper are shown in Fig. 1.

3. Water vapor climatology

The distribution of atmospheric water vapor over China is dominated by latitude, topographical features, and the monsoon.

The graphic software Golden-Surfer was used to produce Figs. 2a, 2b, and 4. Kirging interpolation was used to convert station data to grid data to draw contours.

Fig. 2.

(a) Monthly precipitable water distribution along 113°E, in eastern China. (Data are generated from six stations. These stations are Xisha Island, 16.8°N, 112.3°E; Guangzhou, 23.1°N, 113.3°E; Changsha, 28.2°N, 113.1°E; Zhengzhuo, 34.7°N, 113.7°E; Taiyuan, 37.8°N, 106.2°E; and Erenhot, 43.7°N, 112.0°E.) (b) Same as (a) except along 80°W in the eastern United States. Data are generated from seven stations along approximately 81°W from Reitan (1960). The stations used are Havana, Cuba, 23.2°N, 82.4°W; Miami, Florida, 25.8°N, 80.3°W; Tampa, Florida, 28.0°N, 82.5°W; Charleston, South Carolina, 32.9°N, 80.0°W; Greensboro, North Carolina, 36.1°N, 80.0°W; Pittsburgh, Pennsylvania, 40.5°N, 80.2°W; and Buffalo, New York, 40.9°N, 78.7°W.

Fig. 2.

(a) Monthly precipitable water distribution along 113°E, in eastern China. (Data are generated from six stations. These stations are Xisha Island, 16.8°N, 112.3°E; Guangzhou, 23.1°N, 113.3°E; Changsha, 28.2°N, 113.1°E; Zhengzhuo, 34.7°N, 113.7°E; Taiyuan, 37.8°N, 106.2°E; and Erenhot, 43.7°N, 112.0°E.) (b) Same as (a) except along 80°W in the eastern United States. Data are generated from seven stations along approximately 81°W from Reitan (1960). The stations used are Havana, Cuba, 23.2°N, 82.4°W; Miami, Florida, 25.8°N, 80.3°W; Tampa, Florida, 28.0°N, 82.5°W; Charleston, South Carolina, 32.9°N, 80.0°W; Greensboro, North Carolina, 36.1°N, 80.0°W; Pittsburgh, Pennsylvania, 40.5°N, 80.2°W; and Buffalo, New York, 40.9°N, 78.7°W.

Fig. 4.

Annual mean precipitable water in the surface–200-hPa layer for all stations in China, showing the relationship to station altitude and latitude.

Fig. 4.

Annual mean precipitable water in the surface–200-hPa layer for all stations in China, showing the relationship to station altitude and latitude.

Figure 2a shows the seasonal change of PW along 113°E, and it is clear that the distribution is a strong function of latitude. The cross section shown in Fig. 2a is typical of eastern China. As the latitude increases, the water vapor amount decreases, with a large difference between southern and northern China. The largest value of atmospheric water vapor content in July at the northernmost regions is much lower than the lowest value in January in the south. The least amount of PW is observed in January and the greatest usually in July, except in southernmost China, where the largest values occur in June and August.

Precipitable water in spring is lower than in autumn in China (Fig. 2a, Table 1). In a study of large-scale atmospheric moisture processes, Peixoto et al. (1981) found that water vapor levels in spring and autumn are not identical, but tend to resemble those in their antecedent seasons. This is consistent with our findings for China.

Table 1.

Means, rates of increase b (°C decade−7), and rates of percentage increase Bp (% decade−1) for averaged PW (simple station-by-station arithmetic average) for all of China (mm decade−1). The underlined trends are statistically significant at a 90% confidence level.

Means, rates of increase b (°C decade−7), and rates of percentage increase Bp (% decade−1) for averaged PW (simple station-by-station arithmetic average) for all of China (mm decade−1). The underlined trends are statistically significant at a 90% confidence level.
Means, rates of increase b (°C decade−7), and rates of percentage increase Bp (% decade−1) for averaged PW (simple station-by-station arithmetic average) for all of China (mm decade−1). The underlined trends are statistically significant at a 90% confidence level.

