spatial patterns and processes of bamboo expansion in...

ARTICLE IN PRESS

0143-6228/$ - se

doi:10.1016/j.ap

�CorrespondE-mail addr

Applied Geography 28 (2008) 16–31

www.elsevier.com/locate/apgeog

Spatial patterns and processes of bamboo expansionin Southern China

Benoit Mertensa, Liu Huab, Brian Belcherc,f,�, Manuel Ruiz-Perezd,Fu Maoyie, Yang Xiaoshenge

aWorld Resources Institute, c/o IUCN, Avenue Ebeya, Kinshasa, Democratic Republic of CongobInstitute of Forest Resources Information, Chinese Academy of Forestry, Dongxiaofu 2, Xiangshan Road, Haidian 100091,

Beijing, PR ChinacCenter for International Forestry Research, P.O. Box 6596 JKPWB, Jakarta 10065, Indonesia

dDepartment of Ecologia, Facultad Ciencias, Universidad Autonoma de Madrid, 28049-Madrid, SpaineResearch Institute of Subtropical Forestry, Chinese Academy of Forestry, 73# Daqiao Road, Fuyang, Zhejiang 311400, PR China

fCentre for Non-Timber Resources, Royal Roads University, 2005 Sooke Rd., Victoria, BC, Canada

Abstract

This paper analyzes the spatial distribution and patterns of land use change associated with expansion of bamboo

production in three important bamboo-producing counties (Anji, Pingjiang and Muchuan) along a development gradient

in China. Existing and new bamboo plantations tend to be located at higher elevation and on sloping lands around

formerly established plantations, with different patterns of expansion related to the level of development of the bamboo

industry and general economic development in the county. Distance to factory is more important for low value-added

products where raw material represents a high percentage of production costs.

r 2007 Published by Elsevier Ltd.

Keywords: China; Bamboo; Spatial analysis; Remote sensing; Land use change

Introduction

China has implemented economic and environmental policy reforms over the past two decades, with far-reaching implications. The Household Responsibility System (HRS), reforms in trade, exchange, andinvestment policies, and the ‘‘Natural Forest Protection Programme’’ and the ‘‘Conversion of Arable Land inSteep Slopes’’ policies, have all contributed to encourage rural development and environmental rehabilitation(Hyde, Xu, & Belcher, 2003). The results can be seen in China’s economic statistics, from explosive GDPgrowth of 9%/annum at the national level, to high rural household growth in some areas. The impacts canalso be seen on the landscape, with changes in land use patterns in response to environmental conditions,regulations and economic incentives (Verburg & Chen, 2000).

e front matter r 2007 Published by Elsevier Ltd.

geog.2007.07.012

ing author. Tel.: +62 251 622622; fax: +62 251 622100.

ess: [email protected] (B. Belcher).

www.elsevier.com/locate/apgeog

dx.doi.org/10.1016/j.apgeog.2007.07.012

mailto:[email protected]

ARTICLE IN PRESSB. Mertens et al. / Applied Geography 28 (2008) 16–31 17

Forest products’ demand has grown in parallel with this economic expansion. In the context of newenvironmental and forestry policies, this has brought unprecedented increases in timber imports (Sun,Katsigris, & White, 2004). Increasing land scarcity and competition between alternative landuses has resultedin a reallocation of forest-related activities, enhanced by land rehabilitation and other environmental schemes(Li, 2004). In this paper we focus on the bamboo sector, one of the fastest growing forest landuses in China,analyzing the spatial patters of expanding bamboo plantations in Central China.

With about 7 million hectares (ha), China has the largest bamboo resource in the world. Traditional, small-scale domestic use of bamboo has been overtaken in recent decades by rapid development of commercialbamboo processing, driven by market-oriented reforms. The economic importance of bamboo, its fastexpansion and its potential as a major substitute for timber have been well documented (Gu, 1992; Li & Xu,1998; Lin, 1999; Ruiz-Perez, Belcher, Fu, & Yang, 2003). Consequently, a continuous expansion of bambooplantations has taken place in China during the last 20 years.

However, little is known about the detailed spatial trends of this expansion: How much have bambooplantations expanded and where? What are the major characteristics of these areas of expansion? What landuse changes are being induced by the expansion of bamboo? What factors explain the different patterns ofexpansion? These are essential questions given the fast-changing land use patterns in China and the positiveand negative economic and environmental implications associated with bamboo plantations and relatedindustries (Ruiz-Perez, Belcher, Fu, & Yang, 2004). Answering these questions can also help understandingthe potential and constraints faced by the Chinese forest sector and its growing global implications.

This paper provides a comparative analysis of the spatial patterns and processes associated with theexpansion of bamboo plantations in three counties with contrasting socio-economic and environmentalconditions. It summarizes the key lessons learnt and the importance of understanding such processes whenpreparing new national level policies, local regulations and other measures affecting the forest sector.

The bamboo sector in China

Bamboo occurs naturally throughout the country except in high mountain areas and very dry areas, withthe largest variety of species occurring south of the 401N parallel. Most bamboo species are erect and tall,having woody stems; however, some are herbaceous, being present as forest undergrowth or at high altitude;some bamboo species showing climber-like properties can also be found. The last published forest inventory in1998 recorded 4.2 million ha of bamboo forests and plantations, and 3 million ha of ‘‘mixed and mountainnatural bamboo stands’’ (State Forest Administration, 2000). A significant amount of woody bamboo occursin 17 provinces, with 11 provinces having over 100,000 ha each (Fig. 1). Four contiguous provinces (Fujian,Jiangxi, Zhejiang and Hunan) have more than half a million ha each, accounting for over 60% of totalbamboo plantations in China. The bamboo resource base has increased steadily over the past two decades,both in surface area (31.6%) and in density of stands (40.8%) (State Forest Administration, several years).The combined expansion of plantations and increased density of stands has resulted in an exponential increasein output of bamboo culms and shoots (594% and 1048%, respectively) (State Forest Administration, severalyears). This contrasts sharply with the stagnant output of timber in the country during the 1990s and its sharpdecline following the implementation of the Natural Forest Protection Programme and the accompanyingLogging Ban (Bull & Nilsson, 2004).

