NDVI, temperature and precipitation changes and their relationships with different vegetation types during 1998–2007 in Inner Mongolia, China

Authors

  • X. W. Chuai,

    1. School of Geographic and Oceanic Science, Nanjing University, Nanjing, China
    Search for more papers by this author
  • X. J. Huang,

    Corresponding author
    1. School of Geographic and Oceanic Science, Nanjing University, Nanjing, China
    2. Land Development and Consolidation Technology Engineering Center of Jiangsu Province, Nanjing, China
    Search for more papers by this author
  • W. J. Wang,

    1. School of Geographic and Oceanic Science, Nanjing University, Nanjing, China
    Search for more papers by this author
  • G. Bao

    1. Inner Mongolia Key Laboratory of Remote Sensing and Geographic Information System, Inner Mongolia Normal University, Hohhot, China
    2. International Institute for Earth System Science, Nanjing University, Nanjing, China
    Search for more papers by this author

X. J. Huang, School of Geographic and Oceanic Science, Nanjing University, Nanjing 210093, China. E-mail: hxj369@nju.edu.cn

ABSTRACT

Based on vegetation maps of Inner Mongolia, SPOT-VEGETATION normalized difference vegetation index (NDVI) data, and temperature and precipitation data from 118 meteorological stations, this study analysed changes in NDVI, temperature and precipitation, and performed correlation analyses of NDVI, temperature and precipitation for eight different vegetation types during the growing seasons (April–October) of the period 1998–2007 in Inner Mongolia, China. We also investigated seasonal correlations and lag-time effects, and our results indicated that for different vegetation types, NDVI changes during 1998–2007 showed great variation. NDVI correlated quite differently with temperature and precipitation, with obvious seasonal differences. Lag-time effects also varied among vegetation types and seasons. On the whole, Inner Mongolia is becoming warmer, and drier for most regions, and ecological pressure in Inner Mongolia is increasing, and our focus on such issues is therefore important. Copyright © 2012 Royal Meteorological Society

1. Introduction

Vegetation is the Earth's natural linkage of soil, atmosphere and moisture. It displays obvious seasonal and annual changes (Cui and Shi, 2010; Zhang et al., 2011a) and acts as a sensitive indicator of global climate changes (Schimel et al., 2001; Weiss et al., 2004). Vegetation responds to climate changes in both explicit and subtle ways. Studying these changes has become a global interdisciplinary effort for researchers who seek to understand what is happening and to find the most efficient means of doing so (Meng et al., 2011a, 2011b).

The normalized difference vegetation index (NDVI) was proposed by Rouse et al. (1974) based on differences in pigment absorption features in the red and near-infrared regions of the electromagnetic spectrum (Equation ((1))). The values of NDVI range from − 1.0 to 1.0, increasing positive NDVI values indicate increasing amounts of green vegetation. NDVI values near zero and decreasing negative values indicate nonvegetated features such as barren surfaces (rock and soil) and water, snow, ice and clouds (Schnur et al., 2010). Since it has many advantages such as the simplicity of the algorithm, the capacity to broadly distinguish vegetated areas from other surface types, more sensitive to detect green vegetation than using a single band (Zhang et al., 2005), it can be used to monitor local or global vegetation changes, which can indicate environmental changes brought by natural factor such as climate changes (Qiu and Cao, 2011) and anthropogenic activities such as urban expansion process (Fung and Siu, 2000), to assess crop production (Wardlow and Egbert, 2008) and net primary productivity (NPP) of vegetation (Piao et al., 2006, 2008), and it was also mostly used to indicate climate changes by establishing relationship between climatic factors and NDVI (Nemani et al., 2003; Roerink et al., 2003) and so on:

equation image(1)

where RED and NIR stand for the spectral reflectance measurements acquired in the visible (red) and near-infrared regions, respectively.

Climatic factors, land use changes, the fertilization effect of CO2 and so on could make different impacts on vegetation; among them, temperature and precipitation are the main indicators used to describe climate conditions, and they can affect vegetation growth in an obvious manner (Fang et al., 2004; Ji and Peters, 2004). NDVI, temperature and precipitation data have been used to study the effects of climate change on vegetation for a long time by many scholars. Since the 1980s, NDVI has been widely used for studying climatic effects on changes in vegetation productivity (Ichii et al., 2002; Nemani et al., 2003; Roerink et al., 2003; Meng et al., 2011b; Zhang et al., 2011a). However, the results of these studies varied because of the complexity of vegetation characteristics and regions (Bonan et al., 2003; Crucifix et al., 2005; Ni et al., 2006; Meng et al., 2011b). Most studies considered the different types of vegetation as a whole, and thus did not consider their differential responses to climate change. Schultz and Halpert (1995) conducted a global analysis of the relationships between NDVI, precipitation and land surface temperature and found no significant correlations among them. However, other studies have found significant relationships between these variables. For example, Guo et al. (2008) reported that NDVI variations were significantly correlated with both temperature and precipitation. Rasmusen (1998) found a positive correlation between NDVI and precipitation. Ichii et al. (2002) reported a strong positive correlation between NDVI and temperature in high-latitude districts of the northern hemisphere in both spring and autumn. In recent years, a few studies have analysed the relationships between NDVI, temperature and precipitation for different vegetation types at a regional scale. Luo et al. (2009) reported that there were strong correlations between NDVI, precipitation and temperature for different vegetation types in northeast China, and that the effect of temperature on NDVI was more obvious than that of precipitation. In that study, increased precipitation even inhibited the growth of marsh and scrubs in summer. Zhang et al. (2011a) reported that a positive correlation between NDVI and temperature was most obvious for scrubs and coniferous forests, that the effect of precipitation on NDVI was not as significant as the effect of temperature and that bush NDVI correlated more strongly with precipitation than the NDVI of other vegetation.

