The diurnal temperature range (DTR) is an important indicator of climate change, and it has decreased worldwide since the 1950s, particularly over arid and semiarid regions. This study analyses the effect of meteorological and anthropogenic factors on DTR variation to investigate the possible causes of DTR decreases in semiarid climates. The study region is located in northeast China, and the study period is from 1957 to 2006. There are three main results. First, the rate of decrease in the DTR is −1.24 K per 50 years. This decrease is mainly attributed to the increasing daily minimum temperature rate (Tmin, 2.24 K per 50 years), which is greater than the change in the daily maximum temperature (Tmax, 1.00 K per 50 years). Second, sunshine duration (SD) appears to be the most significant meteorological factor that determines the DTR through downward shortwave radiation (Rsw,d) and surface soil moisture (SM). The effect of Rsw,d is larger for Tmax than for Tmin; therefore, the decrease in Rsw,d results in a smaller increase in Tmax than in Tmin. On the other hand, the increase in SM can strengthen daytime latent heat release, and the increase in Tmax is then slowed because of the cooling effect of evaporation. The precipitation values and the leaf area index show a negative correlation with the DTR, whereas the cloud amount and the relative humidity appear not to be main causes of the DTR decrease in this region. Finally, atmospheric aerosols can reduce the SD by 0.27 h year–1 by decreasing atmospheric transparency, as indicated by an analysis of the Total Ozone Mapping Spectrometer Aerosol Index from 1979 to 2005. The decrease in direct solar radiation is the main cause of decreases in Rsw,d. These findings will provide references for DTR variation studies in similar climates.
The diurnal temperature range (DTR) is considered as a suitable measure of climate change because of its sensitivity to variations in the radiative energy balance (Dai et al., 1999; Przybylak, 2000; Sun et al., 2006; Makowski et al., 2008). The DTR has decreased worldwide since the 1950s, mainly as a result of asymmetric diurnal changes in the daily temperature maximums (Tmax) and minimums (Tmin) (Karl et al., 1991; Karl et al., 1993; Dai et al., 1997; Easterling et al., 1997; Dai et al., 1999; Stone and Weaver, 2002; IPCC, 2007; Zhou et al., 2008; Zhou et al., 2009; Lai and Cheng, 2010; Fan et al., 2011). The warming trend for Tmin is usually stronger than for Tmax. In some regions, the Tmin has increased, but the daytime Tmax has decreased (Karl et al., 1991; Karl et al., 1993; Türkes et al., 1996; Dai et al., 1997; Easterling et al., 1997; Dai et al., 1999; Liu et al., 2004b).
Many studies have determined that the reduction in DTR is a consequence of increases in cloud cover, soil moisture (SM) and precipitation (Karl et al., 1993; Dai et al., 1997; Dai et al., 1999; Stone and Weaver, 2002). Clouds have a negative effect on DTR by reflecting sunlight during the day (decrease Tmax) and enhancing downward longwave radiation (Rlw,d) at night (increase in Tmin) (Dai et al., 1997; Dai et al., 1999; Zhou et al., 2009). SM may reduce the DTR via a surface evaporative cooling effect on Tmax, and precipitation may affect DTR indirectly by increasing SM content (Dai et al., 1997, 1999). Zhou et al. (2009) found that a decline in DTR is generally correlated with a decrease in the leaf area index (LAI), which is determined by climate conditions (e.g. precipitation). Some studies have found that land cover changes can reduce the DTR through the modification of land surface properties (e.g. emissivity) over some regions (Feddema et al., 2005; Zhou et al., 2007). Other factors, such as atmospheric aerosols and greenhouse gases, may also contribute to the decrease in DTR (Stone and Weaver, 2002; Liu et al., 2004b; IPCC, 2007; Zhou et al., 2007). Aerosols may affect the DTR by reflecting solar radiation and by modifying cloud properties, whereas greenhouse gases may play a role in altering the DTR by controlling the surface energy and hydrological balance (Zhou et al., 2007; Zhou et al., 2009).