There are two main categories of use of bamboo: (1) culms (stems) are used like wood in unprocessed formor in a variety of processed forms and (2) edible bamboo shoots are used as a vegetable, fresh, or in a variety ofprocessed forms. Because of this dual utilization, bamboo has economic characteristics of both conventionaltimber-oriented forests and fruit-producing cash trees. In addition, bamboo plantations have several attractiveattributes for farmers: (i) short rotation cycles with yearly or biennial production, (ii) low investment costs,multiple uses that allow for flexible markets and the feasibility of managing small areas make them easilyadapted to the HRS, and (iii) the variable scale and diversity of processing units and the possibility of a degreeof pre-processing by farmers makes local processing a viable option.

They also appear to be more profitable than timber trees (such as Chinese fir or Pinus plantations) andbenefit from an expanding national and international demand, both as a substitute for wood fibre and forsome superior goods like bamboo shoots and bamboo flooring. Modern processing technologies have been

ARTICLE IN PRESS

40N

Xinjiang

Xizang

Neimenggu

Qinghai

Sichuan

Gansu

Yunnan

Jilin

Heilongjiang

Hunan

Hubei

Hebei

Guangxi

Shaanxi

Henan

Anhui

Jiangxi

Shanxi

GuizhouFujian

Liaoning

Guangdong

Shandong

Jiangsu

Zhejiang

Hainan

Taiwan

Ningxiahuizu

Beijing

Tianjin

ShanghaiShanghai

Hongkong

Provinces of China

Bamboo forest area (,000 ha)

0,00 - 22,00

22,01 - 59,20

59,21 - 240,10

240,11 - 355,00

355,01 - 680,70

Fig. 1. Area of bamboo forest by Province.

B. Mertens et al. / Applied Geography 28 (2008) 16–3118

developed to produce a remarkable variety of new products. Consequently, when environmental conditionsallow, farmers find bamboo more attractive than timber plantations, and there is an active process ofreplacement of conifers by bamboo. In combination, these features give an economic attractiveness and aflexibility that have made bamboo a useful resource for adapting to and taking advantage of the opportunitiesoffered by key policy reforms related to land tenure, commercialization, industrialization and exports.

Study area

Three counties along an East-West transect within the subtropical region of China were selected to cover arange of socio-economic and environmental conditions, such as population density, economic structure,industrial and infrastructure development, accessibility, topography and vegetation, particularly the bamboodistribution and its contribution to the local economy. This work is part of a larger study on the changing roleof bamboo in economic development (see for example Ruiz-Perez et al., 2004).1 The survey included countylevel time-series statistics for the period 1980–2001, key informant interviews for the six counties, as well asdetailed farmer and industry questionnaires for three counties included in the current analysis: Anji County inZhejiang Province, Pingjiang County in Hunan Province and Muchuan County in Sichuan Province (Fig. 2).

The 2001 per capita income was significantly higher in the Eastern county (Anji) than in the others (5.1times higher than in Pingjiang, the poorest and 3.8 times higher than in Muchuan). More importantly, Anjihas outperformed the national rate of GDP growth with an average of 12.1% for the period 1980–2001, thuscontributing to an increasing regional differentiation. The poorest county, Pingjiang, experienced the lowest

1The overall study included Anji and Longyou (Zhejiang Province), Pingjiang and Taojiang (Hunan Province) and Changning and

Muchuan (Sichuan Province).

ARTICLE IN PRESS

0N

Muchuan

Pingjiang

Anji

Land Cover in China

needleleaved deciduoud forest

needleleaved evergreen forest

broadleaved evergreen forest

broadleaved deciduoud forest

bush

sparse woods

seaside wet lands

alpine and sub_alpine meadow

slope grassland

plain grassland

desert grassland

meadow

city

river

lake

swamp

glacier

bare rocks

gravels

desert

farmland

alpine and sub-alpine plain gras

Mosaic of cropping

Forest Mosaic / Degraded Forest

Fig. 2. Land cover map of China and study areas.

B. Mertens et al. / Applied Geography 28 (2008) 16–31 19

rate of average annual GDP growth at 7.7%, while Muchuan had a yearly growth rate of 8.7% during thesame period.

Anji County is well connected to large external markets and big cities: the distance from Anji to Hangzhou(the capital of Zheijiang province) is about 70 km, and the distance to Shanghai (the largest Chinese city andmajor port) is about 230 km. The county area is 1887 km2 and the population density is 234 persons/km2.Eighty-seven percent of the population is classified as ‘‘rural’’. The transportation network is well developed,with a road density of 0.34 km/km2. About 60% of the county is covered by forests, of which about half ismonopodial (i.e., with a linear rhizome that expands longitudinally) bamboo. Most forest areas are located atthe periphery of the county, with a large central plain primarily used for agriculture. The contribution ofbamboo industry to total industry output increased dramatically in the 1980s and started to lose ground inrelative terms in the early 1990s, but still shows a strong average annual growth (Table 1). The bambooindustry is highly diversified in the county with many processing units making a remarkable range of products.Bamboo-based industrial development was instrumental in the early stages of rural industrialization but it hasbeen displaced by more advanced manufacturing industries as the industrialization process has matured in thismost developed county.

ARTICLE IN PRESS

Table 1

1985–2001 yearly increase in bamboo local processing, industry output and workers

% Yearly increase in local

processing

% Yearly increase in bamboo industry

output

% Yearly increase in bamboo industry

workers

Anji 23.1 170.7 39.2

Pingjiang �3.4 �1.9 7.2

Muchuan 0 87.9 17.5

Source: County level statistics, several years.


Pingjiang County is located in Hunan province within the Central region, 600 km west from the coast. Thecounty covers 4125 km2, with a population density of 238 persons/km2. Ninety-three percent of the populationis rural. The road network is less developed, with a 600 km of roads, or 0.16 km/km2. Forests cover 51% of thecounty, but only 7.2% of the forest is occupied by monopodial bamboo. Most bamboo is found in higherelevation areas in the southeast of the county. The bamboo industry in Pingjiang declined significantly withthe collapse of the Baiyu bamboo fan factory in the late 1980s. It was one of the largest such factories inChina, and the main consumer of bamboo in the county. Since then most bamboo produced in the county hasbeen sent to other counties to be processed. Pingjiang Forestry authorities have established a modern bamboo-flooring factory that started production in August 2001. Moreover, in the last couple of years the Baiyubamboo fan factory has resumed its activity, having combined with the Changshou fan factory and theMengya bamboo-mat factory. Consequently, the demand for raw material and the relative contribution ofbamboo industry to total industrial output in the county have increased significantly. The 2 million culms ofbamboo raw material harvested in Pingjiang in 2004 can no longer meet the demand of local processingfactories, so the local government has developed policies to prevent bamboo culms from being sold to othercounties.