Regional responses to global warming show wide variation. In China, the strength of the increasing temperature trend decreases from the south to the north (Gao et al., 2009). Inner Mongolia is located in the north of China, the green vegetation types is rich and is covered with China's largest area of grassland, and NDVI can well be used to describe their growth in spatial and the changes in temporal. What is more, most of the area has an arid to semi-arid climate, and the ecological environment in this district is fragile and sensitive to global climate change (Chen and Wang, 2009). Some scholars have studied the relationship between NDVI and climate factors in Inner Mongolia (Liu et al., 2009; Qu et al., 2009; Sun et al., 2010; Xu et al., 2010), but most have only focused on grassland or have not considered different vegetation types. Few studies have analysed the relationships between NDVI, temperature and precipitation in recent years for different vegetation types in Inner Mongolia. Furthermore, different remote-sensing observations provide different results in terms of vegetation responses to climate change (Sivakumar et al., 2005; Zhao et al., 2005; Camberlin et al., 2007). Compared with the NOAA/AVHRR sensor, the SPOT-4 VEGETATION sensor has many advantages, such as being more sensitive to chlorophyll absorption in its red band, the elimination of strong water vapour absorption in its near-infrared band and higher spatial resolution. Therefore, using the SPOT-4 VEGETATION NDVI will increase analytical precision (Xia et al., 2008; Yan et al., 2008; Cui and Shi, 2010).

Based on SPOT-4 VEGETATION NDVI data, a vegetation map of Inner Mongolia, and temperature and precipitation data, this study analysed changes in NDVI, temperature and precipitation and investigated the correlations between NDVI, temperature and precipitation for different vegetation types during the growing seasons of the period 1998–2007 in Inner Mongolia. To avoid spurious NDVI trends caused by winter snow, this study focuses only on growing season NDVI (Zhou et al., 2001; Piao et al., 2004). The growing season was defined as April–October and was further divided into three seasons: April–May (spring), June–August (summer) and September–October (autumn) (Piao et al., 2006; Guo et al., 2007).The objectives of this study were (1) to analyse NDVI, precipitation and temperature changes during the growing seasons for different types of vegetation in Inner Mongolia over the past 10 years; (2) to compare correlations between NDVI, temperature and precipitation for different vegetation types in different growing seasons; and (3) to explain the reasons behind the different relationships identified.

2. Materials and methods

2.1. Study area

Inner Mongolia is located in northwest inland China with an area of 118.3 × 104 km2, and accounts for 12.3% of China's full land area. It lies between latitudes 37° 24′N–53° 20′N and longitudes 77° 10′E–126° 29′E. Most of the area is located on the flat Mongolian Plateau. The climate in Inner Mongolia varies from the west arid and semi-arid climates, to the east wet, half humid monsoon climate. The mean annual precipitation ranges from 35 to 530 mm, and the mean temperature is between − 5 °C and 10 °C (Chen and Wang, 2009). The main vegetation types include cultivated vegetation, marshes, shrubs, steppes, meadows, desert vegetation, coniferous forests and broadleaf forests. There is an obvious zonal vegetation distribution affected by precipitation and temperature (Figure 1). Because of the effects of climate change and human disturbance, the ecological environment in this district is fragile and sensitive to further global climate changes.

Figure 1.

Vegetation map and the locations of meteorological stations in Inner Mongolia, China.

2.2. Data sources

Data used in this study were as follows: (1) NDVI data from 1998 to 2007 used in this study were from SPOT VGT-DN; in March of 1998, SPOT-4 satellite was launched and its vegetation sensor began to receive global vegetation observation data in April, 1998, with 10 day composite (maximum-value) temporal resolution, 1 km × 1 km spatial resolution and stretched values ranging from 0 to 255, available at http://www.geodata.cn, which was provided by the Data Sharing Infrastructure of Earth System Science, where the atmospheric, radiometric and geometric corrections had been made. (2) Monthly mean temperature and monthly precipitation data were obtained from all of the meteorological stations distributed in Inner Mongolia with a total of 118, which was provided by the Inner Mongolia Weather Bureau and Climatic Data Center and National Meteorological Information Center. (3) Information on the distribution of different vegetation types was obtained from the vegetation map of Inner Mongolia with a scale of 1:1 000 000 (Editorial Board of Vegetation Map of China, 2001), provided by the Data Sharing Infrastructure of Earth System Science (http://www.geodata.cn).

