Deforestation and Fragmentation of Natural Forests in the Upper Changhua Watershed, Hainan, China: Implications for Biodiversity Conservation

Environ Monit Assess (2015) 187:4137 DOI 10.1007/s10661-014-4137-3

Deforestation and fragmentation of natural forests in the upper Changhua watershed, Hainan, China: implications for biodiversity conservation

De-Li Zhai & Charles H. Cannon & Zhi-Cong Dai & Cui-Ping Zhang & Jian-Chu Xu

Received: 22 March 2014 /Accepted: 3 November 2014 # Springer International Publishing Switzerland 2014

Abstract Hainan, the largest tropical island in China, change were observed in protected areas, indicating a belongs to the Indo-Burma biodiversity hotspot. The lack of enforcement. Natural forests conversion to rub- Changhua watershed is a center of endemism for plants ber and pulp plantations has a general negative effect on and birds and the cradle of Hainan’s main rivers. How- biodiversity, primarily through habitat fragmentation. ever, this area has experienced recent and ongoing de- The fragmentation analysis showed that natural forests forestation and habitat fragmentation. To quantify hab- area was reduced and patch number increased, while itat loss and fragmentation of natural forests, as well as patch size and connectivity decreased. These land- the land-cover changes in the Changhua watershed, we cover changes threatened local biodiversity, espe- analyzed Landsat images obtained in 1988, 1995, and cially island endemic species. Both natural forests 2005. Land-cover dynamics analysis showed that natu- losses and fragmentation should be stopped by ral forests increased in area (97,909 to 104,023 ha) from strict enforcement to prevent further damage. Pre- 1988 to 1995 but decreased rapidly to 76,306 ha over serving the remaining natural forests and enforcing the next decade. Rubber plantations increased steadily the status of protected areas should be a manage- throughout the study period while pulp plantations rap- ment priority to maximize the watershed’sbiodi- idly expanded after 1995. Similar patterns of land cover versity conservation value.

Keywords Deforestation . Fragmentation . lack of clear understanding of the land-cover changes Connectivity. Plantation forest . Biodiversity and status of fragmentation makes it difficult for conservation management practices to prevent further loss of habitat and biodiversity. Introduction In this study, we analyzed the spatial-temporal patterns of land cover and natural forests cover over three Tropical rainforests are seriously threatened by anthro- time points (1988, 1995, and 2005), using remote sens- pogenic forest degradation and deforestation (Laurance ing techniques, in the upper reaches of the Changhua 1999) which is resulting in a global biodiversity crisis watershed, Hainan, China. The objectives of this study (Anderson and Jenkins 2006). Tropical Asia is an im- were as follows: (1) to determine land-cover changes portant biodiversity hotspot (Myers et al. 2000). How- and associated habitat changes in the study area between ever, tropical Asia suffers serious forest loss and frag- 1988, 1995, and 2005; (2) to evaluate natural forests mentation due to land use and land cover changes, e.g., fragmentation using landscape indices; and (3) to dis- plantations displacing natural forests (Aziz et al. 2010; cuss the implications of these patterns for biodiversity Sodhi et al. 2010). Natural vegetation cover loss, forest conservation. fragmentation, and associated land-use and land-cover changes are the main processes causing global environmental change (Turner et al. 1995)andareresponsible Methods for the loss of biodiversity worldwide (Soule 1991;Sala et al. 2000). Forest fragmentation has become increasing Study area important for global change researches (Gibbs et al. 2010). It has been shown that land-cover analysis can Hainan Island (18° 10′–20° 10′ N and 108° 37′–111° 03′ discriminate specific habitat types and predict species E) (Fig. 1a) is the largest tropical island in China, with distribution at large spatial scales (Wang et al. 2010). an area of 33,920 km2. The island has a tropical mon- Therefore, although it is difficult to directly quantify soon climate, influenced by the southeast monsoon from biodiversity in a landscape, the sensitivity of biological the west Pacific Ocean and southwest monsoon from the communities to land-cover changes and the availability Indian Ocean in the summer and the northeast monsoon of appropriate habitat make it possible to project biodi- in winter, resulting in a rainy season from May to versity loss by land-cover conversion and habitat pa- October and a dry season from November to April rameters, derived from remote sensing studies. (Luo 1985). The annual mean temperature is 22–23 °C Hainan Island is the largest tropical island in China with 1800–2500 mm precipitation and 1800 mm evap- and the only major island in the Indo-Burma biodiver- oration annually (Luo 1985). The study area is the upper sity hotspot (Myers et al. 2000). The island includes two Changhua Watershed of Hainan Island (Fig. 1b), includ- of the World Wide Fund for Nature’s (WWF) Global ing parts of Wuzhishan National Nature Reserve (WZS) 200 ecoregions: the Hainan Monsoon Rain Forests and and Yinggeling Nature Reserve (YGL), with its eleva- the coastal plains, both of which are part of the extensive tion ranging from 169 to 1860 m above sea level South China-Vietnam Subtropical Evergreen Forests. (Fig. 1c). The interior uplands of the island are charac- Both its flora and fauna have high level of overall terized by great habitat diversity and are one of China’s diversity and endemism (Chan et al. 2005;Wangetal. most important biodiversity conservation areas (Feng 2006;Chen2008). Moreover, Hainan shares a strong et al. 1999; Francisco-Ortega et al. 2010). The historical and biotic connection with Southeast Asia, Changhua watershed is the cradle of Hainan’s main especially the Philippines and Malaysia, rather than rivers (Feng et al. 1999) and has been identified as a mainland China, indicated by the distribution patterns center of endemism for plants and birds (Wang et al. and genetic analysis of plants and animals (Su et al. 2006;Chen2008). 2007; Chen 2009;Dong2009; Chen and Sun 2010). Unfortunately, land-cover changes have already resulted Data sources and methods in substantial forest loss in Hainan (Zhang et al. 2000), which then can lead to the loss of biodiversity (Zhang Land-cover classification was performed using three et al. 2000, 2010;Lietal.2007; Cai et al. 2009). But the Landsat Thematic Mapper (TM) images and six Satellite Environ Monit Assess (2015) 187:4137 Page 3 of 12, 4137