Finally, Muchuan County is located in Sichuan province in the southwest region of the country. The countyarea is 1394 km2, population density is 177 persons/km2, and 91% of the population is rural. This county hadthe highest rate of expansion of bamboo area of the three counties studied. It is located at the border ofYunnan province, the province with the fastest bamboo expansion in the country in recent years. The roadnetwork is less developed, with about 145 km of roads, and the lowest road density with less than 0.1 km/km2.About 56% of the county is covered by forests, of which 38% is occupied by bamboo, both monopodial andsympodial (i.e., with a cluster type of growth around the rhizome). Mountains divide the county into two largewatersheds in the southeast and the northwest. As in the other counties, bamboo areas tend to be concentratedat higher elevations. A large majority of the bamboo plots are located in the southern part of the county,which has better transport infrastructure and more bamboo processing factories. The contribution of bambooindustry to total industrial output has been increasing in Muchuan. The paper industry is the main consumerof bamboo in the county, accounting for 99% of total bamboo industrial output. There is a long tradition ofsmall-scale paper manufacturing, and a modern factory with an output capacity of 57,000, using a high-proportion of bamboo raw material, came on line in 2001. There has also been a huge expansion of bambooplantations.

Methods for measuring spatial patterns of bamboo expansion

Landsat TM images with a resolution of 30m� 30m were acquired for three dates for each county: Anji(1984, 1995 and 1999), Pingjiang (1989, 1995 and 2000) and Muchuan (1988, 1996 and 2000). Dates wereselected, constrained by the availability of good quality images (i.e. relatively free of cloud cover), to capturepatterns and processes of land cover change associated with particular periods in terms of socio-economic andpolicy changes (HRS, bamboo trade liberalization, extension of HRS contracts, etc.). The institutionalreforms during the first period of observation (1980s and early 1990s) mostly focused on land reforms andchanges in industry (production and distribution systems), while the reforms in the late 1990s mostlyencouraged investment and export policies.

ARTICLE IN PRESS

Table 2

Classification accuracy

Study areas Time of images Overall accuracy (%) Kappa coefficient

Anji 1984 87.43 0.8573

1995 84.85 0.8184

1999 86.08 0.8393

Muchuan 1988 80.00 0.7744

1992 77.78 0.7530

2000 82.61 0.7792

Pingjiang 1989 73.11 0.6972

1995 75.45 0.7269

2000 74.75 0.7194


The time-series images were georeferenced and integrated in a Geographic Information System (GIS) forfurther combination and analysis with socio-economic and georeferenced environmental information. Imagepre-processing and classification were done using ERDAS IMAGINE. The images were geometricallycorrected by identifying ground control points in the original imagery and on the reference topographicalmaps. Quadratic polynomial transformation equations were then applied. The images were resampled byapplying a nearest-neighbor algorithm. The RMS error for this transformation was less than 0.5 pixel.

Supervised classifications were applied to produce a land cover map for each date. Signatures for each landuse category were generated using manually selected training samples for the most recent images and field dataand contemporaneous forest- and land use maps for older images. We used 10–15 training samples for eachcategory. Using these signatures, each pixel of the image was classified using the Maximum likelihoodParametric Rule and Parallelepiped Non-parametric Rule. The TM original bands 3, 4 and 5 were used asclassification bands (He, Chen, Chen, & Shi, 2001).

The original classification used 10 classes,2 which were then combined into four classes: bamboo, forest,agriculture and other. The minimum mapping unit is 900m2 (30� 30). An accuracy assessment comparedrandom points on the classified images with reference data. For the most recent images we used GCP dataobtained in the field. For the older images we used contemporaneous forest/land use maps and sampling datafrom local forestry bureau records. The overall accuracy and Kappa Coefficient for each images is reported inTable 2:

Based on this, a map of bamboo forest evolution was produced for each county. A mask image wasgenerated for pixels that were obscured by cloud cover for at least one observation, and these were excludedfrom the analyses. This operation was required for a consistent comparison of the bamboo areas throughspace and time. However, it also led to an underestimation of the actual area of bamboo in each countybecause cloud cover, ranging from 1.2% to 10.1% (Table 3) was disproportionately distributed over higherelevation areas where bamboo forests are more likely to occur. Based on the land cover maps, transitionprobability matrices were computed for each county and period of observation to capture the majortrajectories of land cover change.

Maps of roads, settlements, factories, soils and topography at 1:50,000 scale were collected and digitised tobe integrated in the GIS. A digital elevation model was built using contour lines on the topographic maps, andthat was used to derive a slope map.

We then examined the role of the main environmental and human factors expected to have influenced thelocation of the new bamboo plantations. We used binary logistic regression analyses (Eq. (1)) to assessthe relationship between bamboo forest and these independent spatial variables, and their evolution across theperiods of analysis. The parameters of the equations for each period and study area were estimated on the

2(1) Conifer forest; broadleaf forest; saplings; shrub (combined as ‘‘forest’’); (2) Bamboo; (3) Agriculture; (4) Water, bare land, unused,

residential, tea [plantation (combined as ‘‘other’’).

ARTICLE IN PRESS

Table 3

Bamboo cover changea

Remote

sensing

%

Cloudiness

County

statistics

Remote

sensing

%

Cloudiness

County

statistics

Remote

sensing

%

Cloudiness

County

statistics

Anji (Year) 1984 1995 1999

Bamboo area

(km2)

463.44 4.23 543.66 457.61 6.69 579.33 583.57 1.23 620.13

Yearly change

(km2/y)

�0.65 3.24 31.49 10.20

Pingjiang (Year) 1989 1995 2000

Bamboo area

(km2)

57.59 0.37 129.33 106.24 0.90 152.00 93.52 0.45 158.67

Yearly change

(km2/y)

8.11 3.78 �2.54 1.33

Muchuan (Year) 1988 1992 2000

Bamboo area

(km2)

82.65 0.40 73.67 76.07 0.05 133.94 140.38 0.14 300.33

Yearly change

(km2/y)

�1.64 15.07 8.04 20.80

aLocations covered by clouds for at least one date of observation were not considered in computing the values for bamboo area for each

date and yearly change in bamboo cover. The values in the table therefore underestimate the actual value.


basis of a systematic sampling of points:

LogitðpÞ ¼ ln½p=ð1� pÞ� ¼ b0 þX

bjX j and p ¼ PðY ¼ 1Þ, (1)

where P means probability, Y is the binary dependent variable (1 means occurrence), Xj the independentvariables and b the estimated parameters

The dependent variable is the occurrence/absence of bamboo plantations at a given date. The independentvariables were selected to capture environmental conditions (elevation and slope) and market accessibility(computed as the distance from the nearest road, city and bamboo factories).

