International Journal of Biometeorology

, Volume 56, Issue 4, pp 695–706

Phenological responses of Ulmus pumila (Siberian Elm) to climate change in the temperate zone of China

Authors

    • College of Urban and Environmental Sciences, Laboratory for Earth Surface Processes of the Ministry of EducationPeking University
  • Lin Xu
    • College of Urban and Environmental Sciences, Laboratory for Earth Surface Processes of the Ministry of EducationPeking University
Original Paper

DOI: 10.1007/s00484-011-0471-0

Cite this article as:
Chen, X. & Xu, L. Int J Biometeorol (2012) 56: 695. doi:10.1007/s00484-011-0471-0

Abstract

Using Ulmus pumila (Siberian Elm) leaf unfolding and leaf fall phenological data from 46 stations in the temperate zone of China for the period 1986–2005, we detected linear trends in both start and end dates and length of the growing season. Moreover, we defined the optimum length period during which daily mean temperature affects the growing season start and end dates most markedly at each station in order to more precisely and rationally identify responses of the growing season to temperature. On average, the growing season start date advanced significantly at a rate of −4.0 days per decade, whereas the growing season end date was delayed significantly at a rate of 2.2 days per decade and the growing season length was prolonged significantly at a rate of 6.5 days per decade across the temperate zone of China. Thus, the growing season extension was induced mainly by the advancement of the start date. At individual stations, linear trends of the start date correlate negatively with linear trends of spring temperature during the optimum length period, namely, the quicker the spring temperature increased at a station, the quicker the start date advanced. With respect to growing season response to interannual temperature variation, a 1°C increase in spring temperature during the optimum length period may induce an advancement of 2.8 days in the start date of the growing season, whereas a 1°C increase in autumn temperature during the optimum length period may cause a delay of 2.1 days in the end date of the growing season, and a 1°C increase in annual mean temperature may result in a lengthening of the growing season of 9 days across the temperate zone of China. Therefore, the response of the start date to temperature is more sensitive than the response of the end date. At individual stations, the sensitivity of growing season response to temperature depends obviously on local thermal conditions, namely, either the negative response of the start date or the positive response of the end date and growing season length to temperature was stronger at warmer locations than at colder locations. Thus, future regional climate warming may enhance the sensitivity of plant phenological response to temperature, especially in colder regions.

Keywords

Phenological growing seasonUlmus pumilaLinear trendResponse to temperatureSensitivityClimate change

Introduction

A true growing season may be defined as the period of time in a year in which a tree or a crop can grow (Wang 1963). This kind of the growing season, known as the phenological growing season, is usually identified by the number of days between the spring bud burst/leaf unfolding date and the autumn leaf coloration/leaf fall date of trees (Schnelle 1973; Chen 1994; Chmielewski and Rötzer 2001; Menzel 2003). Efforts to determine the growing season can be traced back to the middle of the twentieth century. At that time, the goals of growing season identification were to understand the seasonal growth and development cycles of trees and crops, schedule agricultural activities, as well as introduce new crop varieties from other regions (Nuttonson 1953; Schnelle 1955). Over recent decades, scientists have found new relevance for phenological studies related to global climate change. With respect to interactions between terrestrial biological processes and atmospheric physical processes, since phenological growing season start and end dates are influenced mainly by seasonal climatic conditions, they are known as sensitive, easily observable, and integrated indicators of climate change (Chen 1995; Menzel and Fabian 1999). On the other hand, the phenological growing season could also regulate regional climate through altering albedo, latent and sensible heat exchange, and turbulence, etc. (Schwartz 1996; Peñuelas et al. 2009). Considering interactions between terrestrial biological processes and atmospheric chemical processes, the duration of phenological growing season can influence the seasonal pattern of the atmospheric CO2 concentration (Keeling et al. 1996) and the total annual emission of biogenic volatile organic compounds (BVOCs, Peñuelas et al. 2009), which contribute to many complex processes associated with global climate change. Further, elevated CO2 might also influence the phenological growing season through delaying flowering and greening in grasses (Cleland et al. 2006). On the basis of the above evidence, the phenological growing season driven by climate may influence energy and moisture exchange, seasonal carbon cycle and aerosol formation between land surface and atmosphere. Therefore, revealing spatiotemporal patterns of the phenological growing season at regional scales would be helpful in detecting the responses and feedbacks of vegetation dynamics to climate change.