Although various studies on DTR variation have been conducted, the investigation of DTR change mechanisms is still necessary because of the complexity of the climatology (Dai et al., 1997; Dai et al., 1999; Sun et al., 2006; Martínez et al., 2010). It is difficult to explain DTR changes explicitly based on one parameter alone (Easterling et al., 1997). DTR variation has regional characteristics because of the various changes in local climate due to the complicated interactions of local climate and other anthropogenic factors (Karl et al., 1991; Karl et al., 1993; Easterling et al., 1997; Liu et al., 2004b). In fact, the largest decreases in DTR were observed mostly over arid or semiarid regions (e.g. North China and western African Sahel) where drought has occurred (Zhou et al., 2007).
In this study, the effects of climate elements [e.g. precipitation, sunshine duration (SD), pan evaporation (PE), surface SM, relative humidity (RH), solar radiation and cloud cover], vegetation indicators (LAI) and anthropogenic factors (aerosols) on DTR change are examined using 50 years (1957 to 2006) of daily observation data over the upper Second Songhua River basin (USSR). The USSR borders northeast China, and it is characterized by a temperate, semiarid continental climate; the annual mean precipitation is approximately 700 mm (Wang et al., 2011; Wang et al., 2012). The aim of this work is to investigate the possible causes of DTR variation and the mechanisms behind such variation in a semiarid climate. This work is unique in that the comprehensive observations (including climatology, vegetation and aerosols) are analysed within a semiarid region. The results of this study will provide references for DTR variation studies as well as for regional climate modelling.
In this paper, Section 'Data and methodology' describes datasets and analytical methods. Section 'Results and discussions' presents the results and discussions, which include temperature variation analyses, the possible causes of DTR changes and the role of aerosols in SD changes. Conclusions and future directions are provided in Section 'Conclusions'.
2. Data and methodology
2.1.1. In situ observations
The ground-based meteorological daily observations include Tmax, Tmin, average temperature (Tm), cloud amount (CA), mean surface RH and SD. The data were obtained from the China Meteorological Administration (CMA) National Meteorological Information Center (NMIC) through its website, http://cdc.cma.gov.cn/. The daily precipitation data were provided by Songliao Water Resources, Ministry of Water Resources (SWR MWR). The data from 6 meteorological sites and 15 rain gauges were collected (Figure 1(b)). The available data cover the period from 1957 to 2006 (Table 1). The simple linear interpolation method and the stepwise regression method were used to fill in missing data gaps when gaps are up to 7 days in length and more than 7 days in length, respectively (Liu et al., 2004b).
Table 1. The data for the USSR
The meteorology data include daily maximum, minimum and mean temperature, and daily RH, CA and SD.
The daily downward shortwave radiation (Rsw,d) data were collected from the CMA at the Changchun, Shenyang and Yanji stations (Figure 1(b)). The data period from 1961 to 2006 was selected because the observations for all three stations were available (Table 1). Yanishevsky thermoelectric pyranometers were used to measure surface solar radiation data collected before 1993, whereas DFY-4 pyranometers were used for data collected after 1993. It has been reported that the uncertainty of the measurements is <5% (Shi et al., 2008). All of the above meteorological data were then interpolated to 1000-m cells through inverse-distance weighting to calculate the basin average value and the spatial distributions.
The daily PE was observed using a E601-type evaporation pan at the Wudaogou station (Figure 1(b)) from 1981 to 2006 (Table 1). The data were obtained from SWR MWR. The surface SM observations from the Huadian and Yangzishao stations (Figure 1(b)) were provided by the CMA NMIC. SM was measured every 10 days (i.e. on the 8th, 18th and 28th of every month) in the warm season (May–September) from 1992 to 2008 using the gravimetric technique (Wang and Zeng, 2011). The values for the top 10 cm (mass percentage) from 1993 to 2006 are used here.
2.1.2. Satellite observations
The dynamic vegetation parameter LAI is generally defined as one-sided green leaf area per unit ground surface area (Myneni et al., 1997). The LAI data were obtained from the Advanced Very High Resolution Radiometer (AVHRR) 16-km monthly product (Myneni et al., 1997) through ftp://primavera.bu.edu/pub/datasets. The data are available from July 1981 to May 2001.