Expansion of bamboo plantations in the study areas

The satellite data reveals a high increase of bamboo area in all counties (Table 3). For the whole period,bamboo area increased from 463 to 583 km2 in Anji (25%), from 57 to 93 km2 in Pingjiang (38%) and from 82to 140 km2 in Muchuan (41%). The measured bamboo area closely matches the 1999 county forest bureaustatistics for Anji and Pinjiang. The official Muchuan statistics indicate a much larger bamboo area (255 km2)(Table 4). This discrepancy is due to the fact that bamboo plantations have expanded dramatically in thiscounty during the 1990s, and it takes 3 years for the young plantations, especially of the sympodial bamboosplanted in Muchuan, to be developed enough to be recognizable in the satellite images. The official statisticsfor bamboo area in 1996 are closer to that measured by remote sensing.

During the first period of observation, the amount of land covered by bamboo did not change significantlyat the county level in Anji and Muchuan, while the area doubled in the early 1990s in Pingjiang (Table 3). Inthe second period (late 1990s), bamboo area increased by 25% in Anji, and nearly doubled in Muchuan. Forthe same period (late 1990s), bamboo area decreased slightly in Pingjiang, although large areas of bamboo forthe last date of observation were masked by the accumulated cloud cover image, so this is not definitive: therecould have been a stable or even slightly increasing area of bamboo during this period.

In Anji county, the land cover transition probability matrix shows the major role played by forest inproviding new areas for bamboo extension: about 23% and 30% of the forests shifted to bamboo plantationduring the two periods of observation. But for the same periods a proportion of the bamboo plantationsshifted back to forest cover. While the transitions from agriculture or other non-forest land cover to bamboo

ARTICLE IN PRESS

Table 4

Transition probability matrix of land use types in Anji, Muchuan and Pingjiang County during 1984–2000 (%)

Forest Bamboo Agriculture Others

(a) Anji County during 1984– 1995

Forest 59.0 23.9 13.7 3.4

Bamboo 35.8 60.2 2.5 1.5

Agriculture 8.6 1.7 85.5 4.2

Others 33.4 5.8 32.8 28.0

(b) Anji County during 1995– 1999

Forest 61.8 30.4 6.1 1.7

Bamboo 25.4 73.3 0.6 0.7

Agriculture 13.3 6.3 72.5 7.9

Others 14.4 19.5 9.1 57.0

(c) Muchuan county during 1988– 1992

Forest 64.05 8.48 26.79 0.68

Bamboo 77.11 17.26 5.48 0.15

Agriculture 33.63 0.99 63.06 2.33

Others 31.61 3.05 56.06 9.29

(d) Muchuan county during 1992– 2000

Forest 74.42 10.47 11.77 3.33

Bamboo 71.09 21.51 5.55 1.84

Agriculture 37.75 8.76 43.91 9.57

Others 16.37 4.13 36.47 43.03

(e) Pingjiang county during 1989– 1995

Forest 88.23 2.73 5.75 3.29

Bamboo 55.28 44.50 0.15 0.08

Agriculture 55.12 0.03 30.59 14.27

Others 37.65 0.22 19.61 42.52

(f) Pingjiang county during 1995– 2000

Forest 92.52 0.49 3.84 3.16

Bamboo 31.65 67.82 0.02 0.51

Agriculture 52.42 0.01 43.38 4.19

Others 47.61 0.00 14.19 38.20


plantation was marginal for the first period, respectively, 7% and 20% of agriculture and other non-forestland cover shifted to bamboo during the 1990s.

The trajectory was different in Muchuan county. Most forest conversion was to agriculture. The transitionsduring the second period show a slight increase of forest conversion to bamboo, but a major increase inconversion of agricultural land and other non-forest land cover to bamboo. This finding is consistent with theimplementation of the ‘‘Conversion of Arable Land in Steep Slopes’ policy and other land restorationmeasures in Muchuan for over a decade, which has encouraged afforestation (frequently with bamboo) onsloping lands of more than 251 at the expense of low yield agriculture.

Finally, most of the bamboo expansion in Pingjiang county took place on former forest land during the firstperiod of observation. Bamboo area then decreased slightly during the second period, to be replaced by forestland. Very little agriculture land or other non-forest land shifted to bamboo. The decrease of bamboo areacoincides with the crisis created by the closing of the bamboo fan factory already mentioned, that was themain consumer of bamboo raw material in this county.

As shown in Fig. 3, these differentiated trends are consistent with the evolution of the gross productionvalue of bamboo forestry outputs. The bamboo production increased slowly and almost linearly between 1990and 2000 in the counties of Anji and Muchuan, with a higher increase in the late 1990s, particularly inMuchuan, indicating an acceleration in the expansion of bamboo. For Pingjiang County, which shows a sharp

ARTICLE IN PRESS

Anji

0

200

400

600

800

19

80

19

83

19

86

19

89

19

92

19

95

19

98

20

01

Ba

mb

oo

are

a (

km

2)

Ba

mb

oo

are

a (

km

2)

Ba

mb

oo

are

a (

km

2)

Ba

mb

oo

are

a (

km

2)

Ba

mb

oo

are

a (

km

2)

Ba

mb

oo

are

a (

km

2)

0

50

100

150

200

250

GP

V (

mill

ion Y

yan R

MB

)

0

200

400

600

800

19

80

19

83

19

86

19

89

19

92

19

95

19

98

20

01

0

20

40

60

80

100

Pro

po

rtio

n o

f fo

restr

y o

utp

ut (%

)

Pingjiang

0

20

40

60

80

100

120

19

80

19

83

19

86

19

89

19

92

19

95

19

98

20

01

0

500

1000

1500

2000

2500

3000

3500

GP

V (

10

,00

0 Y

ua

n R

MB

)