Many studies have shown a significant advancement of phenological events in spring and a less pronounced delay of phenological events in autumn on the basis of multi-species observations in temperate and Mediterranean climate regions across Europe and North America (Ahas 1999; Bradley et al. 1999; Menzel and Fabian 1999; Beaubien and Freeland 2000; Chmielewski and Rötzer 2001; Defila and Clot 2001; Fitter and Fitter 2002; Zhao and Schwartz 2003; Gordo and Sanz 2009), whereas a study in Northwestern Russia indicated a nonsignificant advancement or delay of spring phenophases but significant advancement or delay of autumn phenophases (Kozlov and Berlina 2002). At the same time, several studies have focused on the growing season of specific trees and obtained similar results in Europe (Chmielewski and Rötzer 2001; Menzel 2003; Gordo and Sanz 2009), although a contradictory result came from Japan where the advancement rate in the start date of the Ginkgo biloba growing season was less than the delay rate in the end date (Matsumoto et al. 2003). With regard to external factors of phenological variation of trees, most studies have indicated that occurrence dates of spring phenophases were triggered mainly by previous mean temperatures (Chen 1994; Chmielewski and Rötzer 2001; Menzel 2003; Matsumoto et al. 2003; Gordo and Sanz 2010), whereas a few studies have shown that occurrence dates of autumn phenophases were also influenced by previous mean temperatures (Menzel 2003; Matsumoto et al. 2003; Gordo and Sanz 2010). Therefore, the current study is focused only on phenological responses to temperature.

So far, almost all land surface phenological studies in China have been based on discontinuous time series from a few stations in a data set of the Chinese Academy of Sciences (Lu et al. 2006; Zheng et al. 2006). In order to reveal phenological performances of broad geographic coverage and continuous time series in detail, with permission from the Chinese Meteorological Administration (Chen 2009), we have recently digitized the largest phenological data set in China. Using the new data set, we selected a representative plant species, Ulmus pumila (Siberian Elm) to analyze the temporal variability of the growing season and its relation to temperature. The objectives of this study were to: (1) detect linear trends of the beginning date (BGS), end date (EGS) and length (LGS) of the Ulmus pumila growing season at individual stations and across the temperate zone of China; (2) determine spatial dependence of growing season linear trends on temperature linear trends at individual stations; (3) examine the response of the growing season to interannual temperature variation at individual stations and across the temperate zone of China; and (4) identify spatial dependence of the growing season response to temperature at individual stations.

Materials and methods

Study area and plant species

The study area is located in the temperate zone of China, including warm, middle and cold temperate zones from south to north (Domrös and Peng 1988). The dominant vegetation types include deciduous conifer forest, deciduous broad-leaved and coniferous mixed forest, deciduous broad-leaved forest, temperate steppe, temperate desert, etc. (Compilation Committee of the Vegetation of China 1980). Because of the remarkable seasonality in temperature and precipitation, vegetation aspects are rich, colorful, and highly variable. Therefore, this area is suitable for detecting phenological responses of trees to climate change (Chen 2003). Ulmus pumila (Siberian Elm) is one of the major tree species for timber, shelter, food and fodder in East Asia, and has a natural distribution from the northern subtropical zone to the cold temperate zone that ranges from approximately at 32° to 53° north latitude, 74° to 134° east longitude, and 3 m to 3,650 m above see level. In this range, the annual mean temperature is about −4°C to +15°C and annual precipitation is between 16 mm and 1,200 mm (Ma 1989). Thus, Siberian Elm can serve as an “indicator species” for phenological study according to the criteria proposed by Newman and Beard (1962). Recently, Ghelardini and Santini (2009) found a good relationship between bud burst date of some European elm species and winter-spring temperature. This shows that phenology of the genus Ulmus can be used as a sensitive indicator of climate change. Therefore, U. pumila is an appropriate and representative choice of sample species for studying the phenological growing season and its response to climate change.