Compared with the seasonal variation of PW over the eastern United States (Fig. 2b), eastern China’s is more striking. During wintertime, the atmosphere over north China is very dry. Figure 2a shows that PW over eastern China is lower than that over the eastern United States in the wintertime north of 30°N latitude. In eastern China, the 10-mm PW isopleth is found near 33°N, while in the eastern United States, the 10-mm isopleth is located near 38°N. During the summertime, China’s water vapor is influenced by the east Asian summer monsoon. South of 30°N, PW over eastern China is much greater than that over the eastern United States. These results are consistent with those of Peixoto and Oort (1992), who showed that there is a marked annual variation of specific humidity in the monsoon regions.

The January and July distribution of PW were selected to represent winter and summer for China. The zonal distribution of PW over eastern China (Figs. 3a and 3b) mainly reflects the influence of temperature, but in western China reflects the complicated topography.

Fig. 3.

(a) Mean precipitable water, in millimeters, from the surface to 200 hPa over China for Januaries from 1970 to 1990. (b) Same as (a) except for July.

Fig. 3.

(a) Mean precipitable water, in millimeters, from the surface to 200 hPa over China for Januaries from 1970 to 1990. (b) Same as (a) except for July.

Figure 3a shows the distribution of PW over China in January. Influenced by prevailing cold and dry air, PW over northeast China is very limited, ranging from 2.5 to 3.5 mm. Over north China, observed PW is about 3–10 mm, and over the Yangtze River valley and southern China, it is about 10–25 mm. The higher elevation of western China weakens the severe cold air flow from Mongolia, and 4–6 mm of PW are observed over western northwest China, which is greater than the PW observed in northeast China at the same latitude. However, to the south, less than 2.5 mm of PW is observed over the Tibet Plateau. The high, snow-covered surface over the Tibet Plateau results in a colder atmosphere, which can hold less water.

The greatest amount of PW is found over China in the summer, when China is under the influence of a moist summer monsoon. Figure 3b shows the distribution of PW over China in July. More PW is found over the southeast than the northwest. The isopleths of PW closely match China’s topography and reflect the influence of the monsoon. Figure 3b shows 55–60 mm of PW over south and east China near the coast and at locations with low elevations; 50 mm of PW is observed in the Sichuan Basin, east of the Tibet Plateau. Water vapor transported from the Bay of Bengal and adjacent areas, which are low lying, is the primary source of water vapor (Zou et al. 1990). The atmospheric column is much shorter over the Qinghai–Tibet Plateau area because of its high elevation, and only 10–20 mm of PW is observed. Approximately 20 mm of precipitable water is observed over the western part of northwestern China, where the elevation is lower than in Tibet. In the western part of northwest China, the influence of the summer monsoon is weak because of the natural barrier formed by the plateau and mountains to the south.

The calculated values of PW presented in this paper are greater than those of Lu and Gao (1984), although in warm and moist southeastern China, the difference is not large for annual mean PW. Over the west plateau regions, where PW is generally less than 10 mm, our values are about 20% higher. Gaffen et al. (1991) show that the bias δ for vapor pressure caused by using the mean dewpoint temperature Td can be expressed as

 
formula

where δ is always negative. The Clausius–Clapeyron equation is

 
formula

where L is the latent heat of vaporization, Rυ is the gas constant for vapor, and σtd is standard deviation of Td. Obviously, the magnitude of bias is inversely proportional to the square of Td. Elliott et al. (1991) pointed out that the error can be greater than 10%, and as high as 20%, at low temperatures. Errors in calculating vapor pressure result in errors of specific humidity and hence PW. Over the plateau regions, temperature is low and the greater bias in Lu and Gao’s results seems consistent with the above estimations.

