RELATION BETWEEN VEGETATION CHANGES, …rcg.gvc.gu.se/dc/PDF/ostwaldetal07.pdfCLIMATE VARIABLES AND...

RELATION BETWEEN VEGETATION CHANGES, CLIMATE VARIABLES AND LAND-USE POLICY IN SHAANXI

© The authors 2007Journal compilation © 2007 Swedish Society for Anthropology and Geography

223

RELATION BETWEEN VEGETATION CHANGES, CLIMATE VARIABLES AND LAND-USE POLICY

IN SHAANXI PROVINCE, CHINA

BYMADELENE OSTWALD

1,2

, ELISABETH SIMELTON

1

,DELIANG CHEN

1

AND ANLIN LIU

3

1

Department of Earth Sciences, Göteborg University, Sweden

2

Centre for Climate Science and Policy Research, Linköping University, Norrköping, Sweden

3

Shaanxi Remote Sensing Centre for Agriculture, Xi’an, China

Ostwald, M., Simelton, A., Chen D. and Liu, A.,

2007: Relation be-tween vegetation changes, climate variables and land-use policyin Shaanxi Province, China.

Geogr. Ann

., 89 A (4): 223–236.

ABSTRACT. Shaanxi Province in China has been ex-posed to climate variability and dramatic land-usepolicies. The aim here is to examine vegetationchanges in this area on a regional scale from 2000to 2004 in relation to land-use changes and climatetraits. The data in this assessment include remotesensing information from moderate-resolution im-aging spectro-radiometer normalized differencevegetation index from 2000 to 2004, and climatedata (precipitation and temperature) from 1956 to2000. The results show an increase in vegetationproduction from 2000 to 2004, particularly in thenorth, which cannot be explained solely by climateimpacts. Since the vegetation in the north is moredependent on climate variation than the other partsof Shaanxi due to more serious water limitation, theresults suggest that the large-scale land-use policyimplemented over the last decade, with a focus onnorthern Shaanxi, is possibly having an impact onthe overall vegetation.

Key words

: Loess plateau, NDVI, regional climate

Introduction

The starting point for land-use changes in Chinacan be traced back to the 1978 economic reformthat opened up China to the outside world marketin a controlled way, a change forced by years ofeconomic stagnation (Lu and Wang 2002). This re-form, the Open-Door Policy, allowed China to takepart in increased globalization (Yeh and Li 1999;Bao

et al

. 2002) with expanding trade of agricul-tural products. The change was further accentuatedwith the land reform of 1982, the Household Re-sponsibility System, which resolved the com-munes. As a result, land-use rights were distributedto individual farmers with leases of 15 years. The

two reforms have had a clear impact on land use.During the last decade, rising concern about the en-vironmental effects of these reforms and policieshas been highlighted (McElroy

et al

. 1998; Liu

etal

. 2005), particularly in the area of the Loess Pla-teau in Shaanxi, which is the focus here (Hu 1997;Liu 1999; Skinner

et al

. 2001).In 1999 the Chinese government introduced the

Slope Land Conversion Programme at the nationallevel, also known as the Grain for Green Policy (seeFeng

et al

. 2004), and later named the CroplandConversion Programme at the provincial level inShaanxi. The massive policy had a budget of over40 billion US dollars (Xu

et al

. 2004) and can beseen as one of the most ambitious environmentalinitiatives globally. The objective was to halt ero-sion, and indirectly to solve the problem of sedi-mentation in the big rivers, by encouraging farmersto change crop cultivation on slopes to tree andgrass plants. This programme provides compensa-tion to farmers in terms of money and food grainsfor up to eight years. The targeted land for ShaanxiProvince was slopes exceeding 25° and includedthe planting of locust trees (

Robinia pseudoacaciaL

) mainly (Rui

et al

. 2001). Shaanxi has convertedthe most land (0.82 Mha of former agriculturalland) of all the 33 provinces or Autonomous Re-gions covered by the policy. By the end of 2003 thisamounted to 7.2 Mha of former cultivated land inthe whole of China (Xu

et al.

2006), while the totalland conversion, i.e. both former agricultural landsand other lands, was 13.3 Mha (Xie

et al

. 2005).However, earlier studies show that the land-usechange situation before the Cropland ConversionProgramme was dynamic in Shaanxi (Liu

et al

.2005). A total of 44 000 ha of cropland was con-verted to other land uses while almost 57 000 ha of

MADELENE OSTWALD, ELISABETH SIMELTON, DELIANG CHEN AND ANLIN LIU


224

new land areas were taken up for crop productionfrom 1990 to 2000, i.e. a net increase of crop area.

In the wake of China’s economic success duringthe last decade, increasing attention is now beingpaid to the environmental effects. Land use, agri-culture and hence food production is one such area(e.g. Zhao

et al

. 2005). During the last decades,several large-scale land-use policies have been im-plemented causing changes in land use and, hence,land cover (Hu 1997; McElroy

et al

. 1998; Skinner

et al

. 2001). The latest and largest of these reformsis the Slope Land Conversion Programme. At thesame time, climate variability is showing great re-gional differences. For part of Shaanxi Provincethere is a decreasing trend in precipitation charac-terized by erraticism and a significant increase intemperature (e.g. Hageback

et al

. 2005). Several at-tempts have been made to evaluate the impact ofthis policy in terms of economic sustainability(Uchida

et al

. 2005), livelihood impact (Feng

et al

.2004; Xie

et al

. 2005) or the organization of theprogramme itself (Xu

et al

. 2004). However, theimpact of this reform on physical land cover in re-lation to climate variability has not yet been ana-lysed, and is the subject of this study. It has beenstressed that to improve the ability to predict land-use change in the near future and its impact on Chi-na’s food security and environment, there is a needto investigate the mechanisms that drive land-usechange at multiple scales, from landscape to na-tional scales (Liu

et al

. 2005).The simultaneous processes of land-use change,

seen as a direct human impact, and climate varia-bility, seen as a possible indirect human and naturalimpact, are affecting the overall vegetation. Severalstudies from northern China have shown links withvegetation status both from land-use policies andclimate variability (Runnström 2000; Chen