2.3. Methods

2.3.1. Annual and seasonal NDVI, temperature and precipitation

True NDVI was restored with the formula NDVI = DN × 0.004 − 0.1 (Cui and Shi, 2010) in ArcGIS10.0. Each month includes three 10 day NDVI grid images. The maximum values of the three images were calculated as monthly NDVI. NDVI images for spring (April–May), summer (June–August) and fall (September–October) from 1998 to 2007 were generated separately through computed averages of respective monthly NDVI values. In this way, a seasonal NDVI series with 1 km × 1 km spatial resolution was obtained. The annual growing season (April–October) NDVI was defined as the average of NDVI values from April to October.

For the monthly mean temperature and precipitation data from meteorological stations, averages were also calculated to obtain seasonal and annual mean temperature and precipitation data, respectively, for each station from 1998 to 2007 (April–October). Kriging methods were employed using ArcGIS10.0, and then seasonal and annual precipitation and temperature grid maps were produced that cover the whole area of Inner Mongolia.

According to the vegetation map, we extracted NDVI, temperature and precipitation values for each vegetation type during the growing seasons of the period 1998 to 2007. The average values of NDVI, temperature and precipitation associated with a particular vegetation type were calculated from the averages of all grid cells belonging to the same vegetation type.

2.3.2. Response of vegetation NDVI to temperature and precipitation

Pearson's correlations between seasonal NDVI–temperature and NDVI–precipitation were analysed using SPSS (version 11.5), and two-tailed P-values were used to determine significance. Considering the lagged response of NDVI to temperature and precipitation (Braswell et al., 1997; Wen and Fu, 2000; Cui and Shi, 2010), the NDVI–temperature and NDVI–precipitation correlation analyses were also carried out between each seasonal NDVI and the previous season's temperature and precipitation (for each growing season). Since the SPOT-4 VEGETATION sensor just began to receive global vegetation observation data in April, 1998, the time series seems to be a little shorter, which may be a limitation that may influence the accuracy of our analysis at some extent.

3. Results and discussion

3.1. Annual changes in growing season NDVI, precipitation and temperature for different vegetation types

Figure 2 illustrates the changes in growing season NDVI, precipitation and temperature during 1998–2007 in Inner Mongolia for cultivated vegetation, marshes, shrubs, steppes, meadows, deserts, coniferous forests and broadleaf forests.

Figure 2.

Annual changes in growing season NDVI, growing season precipitation (mm) and growing season temperature ( °C) for eight different vegetation types during 1998–2007 in Inner Mongolia, China.

The mean growing season NDVI did not change significantly. There was a moderate increase in marshes, coniferous forests, broadleaf forests and cultivated vegetation (r = 0.59, P = 0.072;r = 0.658, P = 0.037;r = 0.5, P = 0.142;r = 0.31, P = 0.383; respectively), and a weak increase in desert vegetation (r = 0.114, P = 0.753). The mean growing season NDVI of shrubs and meadows both decreased weakly (r = − 0.09, P = 0.807 and r = − 0.029, P = 0.938; respectively), whereas the NDVI of steppes decrease moderately (r = − 0.35, P = 0.319). All mean growing season precipitation conditions during 1998–2007 decreased nonsignificantly with P > 0.05: cultivated vegetation (r = − 0.26, P = 0.474), marshes (r = − 0.413, P = 0.235), shrubs (r = − 0.347, P = 0.326), steppes (r = − 0.56, P = 0.096), meadows (r = − 0.384, P = 0.274), deserts (r = − 0.312, P = 0.381), coniferous forests (r = − 0.19, P = 0.6) and broadleaf forests (r = − 0.497, P = 0.142). All the mean growing season temperature conditions during 1998–2007 increased significantly for cultivated vegetation, shrubs, steppes, meadows and desert vegetation, whose all P-values can meet 0.05 significant test, but the increase trend is not significant for marshes (r = 0.444, P = 0.198), coniferous forests (r = 0.357, P = 0.312) and broadleaf forests (r = 0.493, P = 0.147), respectively. The weak decreasing precipitation and significant increasing temperature trends are consistent with the findings of Ding and Chen (2008) and Gao et al. (2009), who analysed temperature and precipitation changes during the last 47 and 50 years, respectively, in Inner Mongolia.