0

20

40

60

80

100

120

19

80

19

83

19

86

19

89

19

92

19

95

19

98

20

01

0

20

40

60

80

100

Pro

po

rtio

n o

f fo

restr

y o

utp

ut (%

)

Muchuan

0

20

40

60

80

100

120

140

160

19

80

19

83

19

86

19

89

19

92

19

95

19

98

20

01

0

1000

2000

3000

4000

5000

6000

7000

GP

V (

10

,00

0 Y

ua

n R

MB

)

Bamboo area (based on remote sensing)

Bamboo Gross production value (County statistics)

0

20

40

60

80

100

120

140

160

19

80

19

83

19

86

19

89

19

92

19

95

19

98

20

01

0

10

20

30

40

50

60

70

80

90

100

Pro

po

rtio

n o

f fo

restr

y o

utp

ut (%

)

Bamboo area (based on remote sensing)

Bamboo forestry output (County statistics)

Fig. 3. Bamboo cover change and (a) bamboo forestry output and (b) bamboo proportion of total forestry output, at the county level.


increase in bamboo cover in the early 1990s, the evolution of the bamboo production also shows a hugeincrease in 1992, followed by stagnation until 2000, with a temporary decrease in 1996.

A regression analysis (Table 5) illustrates the relationships between bamboo output and value (dependentvariables) and bamboo area (independent variable). All models are highly significant, but Anji shows thelowest regression value, consistent with the fact that bamboo area in Anji has expanded the least, and part ofthe increment in bamboo production has been achieved through increased productivity, through more

ARTICLE IN PRESS

Table 5

Regression models

R R2 Corrected R2 F Sig.

(a) Mosoa bamboo area (independent) to mosoa bamboo output in culms (dependent)

Anji 0.713 0.509 0.484 20.735 0.000

Pingjiang 0.905 0.820 0.811 90.907 0.000

Muchuan 0.974 0.948 0.945 362.996 0.000

(b) Bamboo area (independent) to bamboo forestry value in million yuan (dependent)

Anji 0.867 0.751 0.739 60.328 0.000

Pingjiang 0.631 0.398 0.368 13.218 0.002

Muchuan 0.976 0.953 0.950 401.144 0.000

(c) Bamboo area (independent) to % bamboo on forestry value (dependent)

Anji 0.612 0.375 0.344 12.005 0.002

Pingjiang 0.188 0.035 �0.013 0.736 0.401

Muchuan 0.900 0.811 0.801 85.543 0.000

aMoso is the Chinese name of Phyllostachys heterocycla, var. pubescens, the dominant planted bamboo in the region, highly appreciated

for the size and quality of its culms.


intensive management. The significant positive relationship between expansion of bamboo area and increasein bamboo forestry value (in constant yuan of 1990), with Muchuan showing the strongest relation followedby Anji and Pingjiang, reflects the increase in surface area in Muchuan, while in Anji the increase is due tocombined expansion and intensification (less land available for expansion). In Pingjiang, bamboo plantingcontinued despite the closing of key enterprises in the early 1990s as mentioned in the manuscript. With propermarket mechanisms, spatial expansion of bamboo leads to increased output value and increased share of totalforestry value (Muchuan and Anji) whereas in case of trouble like Pingjiang, increased spatial expansion ofbamboo was not followed by increased share of bamboo in forestry output.

There is a dynamic interaction between forest and bamboo area, with some areas shifting from onedominant cover to the other depending on prevailing policies and markets. Marginal agricultural lands andwastelands have served as reservoirs for forest and bamboo expansion. This is consistent with the findings ofremote sensing-based studies in other regions in China (Wan, Zhang, Zhou, Zhao, & Zhang, 2002; Xu & Li,2003; Zeng, Tang, & Guo, 1999).

Spatial patterns and processes of bamboo expansion

This last section examines the role of the main environmental and infrastructure factors that have influencedthe location of new bamboo plantations in these three counties. We use logistic regression models to analyzethe spatial inertia (or degree of proximity of new bamboo plantations to already existing plantations), theeffect of altitude and height as determining environmental factors and the role of road infrastructure andproximity to urban areas to explain the placement of the new bamboo plantations in each county. The resultsof the logistic regressions for each county are given in Tables 6–8.

Spatial inertia

The most important factors characterising the location of new bamboo areas is proximity to previouslyexisting bamboo plantations. Bamboo expansion tends to be concentrated in areas that already haveestablished bamboo plantations. This is most pronounced in Muchuan, where 96% (period 1) and 98%(period 2) of new bamboo plantation took place within a 1 km radius of existing bamboo plots. In Anji, for theperiods 1984–1995 and 1995–2000, respectively, 87% and 95% of the new bamboo area was established within1 km of existing plots. Almost no bamboo plantations were created more than 4 km from previously existingplantations. In Pingjiang, this spatial inertia is slightly weaker: for the first period (1989–1995), less than 75%of the new bamboo areas were established within 1 km of existing plots and new plantations spread over

ARTICLE IN PRESS

Table 6

Anji county—parameters from logistic regressiona

Variable b S.E. (b) Wald w2 Sig. R Exp (b)

(a) Bamboo occurrence in 1984

Road �0.3834 0.0610 39.4446 0.0000 �0.0682 0.6816

Town 0.1263 0.1042 1.4697 0.2254 0.0000 1.1346

Elevat 0.0013 0.0003 14.8413 0.0001 0.0399 1.0013

Slope 0.1347 0.0057 562.4029 0.0000 0.2637 1.1442

Factory 0.1101 0.0124 78.5044 0.0000 0.0974 1.1164

Constant �3.0431 0.0726 1758.415 0.0000

Pseudo R2¼ 0.344

(b) Bamboo occurrence in 1995

Road �0.2686 0.0524 26.2773 0.0000 �0.0473 0.7644

Town 0.1491 0.0941 2.5087 0.1132 0.0068 1.1608

Elevat 0.0024 0.0003 60.0974 0.0000 0.0731 1.0024

Slope 0.1301 0.0051 659.1855 0.0000 0.2460 1.1390

Factory 0.0678 0.0109 38.9048 0.0000 0.0583 1.0702

Constant �2.5893 0.0616 1766.757 0.0000

Pseudo R2¼ 0.394

(c) Bamboo occurrence in 1999

Road �0.2631 0.0454 33.6327 0.0000 �0.0476 0.7687

Town 0.1852 0.0848 4.7730 0.0289 0.0141 1.2035

Elevat 0.0027 0.0003 88.1056 0.0000 0.0785 1.0027

Slope 0.1120 0.0046 591.0577 0.0000 0.2054 1.1186

Factory 0.0683 0.0096 50.3599 0.0000 0.0589 1.0707

Constant �2.0220 0.0512 1559.851 0.0000

Pseudo R2¼ 0.366

ab: Maximum likelihood estimate of the parameter, S.E. (b): estimated standard error of the parameter estimate, Wald w2: Wald chi-

squared statistic, Sig.: p-value of the Wald chi-squared statistic, R: standardized coefficients, and Exp (b): odd ratio.