Precipitable water in China tends to be dominated by surface elevation and latitude (see Fig. 4). As station elevation increases, PW decreases exponentially. Figure 5 shows the relationship between fractional PW and station elevation. In the surface–700-hPa layer, fractional PW is negatively related to station elevation. Stations in eastern China are generally at low elevations, and 70%–75% of the PW is in the surface–700-hPa layer. At stations above 3000 m, PW is zero in the surface–700-hPa layer simply due to the stations being above 700 hPa. In the 400–200-hPa layer, fractional PW grows linearly as station elevation increases. For stations in the plains (mainly in eastern China), only about 5% of a column’s water vapor is in the 400–200-hPa layer. Over the plateau area, the 400–200-hPa layer contains 10%–20% of the PW. The 700–400-hPa layer contains about 25% to 30% of the PW in the plains, while over the plateau, this layer contains 80% to 90% of the PW. When the station elevation is above 3000 m, the PW fraction in the 700–400-hPa layer decreases slightly as elevation increases. It should be noted that while seasonal variation of PW is very obvious in China, the fractional PW distribution is basically the same for summer and winter.

Fig. 5.

Fractional precipitable water for different layers for all stations in China.

Fig. 5.

Fractional precipitable water for different layers for all stations in China.

4. Variation and trend of PW during 1970–90 over China

The annual mean rate of change of PW of each of our stations throughout China is shown in Fig. 6a. Trends are characterized by increases of PW, with the greatest increases occurring in the northeast, southwest, parts of the northwest, and the southern coastal areas, where they are about 0.5 to 1.0 mm decade−1. Ten of the 44 Chinese upper-air stations, all in the above-mentioned regions, exhibited a statistically significant increase at the 95% confidence level. Over the northern part of north China and central south China, PW decreased slightly. Over some eastern areas, the rate of increase is relatively small.

Fig. 6.

(a) Distribution of the rate of change of annual mean precipitable water in millimeters over China from 1970 to 1990. (b) Same as (a) except for percentage rate.

Fig. 6.

(a) Distribution of the rate of change of annual mean precipitable water in millimeters over China from 1970 to 1990. (b) Same as (a) except for percentage rate.

Because of the great topographical and seasonal differences of PW distribution, trends have also been calculated with respect to climatology. Figure 6b shows the trend as a percentage of annual mean PW. Interestingly, the largest relative trends, 3%–6% decade−1, are in higher latitudes, part of the Tibetan Plateau, and southwest China. Decreasing trends of PW in north China and central southern areas were about 1%–3% decade−1.

Elliott et al. (1995) found increasing trends in PW from 1973 to 1993 over North America, except in northern and eastern Canada, where it decreased slightly. In North America, positive trends tended to increase with decreasing latitude, with a maximum increase of about 3 mm decade−1. Moisture was also found to be increasing at selected tropical stations (Gaffen et al. 1991; Gutzler 1992). The trends in China are similar, with PW also tending to increase with decreasing latitude. Moreover, the maximum rate of increase of PW or trend in China is smaller than in North America.

The largest rates of increase in PW in China occur in summer (Fig. 7a). The increase of PW was about 0.5–1.0 mm decade−1, mainly over northeast, south, and southwest China. In north and eastern northwest China, PW was decreasing by about 0.5 mm decade−1.

Fig. 7.

(a) Same as Fig. 6a except for the rate in summer. (b) Same as Fig. 6b except for winter.

Fig. 7.

(a) Same as Fig. 6a except for the rate in summer. (b) Same as Fig. 6b except for winter.

In China, the trend in PW, expressed as percentages, tends to be greater in winter and spring than in summer and autumn. In winter (Fig. 7b), PW over most of China exhibited increasing trends for the Tibetan Plateau, regions east of the plateau, and over northern China, increasing by 5%–10% decade−1.