The Total Ozone Mapping Spectrometer (TOMS) Aerosol Index (AI) data were obtained from the NASA Goddard Earth Sciences Data Information Services Center (GES DISC, http://acdisc.gsfc.nasa.gov/). AI is a qualitative measure of the presence of UV absorbing aerosols, such as mineral dust and smoke (Herman et al., 1997). The positive AI values and the negative AI values are associated with absorbing (mineral dust, smoke and volcanic aerosols) and non-absorbing (sulphate and sea salt particles) aerosols, respectively (Torres et al., 1998; Torres et al. 2002). Near-zero values indicate cloud presence. The daily TOMS/AI data utilized in the research are the version 8 data, which have a spatial resolution of 1.25° × 1° with global coverage. For the period from 1979 to 1992, we used daily TOMS Nimbus-7 data, and for the period from 1997 to 2005, we used TOMS Earth Probe data.
2.2.1. Trend magnitude calculation
A linear regression model (Tang et al., 2007, 2008) is used to calculate the trend magnitude. The regression function is
where t is the time number and y is the data value at time t; the regression weight β and noise term γ are calculated according to Equations (2) and (3)
where n is the time-series number. Student's t-test is applied to evaluate the statistical significance of the trends. The significance of a correlation coefficient r is tested with
where the distribution of ts is approximately that of the t-distribution with n – 2 degrees of freedom. The trend magnitude ΔY and relative trend magnitude ΔY ′ during the study period are estimated using Equations (5) and (6) if a trend is detected:
The linear trend is removed from the forcing data to compare the observed trends with non-change scenarios.
where y_rm is the time series data after removing tendency and is the mean value of the nonlinear trend time series. The value is set to the mean value of the first 10 years of data (1957–1966).
2.2.2. Variable normalization
The normalized variable (Xnorm) is used to compare the variables with different scales. Xnorm is calculated using Equations (8) and (9):
where Xoi is the observed value, is the mean value of Xoi over the comparison period, n is the total number of time series for comparison, and σ is the standard deviation.
3. Results and discussions
3.1. Temperature variation analysis
3.1.1. Temporal analysis
Figure 2(a)–(d) show the variations in the annual means for the Tmax, Tmin, Tm and DTR over the USSR from 1957 to 2006. In general, the t-test is significant for all four of these variables (Table 2).
Table 2. The change in climate and vegetation conditions in the USSR
The bold values indicate that the t-test is significant.
A significance level of 5% is used to detect the trend.
P (mm year–1)
Rsw,d (W m–2)
The Tmax increases at a rate of 1.00 K per 50 years (Figure 2(a)). This trend is higher than that reported for China as a whole (0.635 K per 50 years) between 1955 and 2000 (Liu et al., 2004b) and for the Northern Hemisphere (0.435 K per 50 years) between 1950 and 1993 (Easterling et al., 1997). Figure 2(b) shows that the Tmin increases at a rate of 2.24 K per 50 years from 1957 to 2006. This trend is also higher than previous analyses for China (1.615 K per 50 years) between 1957 and 2006 (Liu et al., 2004b) and for the Northern Hemisphere (0.92 K per 50 years) between 1950 and 1993 (Easterling et al., 1997). For Tm (Figure 2(c)), we calculate a rate of change equal to 1.68K per 50 years between 1957 and 2006, whereas Liu et al. (2004b) reported a value of 0.59 K per 50 years from 1955 to 1990. For the entire study period, we calculate the rate of change in DTR as −1.24 K per 50 years (Figure 2(d)). This is higher than the rate of −1.01 K per 50 years reported in a previous study on China (Liu et al., 2004b) and of 0.445 K per 50 years for the Northern Hemisphere (Easterling et al., 1997). As previous studies noted, the decreasing trend in DTR is mainly the result of Tmin increasing at a rate outpacing increases in Tmax (e.g. Dai et al., 1997; Easterling et al., 1997; Dai et al., 1999; Liu et al., 2004b). This result further supports the previous conclusions.