et al

.2005; Fang

et al

. 2005; Ostwald and Chen 2006).Correlation has been found between the vegetationpeak and vegetation index from satellite data andrainfall records in northern China (Runnström2000). When studying different biomes, increasedprecipitation gave increased

normalized differ-ence vegetation index (NDVI)

of grassland anddeciduous broadleaf forest, but a decreased NDVIfor deciduous coniferous forest. The various bi-omes also behaved differently depending on pre-cipitation pattern and amount, indicating a non-lin-ear feedback (Fang

et al

. 2005). More complex andvaried relationships between vegetation indicesand climate variables were shown in northernShaanxi when looking at temperature and precipi-

tation in different lagged or pre-seasonal correla-tions (Ostwald and Chen 2006). It was shown thatlow temperatures one to three months before Au-gust were positively correlated with increasingvegetation production in August, which suggeststhat reduced evapotranspiration and hence less wa-ter stress have a positive effect on vegetation pro-duction. Also, a warm spring had a positive effecton vegetation production in August. In general, thevegetation indices showed higher correlation withtemperature than precipitation, which indicatesthat enhanced evaporation due to high tempera-tures has a negative effect on vegetation growth.Similar results regarding the impact of limiting wa-ter supply were shown in a global assessment ofsensitivity of NDVI to climate variability (Boun-oua

et al

. 2000). Correlation with temperature rath-er than precipitation was also determining the startand end dates of growing seasons in temperate Chi-na, including Shaanxi (Chen

et al

. 2005). The tem-poral scale, or lag, for precipitation and soil mois-ture increases from humid to dry climates, demon-strates the role of potential evaporation (Entin

et al

.2000).

This study investigates relationships among landuse, overall vegetation production, and climate var-iability in Shaanxi Province between 2000 and2004. How have the vegetation and climatechanged over the period? What relationships can befound between vegetation, and possibly land-usechanges driven by the policy and climate traits ona regional scale? Are there any regional differenceswithin the province?

Shaanxi Province

Shaanxi Province covers 205 600 km

2

in northerninland China (31°45'–39°35'N and 105°29'–111°15'E), approximately 2% of the total country.Climate and topography divide the province intothree distinct regions: the semi-arid Loess Plateauin the north, the warm-temperate central plain andthe sub-tropical region south of the Qinling Moun-tains (Zhao 1986). These areas have different landuses and are therefore used for the regional analysisin this study (Fig. 1). The main land use in the northis characterized by grassland, rainfed summercrops and sparse woods. Moving to the central area,irrigated croplands, mainly winter wheat andmaize, are dominating (Liu

et al

. 2005). In thesouth, the main crops are maize, wheat and rice incombination with economic forest, sparse woods,and mixed needle- and deciduous broadleaved

VEGETATION CLIMATE AND LAND USE, CHINA


225

types (Zhao 1986; IIASA 2001). The irrigated areaused in agriculture has been fairly constant over theperiod 2000 to 2004, varying between 1 272 000and 1 315 000 ha for the whole province (ChinaAgriculture Yearbook 2001–2005). The province isinhabited by 37 million people, with the greatestdensity in the central, most industrialized part(Knutsson 2005). Approximately 20% of the pro-vincial area is used for agriculture, while half ofShaanxi was considered forested in 2004. The areaused for agriculture decreased by 10% from 2000to 2004, and for forest by 11% from 2003 to 2004(China Statistical Yearbook 2005). The implemen-tation of forest and grass plantation under the Cro-pland Conversion Programme is not part of the for-est category in these statistics. Between 1980 and2000 the cultivated land per farmer changed from0.16 ha to 0.31 ha despite a population increase(Liu and Chen 2005), compared with 0.11 ha as thenational average (Liu

et al

. 2005). Hence, land-use

change is in a constant process. Due to its positionin terms of monsoon extent and hence limits forrainfed agriculture and the sensitive environmentwithin the Loess Plateau, Shaanxi is representativeof northwestern China in terms of climate variabil-ity and vegetation changes.

Data and methods

Remote sensing

Moderate-resolution imaging spectroradiome-ter (MODIS)

13Q1 images were downloaded fromEOS data gateway at a spatial resolution of 250 mas 16-day composites, consisting of the highest val-ue for each pixel over a 16-day period. One imagefor each month from April 2000 (commencement ofMODIS data) to December 2004 was collected.There are several characteristics of temporal com-posited data. They are consistently produced datasets and the cloud contamination is greatly reduced.

Fig. 1. (a) China, showing the posi-tion of Shaanxi Province in theLoess Plateau. (b) Shaanxi Prov-ince with the boundaries of the threeregions, north central and southlyshown by thick lines. The climatestations are 1, Yulin; 2,Yan’an; 3,Luochuan; 4, Baoji; 5, Wugong; 6,Xi’an; 7, Hanzhong; 8, Shangzhou;9, Ankang. The grey scale showsthe NDVI values (× 104) from Au-gust 2001



226

On the other hand the sampling interval may be toolong to detect some phenological events with suffi-cient accuracy (Reed

et al

. 1994). The NDVI is anormalized ratio of the near infra-red (841–876 nm)and red (620–670 nm) bands which is often used asan index to reveal seasonal and/or inter-annualchanges in vegetation. Due to the strong relation-ship between NDVI and vegetation production,NDVI can be used as a proxy for net primary pro-duction (NPP) (Goward

et al

. 1987; Fang

et al

.2005). NDVI is limited in terms of covering vege-tation with rich chlorophyll activities since it satu-rates around 0.7 NDVI (Zhang

et al

. 2004). Thismakes the product suitable for the semi-arid re-gions. The vegetation expressed by NDVI is relatedto the chlorophyll activity, which can be an effect ofincreased density or cover. The vegetation defini-tion used in this paper refers to vegetation activitywithout excluding any primary vegetation produc-tion process. The images were pre-processed (level3) by the producer which includes validation, per-formance assessment, and uncertainty analyses ofseasonal vegetation index response curves and in-ter-annual time series representations (TerrestrialBiophysical and Remote Sensing Lab 2003). TheMODIS resampling tool was used to mosaic the im-ages together and resample the images into a plainprojection. The NDVI values are multiplied by 10