Phenological and climate data

Phenological data for Ulmus pumila were acquired from the Chinese Meteorological Administration (Chen 2009). In order to ensure data quality, the occurrence dates of all Ulmus pumila phenophases (day of year, DOY) at each station in the temperate zone of China were verified systematically according to the inherent sequence and the linear correlation between time series of these phenophases (Chen et al. 2005). Then, we defined the U. pumila growing season as the period between the start date of leaf unfolding and the end date of leaf fall. The start of leaf unfolding was defined as when a few leaves are fully spread in spring, whereas leaf fall end was determined when almost all leaves had fallen to the ground (Chinese Meteorological Administration 1993). Moreover, we chose 46 stations with phenological time series of more than 16 years duration during the period 1986–2005 for the analysis. These stations were distributed mostly across all of the temperate zone of China, except desert and high mountain areas (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00484-011-0471-0/MediaObjects/484_2011_471_Fig1_HTML.gif
Fig. 1

Location of phenological stations in the temperate zone of China

Air temperature data were acquired from the Chinese Meteorological Data Sharing Service System (http://cdc.cma.gov.cn/), including daily mean air temperature at 343 stations in the temperate zone of China from 1986 to 2005. Since several phenological stations were not located close to meteorological stations (Table 1), we used ANUSPLIN 4.2 (Hutchinson 2002) and Digital Elevation Model (DEM) data derived from the United States Geological Survey to interpolate the daily mean air temperature into 8 km × 8 km grids over the temperate zone of China. Thus, we obtained gridded daily mean air temperature data at the phenological stations without meteorological observations.
Table 1

Linear trends (days per decade) and significance levels of parameters of the Ulmus pumila growing season at each phenological station. BGS beginning date, EGS end date, LGS length of growing season