Averaged time series of annual and seasonal mean PW for all Chinese stations are plotted in Fig. 8. In Fig. 8, the all-China values are based on an arithmetic average of the 44 individual stations. Although there is sparse spatial coverage in some regions such as northwest and southwest China, this should not be a serious problem, as the unsampled area is not large compared to the rest of China. An even distribution of stations would require two more stations in southwest China and one more in northwest China. Adding three more stations would provide less than a 10% contribution to the China-wide average. The means and trends in the surface–200-hPa, surface–700-hPa, and 700–400-hPa layers are given in Table 1. Standard deviations calculated from Fig. 8 show that the largest interannual variations occur in autumn and the smallest in winter (Figs. 8a–d). Soden and Lanzante (1996) have shown that radiosondes like the Chinese GZZ, which uses a goldbeater’s skin humidity sensor, are probably biased in the upper troposphere due to sensor time lag, which is a function of temperature. Time series at upper levels generated from sparse samples may also increase the uncertainty in evaluating PW variability. In Table 1, it is shown that the trend of PW is positive in all seasons in all layers. The largest trend was 2.6% decade−1 in winter, with the second largest trend being 1.5% decade−1 in spring and the smallest trend being 0.9% decade−1 in autumn. The annual mean PW trend is 1.2% decade−1. It should be noticed that the percentage increase trend of PW is greater in the 700–400-hPa layer than in the surface–700-hPa layer during the spring and winter. Gutzler (1992) used four tropical stations and found that the greatest increase was near the surface, but in China, the PW increase tended to be greater in the lowest layer only during the summer and autumn.

Fig. 8.

Seasonal and annual averaged (simple arithmetic average of all stations) time series of precipitable water in millimeters for (a) spring, (b) summer, (c) autumn, (d) winter, and (e) the year.

Fig. 8.

Seasonal and annual averaged (simple arithmetic average of all stations) time series of precipitable water in millimeters for (a) spring, (b) summer, (c) autumn, (d) winter, and (e) the year.

5. Water vapor and surface temperature

At most locations during 1970 to 1990, mean surface temperature in China either had no trend or was increasing. The largest increases were in the north, with rates of about 0.5°–1.0°C decade−1 (see Fig. 9). The large increase in surface temperature in northeast China coincided with an increase of PW. However, surface temperatures increased in north China, while there was a slight decrease in PW. In the southwest and the south, the relationship between mean annual surface temperature and PW is not clear. In coastal regions, water vapor increased greatly, but the temperature increase is smaller than that in northern China.

Fig. 9.

Distribution of the rate of change of annual mean surface temperature over China from 1970 to 1990.

Fig. 9.

Distribution of the rate of change of annual mean surface temperature over China from 1970 to 1990.

Figure 10 shows time series (normalized by the standard deviation) of averaged (arithmetic) annual PW, precipitation, and surface temperature, while Table 2 shows that China’s annual mean surface temperature is significantly correlated with PW (0.61, which is significant at the 95% level). The positive correlations of surface temperature with PW and PW with time suggest that positive trends exist for both, although the latter is not statistically significant. Table 3 shows that PW and surface temperatures are significantly correlated in northeast China throughout the year. In southwest China, there was a high correlation between PW and temperature in autumn and winter, although in summer and spring, the correlation was rather poor. Gaffen et al. (1992) found that the relationship between water vapor and surface temperature depends on the temperature range, which supports our findings. In fact, the increase in surface temperature is greatest in winter over China (Ding and Dai 1994), when there is the highest percentage increase of PW (see Table 1).

Fig. 10.

Interannual normalized anomalies (anomaly–standard deviation) of China-wide averaged precipitable water, surface temperature, and precipitation during 1970–90. PW is calculated as in Fig. 8. Temperature and precipitation values are produced by area weighting from provincial averages, which are generated by simple arithmetic averages from each station.

Fig. 10.

Interannual normalized anomalies (anomaly–standard deviation) of China-wide averaged precipitable water, surface temperature, and precipitation during 1970–90. PW is calculated as in Fig. 8. Temperature and precipitation values are produced by area weighting from provincial averages, which are generated by simple arithmetic averages from each station.

