Geophysical Research Letters

Increase in vegetation greenness and decrease in springtime warming over east Asia

Authors


Abstract

[1] This study investigates the impact of increased vegetation greening on the springtime temperature over east Asia for 1982–2000. An analysis of station-based temperature records and satellite-measure normalized difference vegetation index (NDVI) indicates that slight warming (<0.4°C 10-yr−1) occurred over regions that experienced large increase in NDVI (≥0.08 10-yr−1). On the contrary, strong warming (≥0.8°C 10-yr−1) occurred over regions that exhibited minor changes in NDVI (<0.04 10-yr−1). For the most part, this inverse NDVI–temperature relationship observed with the daily maximum temperature. Thus, it is suggested that the decrease in warming was mostly attributable to the increase in evapotranspiration associated with increased vegetation greening. Earlier vegetation growth may have further strengthened the effect of this vegetation–evaporation on spring temperature.

1. Introduction

[2] Approximately half of the solar insolation reaching the top of the atmosphere is absorbed at the earth's surface. In order to compensate for this large amount of surplus energy, the surface emits terrestrial radiation as well as latent and sensible heat fluxes into the atmosphere. Because of the greenhouse effect, the sum of these surface emission fluxes is even greater than the solar radiation that reaches the top of the atmosphere. Thus, the surface acts as a focal parameter that regulates the entire climate system. Vegetation greenness is widely thought to be one of the major factors that alter the surface energy budget and variations in vegetation modify the surface albedo, emissivity, and soil moisture content, which further alter the solar absorption rate and latent and sensible heat fluxes.

[3] To quantify the vegetation greenness, the satellite-measured normalized difference vegetation index (NDVI) is commonly used. NDVI values are calculated from the absorption rates for solar insolation considering that the absorption rates vary with the wavelength for different surface types, e.g., vegetation versus bare soil. Previous studies have reported that NDVI values have increased worldwide in the recent decades because of global and regional warming [Myneni et al., 1997; Zhou et al., 2001; Kaufmann et al., 2003; Gong and Ho, 2003]. It is widely expected that a warmer climate promotes biological activity and thus results in a higher NDVI values. However, this causal assumption may not be valid for some regions. For example, a higher temperature leads to drought stress if there is a lack of available surface water [Barber et al., 2000].

[4] Investigations have also been performed on the strong correlation between temperature and vegetation greenness. In particular, the influences of increased vegetation and earlier spring green-up on temperature variations have been studied [Schwartz, 1996; Kaufmann et al., 2003; Liu et al., 2006]. Currently, the general consensus on the vegetation-temperature feedback is that an increase in vegetation promotes an increase in temperature particularly during summer [Liu et al., 2006]. Nevertheless, there still exist uncertainties regarding this conclusion because the dominant physical and ecological processes affecting vegetation are controlled by many other climatic factors such as moisture availability, radiation, geographical characteristics, and human influences, and these factors vary significantly with region and season [Zhou et al., 2001; Kaufmann et al., 2003].

[5] This study focuses on springtime in east Asia and examines the influence of increased vegetation greenness on the regional warming. Many previous studies have documented an earlier onset and more active growing season over east Asia in the recent decades as a response to regional warming [Schwartz and Chen, 2002; Gong and Ho, 2003; Ho et al., 2006; Schwartz et al., 2006]. Therefore, it is important to evaluate whether the change in spring greening increases or decreases the magnitude of increase in temperature.

2. Data

[6] The NDVI data were obtained using the Advanced Very High Resolution Radiometer on the National Oceanic and Atmospheric Administration polar-orbiting satellites. These data from the NASA Goddard Space Flight Center Distributed Archive Center (http://eosdata.gsfc.nasa.gov), have a spatial resolution of 8 km × 8 km and a 10-day temporal resolution for 1982–2000. The details of the data are given by Zhou et al. [2001].

[7] The daily mean, maximum, and minimum temperatures for 150 Chinese stations and 4 Korean stations were obtained from the China Meteorological Administration and the Korea Meteorological Administration, respectively. The 154 stations are confined to the east of 100E because most of the stations to the west are located over deserts or high altitudes where NDVI values are very low (i.e., almost no vegetation) and exhibit very weak seasonal variations. In addition, urban stations (population is greater than 10 million) are excluded to minimize urbanization effect. For quality control of the station data, we have manually checked the records station by station and ensured that no unreasonable bias or missing exists in the temperature series. Data from those stations have been used in previous climate studies [Zhou et al., 2004; Jeong and Ho, 2005; Ho et al., 2006].