Fig. 1 The Changhua watershed on the island of Hainan, China: a image of China showing the position of Hainan Island; b Hainan with the location of the study area indicated by black; c the upper reaches of Changhua watershed with the nature reserves

Pour l’Observation de la Terre (SPOT 2) images for 2008), which has been preprocessed according to a three separate years: 1988, 1995, and 2005 (Table 1). standardized set of processing parameters (Landsat Sci- The TM and SPOT images of 1995 and 2005 were ence Team 2008), and orthorectified using geodetic and purchased from Beijing Spot Image Co., Ltd. and were elevation control data to correct for positional accuracy corrected for atmospheric and radiometric distortions. and relief displacement (Landsat Science team 2008). The Landsat TM image of the year 1988 was The data were topographically corrected through co- downloaded from USGS (Landsat Science Team registering them to a digital elevation model (DEM) (Ekstrand 1996), derived from topographic maps Table 1 Detailed information about the satellite images used in (scale=1:50,000) from the State Bureau of Surveying this study and Mapping. The orthorectified image matched well Acquisition date Sensor Path/row Spatial resolution (m) with the checked ground control points (GCPs) and their average root mean square error was less than a pixel 08 June 1988 TM4 124/47 28.5 which is 30 m×30 m. To enhance the image resolution, 28 December 1995 TM5 124/47 30 image fusion was made between multispectral TM and 15 January 1996 SPOT 2 278/311 10 the SPOT panchromatic band to produce a 10-m reso- 5 February 1996 SPOT 2 278/312 10 lution multispectral image. The image fusion was done 30 December 1995- SPOT 2 279/312 10 by principal component analysis (PCA) on PCI software 20 December 2005 TM5 124/47 30 (Chavez and Kwarteng 1989; Pohl and van Genderen 23 December 2005 SPOT 2 278/311 10 1998;Antunes2000). 23 December 2005 SPOT 2 278/312 10 Final images for each year were classified into six 7 November 2005 SPOT 2 279/312 10 land-cover types: (1) natural forests, 2) natural shrubs and grasslands, (3) tropical crops, (4) rubber plantations, 4137, Page 4 of 12 Environ Monit Assess (2015) 187:4137