4

for easier interpretation.For images with visible cloud contamination, a

filter was developed that deleted those particularpixels that were placed between two months withhigher NDVI values than itself (lower values indi-cating clouds). Visual qualitative interpretationshowed that the filter did delete cloud-affected ar-eas. No reliable data were found for December2002, leaving the data set with a total of 56 MODISimages. The traditional multi-temporal profileswere produced to indicate the general vegetationphenology (Reed

et al

. 1994) as well as the arealchanges in vegetation dynamics giving increased,decreased or unchanged NDVI between two dates(Ostwald and Chen 2006). This was not conductedin relation to a particular vegetation type or landcover, which is unavailable at this spatial scale, butrather to get overall vegetation changes over differ-ent geographical regions in the province.

Climate data

Nine meteorological stations

*

, three in each region,were selected to provide a maximum cover for eachsub-region (Fig. 1). The meteorological data con-

sisted of daily mean temperature and total precipi-tation for 1956 to 2004, from which values for 1999to 2004 were extracted for the correlation analysis:

monthly mean temperature (T)

,

temperaturesum above 0°C (TΣΣΣΣ

)

,

total precipitation (P)

, and

precipitation frequency (Pfreq)

over various pe-riods. Fang

et al

. (2005) suggested using Pfreq, i.e.number of days with rain in relation to the totalnumber of days considered, to distinguish vegeta-tion patterns. The regional climatic values were de-rived as the mean of the three stations in each re-gion. The meteorological data were provided andquality controlled by China Meteorological Ad-ministration, Beijing. In addition, information onnatural hazards was available for nine agrometeor-ological stations across the province, provided byShaanxi Meteorological Bureau, Xi’an.

Impact of climate

Relationships between NDVI on one side, and T,T

Σ

, P and Pfreq on the other, were evaluated overthe three different areas (Kang

et al

. 2002; Fang

etal

. 2005; Song

et al

. 2005). Pre-seasonal analysis ofthe climate parameters in relation to NDVI wascomputed to identify possible lagging effects as in-dicated in earlier studies (Ostwald and Chen 2006).This was done over the onset of greening (March–April), and the growing season (April–August), inrelation to three different vegetation periods, i.e.start (May–June), peak (July–August) and wilting(September–October) periods and the peak month(August). In testing relationships between climatevariables and vegetation through NDVI, meanNDVI rather than maximum NDVI was correlated(Ostwald and Chen 2006) and was therefore used inthis analysis. The sum of the accumulated NDVIover different periods was also used, as suggestedby Fang

et al

. (2005). To remove the seasonal cycle,anomalies with respect to the seasonal means wereused. Statistical significance tests were performedusing one-tailed Pearson’s test (Ebdon 1985).

Result and discussion

Climate variability and change

Climate change and variability have significant ef-fects on land use, including agriculture and forest-ry. In China, agriculture is strongly dependent onclimate and is thus susceptible to climate change(Smit and Cai 1996). These changes occur both ona large scale related to, for example, the East Asianmonsoon system (Tao

et al

. 2004; Gordon

et al

.



227

2005) and on a regional scale related to human im-pact, such as increased emissions of absorbingblack carbon (Menon

et al

. 2002). In Shaanxi Prov-ince, regional and local climates have shown achange over the last 50 years (Hageback

et al

.2005; Ostwald

et al

. 2004). For example, Hageback

et al

. (2005) showed an increase in annual meantemperature of 0.9°C in Ansai (northern Shaanxi)for the period 1955–2004. Furthermore, our datashow that the number of days with daily mean tem-perature above 0 and 5°C increased by around 20and 15 days respectively in 40 years (1956–2004),indicating a longer potential growing season.

Precipitation generally has a great variability and40–70% falls between May and September. The totalannual amount has a clear decadal variation and thelinear trends are relatively weak (Song

et al

. 2005;Zhao

et al

. 2005). The most significant seasonaltrends occurred during spring (March–May) andwinter (December–February) for both mean tem-perature and for precipitation (Table 1). The winterprecipitation in the north increased significantly, by20 mm in 49 years, and decreased in the other sea-sons. The precipitation decreased by 38.2 mm in 49years, with the greatest decrease in the central region.The annual mean temperature over the whole prov-ince increased by 1°C from 1956 to 2004.

Over the five years considered in this study, meanmonthly total precipitation and mean temperatureshow a cooler and drier northern region, a warmercentral region and a warmer and wetter sothern re-gion (Fig. 2). When looking at the differences overtime, the year 2000 was dry especially in the north,while 2004 was wet (Fig. 3). Agrometeorologicalstations throughout the province recorded droughts,particularly in 2000 and 2002, and the northernmoststations reported droughts in all five years. However,

the natural hazard data bear no relation to the totalannual precipitation since serious droughts havealso been reported for years with normal or above-normal rainfall amount. The annual precipitation inthe five years starting from 2000 were 441, 588, 549,682, 422 mm in the north 538, 432, 444, 911, 531mm in the central region, and 856, 628, 629, 975,713 mm in the south, respectively. The annual meantemperatures were 11.6, 11.9, 12.2, 11.7, 11.8°C inthe north, 15.8, 15.9, 16.5, 15.7, 15.9°C in the centralregion, and 16.7, 16.7, 17.3, 16.7, 16.7°C in thesouth, respectively.

Over the five-year period (Fig. 3) the relative dif-ferences in climate were as follows.

2000 Spring generally dry, especially in the northand the south, and also warm in the south.Summer dry in the north and wet in thesouth. Winter warm and wet in the north.

2001 Spring generally dry, especially in the cen-tral region. Summer warm in the north, dryin the central region. Autumn dry in thesouth. Winter generally warm, especially inthe north and the central region.