[8] An explanation of the temperature–NDVI relationship observed in the present study requires in-depth examination of the variables of surface latent heat flux such as evapotranspiration. Since direct observations of surface latent heat flux are unavailable, we use the data from the Japanese 25-year Reanalysis project (JRA-25) product by Japan Meteorological Agency (JMA) [Onogi et al., 2007] as substitutes. The latest numerical assimilation system and forecast system of JMA and collected observational datasets (i.e., conventional surface and upper air observations, various satellite observations, and snow data) have been used to generate a consistent and high quality reanalysis datasets for climate research. JRA-25 has a spectral resolution of T106 in horizontal and 40 layers in vertical. For land surface analysis, it uses the modified version of the Simple Biosphere (SiB) model [Sellers et al., 1986]. It shows relatively good agreements with in situ observations by the Asian Automatic Weather Station Network. The quality of surface flux data for Asia in JRA-25 is relatively high compared to other global reanalysis data [Onogi et al., 2007].

[9] If we had sufficiently long-term NDVI, temperature, and evapotranspiration observed exactly over the stations, we would easily relate the signals of regional warming to vegetation greening. However, such a fulfilled dataset is rare. Thus, we can only attempt to estimate the vegetation greening effect on regional warming using relatively abundant data sets. Because the above three data sets have different horizontal resolutions, we have to check on the same grid scale by using station temperature and NDVI aggregated onto the 1.25° × 1.25° JRA-25 grid. Compared with the station and grid-mean temperature, the sensitivity of the NDVI/temperature association to grid-mean temperature decreases to some extent. However, the conclusions are overall consistent, although the station temperature is more applicable for vegetation-temperature association in regional scales. The present study uses the data for 1982–2000, a common period for all three data sets. For a direct comparison of the three variables at the same location, the grid (or pixel) values of latent heat flux and NDVI nearest the 154 temperature stations are used.

3. Long-Term Changes in Temperature and Vegetation Over East Asia

[10] Figure 1 shows the spatial distributions of the linear trends of the (a) daily mean temperature and (b) NDVI for spring (March to May) during 1982–2000. The box in Figure 1 outlines statistically significant changes in temperature and NDVI. These trends are determined using the standard least-squares fitting method. A general warming tendency is seen over east Asia (Figure 1a). The temperature information obtained from all 154 stations reveals an increase of 0.54°C per decade. Strong warming (≥0.8°C 10-yr−1), significant at the 95% confidence level, was observed at many stations, particularly over northeastern and southern China and the Yangtze River valley. The NDVI also exhibited generally positive trends (Figure 1b). The most remarkable increases (≥0.08 10-yr−1) occurred over central eastern China (around Beijing), southern China, and South Korea. However, over northeastern China and the Yangtze River valley (around 30°N), small increases (<0.04 10-yr−1) and even decreases were observed in some places.

Figure 1.

Linear trends of (a) the mean temperature of spring (March to May) and (b) NDVI for 1982–2000. The pixel values of NDVI collected at each of the 154 stations are calculated. The units are °C 10-yr−1 for temperature and 10-yr−1 for NDVI. The boxes outlines the stations where the trends are significant at the 95 % significance levels based on the Student's t-tests.

[11] It is notable that the regions that experienced the most warming were also those that experienced the smallest increases in NDVI (e.g., northeastern China and the Yangtze River valley). On the contrary, small increases in temperature associated with large increases in NDVI were observed over central eastern China, south western China (Hubei and Hunan province), and South Korea. Considering that both temperature and NDVI generally show distinctly positive trends, this inverse relationship found on a regional scale seems to contradict the general expectation that the more warming result in the more vegetation greening.