3.1.2. Spatial analysis
Figure 3 depicts the spatial distribution of the annual mean values for the Tmax, Tmin, Tm and DTR trend magnitudes over the USSR from 1957 to 2006. The warming trends appear in all parts of the region. The greatest increase in Tmax is found in the southeastern part of the region, and the rate slows as we move from the southeast to the northwest (Figure 3(a)). The Tmin (Figure 3(b)) and the Tm (Figure 3(c)) show similar spatial distributions of trend magnitude. The largest increasing trend is found in the southeastern part of the region, and the smallest value is found in the northwestern part of the basin. The trend gradually decreases from southeast to northwest. The spatial distribution of the DTR (Figure 3(d)) shows variation trends that are opposite to those of Tmin and Tm. This is reasonable because the decrease in DTR is mainly caused by the greater rate of increase in Tmin relative to Tmax.
From 1957 to 2006 over the USSR, the Tm, Tmax, and Tmin increase, whereas the DTR decreases. All four variables (Tm, Tmax, Tmin and DTR) change at a rate higher than that reported for China as a whole and for the Northern Hemisphere. It should be noted that the USSR is characterized by a semiarid continental climate. As Zhou et al. (2007) noted, the greatest decreases in the DTR were observed mostly over arid or semiarid regions (e.g. North China). Changes in the DTR can result from a number of complicated mechanisms. Because DTR is an important indicator of climate variation, the causes of the DTR variation in this semiarid environment are investigated in the following section.
3.2. The possible causes of DTR decrease
In this section, the causes of DTR variation are investigated by analysing the temporal trend and the correlation between the DTR and climatological factors. These factors include CA, precipitation, RH, SD, Rsw,d, PE, top 10-cm surface SM and LAI.
3.2.1. Trend analysis
Figure 2(e)–(i) illustrate the variation in mean annual CA, precipitation, RH, SD and Rsw,d over the USSR from 1957 to 2006. Figure 4 depicts the change in the mean annual PE at the Wudaogou station (Figure 1(b)), SM at the Huadian (SM_HD) station and the Meihekou (SM_MHK) station, and LAI. In general, the t-test for CA, RH, SD, and Rsw,d is significant, whereas it is not significant for other variables (Table 2).
Figure 2(e) shows the change of CA from 1957 to 2006. The CA decreases at a rate of 0.09 per 50 years between 1957 and 2006. This decreasing trend is consistent with the value of 0.05–0.15 per 50 years for northern China between 1954 and 2001 (Qian et al., 2006), whereas it is slightly lower than the value of 0.10–0.15 per 50 years for northeast China between 1954 and 1994, as reported by Kaiser (2000). Many studies (e.g. Campbell and Vonder Haar, 1997; Dai et al., 1997; Dai et al., 1999; Liu et al., 2004b) noted that cloud cover has a negative effect on DTR. During the day, clouds can decrease Tmax by reducing incident shortwave solar radiation, and they can increase Tmin by intercepting outgoing longwave radiation at night (Campbell and Vonder Haar, 1997; Liu et al., 2004b). Decreasing CA would increase the DTR during this time period, but the observed DTR shows a decreasing trend. This result indicates that the cloud cover change is not the major cause of the DTR variation in this region.
Figure 2(f) plots precipitation trends from 1957 to 2006. A slight increasing trend was detected in this region, but the variation is not significant (Table 2). Previous studies have shown that precipitation is negatively correlated with DTR through surface evaporation cooling (Dai et al., 1999).
Figure 2(g) gives the variation of RH from 1957 to 2006. The RH decreases at a rate of −0.02 per 50 years from 1957 to 2006. The water vapour reduces the DTR by absorbing solar radiation (Dai et al., 1999; Liu et al., 2004b). In this region, the variation trends for RH and the DTR are the same. It is then concluded that the dampening effect of RH is limited in this region.
Figure 2(h) depicts the annual SD variation from 1957 to 2006. The SD decreases at a rate of 0.72 h per 50 years between 1957 and 2006. The SD and the DTR show similar variation trends during the study period. Liu et al. (2004b) noted that the SD may affect DTR through the imbalance effects on Tmax and Tmin. A detailed discussion is provided in the following section.