2002 Spring generally wet. Summer generallywarm and dry, especially in the central re-gion and the south. Autumn generally warmand dry, especially in the central region andthe south. Winter generally wet; cold espe-cially in the north and the central region.

2003 Spring generally cold; wet in the central re-gion. Summer generally cold; wet in the northand the central region while quite dry in thesouth. Autumn generally wet; cold in thesouth. Winter generally dry, especially in thenorth and the central region; warm in the cen-tral region.

Fig. 2. Mean monthly total precip-itation and mean temperature for2000–2004 in north, central andsouthern region of Shaanxi Prov-ince



228

Tab

le 1

. Mea

n, v

aria

bilit

y an

d lin

ear

tren

ds o

f m

onth

ly te

mpe

ratu

re a

nd p

reci

pita

tion

duri

ng 1

956–

2004

on

an a

nnua

l and

sea

sona

l bas

is

Shaa

nxi

Nor

th

Yea

rly

Spri

ngSu

mm

erA

utum

nW

inte

rY

earl

ySp

ring

Sum

mer

Aut

umn

Win

ter

Mea

n te

mpe

ratu

re (

°C)

17.8

19.2

29.8

17.5

4.7

16.1

17.8

28.1

15.9

2.5

Std

of te

mpe

ratu

re (

°C)

0.7

1.0

0.9

1.0

1.4

0.8

1.0

0.9

1.1

1.6

Lin

ear

tren

d (°

C/y

r or

sea

son)

0.02

**0.

02**

0.00

0.02

*0.

04**

0.03

**0.

02*

0.01

0.03

**0.

05**

Tot

al c

hang

e ca

lcul

ated

fro

mth

e tr

end

(°C

/49

yrs)

1.0

1.2

0.1

1.0

1.9

1.3

1.2

0.4

1.2

2.2

Mea

n pr

ecip

itatio

n (m

m)

558.

211

0.6

286.

615

1.3

9.7

494.

782

.128

7.4

118.

76.

5%

of

year

ly to

tal

2051

272

1758

241

Std

of p

reci

pita

tion

(mm

)92

.336

.263

.154

.29.

010

2.9

40.4

74.2

48.7

8.5

Lin

ear

tren

d (m

m/y

r or

sea

son)

–0.8

–0.4

–0.4

–0.4

0.4*

*–0

.50.

1–0

.7–0

.30.

4**

Tot

al c

hang

e ca

lcul

ated

fro

mth

e tr

end

(mm

/49

yrs)

–38.

2–1

8.3

–20.

7–2

0.6

20.1

–22.

76.

2–3

4.7

–14.

820

.0T

otal

cha

nge

calc

ulat

ed f

rom

the

tren

d (%

/49

yrs)

–7.1

–18.

0–7

.5–1

4.7

100.

6–4

.77.

2–1

2.9

–13.

312

0

Cen

tral

Sout

h

Yea

rly

Spri

ngSu

mm

erA

utum

nW

inte

rY

earl

ySp

ring

Sum

mer

Aut

umn

Win

ter

Mea

n te

mpe

ratu

re (

°C)

19.1

20.1

31.0

18.7

6.5

18.2

19.6

30.2

18.0

5.2

Std

of te

mpe

ratu

re (

°C)

0.8

1.2

1.1

1.1

1.4

0.7

0.9

0.8

1.0

1.3

Lin

ear

tren

d (°

C/y

r or

sea

son)

0.02

**0.

03**

0.00

0.02

*0.

03**

0.02

**0.

02*

–0.0

10.

02*

0.04

**T

otal

cha

nge

calc

ulat

ed f

rom

the

tren

d (°

C/4

9 yr

s)1.

11.

40.

21.

01.

60.

80.

9–0

.30.

81.

7

Mea

n pr

ecip

itatio

n (m

m)

588.

112

7.5

270.

817

8.4

11.4

591.

412

2.1

301.

515

6.6

11.2

% o

f ye

arly

tota

l22

4630

221

5126

2St

d of

pre

cipi

tatio

n (m

m)

127.

947

.291

.975

.912

.394

.332

.871

.957

.38.

0L

inea

r tr

end

(mm

/yr

or s

easo

n)–1

.7–0

.7–1

.2–0

.40.

6**

–0.2

–0.5

0.6

–0.6

0.3*

*T

otal

cha

nge

calc

ulat

ed f

rom

the

tren

d (m

m/4

9 yr

s)–8

2.9

–34.

7–5

8.1

–18.

326

.8–8

.9–2

6.2

30.8

–28.

813

.3T

otal

cha

nge

calc

ulat

ed f

rom

the

tren

d (%

/49

yrs)

–15.

2–3

1.6

–24.

1–1

0.8

106.

9–1

.5–2

4.1

9.7

–20.

374

.0

Pear

son’

s co

rrel

atio

n co

effi

cien

ts w

ere

test

ed (

one-

taile

d) w

ith P

ears

on’s

test

for

sig

nifi

canc

e us

ing

SPSS

. Sig

nifi

canc

e le

vel:

**0.

01, *

0.05



229

Fig. 3. Anomalies of monthlymean NDVI, mean temperature(*10°C) and total precipitation(mm) for 2000–2004 in (a) north,(b) central, and (c) south region.**0.01 level one-tailed Pearson’stest. T, P and NDVI central showedno significant trend

A

B

C



230

2004 Spring generally warm, especially the northand the central region; dry in the north. Sum-mer generally quite cold. Autumn dry.

Vegetation analysis

The provincial mean NDVI shows a vegetationchange over the period with 2002 and 2004 indi-cating an earlier spring (Fig. 4). There is also an in-crease in the peak month, August, for 2004. The lastthree years (2002 to 2004) point to a higher activityof chlorophyll during June to October than in theearlier years. Furthermore, 2003 indicates a pro-ductive autumn period compared to the other years.Three of the five years (2000, 2003, 2004) have adiminishing production in June giving the profile abimodal character, which could indicate more thanone distinguished land cover type present in the sig-nature. The multi-temporal profile suggests a grow-ing season from April to November.