Table 2.

Correlation between averaged, China-wide, annual mean precipitable water, surface temperature, precipitation, and time (trend). Time series used in the calculations are shown in Fig. 10. Underlined values are significant at the 95% confidence level.

Correlation between averaged, China-wide, annual mean precipitable water, surface temperature, precipitation, and time (trend). Time series used in the calculations are shown in Fig. 10. Underlined values are significant at the 95% confidence level.
Correlation between averaged, China-wide, annual mean precipitable water, surface temperature, precipitation, and time (trend). Time series used in the calculations are shown in Fig. 10. Underlined values are significant at the 95% confidence level.
Table 3.

Correlation coefficients between precipitable water and mean surface temperature by season. All of the underlined values are significant at the 95% confidence level. The areal values are simply arithmetic averages for data from all radiosonde and colocated surface stations.

Correlation coefficients between precipitable water and mean surface temperature by season. All of the underlined values are significant at the 95% confidence level. The areal values are simply arithmetic averages for data from all radiosonde and colocated surface stations.
Correlation coefficients between precipitable water and mean surface temperature by season. All of the underlined values are significant at the 95% confidence level. The areal values are simply arithmetic averages for data from all radiosonde and colocated surface stations.

The relationship between atmospheric water vapor and surface temperature is complex. Although it was found that the increase of PW over the northeast matches the increase of surface temperature, a slight decrease in PW over north China corresponds to a surface temperature increase. Areas with increasing PW lie within areas of increasing temperature. Long-term changes in surface temperature and PW are interrelated through water vapor in the “natural” greenhouse effect. However, increasing temperature can also be the result of other factors and the regional trends in water vapor might be influenced by other hydrologic or dynamic factors. Relationships between water vapor and precipitation are discussed in the next section.

6. Water vapor and precipitation

To study the relationship between PW and precipitation, averaged, normalized time series of the deviations of PW and precipitation were developed and are shown in Fig. 10. The correlation between these two time series is 0.64, which is significant at the 95% level. Because the trend in precipitation for all of China is very small (Table 2), the high correlation between PW and precipitation is largely the result of interannual variations. Anomalies for summer mean PW and precipitation during 1980 and 1985 are next examined in eastern China, which possesses a dense network of rainfall measuring stations and is typically under the influence of summer monsoon.

In the summer [June–August (JJA)] of 1980, severe flooding occurred in the Yangtze River valley, while there was a drought in northern China. Precipitation totals of 800–1000 mm were observed (about 200 mm more than average) over some areas in the Yangtze River valley. Over northern China, there was only 80–200 mm of precipitation, about 100–180 mm less than normal (see Fig. 11). Meanwhile, northeast China received rainfall totals of 500–700 mm, exceeding normal by 100–300 mm. The summer (JJA) of 1985 was a drought period in the middle and lower reaches of the Yangtze River, with 120–400 mm precipitation observed, a 100–250-mm deficit (see Fig. 12).

Fig. 11.

(a) Departure from normal precipitation in millimeters, and (b) that for precipitable water from the surface to 200 hPa, also in millimeters over eastern China for summer (JJA) 1980.

Fig. 11.

(a) Departure from normal precipitation in millimeters, and (b) that for precipitable water from the surface to 200 hPa, also in millimeters over eastern China for summer (JJA) 1980.

Fig. 12.

Same as Fig. 11 except for summer 1985.

Fig. 12.

Same as Fig. 11 except for summer 1985.

Figures 11b and 12b can be compared to show the difference in PW between the summers of 1980 and 1985. There was an area of positive departure of about 2 mm of PW over the middle and lower reaches of the Yangtze River in the summer of 1980, while most of northern China had negative departures of about 1–2 mm. In the summer of 1985, PW over the middle and lower reaches of the Yangtze River valley was 1 mm below average. Over much of northeast China, there were 2–3 mm more than usual of PW. The positive and negative departure centers of PW correspond closely to the positive and negative departure centers of total precipitation.