longer distance. Still, 95% of the new bamboo plantations are located in areas at less than 8 km from thepreviously existing plantations. New bamboo plantations were more concentrated in the second period, when95% of the new bamboo plantations took place within a distance of 1 km for the existing plots.

The relatively smaller area of established bamboo in Pingjiang resulted in a more dispersed pattern ofexpansion, whereas in Anji and Muchuan, with large areas of existing bamboo plantations, there was acontagious expansion concentrated in established bamboo areas.

Bamboo plantations tend to be concentrated in specific areas, and this concentration is increasing asbamboo keeps expanding around previously established plantations. Now we consider why bambooplantations cluster in these specific areas.

Environmental conditions

The regression analysis (Tables 6–8) shows that higher elevation and/or steeper slope are the mostimportant variables determining the location of bamboo plantations for all counties and all periods ofobservation. Bamboo plantations in Anji are more likely to occur at higher elevation, and the probabilityincreases over time. Between 1984 and 1999, the proportion of bamboo in areas higher than 500m increasedfrom 7% to 13%. This was due to replacement of natural forests by plantations as well as to the establishmentof bamboo plantations in degraded highlands. However, with land scarcity and new environmentalregulations that prohibit converting natural forests, the proportion of new bamboo plantations located inlower areas is increasing. Across the whole period of analysis, the proportion of bamboo plantations in thelarge central plains, which include almost half of the county area, increased from 6% to 13%.

ARTICLE IN PRESS

Table 7

Pingjiang county—parameters from logistic regressiona

Variable b S.E. (b) Wald w2 Sig. R Exp (b)


Road 0.2840 0.0561 25.6464 0.0000 0.1030 1.3284

Town 0.0778 0.0322 5.8223 0.0158 0.0414 1.0809

Elevat 0.0040 0.0003 132.1451 0.0000 0.2416 1.0040

Slope 0.0244 0.0072 11.4280 0.0007 0.0650 1.0247

Factory �0.0453 0.0113 16.1596 0.0001 �0.0797 0.9557

Constant �7.1835 0.3080 543.9922 0.0000

Pseudo R2¼ 0.382


Road 0.2599 0.0317 67.2932 0.0000 0.1128 1.2968

Town 0.0391 0.0203 3.6965 0.0545 0.0182 1.0399

Elevat 0.0032 0.0003 160.6643 0.0000 0.1758 1.0032

Slope 0.0284 0.0042 45.9733 0.0000 0.0925 1.0288

Factory �0.0376 0.0064 34.7987 0.0000 �0.0799 0.9631

Constant �5.3409 0.1578 1145.201 0.0000

Pseudo R2¼ 0.298


Road 0.2828 0.0288 96.2961 0.0000 0.1220 1.3268

Town 0.0469 0.0184 6.4829 0.0109 0.0266 1.0481

Elevat 0.0034 0.0002 207.5381 0.0000 0.1802 1.0034

Slope 0.0270 0.0038 50.5060 0.0000 0.0875 1.0274

Factory �0.0436 0.0059 55.1666 0.0000 �0.0916 0.9574

Constant �5.2166 0.1456 1283.867 0.0000

Pseudo R2¼ 0.351




Elevation is also the main explanatory factor in Pingjiang and Muchuan, but it is becoming less importantover time. In Pingjiang for instance, the proportion of total bamboo area between 100–500m increased from14% to 40%, while the proportion of bamboo area from 500 to 1000m decreased from 76% to 51%.However, with only 11% of the county located in areas higher than 500m, by 1999 more than half of allbamboo plantations were still found in this zone. The same process took place in Muchuan where theproportion of bamboo plantations in the areas between 500 and 1000m and higher than 1000m, respectivelyincreased from 39% to 52%, and decreased from 59% to 44%. The expansion of bamboo on former arableland together with the decreasing yield of bamboo at altitudes above 1000m explains the observed pattern inMuchuan.

Slope is also important, with a significantly higher probability of bamboo occurrence on steeper slopes in allcounties and periods. However, in Pingjiang and Muchuan, this variable is relatively weak in explainingbamboo location as compared to the other variables considered (see Wald values in Tables 7 and 8). In thesetwo counties the proportion of bamboo plantation decreased in the steepest slope areas (more than 251), andincreased slightly in the two other categories (less than 101 and 10–251). In Anji, the proportion of bambooplantations increased slightly on both flat agricultural land and steep forest or wasteland, but decreased in themedium slope category (10–251) where bamboo was already dominant.

The time-trend differences between Anji, on the one hand, and Pingjiang and Muchuan, on the other hand,relate to levels of development and relative importance of bamboo. Anji is a mature bamboo county wherebamboo plantations represent a large component of total forest area. It is also more developed. Therefore,bamboo has to expand in strong competition with other landuses and close to a level of saturation. InPingjiang, bamboo is not a dominant landuse so there is a larger potential for expansion and more

ARTICLE IN PRESS

Table 8

Muchuan county—parameters from logistic regressiona

Variable b S.E. (b) Wald w2 Sig R Exp (b)