The NDVI patterns differ when dividing the datainto the three regions (Fig. 5). The north has thelowest NDVI value while the south has the highest.The difference between the years is greatest in thenorth where NDVI increased during the last threeyears (2002 to 2004). This increase could be due tothe Cropland Conversion Programme effect sincethe north is characterized by slopes, the landformthat is the focus for the programme. The southshows a comparatively smoother growing seasonand little change between years. The central partshows a bimodal pattern signifying a harvest inMay/June.

The percentage of the land area that has experi-enced an increase in vegetation activity in the peakmonth (August) from 2000 to 2004 is 79%, 52%

and 51% for the north, central region and south, re-spectively. The increasing vegetation trend is sig-nificant particularly in the north (R

2 = 0.42) but alsoin the south (R2 = 0.15) based on the anomalies ofmonthly mean of the whole period (Fig. 3a and c).

Over the five-year period (Fig. 5) the relative dif-ferences in the NDVI were as follows.

2000 Early spring in the south. Low productivityin June for the central region. Low produc-tivity in general, particularly in the north.

2001 Productivity peak in September in the cen-tral region. Higher productivity in June in thecentral region compared to all other years.Low productivity dip in July in the central re-gion.

2002 Earlier spring in the south than in previousyears. Highest productivity in June of all theyears in the north. Lowest productivity of allthe years from September to November inthe central region.

2003 High productivity in autumn in the north andthe south. 2004 . Highest productivity inApril and August of all the years in the north.Early spring in the south.

Correlation between climate and vegetation in the three regionsTo determine the relationships between vegetationand climate parameters we will take a closer lookat each region, from north to south, and finish witha synthesis. We start with some particular NDVIvalues or patterns and link these to the climate da-ta.

As seen in Fig. 5, the southern part of Shaanxi

Fig. 4. Mean NDVI values overShaanxi Province from April 2000to December 2004. The start(May–June), growing (May–June), peak (July–August) andwilting (September–October) peri-ods are indicated. NDVI values aregiven (× 104)


© The authors 2007Journal compilation © 2007 Swedish Society for Anthropology and Geography 231

Fig. 5. Mean NDVI values andstandard deviation from April2000 to December 2004 over thethree regions of Shaanxi Provincewith percentage of area with vege-tation change from August 2000 toAugust 2004. (a) North, (b) Cen-tral, and (c) South region. Start ofgrowing period (May–June), peakgrowing activity (July–August)and wilting period (September–October): are indicated. NDVI val-ues are given (× 104)

A

B

C


© The authors 2007Journal compilation © 2007 Swedish Society for Anthropology and Geography232

has changed very little over the five-year period interms of NDVI, while the climate variables in Fig.3 show a great interannual variability. A similarclimate pattern can be observed in the central part,especially in terms of temperature. Correspond-ing similarities between the south and central re-gion are not found in the vegetation pattern. Thearea with increased vegetation from 2000 to 2004is also lower in the south (51%) and central region(52%) than in the north (79%). This implies thatclimate is not the only factor driving the vegeta-tion change. Increased vegetation provides moretranspiration (latent heat) which in turn has acooling effect that can be seen in all the regionsfor the five-year period (Fig. 3), especially thenorth. This would suggest that there is a feedbackbetween the vegetation and the local and regionalclimates.

In the northern region, agriculture consistsmainly of one summer crop (planted in April toJune and harvested in August or September) with orwithout a winter crop. The harvest effect is seen inthe dip of the NDVI anomalies between June andJuly (Fig. 3). However, the harvest cannot be de-tected as absolute NDVI since it constitutes a rela-tively small area in the north as a whole (Fig. 5). Be-tween 2000 and 2004 we see steadily increasingNDVI values in the peak period which are not ev-ident in the other two regions, particularly thesouthern region (Fig. 3). Furthermore, this NDVIvalue is not only related to rain patterns, as 2004was comparatively dry, both in amounts and fre-quency. This could instead be the influence of thetree planting policy, which primarily should be tak-ing place in the north since the Cropland Conver-sion Programme was aimed at slope land which isabundant in the Loess Plateau in the north. Newlyplanted Locust trees may grow 0.5–3 m per yearwith a rapid canopy increase (Singh 1982).

In the central region the dip in the June NDVIvalues reflects agricultural harvests of spring crop

(May–June) and summer crop (September–Octo-ber) (Fig. 5). Furthermore, although precipitation issignificantly correlated with NDVI simultaneously(Table 2), the NDVI values are obscured by irriga-tion, e.g. 2003 (cold spring, wet summer) and 2004(warm spring, dry summer) both produced highNDVI values in August (Figs 5 and 3). SeasonalNDVI development in 2001 is exceptional; highervalues than any other year in May, and then lowerbetween July and August, ascending to average inSeptember. This is most likely related to low pre-cipitation amounts from March to June; this couldentail a delayed harvest, which appears as low JulyNDVI. For the agrometeorological data there is noevidence for a lower yield or later than normal har-vest in terms of winter wheat in 2001; on the otherhand, droughts were frequent. Again, the effect ofirrigation is shown.

The mean monthly temperatures in the centralregion and the south are generally above +5°C,which allows for a long growing season, as mon-itored by the NDVI. For the central region it ispossible to see a combination of temperature andprecipitation (Table 3), in that temperature initi-ates the NDVI in the start period, and then precip-itation takes over NDVI during the peak. In both2001 and 2002 the central region has a high NDVIvalue at the beginning of the year, but not partic-ularly high values in August, compared to 2003and 2004 which have low NDVI values at the be-ginning of the year but high values around August.The summer temperatures for these two groupsare higher in 2001 and 2002 than in 2003 and2004. Apart from weather variability, the hightemperatures may be due to (i) either an increasein evaporation to the extent that not even irrigationcan compensate the loss, or (ii) the plants, mostlymaize, are experiencing moisture stresses leavingthem with lower chlorophyll, i.e. lower NDVI.