(5) pulp plantations, and (6) open areas. “Natural for- the percentage of a class on the map that matches the ests” included primary forests and naturally corresponding class on the ground and measures the regenerating secondary forests, characterized by closed error of omission (1—user’s accuracy). The ground- canopy. “Natural shrub and grasslands” represented a truth data were sampled during field surveys in 2006 transition between abandoned agricultural land and for- and 2007 for the accuracy assessment. The land-cover est or plantations as an alternative climax community map for 2005 was generally reliable (the overall accura- after the removal of natural forests. A data layer for this cy and overall Kappa statistics were 83.5 and 80.8 %). land-cover type was provided by Hainan Environmental The image of 1995 and 1988 were evaluated by creating Science Institute and used in the image classification. random points (100 points) in ERDAS, and then the “Tropical crops” consisted of Cocos nucifera, Areca overall accuracy and overall Kappa statistics were 83 catechu, Anacardium occidentale,andLitchi chinensis and 78.8 % for 1995, 82 and 77.4 % for 1988. plantations and tea gardens. “Rubber plantations” were Changes in land cover among the six classes were identified from the characteristic image texture, land estimated from the resulting raster images for 1988, form, and terracing. These areas were mainly confined 1995, and 2005. Several landscape-level indices of nat- to State Farms. “Pulp plantations” consisted of Euca- ural forests fragmentation and connectivity were calcu- lyptus spp. and Acacia spp. (Barr and Cossalter 2004), lated to quantify and compare land-cover change including Eucalyptus grandis×Eucalyptus urophylla, through the study period, using the software FRAGS E. urophylla×grandis, Eucalyptus 12ABL, Acacia TATS Version 3.3 (McGarigal et al. 2002). These in- mangium,andAcacia crassicarpa. The pulp plantations cluded the following: (1) relative land-type cover: per- were distinguished by their characteristic image color centage of landscape (PLAND) and total class area—an and hue, in combination with a reference layer provided obvious indicator of land use/land-cover change; (2) by Jinhua Forestry Co., Ltd. “Open areas” included number of patches: (NP)—increasing values indicating shifting cultivation, paddy fields, cultivated lands, rivers, greater fragmentation; (3) size of land-type patches: pools, cities, villages, and industrial lands. The land- mean patch size (AREA_MN), patch size standard de- cover classification of the images was performed using viation (AREA_SD); (4) edge characteristics: total edge the supervised maximum likelihood classification meth- (TE), edge density (ED); (5) isolation of patches: Eu- od (Richards and Jia 2006) on ERDAS 9.0. Training clidean nearest neighbors distance mean and Euclidean areas for the TM were generated by the Hainan Environ- nearest neighbor distance standard deviation (ENN_MN mental Science Institute in 2005 and 2006. To minimize and ENN_SD, average distance to the nearest neighbor- the classification errors caused by difference in the res- ing fragment of the same patch type and ENN indices olution of the satellite images, we continued to do trans- were calculated using a 150-m search radius)—greater formation using clump and elimination after classifica- distances between nearest neighbors indicate greater tion in ArcGIS. And the polygons <900 m2 which fragmentation and less connectivity; and (6) total core equaled to 1 cell were eliminated (Li et al. 2007). Large area (TCA) and number of disjunct core areas (NDCA, homogeneous areas were then selected for classification the number of the core areas of each patch of the accuracy assessment (Schowengerdt 2007). For each corresponding patch type), which may indicate quality land-cover class, at least 10 areas (Richards and Jia of patches, since many endangered species are confined 2006;Lietal.2007) were selected to reflect the variation to forest interiors. The definition of edge width, which due to topography and growing condition. The classifi- was selected to calculate the core and edge effect area, cation accuracy was evaluated in terms of producer’s was chosen arbitrarily because different species have accuracy, user’s accuracy, and overall accuracy, which different, mostly unknown, responses to edge effects was commonly calculated by an error matrix (or confu- (McGarigal et al. 2002). Due to the image resolution sion matrix; see Congalton and Green (1999)). Pro- and the edge effects of logging in Hainan (Huang et al. ducer’s accuracy is the percentage of a particular LULC 2004), 30 m was chosen for the edge depth of total core type on the ground that is correctly classified on the map, area (TCA) and number of disjunct core areas (NDCA, and measures the error of omission (1—producer’sac- the number of the core areas of each patch of the curacy). It is calculated as the ratio of the number of corresponding patch type). correctly classified pixels for a