Figure 13 shows the distribution of the precipitation trends. Over the northeast, much of the northwest, southeastern Tibet, the Sichuan Basin, most of the Yangtze River valley, and the southeastern coast of China, there were increasing trends of precipitation of about 5%–10% decade−1. On the other hand, negative trends were present over much of north China, the northeastern part of northwest China, and much of southern China.

Fig. 13.

Same as Fig. 9 except for percentage of precipitation to normal.

Fig. 13.

Same as Fig. 9 except for percentage of precipitation to normal.

Although it is difficult to explain the cause and effect between PW and precipitation, the correlation between them is significant. Higher (lower) PW in China is usually related to more (less) precipitation. It should be noted that precipitation changes may influence PW through surface evaporation, an effect that might be important in dry areas such as north China. However, the observed changes in PW and precipitation may be the result of the same atmospheric dynamics.

7. Conclusions

It was found in this study that the distribution of PW depends mainly on surface temperature. Affected by the east Asian monsoon, PW over China exhibits distinct seasonal variations. Station elevations also strongly influence the spatial distribution of PW, temperature, and precipitation over China. Over 70% of PW in the plains area is in the lower troposphere below 700 hPa, while most PW is in the upper troposphere over the plateau regions.

Precipitable water increased over China during 1970–90, except over northern north China and central south China. Over the northeast and parts of the northwest, southwest, and southern South China, annual mean PW increases were statistically significant at the 95% confidence level. Although absolute increases of PW tended to be the greatest in summer, the increase in percentage relative to climatology was the highest in winter. During winter and spring, the percentage increases in higher layers were greater than those in the lowest layer.

There is a significant correlation between PW and precipitation, and between PW and surface temperature over China. Over northern China, the percentage increase of PW corresponds to a surface temperature increase. The decreases of PW over north China and central southern China seem to be related to a trend of decreasing precipitation.

The relationships between PW and surface temperature, and between precipitable water and precipitation, are complicated. Although the climatological distribution of PW is primarily determined by temperature, a positive feedback should exist between temperature and water vapor in the long term.

Acknowledgments

The CARDS program is supported by the U.S. Department of Energy under Contract DE-AI05-90ER61011, the Climate and Global Change Program of NOAA, and the National Climatic Data Center. We thank Dr. Trevor Wallis for reviewing this paper and two anonymous reviewers for their constructive comments.