Figures 2(i) depicts the change in Rsw,d from 1961 to 2006. Rsw,d shows a decreasing trend with a magnitude equal to −12.03 W/m2 per 50 years. Previous studies have shown that the decreasing trend ranges from −10.5 W/m2 per 50 years to 22.5 W/m2 per 50 years during the period from 1961 to 2000 over China (Che et al., 2005; Shi et al., 2008; Tang et al., 2011). The decline in Rsw,d is also detected worldwide (e.g., Wild et al., 2007). The similarity in the variations of Rsw,d and DTR implies that Rsw,d possibly influences DTR changes. In addition, Tmax could decrease with a decline of Rsw,d if other meteorological variables and the land surface characteristics keep unchanged. However, Tmax displays a significant increasing trend from 1957 to 2006. Greenhouse gases, such as water vapour, carbon dioxide, and nitrous oxide, may affect Tmax by changing the distribution of net radiation (Liu et al., 2010). It has been reported that the radiative forcing generated by an increase in the concentration of greenhouse gases may be the main reason for global warming (IPCC, 2007). Tmax may be affected by increasing greenhouse gas concentration. Thus, an increasing trend of the Tmax is detected from 1957 to 2006 in this region.
Figure 4(a)–(d) depict the variation in PE from 1981 to 2006, in SM_HD and SM_MHK from 1993 to 2006, and in LAI from 1982 to 2000. All four of the variables show an increasing trend, although it is not significant. The correlations between these variables and the DTR are discussed in the following section.
3.2.2. Correlative analysis
Figure 5 shows the relationship between the normalized anomaly of annual DTR and the normalized anomalies of other variables over the USSR. These variables include SD, PE, SM_HD and SM_MHK, Rsw,d, precipitation and LAI. Figure 6(a) and (b) provide the spatial distribution of the correlation coefficient (R) between DTR and SD and between DTR and precipitation, respectively.
DTR is strongly correlated with SD (Figures 5(a) and 6(a)), with R equal to 0.8173. This relationship was also found in Northeast India (Jhajharia and Singh, 2011) and in lower-elevation sites in the Swiss Alps (Rebetez and Beniston, 1998). SD is directly related to Rsw,d. One of the factors that contributes to this phenomenon in this region is the unbalanced effect of Rsw,d on Tmax and Tmin. Liu et al. (2004b) reported that Rsw,d is closely related to DTR during the period from 1955 to 2000 in China. The effect of Rsw,d is greater for daytime Tmax than for night-time Tmin. This results in a larger increase in Tmin than in Tmax and thus a lower DTR. In this region, the R between Rsw,d and DTR is equal to 0.5089 (Figure 5(e)).
Figure 5(b) shows the R between the DTR and PE from 1981 to 2006. There is a high positive correlation between the DTR and PE, with R equal to 0.6701. The similarity between the variation in DTR and that in PE is consistent with previous reports (e.g. Peterson et al., 1995; Roderick and Farquhar, 2002; Liu et al., 2004a).
Figure 5(c) and (d) plot the correlation between DTR and SM_HD and between DTR and SM_MHK (top 10 cm of soil) during May and September from 1993 to 2006. The SM is negatively correlated with the DTR, with R equal to −0.5348 and −0.8167 for the Huadian station and the Meihekou station, respectively. Many studies (Dai et al., 1999; Stone and Weaver, 2002) have noted that SM can decrease the DTR through evaporative cooling during the day, when the planetary boundary layer is unstable and the potential for evapotranspiration is high. The evaporative cooling effect on Tmax is larger than on Tmin, especially under dry conditions. Therefore, the DTR decreased with increased SM.
Figure 5(f) shows the R between precipitation and DTR from 1957 to 2006. Precipitation has an inhibiting effect on the DTR, with R equal to −0.4359. Many previous studies show that the evaporative cooling effect of precipitation may reduce the DTR (Dai et al., 1999; Liu et al., 2004b). The spatial distribution (Figure 6(b)) of R also indicates that precipitation and the DTR are negatively correlated over the USSR region.