The southern region NDVI has a smoother annualprofile in general and a two to three month peak,

Table 2. Correlation coefficients between monthly anomalies of NDVI and T and P for the period from April 2000 to December 2004

Mean NDVI Simultaneous 1 month lag 3 months lag 6 months lag 1 year lag

P T P T P T P T P T

North 0.16 –0.20 0.32** –0.22 0.42** –0.16 0.31** 0.05 0.69** –0.22Central 0.34** –0.08 0.18 –0.19 0.18 0.00 0.03 0.03 0.06 –0.05South –0.10 0.20 0.09 –0.04 0.22 –0.03 0.18 –0.07 0.25* –0.25*

Four time lags (sum of monthly values) are tried for the three regions. Significance level of correlations using one-tailed Pearson’s test:**0.01, and * 0.05



which probably reflects the sub-tropical influence,regardless of summer bi- or trimodal precipitationdistribution. Interestingly there is no NDVI decreasecorresponding to the ‘harvest effect’ seen in the northand central region. The dip between May and June2000 is probably not a harvest effect but rather an ef-fect of the heavy rainfall during June (211 mm)flooding the paddy fields and leaving little green-ness.

Pre-seasonal climate effects on vegetation in the three regionsTo evaluate the impact of climate variability onvegetation, seasonal cycles were removed by usingmonthly anomalies. Further, correlation analysiswas conducted using simultaneous and lagged datasets (Table 2) for the whole time period.

Contrary to the previous findings in this part ofChina by Chen et al. (2005) and Ostwald and Chen(2006) and similarly to Runnström (2000), there isa stronger correlation between precipitation andNDVI than between temperature and NDVI. Oneexplanation is the water dependency. At the region-al level water stress is more pronounced in the aridnorth compared to the greatly irrigated central part

and the sub-tropical south. The negative correla-tions (both significant and non-significant) duringApril to August in the north between NDVI andtemperature indicate that during warm periodsthere is a higher evaporation and water stress caus-ing less vegetation. In the central region, NDVI iswell correlated with precipitation from the samemonth, hence grain crop biomass growth reacts di-rectly to rainfall as seen by Kang et al. (2002),while lagged precipitation has little effect due to ir-rigation. In the north, precipitation and NDVI cor-relate in all temporal lags except the simultaneousprecipitation, confirming that stored soil water iscrucial for vegetation, as suggested by Entin et al.(2000) and Fang et al. (2005). In the south, on theother hand, water is not limiting growth. Tempera-ture correlates significantly with NDVI in only onecase, i.e one year lag in the south (Table 2).

For mean temperature (T), temperature sum(TΣ), precipitation (P) and precipitation frequen-cy (Pfreq) during the periods March–April andApril to August, few correlations were found withthe sum of accumulated NDVI for the differentvegetation periods (Table 3). Furthermore, corre-lations were generally highest for the growing

Table 3. Correlation coefficients between sums/means of temperature and precipitation over various periods including climate data from1999 to 2004 and NDVI data from 2000 to 2004.

NDVI March–April April–August

T mean T ∑ >0 P∑ Pfreq T mean T ∑ >0 P∑ Pfreq

North∑ growing period 1 0.54 0.56 –0.26 –0.17 –0.76 –0.75 0.38 0.63∑ start period 2 0.34 0.37 –0.05 0.04 –0.84* –0.83* 0.57 0.75∑ peak period 3 0.60 0.62 –0.32 –0.23 –0.73 –0.71 0.29 0.59∑ wilting period 4 –0.04 –0.01 0.09 0.38 –0.92* –0.93* 0.66 0.87*August 0.79 0.82* –0.52 –0.52 –0.46 –0.44 0.06 0.29

Central∑ growing period 1 0.13 0.12 0.81* 0.42 –0.26 –0.26 0.41 0.56∑ start period 2 –0.17 –0.16 0.41 0.26 –0.06 –0.08 –0.12 0.18∑ peak period 3 0.04 0.02 0.65 0.44 –0.44 –0.43 0.67 0.64∑ wilting period 4 –0.46 –0.47 0.76 0.70 –0.70 –0.71 0.52 0.84*August 0.13 0.12 0.60 0.35 –0.36 –0.34 0.58 0.57

South∑ growing period 1 0.13 0.12 0.71 0.29 –0.02 –0.02 –0.77 0.17∑ start period 2 –0.56 –0.56 0.93* 0.79 –0.60 –0.62 –0.41 0.76∑ peak period 3 –0.33 –0.33 0.82* 0.42 –0.48 –0.49 –0.80 0.56∑ wilting period 4 –0.40 –0.40 0.53 0.82* –0.31 –0.29 0.44 0.47August 0.21 0.20 0.42 0.03 0.13 0.10 –0.72 0.02

1 The growing period = April–Aug2 The start period = May–June3 The peak period = July–August4 The wilting period = September–October. Significance level of 0.05 is indicated by * with one-tailed Pearson’s test



season period (April to August). Mean T and TΣgave greater differences in correlation in the norththan in the other regions. Low temperature duringApril to August in the north is correlated with theaccumulated NDVI during the start of the growingperiod (May–June) and during the wilting period(September–October), again indicating that a wa-ter stress situation is soothed by less evapotran-spiration. The peak vegetation period is not affect-ed by evaporation to the same extent since this isthe wettest period. On the other hand, Pfreq in thenorth is more important for vegetation than Pamount. This suggests that frequent rainfall makesmore water available for plants, hence increasingNDVI in the north. In rain storms, however, wateris lost quickly as runoff rather than transpiration.This is not seen in the south and the central region.Both the central region and particularly the southshow that for a high peak NDVI, the P amount iscomparatively more important than both Pfreqand temperatures, especially at the start and peakvegetation.

Even if there are few statistically significant cor-relations between the NDVI and climate variables,notable patterns emerge. If we compare the differ-ent regions it is remarkable that the north and alsothe south have higher correlations with the differentclimate variables than the central parts. This can bean indication that irrigation is playing a central rolefor agriculture in the central region, which is notsurprising since this is where most of the irrigationis found. There is no pattern of trend in terms of ir-rigated land area for the province during 2000 to2004 (China Agriculture Yearbook 2001–2005); infact the irrigated area has remained within the samerange over the last 25 years (Deng et al. 2005). Thecentral region has more easily available irrigationwater because of its topography; it also has moreflatlands which allows for more intensive agricul-ture compared to the two other regions, which arecharacterized by rainfed agriculture in the northand sub-tropical paddy-dominated farming in thesouth.