Figure 2 shows that the NDVI fluctuations corresponded well with those of precipitation for cultivated vegetation, shrubs, steppes, meadows and desert vegetation. For cultivated vegetation, growing season precipitation was relatively high in 1998 and 2003, and low in 1999, 2000 and 2001, and coincided with peaks and troughs in growing season NDVI. For shrubs, the growing season precipitation maximum in 1998 and 2003 and minimum in 2000, 2001 and 2007 corresponded with maximum and minimum NDVI, respectively. For steppes, peak NDVI and precipitation values were reached in 1998 and 2003. For meadows, all low NDVI and precipitation values appeared in 2000, 2001 and 2007, and peak values appeared in 1998. For desert vegetation, peak NDVI and precipitation values both appeared in 1998, 2003 and 2007 and low values appeared in 2001. The NDVI and precipitation patterns for marshes and coniferous forests were reversed, as indicated in Figure 2. For marshes, maximum NDVI values occurred in 1999, 2006 and 2007, but the values of precipitation in these years were much lower than those in other years. Minimum values occurred in 1998 and 2003, but values of precipitation in those 2 years reached their maximum. Similarly, for coniferous forests, low NDVI values corresponded with high precipitation values in 1998 and 2008, and high NDVI values corresponded with low precipitation values in 1999, 2002, 2005, 2006 and 2007. Fluctuations in temperature corresponded well with NDVI for marshes and coniferous forests except in 1998. For marshes, high NDVI and temperature values occurred in 2000, 2002, 2004, 2005 and 2007 and low values occurred in 2003 and 2006. For coniferous forests, high NDVI and temperature values occurred in 2000, 2001, 2002, 2004, 2005 and 2007 and low values occurred in 2003. For broadleaf forests, fluctuations in precipitation and temperature did not obviously correspond with fluctuations in NDVI.

3.2. Annual correlations between growing season NDVI, precipitation and temperature

It can be seen from the above analysis that both precipitation and temperature can affect NDVI, and the effect of precipitation seems more significant than that of temperature. The effects of precipitation and temperature on NDVI varied for different vegetation types; therefore, we conducted separate correlation analyses for each vegetation type to compare the differences between them (Table 1).

Table 1. Annual correlation coefficients (r) and two-tailed significance test values (P) between growing season NDVI, precipitation and temperature for different vegetation types during 1998–2007 in Inner Mongolia, China
Vegetation typeNDVI and temperatureNDVI and precipitation
 rPrP
  1. Significant at

  2. a

    P = 0.05 and

  3. b

    P = 0.01 levels.

Cultivated vegetation0.3900.2650.638a0.047
Marshes0.1540.672− 0.791b0.006
Shrubs0.0010.9980.722a0.018
Steppes− 0.2450.4950.706a0.022
Meadows− 0.0410.9110.5540.097
Desert vegetation− 0.0030.9920.510.132
Coniferous forests0.2500.487− 0.728a0.017
Broadleaf forests0.1540.671− 0.3030.395

Table 1 shows that the correlation coefficients for growing season NDVI and precipitation were all much higher than those for NDVI and temperature. For cultivated vegetation, shrubs and steppes, growing season NDVI correlated strongly with precipitation (r = 0.638, P = 0.047;r = 0.722, P = 0.018;r = 0.706, P = 0.022, respectively). The correlations were moderate for meadows and desert vegetation (r = 0.554, P = 0.097;r = 0.51, P = 0.132, respectively). However, for marshes and coniferous forests, there were strong negative correlation coefficients for growing season NDVI and precipitation (r = − 0.791, P = 0.006 and r = − 0.728, P = 0.017, respectively). For broadleaf forests, the correlation between growing season NDVI and precipitation was not significant (r = − 0.303, P = 0.395). The correlations between growing season NDVI and temperature were weak for all vegetation types when compared with precipitation. As indicated in Table 1, the correlation coefficients were positive for cultivated vegetation, marshes, shrubs, desert vegetation and coniferous forests, but negative for others.

The different correlations can be explained by their different temporal and spatial growth environments, and differences in their degree of human disturbance. Shrubs, steppes, meadows and desert vegetation are mainly located in dry climate districts where precipitation is the limiting factor for vegetation growth. Higher temperatures accelerate the evaporation process, which leads to water scarcity and prohibits vegetation growth (Jobbagy et al., 2002; Chen and Wang, 2009; Li et al., 2009; Xu et al., 2010). There was a significant increase in temperature and a weak decrease in precipitation during 1998–2007, as discussed above, which led to weak decreases in NDVI for shrubs, steppes and meadows. The weak increase in NDVI for desert vegetation may be attributed to the effects of the Three-North Shelterbelt Construction and the Returning Farmland to Forest and Grassland project, which protected the growth environment (Yin et al., 2011). However, for marshes, coniferous forests and broadleaf forests, which are located in northeast Inner Mongolia under conditions of abundant precipitation but low temperatures, temperature increases are beneficial for vegetation growth (Piao et al., 2006). Marshes grow in rich water environments, and two much water will prohibit growth (Luo et al., 2009; Zhang et al., 2011a). Precipitation will also decrease temperatures, which is not good for vegetation growth in high, cold northeast Inner Mongolia. Therefore, the moderate increase in temperature and weak decrease in precipitation led to a moderate NDVI increase for marshes, coniferous forests and broadleaf forests. However, there are also studies that show positive correlations between precipitation and NDVI for coniferous forests, and negative correlations between NDVI and temperature for broadleaf forests (Guo et al., 2007; Luo et al., 2009); these discrepant results may be caused by different study areas with different climate environments and degrees of human disturbance. Cultivated vegetation is mainly located in the east of Inner Mongolia where there is better water and heat conditions, but because the growth of cultivated vegetation requires more water and heat compared with other vegetation types, adequate precipitation and temperature is crucial, and especially the demand for precipitation, which is consistent with the findings of Luo et al. (2009) who conducted a study of cultivated vegetation in northeast China.