[12] To explain the inverse relationship between temperature and NDVI shown in Figure 1, we present a scatter plot of the linear trend of temperature and evapotranspiration against that of NDVI for the 154 stations (Figure 2). Here, three temperature values and evapotranspiration are analyzed: (a) daily maximum temperature, (b) daily mean temperature, (c) daily minimum temperature, and (d) evapotranspiration. In general, most warming trends were distributed in the range of temperature change between 0.2°C per decade and 1.0°C per decade, and most of the corresponding NDVI increases were less than 0.05 per decade. In contrast, relatively small increases in warming were found over the stations that exhibit large increases in NDVI (>0.05 per decade). The temperature values, particularly the maximum temperature values, at some locations even decreased during the analysis period.

Figure 2.

Scatter plots of the linear trends of (a) daily maximum temperature, (b) mean temperature, (c) minimum temperature, and (d) evapotranspiration against the linear trend of NDVI for the 154 stations. Closed circles denote the NDVI changes that are significant at the 95% confidence level based on the Student's t-tests.

[13] More specifically, the plots of the linear trends of the maximum and mean temperatures against NDVI show negative slopes (Figures 2a and 2b). Highly negative slopes correspond to the stations where NDVI increases significantly (closed circles). The slopes, −4.2°C and −3.9°C per unit NDVI for maximum and mean temperature, respectively, are both significant at the 95% confidence level based on Student's t-test. On the contrary, a negligible relationship was observed between the minimum temperature and NDVI (Figure 2c). Thus, it can be concluded that the local inverse temperature–NDVI relationship shown in Figure 1 is due mostly to the changes in the maximum temperature.

[14] This observation leads to the physical speculation that the daytime temperature range is reduced by increased vegetation via evapotranspiration. Namely, increased evapotranspiration lowers the local maximum temperature. Although there is counter effect the enhanced vegetation rather warms the surface through secreased albedo effect, however, present results indicate that vegetation–evapotranspiration effect is more dominant in east Asia as simulated by Bounoua et al. [2000]. A similar conclusion has also been drawn by other observational analyses [Bonan, 2001; Kaufmann et al., 2003]. Presumably the increase in evaporation induces an increase in precipitation through local moisture recycling, which may strengthen the more greenness and less warming relationship shown in the present results. The coupled land-atmosphere model result of Bounoua et al. [2000] has indicated that an increase in NDVI results in both evapotranspiration and precipitation over east Asia during the growing season. To show the dominance of the evapotranspiration process, we show a scatter plot of evapotranspiration against the changes in NDVI (see Figure 2d). There is a clear relationship between the two as shown by the slope of 3.2 W m−2 per unit NDVI, which is significant at the 95% confidence level. Interestingly, positive trends are shown in all evapotranspiration values taken from the areas where changes in NDVI are significant.

[15] One thing to note in the temperature and NDVI changes (Figures 1 and 2) is the possible effect of land-use changes. The minor increase in NDVI over northeastern China and the Yangtze River valley partly results from increased cultivated area [Piao et al., 2003] and it has been reported that the urbanization in the northern and the east coast of China has caused a sharp decrease or slight increase in NDVI [Piao et al., 2003; Zhou et al., 2004]. In spite of excluding urban stations from analysis, still these effects might have contributed the NDVI-temperature changes to some degree. Also it can be speculated that weak warming over some coastal regions is showing an effect of cooling due to increased aerosols rather than mainly an effect of increased vegetation. However, previous studies suggested that neither the direct effect [Gong et al., 2006] nor the indirect effect [Choi et al., 2008] of aerosols have caused maximum temperature change at a significant level. Admitting some minor effect of aerosol on temperature variation, it seems not to overrule the NDVI-temperature relationship in the present result.

4. Vegetation Changes and the Seasonal Cycle of Temperature in Spring

[16] To further verify whether the effect of the vegetation–evapotranspiration effect is more dominant than that of vegetation-albedo effect, we analyze the seasonal cycle of temperature. In spring, vegetation usually starts to grow as the cumulated winter heat exceeds the minimum level that activates vegetation growth. As new leaves emerge and grow, the amount of evapotranspiration increases if sufficient moisture is available affecting the variation of surface temperature. Hence, the vegetation–evapotranspiration effect on temperature may be strongly reflected in the seasonal cycle of temperature.