Figure 5(g) indicates a negative correlation between the DTR and LAI, with R equal to −0.2702. This result is the opposite of that reported by Zhou et al. (2009). They found a strong positive correlation between the LAI and DTR over global land from 1950 to 2004. The result found in this study is probably caused by enhancing vegetation evaporative cooling effect through evapotranspiration. Increased evapotranspiration causes daytime cooling during the green season (Scheitlin and Dixon, 2010). Therefore, increases in LAI results in a decrease in the DTR.
The R between the DTR and CA and between the DTR and RH are 0.1581 and −0.0949 (not shown), respectively. These results indicate that the effects of CA and RH are small in this region.
The decrease in the DTR is mainly controlled by SD, PE and surface SM in the semiarid USSR. Other factors, such as precipitation and LAI, also affect the DTR. The R is low between the DTR and CA and between the DTR and RH. We found that SD is negatively correlated with SM, with R equal to −0.6442 and −0.6834 at the Huadian station and the Meihekou station, respectively. Thus, we suspect that SD may also affect the DTR through SM. Increased surface SM can increase latent heat release and slow down the increase in daytime Tmax (Dai et al., 1999). The DTR is then decreased as a result of the slower rate of increase in Tmax relative to Tmin. It is important to investigate the possible causes of the SD (or solar radiation) decrease because SD is the variable most closely correlated with the DTR.
3.3. The causes of SD change
Several factors can affect the variation in SD (or solar radiation), such as changes in cloud optical properties, radiative active gases and the mass and optical properties of aerosols (Streets et al., 2006; Qian et al., 2007; Xu et al., 2011). It has been suggested that clouds and aerosols are the most important of these factors (Xu et al., 2011). Some studies relate a decrease in SD to a decline in CA (e.g. Wild et al., 2005; Biggs et al., 2007). However, cloud cover has been decreasing with solar irradiance in this region (see Figure 2(e) and (i)). Our analysis indicates that the SD trend is evident even when only clear-sky days are considered (Figure 7). Thus, aerosols are the primary contributors to this trend.
3.3.1. The role of aerosols on SD
Figure 8 plots the variation in the basin-averaged TOMS AI, the basin-averaged SD, and the ratio of diffuse to direct solar radiation (Df/Di). Df and Di constitute total Rsw,d on a horizontal surface. The periods from 1979 to 1992 and from 1997 to 2005 (23 years total) are selected here because of the availability of the TOMS AI data.
Figure 8(a) shows the AI time series during the observation period. The annual mean AI varies from 0.22 (in 2005) to 0.97 (in 2001), with the average value equal to 0.50. The positive AI value indicates that the mineral dust, smoke and volcanic aerosols are the major aerosols in this basin. Zhang et al. (2011) also noted that a high mineral fraction exists in the northeast part of China.
Figure 8(b) depicts the mean annual SD under all-sky (SD) and aerosol-low sky (SD_Aero_rm) from 1979 to 2005. Aerosol-low sky is defined as a daily mean AI of <0.50 (23-year average AI). The average SD under an aerosol-low sky is 6.65 h, whereas it is 6.38 h under an all-sky condition during the 23-year period. This is reasonable because aerosols intercept sunshine (or solar radiation) on the way to the Earth's surface. Increased atmospheric aerosols resulting from industrial aerosols are known to reduce solar irradiance by reducing the amount of sunlight reaching the ground (Liu et al., 2004b).
Figure 8(c) shows the variation in annual mean diffuse surface solar radiation (Df) and the direct solar radiation (Di) ratio (Df/Di) under all-sky and aerosol-low sky conditions from 1979 to 2005. The 23-year averages Df/Di under aerosol-low sky and under all-sky conditions are 0.85 and 0.91, respectively. Previous studies (e.g. Qian et al., 2007; Xu et al., 2011) noted that increasing of atmospheric aerosols can enhance scattering in the atmosphere, whereas they result in a decrease in atmospheric transparency. This mechanism likely explains the corresponding lower value of (Df/Di) under aerosol-low sky compared to the value under all-sky in this region.