3.3. Seasonal correlations between NDVI, precipitation and temperature

The effect of climate factors on NDVI may differ according to growth phase (Piao et al., 2006). Therefore, to recognize seasonal differences, we performed correlation analyses between NDVI, precipitation and temperature in spring (April–May), summer (June–August) and autumn (September–October) for different vegetation types. In addition, because some studies have shown an obvious lag-time effect (Luo et al., 2009; Cui and Shi, 2010; Xu et al., 2010), we performed correlation analyses between summer NDVI and spring precipitation and temperature, and between autumn NDVI and summer precipitation and temperature.

Table 2 shows that in spring, NDVI correlated positively with temperature for all vegetation types. NDVI also correlated positively with precipitation for all vegetation types with the exceptions of marshes and coniferous forests. These findings indicate that in spring, the warmer environment and increased precipitation in Inner Mongolia are critical for the growth of most of the vegetation types. For marshes and coniferous forests, higher temperatures and less precipitation are beneficial for vegetation growth in spring, because they are located under conditions of abundant precipitation, and more precipitation can decrease temperatures, as discussed above. For cultivated vegetation, desert vegetation and broadleaf forests, the effects of temperature and precipitation on NVDI seem equivalent. For shrubs, steppes and meadows, the positive effects of precipitation (r = 0.561, P = 0.091; r = 0.561, P = 0.092; r = 0.555, P = 0.096, respectively) on NDVI were more obvious than the effects of temperature (r = 0.241, P = 503; r = 0.365, P = 0.300; r = 0.398, P = 0.254, respectively).

Table 2. Correlation coefficients (r) and two-tailed significance test values (P) for NDVI, precipitation and temperature, for different vegetation types in spring in Inner Mongolia, China
Spring Vegetation typeNDVI and temperatureNDVI and precipitation
 rPrP
Cultivated vegetation0.4260.2190.4310.214
Marshes0.4850.155− 0.3290.353
Shrubs0.2410.5030.5610.091
Steppes0.3650.3000.5610.092
Meadows0.3980.2540.5550.096
Desert vegetation0.5270.1180.5470.098
Coniferous forests0.5950.07− 0.5430.105
Broadleaf forests0.4360.2080.4220.225

Table 3 shows the results of correlation analyses between NDVI, temperature and precipitation in summer, and between summer NDVI, and temperature and precipitation in spring for different vegetation types. It shows that for cultivated vegetation, summer NDVI correlated positively with summer precipitation, but negatively with temperature. This can be explained by the fact that under the background of global warming, the heat conditions are sufficient for the growth of most cultivated vegetation planted in Inner Mongolia, high temperatures lead to evaporation, which prohibits growth. NDVI correlated positively and more strongly with precipitation for shrubs, steppes and desert vegetation in summer than that in spring (r = 0.682, P = 0.030;r = 0.638, P = 0.047;r = 0.81, P = 0.004, respectively), which indicates that precipitation in summer is not abundant and is still the limiting factor for the growth of these vegetation types. However, unlike in spring, NDVI had a strong negative correlation with temperature (r = − 0.728, P = 0.017;r = − 0.692, P = 0.027;r = − 0.711, P = 0.021, respectively). This can be explained by the fact that the much warmer environment in summer provided enough heat for vegetation growth and that increased temperatures accelerate water evaporation and restrict vegetation growth (Jobbagy et al., 2002; Piao et al., 2006). For meadows, NDVI correlated weakly with temperature (r = − 0.15, P = 0.679) and precipitation (r = 0.002, P = 0.995), which indicates that the effect of temperature on NDVI was minimal and that the summer-growth demand for precipitation is low for meadows (Luo et al., 2009). Compared with desert vegetation, meadows may have a much higher precipitation demand, but meadows in Inner Mongolia are located under conditions with much more precipitation than desert vegetation, so there is a strong and positive correlation coefficient between NDVI and precipitation for desert vegetation but not for meadows in summer. For marshes, broadleaf forests and coniferous forests, as in spring, NDVI also correlated positively with temperature and negatively with precipitation in summer. This can be explained by the fact that precipitation for the three vegetation types is abundant in summer, and increases in precipitation will decrease temperatures, which is not good for vegetation growth. What is more, according to our calculation, NDVI of the three vegetation types are higher than that of others (Figure 2); they have high biomass in Inner Mongolia, and in the mid-growing season of summer, NDVI values will reach maximum compared with those in other seasons (Sun et al., 2010; Xu et al., 2010), which is close to saturation, and the apparent saturation of NDVI in high biomass region during the mid-growing season could also have contributed to this negative relationship (Gitelson, 2004).