Station number

Number of years

BGS

Number of years

EGS

Number of years

LGS

1a

18

0.28

18

−4.62

18

−4.90

2a

20

−0.04

20

−0.14

20

−0.11

3a

20

−3.26*

20

0.35

20

3.62

4

19

1.05

19

0.25

19

−0.81

5a

20

−0.17

20

−0.75

20

−0.58

6

18

8.30

20

5.80**

18

−2.50

7

20

−4.38

20

5.12

20

9.42

8

20

−4.17*

20

1.09

20

5.26

9a

20

−8.51*

20

4.29

20

12.80*

10a

20

0.92

19

−0.90

19

−2.23

11a

20

−6.52**

20

−5.34*

20

1.18

12a

19

−8.47***

19

3.75

19

12.23**

13a

20

−3.44

19

5.30**

19

9.30**

14a

20

−4.05

19

−8.35*

19

−3.69

15a

20

−5.03

20

3.32

20

8.35*

16a

16

−11.99***

16

1.91

16

13.90**

17a

20

−7.38***

20

0.05

20

7.44*

18a

20

−3.69

20

3.85

20

7.54

19a

20

−0.65

20

−1.46

20

−0.80

20a

20

3.47

20

1.44

20

−2.02

21a

20

−0.26

20

−6.20

20

−5.93

22a

19

−4.23

19

3.56*

19

7.79*

23a

19

−7.77**

19

−3.89

19

3.88

24a

20

−5.98*

20

−2.36

20

3.62

25a

19

−5.11*

19

−0.58

19

4.53

26

20

1.77

20

−2.38

20

−4.15

27

20

2.14

20

6.30*

20

4.16

28a

20

−2.50

20

2.54

20

5.05

29a

20

3.58

20

6.92***

20

3.35

30a

20

−5.92**

20

−0.43

20

5.49

31a

20

−5.92*

20

3.90*

20

9.83**

32a

20

−8.63***

20

−0.55

20

8.08**

33a

20

−2.41

20

2.17

20

4.59

34a

20

−5.30*

20

5.56*

20

10.86**

35a

20

−16.88***

19

4.82*

19

21.46***

36a

19

2.46

19

6.90

19

4.45

37

19

2.33

19

1.47

19

−0.86

38

16

−2.97

16

2.96

16

5.93

39

20

1.21

20

8.91*

20

7.70

40

16

−5.01

16

3.81

16

8.78

41a

20

−1.44

20

−4.63

20

−3.19

42

16

−11.97***

16

10.82***

16

22.79***

43a

20

−9.80***

20

2.52

20

12.32*

44a

20

−7.74***

20

6.33**

20

14.08***

45

16

−6.28**

16

2.97

16

9.25*

46a

20

−6.47**

19

0.46

19

7.08*

* P < 0.05; ** P < 0.01; *** P < 0.001

aPhenological station with meteorological observations

Statistical model

The traditional method for detecting phenological responses to temperature is to compute mean temperature during several months, normally including the month in which the mean phenological event occurred as well as the preceding months (Chen 1994; Sparks et al. 2000; Chmielewski and Rötzer 2001; Menzel 2003; Gordo and Sanz 2010). Using preceding monthly mean temperature as the independent variable is simple and practical but may not be precise enough because the phenological event is not likely induced by the integral monthly mean temperature exactly, but by daily mean temperature during a certain length period (LP, days). Matsumoto et al. (2003) used daily mean temperature during an optimum LP prior to the mean occurrence date of a phenological event to estimate phenological response to temperature. This method is detailed and precise but may not be sufficiently rational because using the mean occurrence date of the phenological event, as the end date of LP, does not consider temperature effects in the period after the mean occurrence date of the phenological event, and excludes temperature effects in the period before the flexible start date of LP on earlier phenological occurrence dates. As an alternative method, we propose a new approach to identify phenological response to daily mean temperature.

The basic hypothesis of our statistical model is that the occurrence dates of a phenological event are influenced mainly by mean temperature within a particular LP during and before its occurrence. In order to determine the optimum LP during which mean temperature affects the U. pumila BGS/EGS most markedly at a station, we first calculated the interval between the earliest and latest date in BGS/EGS time series at the station, respectively, as the basic LP (bLP). Then, we computed the mean temperature time series during the basic LP plus a moving LP (mLP) prior to the earliest date in BGS/EGS time series by a step length of 1 day at the station from 1986 to 2005, namely, during bLP + 1 day, bLP + 2 day, bLP + 3 day, etc. The maximum mLP was limited to 60 days. Thus, the LP is defined as follows:
$$ LP = bLP + mLP $$
Moreover, we calculated correlation coefficients between BGS/EGS time series and mean temperature time series during the different LP at the station, respectively. Finally, we obtained the optimum LP with the largest correlation coefficient between BGS/EGS and mean temperature at each station. Figure 2 shows an example of determining the optimum spring and autumn LP at station 28. The curves illustrate variation of correlation coefficients between the BGS/EGS time series and mean temperature time series during the different LP. The optimum LP with the largest correlation coefficient is 32 days for BGS (negative correlation) and 63 days for EGS (positive correlation). Figure 3 shows the optimum LP (bLP + mLP) at the 46 stations. The above four-step procedure for looking for the optimum LP at individual stations can also be applied to the entire region. In the latter case, the regional mean BGS/EGS time series and regional daily mean temperature time series (based on the 46 stations) from 1986 to 2005 were used. Furthermore, the annual mean temperature at individual stations and in the entire region was used as the independent variable in the statistical model of LGS.
https://static-content.springer.com/image/art%3A10.1007%2Fs00484-011-0471-0/MediaObjects/484_2011_471_Fig2_HTML.gif
Fig. 2

Schematic demonstration of determination of the optimum a spring length period (LP) and b autumn LP based on correlation coefficients between BGS/EGS time series and mean temperature time series during the different LPs at station 28

https://static-content.springer.com/image/art%3A10.1007%2Fs00484-011-0471-0/MediaObjects/484_2011_471_Fig3_HTML.gif
Fig. 3

Optimum spring LP [black bars basic LP (bLP), gray bars moving LP (mLP)] and autumn LP (black bars bLP, gray bars mLP) at each station

Results

Linear trend of the growing season

At individual stations, we found a significant advancing trend in BGS (P < 0.05) at 20 stations (44%) during 1986 to 2005, but no significant delay. The significant trends of BGS ranged from −3.26 days per decade at station 3 to −16.88 days per decade at station 35. However, a significant delayed trend in EGS (P < 0.05) appeared only at 11 stations (24%) while a significant advancing trend (P < 0.05) occurred at 2 stations (4%). The significant trends of EGS ranged from −8.35 days per decade at station 14 to 10.82 days per decade at station 42. Furthermore, we identified a significant lengthening in LGS at 16 stations (35%) but found no significant shortening. The significant trends of LGS ranged from 7.08 days per decade at station 46 to 22.79 days per decade at station 42 (Table 1).