class to the total number The main features of forest habitat fragmentation are of ground truth pixels for that class. User’s accuracy is the decrease of habitat area, increase of patch number, Environ Monit Assess (2015) 187:4137 Page 5 of 12, 4137 decrease of patch size, and increase of patch isolation. In the protected area, comprised of Wuzhishan Na- To analyze the degree of human interference on forest tional Nature Reserve and Yinggeling Nature Reserve fragmentation and changes of forest landscape patterns (Fig. 1), the rate of increase in natural forests (0.94 %/ in the region, only the natural forests and plantations year) was higher than in unprotected areas (0.89 %/year) were compared. The vector layer of natural forests was between 1988–1995 (Table 3). In 1995–2005, the loss rasterized to a cell size of 30-m resolution, and then of natural forests in protected areas was slower than in landscape indices were calculated using software the unprotected areas (−0.38 vs. −2.68 %/year, respec- FRAGSTATS 3.3 (McGarigal et al. 2002). tively). A similar but even more dramatic loss of natural shrubs and grasslands during this period was observed in both protected and unprotected areas (−2.2 %/year vs. − Results and unprotected area 5.08 %/year, respectively). While proportionally more rubber plantation (79.04 %/year), Land-cover changes in and outside nature reserves open areas (26.71 %/year), and tropical crops (22.88 %/ year) were increased in protected areas, more total area Natural forests expanded slightly between 1988 and of pulp plantation (507.6 ha/year), rubber plantation 1995 (Table 2), soon after Hainan became a separate (1383.6 ha/year), and open areas (27.4 ha/year) was province. But the increase of the natural forests was increased outside of protected areas compared to inside accompanied by an almost equivalent loss of natural of them (Table 3). Compared with land cover changes of – – shrubs and grasslands. During 1988–1995, only natural 1995 2005, the period of 1988 1995 showed quite shrubs and grasslands and open areas decreased in area, different pattern between protected and unprotected while the other four land-cover types increased. The areas, with only natural forests and tropical crops ex- transition analysis found that only a small fraction of pansion in protected areas, but in unprotected areas, it natural forests was converted into other land covers, and was natural shrubs and grasslands and open areas that the loss of natural shrubs and grasslands was actually decreased their areas, and the other four (natural forest, due to its conversion into natural forests (Fig. 2). tropical crops, rubber plantation, and pulp plantation) – This pattern changed dramatically in the next decade increased their areas. During 1988 1995, pulp and rub- when natural forests, natural shrubs, and grasslands both ber plantations expanded outside the protected areas, but decreased in cover and all of the other land-cover types decreased in the protected areas. expanded. Among these, pulp plantation showed the highest increase (1527.9 ha/year), followed by rubber Forest-associated habitat quality changes plantation (1385.1 ha/year) and tropical crops (507.6 ha/ year) (Table 2). During this time period, natural forest was The metric total core area (TCA) can be used as an converted into rubber and pulp plantations, tropical crops, indicator of habitat quality. TCA in the landscape in- and open area. For example, a large proportion of the pulp creased 8.37 % during the first period (84,466 ha in plantation (9930.8 ha out of 15,301.0 ha) was established 1988 to 91,539 ha in 1995) but then decreased in areas that was originally natural forests (Fig. 2). 29.38 % (91.539 ha in 1995 to 64,646 ha in 2005)

Table 2 Area, annual change, and change rates of different land cover types in Changhua watershed in 1988, 1995, and 2005

Land-cover types Area (ha) Annual change ha/year and change rate (%)

1988 1995 2005 1988–1995 1995–2005

Natural forests 97,909 104,023 76,305.5 873.4 (0.9) −2771.8 (−2.7) Natural shrubs and grassland 18,892.2 13,344.8 6578.8 −792.5 (−4.2) −676.6 (−5.1) Tropical crops 3151.9 4035.3 9111.6 126.2 (4.0) 507.6 (12.6) Rubber plantation 8326.2 12,925.9 26,776.5 657.1 (7.9) 1385.1 (10.7) Open area 19,233.1 13,177.4 13,455.2 −865.1 (−4.5) 27.8 (0.2) Pulp plantation 2.4 21.8 15,300.7 2.8 (115.5) 1527.9 (7008.7) 4137, Page 6 of 12 Environ Monit Assess (2015) 187:4137

Fig. 2 Land-cover change in the Changhua watershed from 1988 d The area distribution of each land cover and natural forests to 2005, with the nature reserves shaped in red lines.Mapsshown transition which indicate the conversion from natural forests to in the left column illustrate the distribution of land-cover classes: a other land-cover types, with the thickness of the arrows indicating 1988; b 1995; and c 2005. Pie charts in the right column indicate the relative amount of change. Color coding follows the labels on the relative proportion of each land-use class for each time period. the histogram