Table 3. Correlation coefficients and two-tailed significance test values between NDVI, precipitation and temperature for different vegetation types in summer
Summer Vegetation typeNDVI and temperatureNDVI and precipitationNDVI and previous season temperatureNDVI and previous season precipitation
 rPrPr−1P−1r−1P−1
  1. Significant at

  2. a

    P = 0.05 and

  3. b

    P = 0.01 levels.

    r: correlation coefficient between NDVI and temperature or precipitation in the same season; r−1: correlation coefficient between NDVI and temperature or precipitation in the previous season; P: two-tailed significance test values between NDVI and temperature or precipitation in the same season; P−1: two-tailed significance test values between NDVI and temperature or precipitation in the previous season.

Cultivated vegetation− 0.681a0.030.3320.3480.696a0.0250.5220.122
Marshes0.6090.062− 0.71a0.0210.2610.446− 0.5510.099
Shrubs− 0.728a0.0170.682a0.030.470.1710.6280.052
Steppes− 0.692a0.0270.638a0.0470.2520.4820.71a0.021
Meadows− 0.150.6790.0020.995− 0.0850.8160.4150.233
Desert vegetation− 0.711a0.0210.81b0.0040.4530.1880.5540.096
Coniferous forests0.5980.068− 0.5710.0850.3070.389− 0.5270.118
Broadleaf forests0.3290.353− 0.654a0.040.5020.1170.747a0.015

Table 4 shows the correlations between NDVI, temperature and precipitation in autumn, and the lag-time effect between autumn NDVI, and temperature and precipitation in summer for different vegetation types. It shows that in autumn, the effect of precipitation on NDVI weakened compared with that in spring and summer for cultivated vegetation, steppes and desert vegetation, and for shrubs, the direction of the effect changed from positive to negative. This indicated that the demand for precipitation by these vegetation types decreased, and that precipitation does not limit vegetation growth in autumn. However, for meadows, precipitation correlated with NDVI more strongly in autumn (r = 0.389, P = 0.267) than in summer (r = 0.002, P = 0.995). Piao et al. (2006) found that the correlation co-efficient between NDVI and precipitation reached its maximum at about 200 mm annual precipitation for temperate grassland (including meadows) during the growing season. Too little or too much precipitation both failed to raise NDVI, so the low demand for precipitation and the decrease of precipitation in autumn led to the correlation coefficient in autumn being much higher than that in summer. For marshes and coniferous forests, the correlation coefficient for NDVI and precipitation was still negative, but for broadleaf forests, the negative effect of precipitation in summer on autumn NDVI decreased and became weakly positive (r = 0.012, P = 0.973) because of the decreased amount of precipitation in autumn. For shrubs, steppes and meadows, the effects of temperature on NDVI were positive (r = 0.455, P = 0.187;r = 0.458, P = 0.183;r = 0.585, P = 0.076, respectively), as in spring, which indicated that temperature can boost vegetation growth. This may be because temperatures decreased in autumn, and evaporation was weak, and not sufficient enough to cause water scarcity. For broadleaf forests, the correlation was stronger (r = 0.431, P = 0.213) in autumn than in summer (r = 0.239, P = 0.353). The correlation coefficients for the other vegetation types all decreased and did not seem as obvious as those found in spring and summer.

Table 4. Correlation coefficients and two-tailed significance test values between NDVI, precipitation and temperature for different vegetation types in autumn
Autumn Vegetation typeNDVI and temperatureNDVI and precipitationNDVI and previous season temperatureNDVI and previous season precipitation
 rPrPr−1P−1r−1P−1
  1. Significant at

  2. a

    P = 0.05 and

  3. b

    P = 0.01 levels.

    r: correlation coefficient between NDVI and temperature or precipitation in the same season; r−1: correlation coefficient between NDVI and temperature or precipitation in the previous season; P: two-tailed significance test values between NDVI and temperature or precipitation in the same season; P−1: two-tailed significance test values between NDVI and temperature or precipitation in the previous season.