Road 0.1055 0.0165 40.8006 0.0000 0.0799 1.1113

Town �0.0453 0.0231 3.8419 0.0500 �0.0174 0.9557

Elevat 0.0037 0.0002 256.0320 0.0000 0.2044 1.0037

Slope 0.0165 0.0034 23.9067 0.0000 0.0600 1.0166

Factory �0.1056 0.0094 127.0775 0.0000 �0.1435 0.8998

Constant �5.5374 0.1918 833.1703 0.0000

Pseudo R2¼ 0.261


Road 0.1050 0.0136 59.6334 0.0000 0.0837 1.1107

Town �0.0298 0.0200 2.2218 0.1361 �0.0052 0.9706

Elevat 0.0034 0.0002 282.2885 0.0000 0.1846 1.0034

Slope 0.0207 0.0028 54.5118 0.0000 0.0799 1.0209

Factory �0.1044 0.0075 191.4996 0.0000 �0.1518 0.9009

Constant �4.9126 0.1607 934.1663 0.0000

Pseudo R2¼ 0.269


Road 0.0784 0.0111 49.9978 0.0000 0.0678 1.0816

Town �0.0357 0.0181 3.8699 0.0492 �0.0134 0.9650

Elevat 0.0029 0.0002 269.8643 0.0000 0.1601 1.0029

Slope 0.0176 0.0024 55.0720 0.0000 0.0713 1.0178

Factory �0.0860 0.0058 219.3915 0.0000 �0.1443 0.9176

Constant �4.0261 0.1336 908.2049 0.0000

Pseudo R2¼ 0.214




opportunity for expansion on less marginal land. In Muchuan, the policy of conversion of arable land toforestry implemented for over a decade has also allowed bamboo to expand in less marginal areas.

Accessibility

Proximity to main towns had the least influence of the variables tested in all counties and periods. InPingjiang, there is some increased likelihood that new bamboo plantations will be established in locationsfarther from the cities. The trends are similar for all counties and periods, although in Pingjiang, the countywith the lowest cover of bamboo, the trend is becoming weaker, with an increasing proportion of new bambooplantations in areas at shorter distance from the main cities in the period between 1989 and 2000.

In examining the role of road infrastructure, we must first emphasize the differences in road networks anddensity in the three counties (Fig. 4). The road density (km/km2) is 0.36 in Anji, 0.16 in Pingjiang and 0.10 inMuchuan. Fig. 4 shows that the cumulative proportion of the total bamboo area increased sharply with thedistance from roads in Anji. This pattern is also positive but less pronounced in Pingjiang and Muchuan(where about 20% of the county is located at more than 10 km from the roads). Moreover, the road network ismuch more developed across the whole county in Anji, while still limited in Pingjiang and Muchuan.

The results indicate that new bamboo plantations in Anji tend to be relegated to more remote areas, while inPingjiang and Muchuan the new plantations tend to occupy more accessible areas. Because of the high roaddensity in the Anji, bamboo plots tend to be located near the roads. Still, the probability of occurrence ofbamboo plantations, as compared to other landuses, decreases rapidly with the distance from roads(b ¼ �0.38). And in recent years bamboo expansion has tended to take place in less accessible areas, at alonger distance from roads. The opposite conditions and trends are observed in Pingjiang and Muchuan. Here

ARTICLE IN PRESS

Road network County

Length (km) Density (km/km2)

Anji 649.38 0.36

Pingjiang 654.17 0.16

Muchuan 145.37 0.10

Log Distance (km)

2010864210.80.60.4

Accu

mu

late

d c

ou

nty

are

a (

%)

100

80

60

40

20

0

Anji

Pingjiang

Muchuan

Fig. 4. Accumulated proportion of the counties’ area and distance from the road network.


bamboo plantations are generally found in less accessible areas, far from the main roads, but new bambooplantations are slowly increasing in more accessible areas, closer to the main roads. This is consistent with theabove-mentioned differences in county conditions that result in a differentiated pattern of expansion, towardsthe marginal, remote areas in Anji and towards more middle-level altitude areas in Pingjiang and Muchuan(albeit for different reasons in each county).

The conclusions on the role of proximity to the bamboo-based factories are more difficult to draw becausewe do not have good data on the size or age of bamboo processing factories in each county. Therefore, it isimpossible to fully assess the importance of this variable, or whether concentrations of bamboo plantationshave actually attracted investment in the processing sub-sector, or if installed processing capacity (anddemand for raw material) has provided incentives for the expansion of bamboo plantations nearby. Theresults indicate that factories tend to be located within or nearby the bamboo production areas in Pingjiangand Muchuan, while bamboo plantations tend to be located at longer distance from factories in Anji, ascompared to other landuses. Across the time periods, these patterns tend to reverse, as bamboo plantationstend to get closer to the factories in Anji (a concentration of the new bamboo plots is observed around two ofthe three main bamboo-related factories in the extreme south of this county) and at longer distance inMuchuan. The pattern is virtually static in Pingjiang, although it can be expected to change in response todemand from the new pulp mill.

Again the type of industry and general level of development in the county seem to influence this pattern. TheAnji bamboo industry is very large (around 1000 processing units, from family workshops to large factories),diverse and sophisticated, including low and high value-added products. Anji has also more developedtransportation infrastructure, which facilitates the movement of the abundant bamboo raw material in thecounty. This offers more opportunities to the bamboo industry and reduces the raw material transportationcosts. The combination of many factories distributed widely in the county along with good infrastructureconditions explains the regression coefficients found in Table 6.

Pingjiang and especially Muchuan bamboo processing is very specialized around very few products. Bothhave fewer factories than Anji and tend to be more concentrated around few towns. Muchuan is an extreme

ARTICLE IN PRESSB. Mertens et al. / Applied Geography 28 (2008) 16–3130

case where most bamboo is used as raw material for a low value-adding paper industry that is very sensitive totransportation costs. Moreover, both counties have less developed transportation infrastructure than Anji. Inthese conditions transportation costs represent a significant part of total costs, thus binding the industry andbamboo plantations closer to each other.

Conclusion

There has been a large expansion in the bamboo sector in China in the past two decades, and this hastranslated into significant land use and land cover changes. Examining the details of these changes in threebamboo counties, each with very different economic contexts, provides us with some important insights on thepatterns and the determinants of change.

All three counties experienced an expansion in bamboo area during the study period, driven by increasingdemand for raw material by local processing industries. This expansion came at the expense of forest area(especially in Anji and Pingjiang) and, in recent periods, from sloping agricultural lands and degraded lands.The trend to clear forest to establish bamboo has been slowed markedly by regulations and strongerenforcement to prevent conversion of natural forests. In all three counties, new bamboo plantations wereconcentrated in and around already-established bamboo areas. This is logical in that these areas haveappropriate environmental and economic conditions—the analysis showed a concentration of bamboo nearprocessing centers—and local people have the appropriate skills. It is much more common to have bambooplantations at higher elevations and on sloping lands that are less suitable for other uses, especially agriculture.This trend has been strongly reinforced by regulations and incentives that encourage afforestation andreforestation on sloping lands.