Across the temperate zone of China, BGS advanced significantly at a rate of −4.0 days per decade (P < 0.01), whereas EGS was significantly delayed at a rate of 2.2 days per decade (P < 0.01) and LGS was significantly prolonged at a rate of 6.5 days per decade (P < 0.001) during 1986 to 2005.

Response of the growing season to interannual temperature variation

At individual stations, a significantly negative correlation between BGS and the optimum spring LP temperature was apparent at 41 stations (89%), namely, the higher the spring LP temperature in a year, the earlier the BGS. Slopes of significant linear regression lines between BGS and spring LP temperature ranged from −1.02 days °C−1 at station 10 to −7.63 days °C−1 at station 35 (Table 2). However, a significantly positive correlation between EGS and the optimum autumn LP temperature was found at only 12 stations (26%), where the higher the autumn LP temperature in a year, the later the EGS. Slopes of significant linear regression lines between EGS and autumn LP temperature ranged from 1.96 days °C−1 at station 3 to 7.65 days °C−1 at station 12 (Table 2). Moreover, a significantly positive correlation between LGS and annual mean temperature existed at 21 stations (46%), at which the higher the annual mean temperature in a year, the longer the growing season. Slopes of significant linear regression lines between LGS and annual mean temperature ranged from 5.75 days °C−1 at station 27 to 19.69 days °C−1 at station 35 (Table 2).
Table 2

Correlation and regression analysis between BGS-spring LP temperature, EGS-autumn LP temperature, and LGS-annual mean temperature at each phenological station

Station number

BGS

EGS

LGS

r

Slope (days °C-1)

r

Slope (days °C-1)

r

Slope (days °C-1)