In general, the SD was reduced by 0.27 h year–1 because of the aerosol effect. It should be noted that the threshold for AI was set to 0.50 because only 23 years of data are available for TOMS. The aerosol effect may not have been completely eliminated. This issue can be addressed in future studies if both aerosol data and SD (or solar radiation) data are available for a long-term period.
The DTR is an important variable for detecting climate change (Sun et al., 2006; Makowski et al., 2008). The decreasing trend in DTRs has been observed worldwide since the 1950s (Karl et al., 1991; Karl et al., 1993; Dai et al., 1997; Easterling et al., 1997; Dai et al., 1999; IPCC, 2007), especially over arid and semiarid regions (Zhou et al., 2008; Zhou et al., 2009). However, the mechanisms of DTR variation remain somewhat ambiguous due to the complexity of the relevant climatology. In this study, the effects of climate elements (e.g. precipitation, SD), vegetation indicators (LAI) and anthropogenic factors (e.g. aerosols) over a semiarid region (northeast part of China) were examined using 50 years (1957–2006) of daily observation data. The main conclusions are described below.
First, the Tmax, Tmin and Tm increased, whereas the DTR decreased over all parts of the USSR, as indicated by temporal and spatial analyses of the period from 1957 to 2006. The rates of increase in Tmax and Tmin are 1.00 K per 50 years and 2.24 K per 50 years, respectively, whereas the rate of decrease in the DTR is −1.24 K per 50 years. The decrease in the DTR is mainly attributed to the higher rate of increase for Tmin relative to Tmax. This result is consistent with many previous studies (Karl et al., 1991; Karl et al., 1993; Dai et al., 1997; Easterling et al., 1997; Dai et al., 1999; Liu et al., 2004b).
Second, the causes of the DTR decrease were investigated by analysing the temporal trends and R of climatological and vegetation parameters. These parameters include CA, precipitation, RH, SD, Rsw,d, PE, SM and LAI. The results show that CA and RH appeared not to be the main cause of the decrease in DTR in this region. SD, PE and Rsw,d have positive correlations with the DTR, whereas the SM, precipitation and LAI show negative correlations with the DTR.
SD has the most significant relationship with the DTR, with R equal to 0.8173. SD may affect the DTR in two ways. Through the unbalanced impact of Rsw,d on Tmax and Tmin, the daytime Tmax is more sensitive than the night-time Tmin to Rsw,d. The decrease in Rsw,d results in a relatively lower increase in Tmax compared to the increase observed in Tmin. The second possible influence comes from surface SM. Increasing SM can increase daytime latent heat release. An increase in daytime Tmax is then slowed due to a cooling effect of evaporation (Dai et al., 1999). Therefore, DTR decreases because Tmax increases more slowly than Tmin.
Precipitation and LAI reduces the DTR through evaporative cooling. Therefore, increases in precipitation and the LAI also result in a decrease in the DTR.
Third, the role of aerosols on solar radiation reduction was determined by analysing the TOMS AI from 1979 to 1992 and from 1997 to 2005. Mineral dust and smoke are the major aerosols in the study region. SD is reduced by 0.27 h year–1 as a result of decreasing atmospheric transparency induced by aerosols. Because aerosols can enhance scattering in the atmosphere, the ratio of annual mean diffuse surface solar radiation (Df) and direct solar radiation (Di) is higher under all-sky conditions than under aerosol-low sky conditions in this region.
In general, DTR variation is controlled by a number of factors. In addition to the meteorology and the anthropogenic factors analysed in this study, other factors (e.g. greenhouse gasses) may also affect DTR changes (IPCC, 2007). Further analysis regarding the causes of DTR variation is necessary. Given the importance of climate change (Piao et al., 2010), further research is also encouraged to assess the impact of climate variation on the water and energy cycles in this region.
This study was supported by National Basic Research and Development Program of China (973, Grant No. 2013CB036401) and National Natural Science Foundation of China (Grant No. 51079014). The authors are deeply indebted to anonymous reviewers for their valuable comments and suggestions that greatly improve the quality of this manuscript.