(Table 4). Meanwhile, the number of disjunct core areas 2005; Table 4). Total edge length increased 6.64 % from (NDCA) increased from 1237 in 1988 to 1,383 in 1995, 422 m in 1988 to 450 m in 1995 and then increased with overall fragmentation continuing to increase to 6.67 % to 480 m in 2005, while the edge density 1676 in 2005 (Table 4), at an annual rate of 2.09 % over increased from 28.6 m/ha in 1988 to 30.5 m/ha in the entire period. The decrease of TCA combined with 1995 and to 32.6 m/ha in 2005 (Table 4). The mean the increase of NDCA indicates decreased the overall nearest neighbor distance (ENN_MN) of natural forests habitat quality and increased fragmentation of these core increased from 104.8 m in 1988 to 115.3 m in 1995 and areas in Changhua watershed. to 110.2 m in 2005, with an annual increase rate of 0.30 %. The increase of ENN_MN indicates that large natural forests patches tended to become more isolated, Forest-associated habitat fragmentation while the decrease in ENN_MN in 1995–2005 may be the result of the overall decrease of large patches. Natural forests cover decreased by 15 % (1.30 %/year) for 2005 compared to 1988 (Table 4), as shown by the percentage of landscape (PLAND). Besides the loss of natural forests, the number of patches (NP) of natural Discussion forests increased 97.27 % (403 in 1988 to 795 in 1995) and then increased 48.30 % (795 in 1995 to 1179 in Conversion between natural forests and plantations 2005), while the patch density increased (0.07/100 ha in and its implications to conservation 1988 to 0.13/100 ha in 1995 and 2005). At the same time, the mean patch size (AREA_MN) decreased Land-cover changes during the study periods lead to the (243 ha in 1988 to 131 ha in 1995 and to 64 ha in overall loss, isolation, and fragmentation of the natural Environ Monit Assess (2015) 187:4137 Page 7 of 12, 4137

Table 3 The area (ha) of proportion of each land-use/land- Reserve) and unprotected area (%), the change of absolute cover type to the total area both in protected areas (including amount (ha/year), and change rate (%/year) in each period and Wuzhishan National Nature Reserve and Yinggeling Nature in the study periods

Area (ha) Proportion to the Absolute amount (AA) (ha/year) Change rate (CR) (%/year) total area (%)

1988 1995 2005 1988 1995 2005 1988–1995 1995–2005 1988–2005 1988–1995 1995–2005 1988–2005

Protected area NF 548.2 584.2 562.2 86.3 92.2 88.8 5.16 −2.21 0.83 0.94 −0.38 0.15 NS&G 79.2 45.1 35.1 12.5 7.1 5.5 −4.87 −0.99 −2.59 −6.15 −2.2 −3.28 TC 0.4 0.9 3.1 0.1 0.1 0.5 0.07 0.22 0.16 17.28 22.88 39.71 RP 2 1.8 15.9 0.3 0.3 2.5 −0.03 1.41 0.82 −1.31 79.04 40.88 OA 5.6 1.3 4.8 0.9 0.2 0.8 −0.61 0.35 −0.05 −10.97 26.71 −0.84 PP 0.08 0 12.2 0.01 0 1.93 −0.01 1.22 0.71 −14.29 – 891.18 Unprotected area NF 97,360.8 103,437.8 75,742.8 66.3 70.4 51.6 868.1 −2769.5 −1271.6 0.89 −2.68 −1.31 NS&G 18,812.8 13,299.9 6542.9 12.8 9.1 4.5 −787.6 −675.7 −721.8 −4.19 −5.08 −3.84 TC 3150.6 4034.1 9109.9 2.1 2.7 6.2 126.2 507.6 350.5 4.01 12.58 11.13 RP 8324 12,924.1 26,760.1 5.7 8.8 18.2 657.2 1383.6 1084.5 7.89 10.71 13.03 OA 19,227.4 13,176.1 13,450.2 13.1 9 9.2 −864.5 27.4 −339.8 −4.5 0.21 −1.77 PP 2.3 22 15,288.8 0.002 0.015 10.4 2.8 1526.7 899.2 121.42 6939.4 38,825.7

AA= (Aye-Ayb)/T, CR=(Aye-Ayb)/Ayb×1/T×100 %, Ayb is the area of a land cover in the beginning year of a study period; Aye is the area of a land cover in the end of a study period., T is the time length NF natural forests, NS&G natural shrubs and grassland, TC tropical crops, RP rubber plantation, OA open area, PP pulp plantation. forests, particularly during the most recent time period and the losses of natural forests were caused by changes of our study. This loss of forest quantity and quality will in forest management policies (Zhai et al. 2012, 2014). probably cause decline in the distribution and abun- Natural forests losses and the expansion of rubber dance of individual species (Lindenmayer and Fischer and pulp plantations were two main features of the land 2006). Although, natural forests made a mild recovery cover in this area. The expansion of both rubber and during 1988–1995, which agrees with results from Lin pulp plantations took place mainly through displacing and Zhang (2001), they decreased greatly, primarily due natural forests, which indicated they were the main to and the dramatic expansion of plantation forests causes of deforestation. This result does not agree with (Fig. 2). The recovery of natural forests in 1988–1995 Zhou (1995), where swidden agriculture was found to

Table 4 The landscape indices at the class level for natural forests in the upper Changhua Watershed from 1988 to 2005

Year PLAND NP TE (m) ED (m/ha) AREA_MN AREA_SD TCA (ha) NDCA ENN_SD ENN_MN (%) (ha) (m)

1988 66.3 403 422 28.6 243 3475 84,466 1237 87.7 104.8 1995 70.5 795 450 30.5 131 1736 91,539 1383 94.8 115.4 2005 51.7 1179 480 32.6 65 1063 64,646 1676 101.2 110.2 1988–1995 (%) 6.33 97.27 6.64 6.64 −46.09 −50.04 8.37 11.80 8.10 10.11 1995–2005 (%) −26.67 48.30 6.67 6.89 −50.38 −38.77 −29.38 21.19 6.75 −4.51 1988–2005 (%/year) −1.30 11.33 0.81 0.82 −4.31 −4.08 −1.38 2.09 0.91 0.30