1

−0.76

−1.95***

−0.16

−0.88

−0.37

−4.11

2

−0.37

−1.28

−0.36

−0.70

0.20

1.30

3

−0.86

−2.47***

0.50

1.96*

0.40

3.57

4

−0.56

−1.65*

0.35

1.83

−0.25

−2.00

5

−0.59

−1.46**

0.35

1.83

0.30

2.87

6

−0.42

−2.54*

0.55

2.08*

0.29

6.25

7

−0.65

−4.12**

0.46

3.17*

0.61

13.78**

8

−0.82

−2.28***

0.41

3.13

0.53

9.21*

9

−0.74

−4.35***

−0.19

−0.83

0.51

12.46*

10

−0.46

−1.02*

0.60

4.50**

0.47

8.03*

11

−0.85

−4.63***

0.20

1.78

0.58

6.41**

12

−0.56

−2.79*

0.47

7.65*

0.46

9.72*

13

−0.68

−2.62**

0.41

2.66

0.46

9.00*

14

−0.69

−3.18***

−0.43

−2.09

0.12

2.23

15

−0.51

−4.61*

−0.28

−1.92

0.54

10.14*

16

−0.82

−6.37***

0.39

3.92

0.73

14.12**

17

−0.42

−2.25

0.25

1.21

0.34

4.72

18

−0.83

−3.99***

0.42

3.18

0.49

8.02*

19

−0.57

−2.48**

0.34

0.56

0.38

4.97

20

−0.50

−4.38*

0.27

3.26

0.37

8.85

21

−0.52

−2.27*

0.29

2.86

0.39

6.58

22

−0.79

−3.36***

0.39

1.72

0.65

7.96**

23

−0.80

−4.00***

−0.12

−1.35

0.14

2.03

24

−0.75

−4.20***

−0.24

−2.05

0.32

2.45

25

−0.76

−2.73***

0.24

1.97

0.42

7.12

26

−0.60

−1.61**

−0.20

−0.71

0.05

0.77

27

−0.61

−2.33**

0.42

2.77

0.50

5.75*

28

−0.69

−2.51***

0.51

2.67*

0.55

7.62*

29

−0.38

−2.18

0.71

5.20***

0.40

7.68

30

−0.75

−3.55***

0.30

1.36

0.34

8.80

31

−0.74

−3.97***

0.53

2.22*

0.49

9.62*

32

−0.76

−4.17***

0.31

1.89

0.39

5.15

33

−0.79

−3.94***

0.25

0.84

0.41

5.29

34

−0.81

−2.98***

0.61

4.34**

0.54

9.15*

35

−0.66

−7.63**

0.18

1.52

0.60

19.69**

36

−0.46

−2.64*

0.54

4.76*

0.60

10.09**

37

−0.67

−3.39**

0.19

0.95

0.52

7.27*

38

−0.65

−3.29**

0.31

2.54

0.51

10.21*

39

−0.42

−4.17

0.29

2.75

0.09

3.12

40

−0.81

−5.10***

0.53

2.30*

0.48

11.34

41

−0.33

−2.69

0.09

0.58

0.08

−0.48

42

−0.82

−6.58***

0.31

4.35

0.33

15.29

43

−0.85

−5.88***

0.40

6.29

0.43

14.44

44

−0.86

−4.46***

0.41

2.12

0.65

13.99**

45

−0.89

−2.94*

0.63

6.42**

0.68

13.30**

46

−0.54

−2.68*

−0.25

−1.74

0.38

5.65

* P < 0.05; ** P < 0.01; *** P < 0.001

For all of the temperate zone of China, we calculated regional linear regression equations between the regional mean BGS/EGS time series and the regional optimum spring/autumn LP (73 days for BGS, 65 days for EGS) temperature time series, and between the regional mean LGS time series and the regional annual mean temperature time series during 1986 to 2005. The result shows that the regional mean BGS correlates negatively with the regional optimum spring LP temperature, and the regional mean EGS and LGS correlate positively with the regional optimum autumn LP temperature and regional annual mean temperature, respectively. It is worth noting that the correlation coefficient for BGS is much larger than those for EGS and LGS. The regional regression equations indicate that a 1°C increase in the regional optimum spring LP temperature may induce an advancement of 2.8 days in the regional mean BGS; a 1°C increase in the regional optimum autumn LP temperature may cause a delay of 2.1 days in the regional mean EGS; while a 1°C increase in the regional annual mean temperature may result in an extension of 9.0 days in the regional mean LGS (Fig. 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs00484-011-0471-0/MediaObjects/484_2011_471_Fig4_HTML.gif
Fig. 4

Correlation and regression analysis between a regional mean beginning date (BGS) and regional spring length period (LP) temperature (Ts), b regional mean end date (EGS) and regional autumn LP temperature (Ta), and c regional mean length of growing season (LGS) and regional annual mean temperature (T) from 1986 to 2005

Spatial dependence of growing season linear trends on temperature linear trends

Taking the linear trends of BGS, EGS and LGS (Table 1) and the linear trends of the corresponding temperature indicators (not shown) at each station as correlative variables, we found that the linear trend of BGS correlates negatively with the linear trend of the optimum spring LP temperature (r = −0.46, P < 0.01, n = 46). The dots below the dashed horizontal line in Fig. 5 indicate that the quicker the optimum spring LP temperature increased at a station, the quicker the BGS advanced. In contrast with BGS, linear trends of EGS and LGS did not correlate significantly with linear trends of the optimum autumn LP temperature and the annual mean temperature, respectively (P > 0.1, not shown).
https://static-content.springer.com/image/art%3A10.1007%2Fs00484-011-0471-0/MediaObjects/484_2011_471_Fig5_HTML.gif
Fig. 5

Relationship between linear trend of BGS and linear trend of spring LP temperature at 46 stations

With respect to the dependence of growing season linear trends on temperature linear trends across the temperate zone of China, a larger positive trend in the optimum spring LP temperature (1.23°C per decade, P < 0.01) corresponds to a larger negative trend in BGS (−4.0 days per decade, P < 0.01), and a smaller positive trend in the optimum autumn LP temperature (0.5°C per decade, P < 0.1) matches a smaller positive trend in EGS (2.2 days per decade, P < 0.01). In addition, a significantly positive trend in annual mean temperature (0.44°C per decade, P < 0.01) is linked to a significantly positive trend in LGS (6.5 days per decade, P < 0.001) (not shown).