Cultivated vegetation0.1750.6290.3360.3430.4270.2180.2310.52
Marshes0.1080.767− 0.2860.4220.0610.866− 0.3020.396
Shrubs0.4550.187− 0.1160.749− 0.5230.1210.6150.059
Steppes0.4580.1830.4460.196− 0.721a0.0190.78b0.008
Meadows0.5850.0760.3890.267− 0.5480.1010.5330.113
Desert vegetation− 0.2920.4140.4110.238− 0.757a0.0110.5660.088
Coniferous forests0.0520.886− 0.5960.0690.2080.564− 0.3980.255
Broadleaf forests0.4310.2130.0120.9730.0960.792− 0.1560.667

Table 3 shows the lag-time effect of spring temperature on summer NDVI was not significant (P > 0.05) for any vegetation types except cultivated vegetation (r = 0.696, P = 0.025), with the r-values all lower than those for the effects of summer NDVI and summer temperature. But there was a significant lag-time effect of spring precipitation on summer NDVI, especially for cultivated vegetation, steppes and meadows, which had higher r-values for this relationship than for that of summer NDVI and summer precipitation. Unlike summer precipitation, spring precipitation correlated positively and strongly with summer NDVI for broadleaf forests (r = 0.747, P = 0.015), which indicated that spring is the limiting growing phase for broadleaf forests and that vegetation growth in spring can greatly affect growth conditions in summer. Table 4 shows that there also exists lag-time effect of summer temperature on autumn NDVI, and that the effect was significant for steppes and desert vegetation (r = − 0.721, P = 0.019;r = − 0.757, P = 0.011, respectively), moderate for shrubs, meadows and cultivated vegetation (r = − 0.523, P = 0.121;r = − 0.548, P = 0.101; r = 0.427, P = 0.218; respectively), and weak for marshes, coniferous forests and broadleaf forests, which was not obvious (r = 0.061, P = 0.866;r = 0.208, P = 0.564; r = 0.096, P = 0.792; respectively). The lag-time effect of summer precipitation on autumn NDVI was even more obvious than the effect of autumn precipitation, with higher r-values for most of the vegetation types. And it was significant for steppes (r = 0.78, P = 0.008), moderate for shrubs, meadows and desert vegetation (r = 0.615, P = 0.059;r = 0.533, P = 0.113;r = 0.566, P = 0.088; respectively), but the effect is weak for cultivated vegetation (r = 0.231, P = 0.52), marshes (r = − 0.302, P = 0.396), coniferous forests (r = − 0.398, P = 0.255) and broadleaf forests, which was not obvious (r = − 0.156, P = 0.667). Temporal lags in vegetation response to climate change, which has been widely observed in other regions (Braswell et al., 1997; Los et al., 2001; Wang et al., 2003; Piao et al., 2006). However, the range of these lags is likely to vary spatially and temporally, the 2–3 months lag time of NDVI response to temperature and precipitation can be tested in many studies (Piao et al., 2006; Li et al., 2007), and results of their studies are very similar to those of our study.

As discussed above, the correlations between NDVI, temperature and precipitation differed according to vegetation type and season. For cultivated vegetation, temperature and precipitation both had positive effects on NDVI in spring and autumn, and the effects of temperature and precipitation were equivalent in spring. The effect of temperature was negative and more obvious than precipitation in summer, but in autumn the effect of temperature was not obvious as that of precipitation. The previous season's temperature and precipitation both had lag-time effects on NDVI in summer and autumn, and especially obvious for the effect of spring temperature on summer NDVI, significant at P < 0.05 level. For marshes and coniferous forests, temperature had positive and precipitation negative effects on NDVI during the whole growing season. The positive effect of temperature was moderate in spring, more stronger in summer than spring, but weak in autumn. The negative effect of precipitation was also the most obvious in summer, and similar in spring and autumn. The lag-time effect was the most obvious for spring precipitation and summer NDVI but still meet significant test at P < 0.05 level. For shrubs, steppes and meadows, temperature had a moderate positive effect on NDVI in spring and autumn, but negative effect in summer. Precipitation had a positive effect on NDVI during the whole growing season for steppes, and it was significant at P < 0.05 level in summer. The effect of precipitation on NDVI decreased and became weakly negative in autumn for shrubs. Precipitation had moderate positive effect on NDVI in spring, but the effect was weak in summer and autumn for meadows. There exists lag-time effect of the previous season's temperature and precipitation on NDVI both in summer and in autumn, and the effect was more obvious in autumn than that in summer and more obvious for precipitation than temperature in general. For desert vegetation, precipitation is critical to boost vegetation growth, especially in summer. Increased temperatures in spring are beneficial for vegetation growth, but in summer and autumn, it will prohibit vegetation growth, especially in summer. A lag-time effect of the previous season's precipitation and temperature on NDVI was moderate both in summer and in autumn, and in autumn the lag-time effect was even stronger than the effects brought by autumn temperature and precipitation, especially for the significant negative effect of summer temperature on autumn NDVI (r = − 0.757, P = 0.011). For broadleaf forests, temperature had a positive effect on NDVI during the whole growing season, although this was not obvious. The effect of precipitation was positive in spring and autumn but negative in summer. A lag-time effect of the previous season's precipitation and temperature on NDVI was only obvious in summer, and its effect was even stronger than that of spring temperature and precipitation.