PLAND percentage of landscape, NP the number of patches, TE total edge, ED edge density, AREA_MN mean patch size, AREA_SD patch size standard deviation, ENN_MN, ENN_SD Euclidean nearest neighbors distance_mean and Euclidean nearest neighbor distance_standard deviation, TCA total core area, NDCA number of disjunct core areas 4137, Page 8 of 12 Environ Monit Assess (2015) 187:4137 be a main reason for deforestation and forest degrada- plantation expansion, and natural forests habitat frag- tion in Hainan. The primary difference between these mentation has also occurred in the nature reserves of the two periods of time was the large-scale introduction of Xishuangbanna, another tropical region of China (Li pulp plantations and a greater demand of rubber after et al. 2007, 2009). Fortunately, during the study period, 1995 (Li et al. 2007;Zhaietal.2012). the Hainan Provincial Forestry Department has in- Deforestation will strongly impact both floral and creased its protected areas, with the aim to create a total faunal diversity in this area, especially for rare and of 3500 km2 of protected area for forest and wildlife in endangered species, and also will greatly threaten the Hainan by 2010 (Zhang et al. 2010). Recently, the habitat restricted and habitat alteration sensitive species, Hainan government has passed the Hainan Non- most of which are endemic to the island. For example, commercial Forest Development Program (2010– the hill-partridge (Arborophila ardens), which is quite 2020), aiming to increase the area of non-commercial sensitive to habitat changes and can only be found in natural forest (Zhai et al. 2014). Despite these efforts primary forests, is seriously threatened by natural forest and without effective enforcement, the natural forest losses (Gao 1998; BirdLife International 2010). The may still face further reductions as rubber and pulp survival and reproduction (World Conservation plantations continue to encroach on these protected Monitoring Centre 1998) and genetic diversity (Dai areas in the future (Zhai et al. 2012, 2014). The decrease et al. 2013) of many other species, including Madhuca of natural forests habitat at a rate of ∼1.3 % (∼1270 ha/ hainanensis (an endangered timber species), Hainan year) and expansion of pulp and rubber plantation in the black-crested gibbons (Nomascus hainanus)(Zhou region should be of major conservation concern. et al. 2005), Hainan Eld’sdeer(Cervus eldi hainanus) (Zeng et al. 2005), Reevese’s butterfly Lizard (Leiolepis Forest fragmentation and its implication to biodiversity reevesii)(Lin2008), and other bird species (Zhou 1995), conservation are all severely and negatively affected by deforestation. Conversion from natural forests to rubber plantation In the decade from 1995 to 2005, both nature reserves also decreased the genetic diversity of wild rice in became almost completely isolated from the interior of Yunnan and Hainan (Qian et al. 2001). the watershed (Fig. 2). The Yinggeling Nature Reserve Overall, the conversion from natural forests to other is now separated by an expanding area of rubber mono- land types may have many negative effects, including culture plantation, while an additional region of natural changes in forest community structure, composition, shrubs and grasslands within the core area of reserve has and species richness (Ding and Zang 2008), particularly been converted to pulp plantation. Although the forest in the decline of local avian biodiversity (Cai et al. loss within the Wuzhishan Nature Reserve was not 2009). The intensification of agroforestry through the substantial, its margins within the watershed have be- continued expansion of plantations and tropical crops come completely converted into pulp plantations. With remains a main hazard to Hainan’s eco-environment the expansion of the rubber plantations, the natural (Huang et al. 2009). The loss and disturbance to natural forests in the watershed have in fact been divided into forests was also associated with the appearance and three separate blocks with very limited migration among expansion of invasive species in Hainan Island which them (Fig. 2c). Increasing fragmentation and isolation may be a threat to the survival of native species (Chen has also found in most nature reserves of Hainan 2008). Zhang et al. (2007) found that in Yinggeling (>60 %) (Wang et al. 2013). A similar level of fragmen- Nature Reserve, the disturbed natural forests were more tation was also found in the Xishuangbanna prefecture easily threatened and also more easily invaded by exotic (Li et al. 2009), and these small fragments had lower species compared to undisturbed or less disturbed natu- abilities to support enough species, especially endemic ral forests (Qin et al. 2008). Other studies also indicated ones (Chang et al. 2013). One of the ecological effects that plantations negatively affect biodiversity by of fragmentation is increasing patch numbers which can displacing natural forests or natural ecosystems (Bremer lead to larger distances between habitat patches and and Farley 2010). hence greater habitat isolation (Hanski 1994). As point- The deforestation also occurred in the nature re- ed out by Lindenmayer and Fischer (2006), habitat serves, with plantations also displacing natural forests isolation may limit the movement of propagules for within protected areas. This type of illegal deforestation, plants and may impair movements of plants (dispersal Environ Monit Assess (2015) 187:4137 Page 9 of 12, 4137 and pollination) and animals at several spatial and tem- Conclusions poral scales. These barriers to migration and gene flow are a particular concern given the expected need for Our study provides the first quantitative descrip- plants and animals to track their appropriate habitat in tion of forest deforestation and fragmentation of the coming decades due to climatic change. Additional- natural forests in a typical watershed of Hainan, ly, edge length has increased in the study area which can China. The study area has undergone dramatic affect the remaining natural vegetation due to adverse land-use changes during the study period, with edge effects such as structural damage to the vegetation, deforestation of natural forests and expansion of disruption of forest floor and soil layer, altered nutrient rubber and pulp plantations the main two features. cycling and decomposition, and altered pollen and seed These two features of local land use have serious dispersal within the edge (Harper et al. 2005; Laurance ecological consequences in the landscape, includ- et al. 2006; Broadbent et al. 2008). ing habitat loss, isolation, and habitat fragmenta- Besides deforestation causing habitat loss, fragmen- tion, which strongly and negatively impacts local tation was a major threat to the survival of many species flora and faunal diversity and survival, especially (Huxel and Hastings 1999), primarily due to the poor for rare and endangered species. Therefore, we connectivity among habitats caused by fragmentation. recommend that natural forest protection and con- Habitat connectivity has a significant impact on the flow servationbeacentralpriorityinthefuture,espe- of nutrients (Forman 1995), individuals and genes be- cially efficient management planning in key areas tween separated populations (Hobbs and Hopkins with high biodiversity with strict and clear policies 1991). Forest connectivity will become more important on deforestation and plantation expansions. We in the future, particularly in these interior watersheds, as also suggest strict enforcement of existing conser- global warming will lead to widespread habitat modifi- vation laws by local governments to prevent natu- cation and habitat loss (Corlett 2014). Therefore, the ral forest conversion in protected areas. However, loss, isolation, and fragmentation of natural forests hab- large forces beyond the scope of the local govern- itat pose serious threats on the local biodiversity, espe- ment still threaten natural forests with conversion cially for the survival of macro-vertebrates that require into plantations, e.g., overcapacity of pulp mills in large home ranges. For example, the Hainan gibbon Hainan, great demand and higher price of rubber (N. hainanus) needs a territory size of 200–500 ha and plantation, and a lack of distinction between natu- once had a wide distribution on Hainan Island. They are ral and industrial forestry in some forest laws (Li currently restricted to a single narrow tropical rain forest et al. 2007;Zhaietal.2012, 2014). Forest con- (Bawangling Nature Reserve) (Zhou et al. 2005). As Li version in the two nature reserves should be et al. (2010) showed that both fragmentation of natural completely halted and environmentally friendly forests and expansion of lowland plantation lead to corridors established to maintain ecosystem connectivi- hindered population rejuvenation of Hainan gibbons. ty. To limit the negative impacts of fragmentation and Habitat fragmentation was also one of the two main habitat isolation, restoring and increasing the quantity threats to the survival of the endangered tropical tree and quality of natural vegetation corridors are required. Vatica mangachapoi (Dai et al. 2008)and The recently passed Hainan Non-commercial Forest M. hainanensis (Dai et al. 2013). Habitat fragmentation Development Program (2010–2020) by the Hainan has many significant ecological and genetic conse- Government, which aims to increase the area of non- quences for endemic plant populations (Culley et al. commercial natural forest, is an opportunity to protect 2007). Both habitat loss and fragmentation will put local remaining natural forests and to increase natural forest endemic taxa at risk of extinction (Sodhi et al. habitat connectivity. To increase its connectivity, we 2006). Maintaining or increasing natural forests particularly suggest that local governments improve patches’ connectivity is a way to counter fragmen- their role and activities on natural forests monitoring tation and isolation (Li et al. 2010;Zhangetal. and systematic planning. With its biological importance 2010). Given these trends, the future increase in of Hainan province, more research on the conservation the quantity and quality of habitat corridors and and management of these threatened natural forests, expansion of forest restoration projects will be forming a major watershed on the island, is required in crucial for biodiversity conservation. the future. 4137, Page 10 of 12 Environ Monit Assess (2015) 187:4137