To increase the understanding of relation be-tween vegetation change, climate variability andland-use policy, the future is promising with re-gard to longer time series of data, which has beena limitation to this study. Of particular interest isthe impact of the termination of the subsidies tofarmers in 2008, which has been a central part ofthe design of the land-use policy. Further, with theeventual increase in vegetation in the province, thecoupling between climate variables and vegeta-

tion will be one important research area in theyears to come.

Conclusions– The climatic trend in Shaanxi between 1956 and

2004 can be summarized as a significant increasein seasonal mean temperature accompanied bydecreasing precipitation in three of the seasons.

– Vegetation production, expressed as anomaliesof NDVI, has increased in Shaanxi from 2000 to2004. Despite variability, the trend is statistical-ly significant in the north (R2 = 0.43, p < 0.01)and the south (R2 = 0.15, p < 0.01).

– Pre-seasonal analysis based on monthly anom-alies of climate and NDVI data over the wholeyear gave significant correlations mainly withprecipitation in the north, indicating the impor-tance of water supply in that region.

– The north shows a general temperature de-crease from 2000 to 2004 in July, the hottestmonth. Further, correlation analysis indicatesthat vegetation is more closely coupled to cli-mate in the north as compared to the other tworegions.

– The general vegetation increase (79% in thenorth), where more slope land has been convert-ed than in the central and the southern regions,suggests an impact of the Cropland Conversa-tion Programme in addition to climate change.

AcknowledgementsThe project was supported by the Knut and AliceWallenberg Foundation, the Swedish InternationalDevelopment and Cooperation Agency, and theSwedish Foundation for International Cooperationin Research and High Education. The authors thankProfessor Hanqin Tian from Auburn University,USA, for useful comments. The reviewers of thismanuscript are also acknowledged for their fruitfulcomments.

Note* North: Yulin (109°42'E, 38°14'N), Yan’an (109°30'E,

36°36'N), Luochuan (109°30'E, 35°49'N). Central: Baoji(107°13'E, 34°35'N), Xi’an (108°93'E, 34°30'N), Wugong(108°13'E, 34°15'N). South: Shangzhou (109°58'E, 33°52'N),Hanzhong (107°03'E, 33°07'N), Ankang (109°02'E, 32°43'N).

Madelene Ostwald, Department of Earth Sciences,Göteborg University, 405 30 Göteborg, SwedenE-mail: [email protected]/



Also at:Centre for Climate Science and Policy Research,Department of Water and Environmental Studies,Linköping University, 601 74 Norrköping, SwedenE-mail: [email protected]

Elisabeth Simelton, Department of Earth Sciences,Göteborg University, 405 30 Göteborg, SwedenE-mail: [email protected]

Deliang Chen, Department of Earth Sciences,Göteborg University, 405 30 Göteborg, SwedenE-mail: [email protected]

Anlin Liu, Shaanxi Remote Sensing Centre for Ag-riculture, 710015 Xi’an, ChinaE-mail: [email protected]

ReferencesBao, S., Chang, G., Sachs, J. and Woo, W., 2002: Geographic fac-

tors and China’s regional development under market reforms,1978–1998. China Economic Review, 13: 89–111.

Bounoua, L., Collatz, G.J., Los, S.O., Sellers P.J., Dazlich, D.A.,Tucker, C.J. and Randall, D.A., 2000: Sensitivity of climateto changes in NDVI. Journal of Climate 13: 2277–2292.

Chen, X., Hu, B. and Yu, R., 2005: Spatial and temporal variationof phenological growing season and climate change impactsin temperate eastern China. Global Change Biology, 11:1118–1130.

China Agriculture Yearbook, 2001–2005: China AgriculturePress. Beijing. China Statistical Yearbook, 2005: NationalBureau of Statistics. Available at: www.stats.gov.cn (down-loaded 8 January 2006).

Deng, X., Luo, Y., Dong, S. and Yang, X. 2005: Impact of resourcesand technology on farm production in northwestern China.Agricultural Systems, 84: 155–169.

Ebdon, D., 1985: Statistics in geography. Blackwell. Oxford. 232p.

Entin, J.K., Robock, A., Vinnikov, K.Y., Hollinger, S.E., Liu, S.and Namkhai, A., 2000: Temporal and spatial scale of soilmoisture variations in the extratropics. Journal of Geophysi-cal Research, 105: 11 865–11 877.

Fang, J., Piao, S., Zhou, L., Wei, F., Myneni, R.B., Tucker, C. andTan, K., 2005: Precipitaion patterns alter growth of temperatevegetation. Geophysical research Letters, 32: DOI:10.10292005GL024231

Feng, Z., Yang, Y., Zhang, Y., Zhang, P. and Li, Y., 2004: Grain-for-green policy and its impacts on grain supply in West Chi-na. Land Use Policy, 22: 301–312.

Gordon, L.J., Steffen, W., Jönsson, B., F., Folke, C., Falkenmark,M. and Johannessen, Å., 2005: Human modification of globalwater vapour flows from the land surface. Proceedings of theNational Academy of Science of the United States of America102: 7612–7617.

Goward, S.N., Dye, D., Kerber, A. and Kalb, V., 1987: Compar-ison of north and south American biomes from AVHRR ober-vation. GeoCarto International, 1: 27–39.

Hageback, J., Sundberg, J., Ostwald, M., Chen, D., Xie, Y. andKnutsson, P., 2005: Climate variations and land use in Dan-angou watershed, China – Examples of small-scale farmers’adaptation. Climatic Change, 72: 189–212.

Hu, W., 1997: Household land tenure reform in China: its impacton farming land and agroenvironment. Land Use Policy, 14:175–186.