Spatial dependence of the growing season response to temperature

As mentioned above, the regression slopes (days °C−1) between growing season parameters (BGS, EGS, LGS) and temperatures show obvious differences among the stations (Table 2), which may relate to thermal condition differences among the stations. In order to detect the spatial dependence of the growing season response to temperature, we used the long-term annual mean temperature from 1986 to 2005 as the indicator of thermal conditions at individual stations, and carried out a correlation analysis between growing season-temperature regression slopes (Table 2) and long-term annual mean temperatures at the 46 stations. The result shows that a negative correlation was found between BGS-spring LP temperature regression slope and long-term annual mean temperature at individual stations (P < 0.1), whereas a positive correlation was detected between EGS-autumn LP temperature regression slope and long-term annual mean temperature at individual stations (P < 0.05). Generally speaking, either the negative response of BGS to spring LP temperature or the positive response of EGS to autumn LP temperature (dots above the dashed horizontal line) was stronger at warmer locations than at colder locations. In addition, a positive correlation appeared between LGS-annual mean temperature regression slope and long-term annual mean temperature at individual stations (P < 0.001), namely, the positive response of LGS to annual mean temperature (dots above the dashed horizontal line) was also stronger at warmer locations than at colder locations (Fig. 6).
https://static-content.springer.com/image/art%3A10.1007%2Fs00484-011-0471-0/MediaObjects/484_2011_471_Fig6_HTML.gif
Fig. 6

Relationship between growing season response to temperature and annual mean temperature at 46 stations. a BGS, b EGS, c LGS

Discussion

In this study, the Ulmus pumila BGS indicates a significant advancement at 44% of all stations from 1986 to 2005, whereas the U. pumila EGS shows a significant delay at only 24% of these stations. In addition, the advancement rate in the regional mean BGS (−4.0 days per decade) is much larger than the delay rate in the regional mean EGS (2.2 days per decade) across the temperate zone of China. Thus, the growing season extension was induced mainly by the advancement of BGS. This result is consistent with field-based phenological observations of temperate zone trees worldwide, especially the significant advancement of phenological events in spring and the less pronounced delay of phenological events in autumn across Europe and North America (Ahas 1999; Bradley et al. 1999; Menzel and Fabian 1999; Beaubien and Freeland 2000; Chmielewski and Rötzer 2001; Fitter and Fitter 2002; Menzel 2003; Gordo and Sanz 2009). Therefore, our results provide new evidence for the estimate on an overall spring advancement across the northern hemisphere (Parmesan and Yohe 2003; Root et al. 2003; Parmesan 2007).

In order to explain the spatial variation of growing season linear trends among the stations, we conducted a correlation analysis between phenological trends and temperature trends at the 46 stations. A significantly negative relationship was found between BGS trends and temperature trends, but there was no significant relationship between EGS trends and temperature trends. This finding supports the result from Ginkgo biloba phenology in Japan (Matsumoto 2009). Therefore, phenological linear trends in spring are influenced mainly by spring temperature linear trends at individual stations, namely, the quicker the spring temperature increased at a station, the quicker the spring phenophase advanced.

The correlation analysis between U. pumila BGS/EGS and optimum spring/autumn LP temperature from 1986 to 2005 across the temperate zone of China showed results consistent with previous studies in other regions (Chen 1994; Chmielewski and Rötzer 2001; Menzel 2003), namely, the response of spring phenophases (such as leaf unfolding and flowering) to temperature is stronger than that of autumn phenophases (such as leaf coloration and leaf fall). At individual stations, we found a significantly negative correlation between U. pumila BGS and spring temperature at 89% of all stations, whereas a significantly positive correlation between EGS and autumn temperature occurred only at 26% of these stations. This implies that leaf unfolding was triggered mainly by seasonal temperature in most parts of the temperate zone of China but leaf fall was influenced also by other climatic factors, such as continuous low temperature (Heide and Prestrud 2005; Dufrêne et al. 2005), photoperiod (Kramer 1994; Delpierre et al. 2009), precipitation (Chen 1994; Chen and Pan 2002; Gordo and Sanz 2009) and wind (Chen et al. 2005), which could also explain why linear trends of EGS did not correlate significantly with linear trends of autumn temperature at the 46 stations. More detailed studies on the mechanism of autumn phenophases are required.