4. Conclusions

Correlation analyses between NDVI and climate variables are powerful tools for probing ecosystem function responses to global climate change (Potter and Brooks, 1998; Piao et al., 2006).This study analysed NDVI, temperature and precipitation changes and investigated the correlations between NDVI, temperature and precipitation for eight different vegetation types during the growing seasons (April–October) of the period 1998–2007 in Inner Mongolia, China. We also analysed seasonal correlations and lag-time effects. Our main findings are summarized below.

During the growing seasons of 1998–2007, in the east wet, half humid monsoon climate district, average NDVI increased moderately for marshes, coniferous forests, broadleaf forests and cultivated vegetation. However in the west arid and semi-arid climates district, average NDVI of shrubs and meadows decreased weakly, decreased moderately for steppes, but increased weakly for desert vegetation. The average precipitation conditions of all vegetation types decreased slightly, but the average temperature conditions of all vegetation types increased and the increase is significant for shrubs, steppes, meadows and desert vegetation in arid and semi-arid climates district.

Correlation coefficients between growing season NDVI and precipitation were high during the whole growing season, and the NDVI of shrubs, steppes and cultivated vegetation was more sensitive to precipitation than temperature. The effect of precipitation on NDVI was negative for marshes, coniferous forests and broadleaf forests, but an effect was not obvious for broadleaf forests. The effect of temperature on NDVI was not as obvious as that of precipitation for all vegetation types during the whole growing season. This finding is quite different from those studies from the 1990s (Piao et al., 2006), which implies that the sensitivity of vegetation in high, cold regions to temperature may decline under global warming conditions (Jobbagy et al., 2002).

The effects of precipitation and temperature on NDVI varied among different vegetation types and seasons. In spring, both temperature and precipitation correlated with NDVI moderately, but precipitation correlated with NDVI negatively for marshes and coniferous forests. In summer, NDVI correlated with temperature negatively and precipitation positively for cultivated vegetation, shrubs, steppes, meadows and desert vegetation. However, NDVI correlated positively with temperature and negatively with precipitation for marshes, coniferous forests and broadleaf forests. In autumn, temperature had positive effects on NDVI for all vegetation types except deserts, and precipitation correlated with NDVI positively for all vegetation types with the exceptions of marshes, shrubs and coniferous forests. The differences in the observed relationships can be explained by the fact that the demand for heat and water varied among different vegetation types and seasons. Precipitation that is too much or too little, and temperatures that are too high or too low all prohibit vegetation growth. The responses to climate change were more sensitive for shrubs, steppes, meadows and desert vegetation.

Lag-time effects of the previous season's precipitation on NDVI were moderate or significant in summer for all vegetation types. In autumn, they were only obvious for shrubs, steppes, meadows and desert vegetation, and the effects were negative for marshes and coniferous forests both in summer and autumn. There were no significant lag-time effects of spring temperature on summer NDVI for any vegetation types except cultivated vegetation. There exists moderate or significant lag-time effects of the previous season's temperature on NDVI, which were obvious in autumn for most of the vegetation types, with the exceptions of marshes, coniferous forests and broadleaf forests.

The interaction effect between NDVI and climate factors is obvious. Since precipitation decreased and temperature increased more rapidly compared with the early 1990s (Piao et al., 2006; Gao et al., 2009), climate changes during 1998–2007 in Inner Mongolia seems more obvious in recent years. What is more, some studies showed NDVI correlated positively with evapotranspiration in growing seasons of north Asia (Suzuki et al., 2007). The moderate decrease NDVI for shrubs, steppes and meadows will decrease evapotranspiration into the atmosphere, which will make the arid and semi-arid climates district drier. For marshes, coniferous forests, broadleaf forests and cultivated vegetation covered district, the moderate NDVI increase will accelerate the evaporation process and may make local atmosphere wetter. For desert vegetation, although its NDVI is weakly increased, due to its low vegetation cover density and biomass, the effect to evapotranspiration is negligible. On the whole, Inner Mongolia is becoming warmer, the west arid and semi-arid climates district may become drier, air of the east wet, half humid monsoon climate district may become more moist, but since most of the area in Inner Mongolia is covered with arid and semi-arid climates, the ecological pressure in Inner Mongolia is increasing.

Acknowledgements

This research was supported by the National Social Science Foundation of China, No.10ZD&M030; the Nonprofit Industry Financial Program of Ministry of Land and Resources of China, No.200811033; a Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (geography of Nanjing University); and The National Natural Science Foundation of China (No.40801063, 40971104, 41101160).

Ancillary