IIASA, 2001: Land use data base China. http://www.iiasa.ac.at/Research/LUC/GIS/landuse_china.htm

Kang, S., Zhang, L., Liang, Y., Hu, X., Cai, H. and Gu, B., 2002:Effects of irrigation on yield and water use efficiency of win-ter wheat in the Loess Plateau of China. Agricultural WaterManagement, 55: 203–216.

Knutsson, P., 2005: Interdisciplinary knowledge integration andthe sustainable livelihoods approach. Dissertation, School ofGlobal Studies, Human Ecology Section, Göteborg Universi-ty.

Liu, G., 1999: Soil conservation and sustainable agriculture on theLoess Plateau: challenges and prospects. Ambio, 28: 663–668

Liu, X. and Chen, B., 2005: Efficiency and sustainability analysisof grain production in Jiangsu and Shaanxi Provinces of Chi-na. Journal of Cleaner Production, 15: 313–322.

Liu, J., Liu, M., Tian, H., Zhuang, D., Zhang, Z., Zhang, W.,Tang, X. and Deng, X., 2005: Spatial and temporal pattern ofChina’s cropland during 1990–2000: An analysis based onLandsat TM data. Remote Sensing of Environment, 98: 442–456.

Lu, M. and Wang, E., 2002: Forging ahead and falling behind:Changing regional inequalities in post-reform China. Growthand Change, 33: 42–71.

McElroy, M.B., Nielsen, C.P. and Lydon, P., 1998: EnergizingChina: Reconciling Environmental Protection and EconomicGrowth. Harvard University Press.

Menon, S., Hansen, J., Nazarenko, L. and Luo, Y., 2002: Climateeffects of black carbon aerosols in China and India. Science,297: 2250–2253.

Ostwald, M. and Chen, D., 2006: Land-use change: Impact of cli-mate variations and policies among small-scale farmers in theLoess Plateau, China. Land Use Policy, 23: 361–371.

Ostwald, M., Chen, D., Xie Y., Knutsson, P., Brogaard, S., Borne,K. and Chen, Y., 2004: Impact of climate change and varia-bility on local-scale land use – Shaanxi Province, China. Re-search report. Earth Science Centre Series C61. GöteborgUniversity.

Reed, B.C., Brown, J.F., VanderZee, D., Loveland, T. R., Mer-chant, J.W. and Ohlen, D.O., 1994: Measuring phonologicalvariability from satellite imagery. Journal of Vegetation Sci-ence, 5: 703–714.

Rui, L., Zhongming, W., Fei, W., Yue, W. and Mingliang, Z., 2001:Case study on conversion of farmland to forest and grasslandin Ansai County, Shaanxi Province. Prepared for the CCICEDWestern China Forest Grassland Task Force. 22 p.

Runnström, M., 2000: Is Northern China winning the battleagainst desertification? Ambio, 29: 468–476.

Singh, R.V., 1982: Fodder Trees of India. Oxford and IBA. NewDelhi.

Skinner, M.W., Kuhn, R.G. and Joseph, A.E., 2001: Agriculturalland protection in China: a case study of local governance inZhejiang Province. Land Use Policy, 18: 329–340.

Smit, B. and Cai, Y., 1996: Climate change and agriculture in Chi-na. Global Environmental Change, 6: 205–214.

Song, Y., Simelton, E., Chen, D. and Dong, W., 2005: Influence ofclimate change on winter wheat growth in north China during1950–2000. ACTA Meteorologica Sinica, 19: 501–510.

Tao, F., Yokozawa, M., Zhang, Z., Hayashi, Y., Grassl, H. and Fu,C., 2004: Variability in climatology and agricultural produc-tion in China in association with the East Asian summer mon-soon and El Niño Southern Oscillation. Climate Research, 28:23–30.

Terrestrial Biophysical and Remote Sensing Lab. 2003: MODIS



VI User Guide, Terrestrial Biophysical and Remote SensingLab, Arizona University.

Uchida, E., Xu, J. and Rozelle, S., 2005: Grain for green: Cost-ef-fectiveness and ustainability of China’s conservation set-aside program. Land Economics, 81: 247–264.

Xie, C., Zhao, J., Liang, D., Bennet, J., Zhang, L., Dai, G. andWang, X., 2005: Livelihood impacts of the conversion of cro-pland to forest and grassland program. Australian Centre forInternational Agricultural Research (ACIAR) Project. Re-search Report No. 3.

Xu, Z., Bennet M.T., Tao, R. and Xu, J., 2004: China’s sloping landconversion programme four year on: current situation, pend-ing issues. International Forestry Review, 6: 317–326.

Xu, A., Xu, J., Deng, X, Huang, J.,Uchida, E. and Rozelle, S.,2006: Grain for green versus grain: conflict between food se-curity and conservation set-aside in China. World Develop-ment, 34: 130–148.

Yeh, A.G. and Li, X., 1999: Economic development and agricul-tural land loss in the Pearl river delta, China. Habitat Inter-national, 23: 373–390.

Zhang, P., Barlow, M., Tan., B. and Myneni, B., 2004: Climate-related vegetation characteristics derived from Moderate Res-olution Imaging Spectroradiometer (MODIS) leaf area indexand normalized difference vegetation index. Journal of Geo-physical Research, 109: 1–13.

Zhao, S., 1986: Physical Geography of China. Science Press. Be-ijing.

Zhao, W. Z, Xiao, H.L., Liu, Z. M. and Li, J., 2005: Soil degrada-tion and restoration as affected by land use change in the sem-iarid Bashang area, northern China. Catena, 59: 173–186.

Manuscript received Nov. 2006, revised and accepted May 2007.

RELATION BETWEEN VEGETATION CHANGES, …rcg.gvc.gu.se/dc/PDF/ostwaldetal07.pdfCLIMATE VARIABLES AND...

Documents

Transcript of RELATION BETWEEN VEGETATION CHANGES, …rcg.gvc.gu.se/dc/PDF/ostwaldetal07.pdfCLIMATE VARIABLES AND...