REFERENCES

REFERENCES
Alduchov, A. O., and R. E. Eskridge, 1996a: Improved Magnus form approximation of saturation vapor pressure. J. Appl. Meteor., 35, 601–609.
——, and ——, 1996b: Complex quality control of upper-air variables (geopotential heights, temperature, wind, and humidity) at mandatory and significant levels for the CARDS dataset. NCDC Rep., 135 pp. [NTIS PB97–132286.]
.
Angell, J. K., W. P. Elliott, and M. E. Smith, 1984: Tropospheric humidity variations at Brownsville, Texas, and Great Falls, Montana, 1958–80. J. Climate Appl. Meteor., 23, 1286–1296.
Bannon, J. K., and L. P. Steel, 1960: Averaged water vapor contents of the air. Geophysical Memo. 102, U.K. Meteorological Office M.O. 631b, 38 pp.
Ding, Y., and X. Dai, 1994: Temperature change in China during the recent century. Meteor. Mon., 12, 19–26.
Easterling, D. R., and T. C. Peterson, 1995: A new method for detecting and adjusting for undocumented discontinuities in climatological time series. Int. J. Climatol., 15, 369–377.
Elliott, W. P., and D. J. Gaffen, 1991: On the utility of radiosonde humidity archives for climate study. Bull. Amer. Meteor. Soc., 72, 1507–1519.
——, M. E. Smith, and J. K. Angell, 1991: Monitoring tropospheric water vapor changes using radiosonde data. Greenhouse-Gas-Induced Climate Change: A Critical Appraisal of Simulations and Observations, M. E. Schessinger, Ed., Elsevier, 311–327.
——, R. J. Ross, and D. J. Gaffen, 1995: Water vapor trends over North America.Preprints, Sixth Symp. on Global Change Studies, Dallas, TX, Amer. Meteor. Soc., 185–186.
Eskridge, R. E., O. A. Alduchov, I. V. Chernykh, P. Zhai, S. R. Doty, and A. C. Polansky, 1995: A Comprehensive Aerological Reference Data Set (CARDS): Rough and systematic errors. Bull. Amer. Meteor. Soc., 76, 1759–1775.
Gaffen, D. J., 1993: History changes in radiosonde instruments and practices. WMO Instruments and Observing Method Rep. 50, WMO/TD 541.
——, T. P. Barnett, and W. P. Elliott, 1991: Space and time scales of global tropospheric moisture. J. Climate, 4, 889–1008.
——, W. P. Elliott, and A. Robock, 1992: Relationship between tropospheric water vapor and surface temperatures as observed by radiosondes. Geophys. Res. Lett., 19, 1839–1879.
Gutzler, D., 1992: Climatic variability of temperature and humidity over the tropical western Pacific. Geophys. Res. Lett., 19, 1595–1598.
Hense, A., P. Krahe, and H. Flohn, 1988: Recent fluctuations of tropospheric temperature and water vapor in the tropics. Meteor. Atmos. Phys., 38, 215–227.
IPCC, 1996: Climate Change 1995: The Science of Climate Change. J. T. Houghton, L. G. Meria Filho, B. A. Callander, N. Harris, A. Kattenberg, and K. Maskell, Eds., Cambridge University Press, 573 pp.
Liu, X., and A. Sun, 1995: An inhomogeneity test study of annual precipitation series. Meteor. Mon., 21, 3–6.
Lu, Y., and G. Gao, 1984: Atlas of Water Climatology of China. Meteorological Press, 188 pp.
Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate. American Institute of Physics, 520 pp.
——, D. A. Salstein, and R. D. Rosen, 1981: Intra-annual variation in large-scale moisture fields. J. Geophys. Res., 86, 1255–1264.
Reitan, C. H., 1960: Distribution of precipitable water vapor over the continental United States. Bull. Amer. Meteor. Soc., 41, 79–87.
Soden, B. J., and I. R. Lanzante, 1996: An assessment of satellite and radiosonde and climatologies of upper-tropospheric water vapor. J. Climate, 9, 1235–1250.
Song, C., 1995: A study of testing methods on inhomogeneity of temperature series. Quart. J. Appl. Meteor., 6, 289–296.
Starr, D. O’C., and S. H. Melfi, 1990: The role of water vapor in climate: A strategic research plan for the proposed GEWEX water vapor project (GVap). NASA Rep. NAS 1.55:3120, 50 pp.
Starr, V. P., J. P. Peixoto, and A. R. Crisi, 1965: Hemispheric water balance for IGY. Tellus, 17, 463–471.
Wang, S. T., and B. M. Wang, 1987: Methods for Summarization and Statistics of Meteorological Data. Meteorological Press, 327 pp.
Xu, S., 1958: China’s water vapor transfer and budget. Acta Meteor. Sin., 29, 33–43.
Zhai, P., and R. E. Eskridge, 1996: Analyses of inhomogeneities in radiosonde temperature and humidity time series. J. Climate, 9, 884–894.
Zou, J. S., J. Jiang, and M. H. Huang, 1990: Upper-Air Climatology. Meteorological Press, 239 pp.

Footnotes

* Part of this work done while the author was a visiting scientist at the National Climatic Data Center.

Corresponding author address: Dr. Robert E. Eskridge, National Climatic Data Center, 151 Patton Ave., Asheville, NC 28801-5001.