[17] Figure 3a shows the mean seasonal cycles of maximum (upper line) and minimum (lower line) temperatures from 40 days before (−40 days) to 40 days after (+40 days) the date of leaf growth, which is averaged for all the stations. The date of the onset of leaf growth at each station in each year is determined by the method proposed by Reed et al. [1994], detecting the inflection point (date) when the NDVI time-series begins to ascend for the specific year. There is a distinct change in the maximum temperature time-series on 10 April (average date of the onset of leaf growth for all the stations), while the minimum temperature shows a nearly continuous increase. This observation agrees with the expectation that the minimum temperature progressively increases with radiation because vegetation is inactive at night, whereas after the leafing out of the vegetation, the increase in the daily maximum temperature is limited by the transpirational cooling of the plant [Schwartz and Karl, 1990; Schwartz, 1996; Levis and Bonan, 2004]. The maximum surface temperature increases rapidly until the date of the onset of leaf growth (at least 0.25°C per day) and then drops (to below 0.11°C per day). The observed reduction about 0.14°C per day is consistent with other observational studies, which have revealed maximum temperature reduction after leaf emergence at diverse vegetated regions [Schwartz and Karl, 1990; Schwartz, 1996; Levis and Bonan, 2004]. Thus, it can be claimed that the vegetation–evapotranspiration effect dominates over the vegetation–albedo effects. In addition, the spring maximum temperature is also influenced by other factors such as snow cover, cloud, vegetation type, and geographical characteristics. These factors can be the potential reasons for springtime maximum temperature discontinuity. However, they only play a minor role in the maximum temperature discontinuity over east Asia as shown in the model simulation by Levis and Bonan [2004] and the current study.

Figure 3.

(a) Averaged time-series of daily maximum (upper line) and minimum (lower line) temperatures from 40 days before to 40 days after the onset of leaf growth averaged for all the 154 stations. Dashed and dotted lines represent the linear regression slopes of the temperatures for the 40 days before and the 40 days after the date of the onset of leaf growth, respectively. (b) The same plot as Figure 3a except the daily maximum temperature time-series is averaged over the first five years of the analysis period (1982–1986) (black lines) and over the last five years of the analysis period (1996–2000) (gray lines).

[18] We further investigate the effects of temperature warming trends on the seasonal march of temperature over the entire analysis period (1982–2000). Figure 3b shows the comparison of the averages of the time-series of daily maximum temperature from −40 days to +40 days over the first five years (1982–1986) with those averaged over the last five years (1996–2000). It is clear that the onset of the vegetation growing season has advanced by 7 days. This feature is consistent with the results of previous studies that have reported an advanced onset of the vegetation growth in east Asia [Schwartz and Chen, 2002; Gong and Ho, 2003; Ho et al., 2006; Schwartz et al., 2006]. In addition to this earlier onset, the seasonal cycle in the last five years shows a lower rate of temperature increase after the onset of leaf growth than does the seasonal cycle in the first five years. However, there is no notable difference between the rates of temperature increase before the onset of leaf growth. This feature suggests that the recent warming has led to an increase in vegetation greenness and earlier onset of leaf growth, which in turn results in a stronger influence on temperature via the vegetation–evapotranspiration effect.

5. Summary and Discussion

[19] In this study, we have analyzed the long-term trends of daily mean, maximum, and minimum temperatures over east Asia with respect to the NDVI change in spring during 1982–2000. Low warming rates were observed over regions that experienced significant changes in vegetation, whereas higher warming rates were found over regions that vegetation changes were insignificant. This inverse vegetation–temperature relationship is mainly linked to the changes in daily maximum temperature. It is suggested that the decrease in warming rates is due to increased vegetation greening through increase in evapotranspiration. In addition, an advanced onset of vegetation growth appears to strengthen the inverse vegetation–temperature relationship. Considering the recent greening trends in the Northern Hemisphere, we can apply our results to future climate simulations. Inclusion of the vegetation greening trends in model simulations for future climate simulations may decrease the amplitude of global and/or regional warming through the cooling effect by vegetation greening. In a future study, we will further verify the results of the present study using a climate model coupled with a dynamic global vegetation model.

Acknowledgments

[20] This research was funded by the Korea Meteorological Administration Research and Development Program under grant CATER 2006-4204. The first author is also supported by the BK21 project of the Korean government. Jee-Hoon Jeong's contribution is the contribution No. 23 from TELLUS, the Centre of Earth System Science at University of Gothenburg. The authors sincerely appreciate the critical and valuable comments made by Praveen Kumar and two anonymous reviewers.

Ancillary