Acknowledgments This work is part of the CGIAR Congalton, R., & Green, K. (1999). Assessing the accuracy of Research Program 6: Forests, Trees, and Agroforestry. This work remotely sensed data: principles and practices.NewYork: is funded by the National Natural Science Foundation of China Lewis Publishers. (Grant 31300403) and the China Postdoctoral Science Foundation Corlett, R.T., (2014). Forest fragmentation and climate change in (Grant 2013M540722). We thank William D. Dijak (Northern global forest fragmentation. eds Kettle, C.J., Koh, L.P.,CABI Research Station, U.S. Forest Service, Columbia, USA) for his Publishing advice in landscape indices selection and indices analyses. We also Culley, T. M., Sbita, S. J., & Wick, A. (2007). Population genetic appreciate constructive comments and suggestions provided by effects of urban habitat fragmentation in the perennial herb the anonymous reviewers. Viola pubescens (Violaceae) using ISSR markers. Annuals of Botany, 100,91–100. Dai, Z. C., Zhong, Q. X., Si, C. C., Lin, Y., Wang, K. R., Zhang, B., & Du, D. L. (2008). A review of study on endangered mechanism and conservation ecology of endangered Vatica References mangachapoi. Journal of Hainan Normal University(Natural Science), 21,82–86 (in Chinese with English abstract). Anderson, A., & Jenkins, C. (2006). Applying nature’sdesign: Dai, Z. C., Si, C. C., Zhai, D. L., Huang, P., Qi, S. S., Zhong, Q. X., corridors as a strategy for biodiversity conservation.New Hu, X., Li, H. M., & Du, D. L. (2013). Human impacts on York: Columbia University Press. genetic diversity and differentiation in six natural populations Antunes, A. (2000). Thematic resolution assessment merging of Madhuca hainanensis, an endemic and endangered timber Landsat and SPOT 10m. International Archives of species in China. Biochemical Systematics and Ecology, 50, Photogrammetry and Remote Sensing, 33,52–55. 212–219. Aziz, S. A., Laurance, W. F., & Clements, R. (2010). Forests Ding, Y., & Zang, R. G. (2008). Changes in decidous trees during reserved for rubber? Frontiers in Ecology and the recovery of tropical lowland rain forests on abandoned Environment, 8,178–178. shifting cultivation lands in Hainan Island, South China. Barr, C., & Cossalter, C. (2004). China’s development of a Biodiversity Science, 16,103–109 (in Chinese with English plantation-based wood pulp industry: government policies, abstract). financial incentives, and investment trends 1. Inernational Dong, S. Y. (2009). Hainan tree ferns (Cyatheaceae), morpholog- Forestry Review, 6, 267–281. ical, ecological and phytogeographical observations. Annales BirdLife International. (2010). Species factsheet: Arborophila Botanici Fennici, 46,381–388. ardens. http://www.birdlife.org. Accessed on 21/05/2011. Ekstrand, S. (1996). Landsat TM-based forest damage assessment, Bremer, L. L., & Farley, K. A. (2010). Does plantation forestry correction for topographic effects. Photogrammetric restore biodiversity or create green deserts? A synthesis of the Engineering and Remote Sensing, 62,151–161. effects of land-use transitions on plant species richness. Feng, Z. W., Wang, X. K., & Ouyang, Z. Y.(1999). Distribution of Biodiversity and Conservation, 19,3893–3915. Eucalyptus forest and ecological regionalization in Hainan Broadbent, E. N., Asner, G. P., Keller, M., Knapp, D. E., Oliveira, P. Province. Soil and Environmental Sciences, 8,168–173 (in J. C., & Silva, J. N. (2008). Forest fragmentation and edge Chinese with English abstract). effects from deforestation and selective logging in the Forman, R. (1995). Land mosaics: the ecology of landscapes and Brazilian Amazon. Biological Conservation, 141,1745–1757. regions. Cambridge: Cambridge University Press. Cai, Y., Yang, C. C., & Liang, W. (2009). Negative effects of Francisco-Ortega, J., Wang, Z. S., Wang, F. G., Xing, F. W., Liu, plantations on bird diversity in Yinggeling Nature Reserve, H., Xu, H., Xu, W. X., Luo, Y. B., Song, X. Q., Gale, S., Hainan Island. Sichuan Journal of Zoology, 28,764–767 (in Boufford, D. E., Maunder, M., & An, S. Q. (2010). Seed Chinese with English abstract). plant endemism on Hainan Island: a framework for conser- Chan, B. P. L., Lee, S. K., Zhang, J. F., & Su, W. B. (2005). vation actions. Botanical Review, 76,346–376. Notable bird records from Bawangling National Nature Gao, Y. R. (1998). Conservation status of endemic Galliformes on Reserve, Hainan Island, China. Forktail, 21,33–41. Hainan Island, China. Bird Conservation International, 8, Chang, X., Quan, R. C., & Wang, L. (2013). Bird conservation in 411–416. extremely small tropical rainforest patches in southwest Gibbs, H. K., Ruesch, A. S., Achard, F., Clayton, M. K., China. Biological Conservation, 158,188–195. Holmgren, P., Ramankutty, N., & Foley, J. A. (2010). Chavez, P., & Kwarteng, A. (1989). Extracting spectral contrast in Tropical forests were the primary sources of new agricultural Landsat thematic Mapper image data using selective princi- land in the 1980s and 1990s. Proceedings of the National pal component analysis. Photogrammetric Engineering and Academy of Sciences of the United States of America, 107, Remote Sensing, 55,339–348. 16732–16737. Chen, Y. H. (2008). Avian biogeography and conservation on Hanski, I. (1994). Patch-occupancy dynamics in fragmented Hainan Island, China. Zoological Science, 25,59–67. landscapes. Trends in Ecology and Evolution, 9,131– Chen, Y. H. (2009). Distribution patterns and faunal characteristic 135. of mammals on Hainan Island of China. Folia Zoology, 58, Harper, K. A., Macdonald, S. E., Burton, P. J., Chen, J. Q., 372–384. Brosofske, K. D., Saunders, S. C., Euskirchen, E. S., Chen, G., & Sun, W. B. (2010). Ploidy variation in Roberts, D., Jaiteh, M. S., & Esseen, P. A. (2005). Edge Trigonobalanus verticillata (Fagaceae). Plant Systematics influence on forest structure and composition in fragmented and Evolution, 284,123–127. landscapes. Conservation Biology, 19,768–782. Environ Monit Assess (2015) 187:4137 Page 11 of 12, 4137