The expansion of bamboo, replacing forests and, more recently, on marginal, sloping, agricultural land inthe three counties, epitomizes the general trend observed in the bamboo sector in China. As in the case of thestudy counties, a combination of increasing demand, fast growth, quick returns and appropriate policies haveaccelerated an already dynamic sector. The paper has identified some of the spatial characteristics of thisprocess and their relationship with local conditions that could be relevant in future bamboo (and forestrymore generally) sector planning and development. We stress the close link between plantations and processingindustry for a normally low value raw material. Likewise, the importance of ‘seed areas’ where plantations andsome associated industry already exists seem to be key in the success and potential expansion of the resource.

The more detailed patterns of distribution and expansion are determined by environmental factors,infrastructure and level of economic development, as well as by policies that influence land use decisions andthat have driven the expansion of the bamboo industry (the latter discussed by Ruiz-Perez et al., 2003). Wherebamboo is an economically competitive land use option, some expansion has taken place even in low-lying andlevel areas suitable for irrigated agriculture. This was true, for example, during the bamboo boom period inAnji county in the 1990s. More recently, with a mature bamboo sector where bamboo plantations havealready taken most of the available land, there is less opportunity to expand the bamboo area and much of thegrowth in output has come through intensification and increased productivity. And, as Anji has developed andbamboo growing has become relatively less attractive compared to the available alternatives, new expansionhas occurred in less accessible, less valuable lands, further from roads. In Muchuan and especially Pingjiang,both of which have much less developed economies and which have large areas available for bambooexpansion, the pattern is quite different. New plantations are more likely to be established in more accessibleareas. Policies aimed at converting steeply sloping agricultural land and degraded lands to forests reinforcethis pattern.

This comparison shows clearly the importance of local conditions in determining the specific spatial patternsof land use change. There is a strong element of spatial inertia. In the bamboo sector, the availability of rawmaterial supplies attracts investment in processing capacity, and industrial demand coupled with localexperience with bamboo raw material production encourages local expansion. General levels of economicdevelopment affect the opportunity cost of land and labour. Where bamboo growing remains an economicallysuperior option, bamboo expansion takes place on higher quality and/or more accessible lands. As the relativeimportance of the sector declines, and as the physical availability of land is reduced, as in Anji, newplantations are squeezed onto more remote and less valuable land. The policy environment provides the larger

ARTICLE IN PRESSB. Mertens et al. / Applied Geography 28 (2008) 16–31 31

context, affecting investment in the processing sector and, as shown strongly in China, directly influencingland use choices, in this case on sloping lands. These findings support the need for a careful understanding ofspatial processes when preparing new national level policies, local regulations and other measures to reinforceand to expand China’s very dynamic forest sector.

Acknowledgements

This research was made possible by support from the Spanish Cooperation Agency (AECI) and fromCIFOR’s core supporters. The authors equally shared the effort and responsibility of conducting andreporting this research.

References

Bull, G. Q., & Nilsson, S. (2004). An assessment of China’s forest resources. International Forestry Review, 6(3–4), 210–220.

Gu, D. Q. (1992). The influence of bamboo industry on bamboo growing areas. In S. Zhu, W. Li, X. Zhang, & Z. Wang (Eds.), Bamboo

and its uses. International symposium on industrial use of bamboo (pp. 329–332). Beijing, China: ITTO-CAF.

He, Chun-yang, Chen, Jin, Chen, Yun-hao, & Shi, Pei-jun (2001). Land use/cover change detection based on hybrid method. Journal of

Natural Resources, 16(3), 255–262.

Hyde, W. F., Xu, J., & Belcher, B. (2003). Introduction. In W. F. Hyde, B. Belcher, & J. Xu (Eds.), China’s forests: Global lessons from

market reforms (pp. 1–21). Washington, DC: Resources for the Future Press.

Li, S., & Xu, C. (1998). China’s bamboo development process and development strategies towards the 21st century. Journal of Bamboo

Research, 17(1), 1–5 (in Chinese).

Li, W. (2004). Degradation and restoration of forest ecosystems in China. Forest Ecology and Management, 201, 33–41.

Lin, S. (1999). Wood supply channels after logging ban. Forestry Economics, 113(1), 41–42 (in Chinese).

Ruiz-Perez, M., Belcher, B., Fu, M., & Yang, X. (2003). Forestry, poverty and rural development: Perspectives from the bamboo sector. In

W. F. Hyde, B. Belcher, & J. Xu (Eds.), China’s forests: Global lessons from market reforms (pp. 151–176). Washington, DC: Resources

for the Future Press.

Ruiz-Perez, M., Belcher, B., Fu, M., & Yang, X. (2004). Looking through the bamboo curtain: An analysis of the changing role of forest

and farm income in rural livelihoods in China. International Forestry Review, 6(3–4), 306–316.

State Forest Administration. (2000). Forest resources statistics of China (1994– 1998). Beijing: Department of Forest Resources

Management (in Chinese).

State Forest Administration. (several years). China forestry yearbook. Beijing: China Forestry Publishing House (in Chinese).

Sun, X., Katsigris, E., & White, A. (2004). Meeting China’s demand for forest products: An overview of import trends, ports of entry, and

supplying countries, with emphasis on the Asia-Pacific region. International Forestry Review, 6(3–4), 227–236.

Verburg, P. H., & Chen, Y. (2000). Multiscale characterization of land-use patterns in China. Ecosystems, 3(4), 369–385.

Wan, S., Zhang, Z., Zhou, Q., Zhao, X., & Zhang, Z. (2002). Dynamic change of forest land and grassland based on RS and GIS. Ziyuan

Kexue (Resource Science), 24(5), 64–69.

Xu, Y., & Li, X. (2003). Dynamic change of land use in Hebei southern plain based on GIS. Ziyuan Kexue (Resources Science), 25(2),

77–84.

Zeng, H., Tang, J., & Guo, Q. (1999). A landscape study of element transferring pattern and changing stage of Changping to WN in

eastern part of Zhujiang delta area. Dili Xuebao (Scientia Geographica Sinica), 19(1), 73–77.

spatial patterns and processes of bamboo expansion in...

Documents