To explain the spatial variation of phenological response to temperature among the stations, we used long-term annual mean temperature as the indicator of local thermal condition to analyze the relationship between growing season-temperature regression slopes (Table 2) and long-term annual mean temperatures at the 46 stations, and found that both the negative response of BGS to spring temperature and the positive response of EGS to autumn temperature were stronger at warmer locations than at colder locations. This finding supports the results that the temporal response of plant phenology to temperature in warmer countries (spring phenophases) or at lower latitudes (autumn phenophases) is stronger than in colder countries or at higher latitudes (Menzel et al. 2006; Doi and Takahashi 2008). It is worth noting that using long-term annual mean temperature to represent local thermal condition is more convenient to monitor the climatic control of spatial variation of phenological response to temperature than using mean dates of plant phenophases (Menzel et al. 2006) and latitudes (Doi and Takahashi 2008). The spatial variation of phenological response to temperature may essentially be attributed to the long-term adaptation of plant phenology to regional climate. However, as rapid climate change is likely to disturb the adaptation (Kramer 1996), future regional climate warming may enhance the sensitivity of plant phenological response to temperature, especially in the colder regions where the greatest climate warming is expected (IPCC 2007).

Conclusions

The current study analyzed temporal variability of the U. pumila growing season and its response to temperature, based on leaf unfolding and leaf fall data at 46 stations in the temperate zone of China from 1986 to 2005. The main conclusions are as follows:
  1. 1.

    A significant advancing trend in BGS and a less pronounced delayed trend in EGS were detected at individual stations. On average, BGS significantly advanced at a rate of −4.0 days per decade, whereas EGS was significantly delayed at a rate of 2.2 days per decade, and LGS was significantly prolonged at a rate of 6.5 days per decade across the temperate zone of China. Therefore, the growing season extension was induced mainly by the advancement of BGS.

     
  2. 2.

    At individual stations, the linear trend of BGS correlates negatively with the linear trend of the spring temperature, namely, the quicker the spring temperature increased at a station, the quicker the BGS advanced. However, linear trends of EGS and LGS did not correlate significantly with linear trends of the autumn temperature and annual mean temperature, respectively. Across the temperate zone of China, a larger positive trend in spring temperature corresponds to a larger negative trend in BGS, whereas a smaller positive trend in autumn temperature matches a smaller positive trend in EGS. In addition, a significantly positive trend in the annual mean temperature is linked to a significantly positive trend in LGS.

     
  3. 3.

    With respect to the growing season response to interannual temperature variation, 89% of the stations show a significantly negative correlation between BGS and the spring temperature, whereas only 26% of the stations indicate a significantly positive correlation between EGS and the autumn temperature. On average, a 1°C increase in the spring temperature may induce an advancement of 2.8 days in BGS; a 1°C increase in the autumn temperature may cause a delay of 2.1 days in EGS; while a 1°C increase in the annual mean temperature may result in a lengthening of LGS by 9 days across the temperate zone of China. Therefore, the response of BGS to temperature is more sensitive than the response of EGS.

     
  4. 4.

    Generally speaking, both the negative response of BGS to spring temperature and the positive response of EGS/LGS to autumn temperature/annual mean temperature were stronger at warmer locations than at colder locations. So, future regional climate warming may enhance the sensitivity of plant phenological response to temperature.

     

Acknowledgments

The authors wish to thank the Meteorological Information Center of the Chinese Meteorological Administration for providing phenological data. This research is funded by the National Natural Science Foundation of China under Grant No. 40871029 and 41071027.

Copyright information

© ISB 2011