Hobbs, R. J., & Hopkins, A. J. M. (1991). The role of conservation Qian, W., Ge, S., & Hong, D. Y. (2001). Genetic variation within corridors in a changing climate. In D. A. Saunders & R. J. and among populations of a wild rice Oryza granulata from Hobbs (Eds.), Nature conservation 2: the role of corridors China detected by RAPD and ISSR markers. Theoretical and (pp. 281–290). Chipping Norton, New South Wales: Surrey Applied Genetics, 102,440–449. Beatty and Sons. Qin, X. S., Zhang, R. J., Chen, H. F., Yan, Y. H., Zheng, X. L., & Huang, S. N., Wang, B. S., & Li, Y. D. (2004). Edge effects in two Xing, F. W. (2008). Alien plants in limestone regions of secondary tropical Montane rainforests at Jianfengling, Hainan Island, China. Chineses Journal of Ecology, 27, Hainan Island of China. Forest Research, 17,693–699 (in 1861–1868 (in Chinese with English abstract). Chinese with English abstract). Richards, J. A., & Jia, X. (2006). Remote sensing digital image Huang, B. R., OuYang, Z. Y., Zhang, H. Z., & Zheng, H. (2009). A analysis: an introduction (4th ed.). Berlin: Springer-Verlag. graph-theoretic analysis of relationships among anthropogen- Sala, O. E., Chapin, F. S., Armesto, J. J., Berlow, E., Bloomfield, ic stressors on natural forest in Hainan Island. Journal of J., Dirzo, R., Huber-Sanwald, E., Huenneke, L. F., Jackson, Natural Resources, 20,154–161 (in Chinese with English R. B., Kinzig, A., Leemans, R., Lodge, D. M., Mooney, H. abstract). A., Oesterheld, M., Poff, N. L., Sykes, M. T., Walker, B. H., Huxel, G. R., & Hastings, A. (1999). Habitat loss, fragmentation, Walker, M., & Wall, D. H. (2000). Biodiversity - global and restoration. Restoration Ecology, 7,309–315. biodiversity scenarios for the year 2100. Science, 287, Landsat Science Team. (2008). Free access to Landsat data, letter. 1770–1774. Science, 320,1011. Schowengerdt, R. A. (2007). Remote sensing: models and Laurance, W. F. (1999). Reflections on the tropical deforestation methods for image processing. San Diego, California: crisis. Biological Conservation, 91,109–117. Academic Press. Laurance, W. F., Nascimento, H. E. M., Laurance, S. G., Andrade, Sodhi, N. S., Koh, L. P., & Brook, B. W. (2006). Southeast Asian A. C., Fearnside, P. M., Ribeiro, J. E. L., & Capretz, R. L. birds in peril. Auk, 123,275–277. (2006). Rain forest fragmentation and the proliferation of Sodhi, N. S., Posa, M. R. C., Lee, T. M., Bickford, D., Koh, L. P., successional trees. Ecology, 87,469–482. & Brook, B. W. (2010). The state and conservation of Li, H. M., Aide, T. M., Ma, Y. X., Liu, W. J., & Cao, M. (2007). Southeast Asian biodiversity. Biodiversity and Demand for rubber is causing the loss of high diversity rain Conservation, 19,317–328. forest in SW China. Biodiversity and Conservation, 16, Soule, M. E. (1991). Conservation: tactics for a constant crisis. 1731–1745. Science, 253,744–750. Li, H. M., Ma, Y. X., & Liu, W. J. (2009). Clearance and frag- Su, Y. C., Chang, Y. H., Lee, S. C., & Tso, I. M. (2007). mentation of tropical rain forest in Xishuangbanna, SW, Phylogeography of the giant wood spider (Nephila pilipes, China. Biodiversity and Conservation, 18,3421–3440. Araneae) from Asian–Australian regions. Journal of Li,Z.G.,Wei,F.W.,&Zhou,J.(2010).Mitochondrial Biogeography, 34,177–191. DNA D-loop sequence analysis and population rejuve- Turner, B., Skole, D., Sanderson, S., Fischer, G., Fresco, L., nation of Hainan gibbons (Nomascus hainanus). Leemans, R., (1995). Land-use and land-cover change. Biodiversity Science, 18, 523–527 (in Chinese with Science/Research Plan. IGBP Global Change Report English abstract). (Sweden), Report no. 35/7. (Available from International Lin, L.H., (2008). Population density, thermal requirement and Geosphere-Biosphere Program Secretariat, Stockholm) geographic pattern of genetic structure in the Reevese’sbut- Wang, F. G., Qin, X. S., Chen, H. F., Zhang, R. J., Liu, D. M., & terfly Lizard (Leiolepis reevesii) from Hainan, southern Xing, F. W. (2006). Endemic plants in limestone region on China. PhD dissertation, Nanjing Normal University, Hainan Island. Journal of Tropical and Subtropical Botany, Nanjing (in Chinese with English abstract) 14,45–54 (in Chinese with English abstract). Lin, M. Z., & Zhang, Y. L. (2001). Dynamic change of tropical Wang, K., Franklin, S. E., Guo, X. L., & Cattet, M. (2010). forest in Hainan Island. Geographical Research, 20,703– Remote sensing of ecology, biodiversity and conservation: 712 (in Chinese with English abstract). a review from the perspective of remote sensing specialists. Lindenmayer, D., & Fischer, J. (2006). Habitat fragmentation and Sensors, 10,9647–9667. landscape change: an ecological and conservation synthesis. Wang, W., Pechacek, P., Zhang, M., Xiao, N., Zhu, J., & Li, J. Washington, D.C.: Island Press. (2013). Effectiveness of nature reserve system for conserving Luo, K. F. (1985). Collection of Hainan tropical agricultural tropical forests: a statistical evaluation of Hainan Island, resources zoning. Beijing: Science Press (in Chinese). China. Plos One, 8, e57561. McGarigal, K., Cushman, S., Neel, M., Ene, E. (2002). FRAGS World Conservation Monitoring Centre. (1998). Madhuca TATS: spatial pattern analysis program for categorical maps hainanensis. In: IUCN red list of threatened species version computer software program produced by the authors at the 20104 (http://www.iucnredlist.org/). International Union for University of Massachusetts, Amherst. Amherst. http://www. Conservation of Nature and Natural Resources (IUCN). umass.edu/landeco/research/fragstats/fragstats.html. Zeng, Z. G., Song, Y. L., Li, J. S., Teng, L. W., Zhang, Q., & Guo, Accessed on 2010-04-20. F. (2005). Distribution, status and conservation of Hainan Myers, N., Mittermeier, R. A., Mittermeier, C. G., da Fonseca, G. Eld’sdeer(Cervus eldi hainanus)inChina.Folia Zoologica, A. B., & Kent, J. (2000). Biodiversity hotspots for conserva- 54,249–257. tion priorities. Nature, 403,853–858. Zhai, D. L., Cannon, C. H., Slik, J. W. F., Zhang, C. P., & Dai, Z. Pohl, C., & van Genderen, J. L. (1998). Multisensor image fusion C. (2012). Rubber and pulp plantations represent a double in remote sensing: concepts, methods and applications. threat to Hainan’s natural tropical forests. Journal of International Journal of Remote Sensing, 19,823–854. Environmental Management, 96,64–73. 4137, Page 12 of 12 Environ Monit Assess (2015) 187:4137

Zhai, D. L., Xu, J., Dai, Z. C., Cannon, C. H., & Edward, G. R. Zhang, M. X., Fellowes, J. R., Jiang, X. L., Wang, W., Chan, B. P. (2014). Increasing tree cover while losing diverse natural L., Ren, G. P., & Zhu, J. G. (2010). Degradation of tropical forests in tropical Hainan, China. Regional Environmental forest in Hainan, China, 1991–2008: conservation implica- Change, 14,611–621. tions for Hainan Gibbon (Nomascus hainanus). Biological Zhang, Y. Q., Uusivuori, J., & Kuuluvainen, J. (2000). Conservation, 143,1397–1404. Econometric analysis of the causes of forest land use changes Zhou, G. Y. (1995). Influences of tropical forest changes on in Hainan, China. Canadian Journal of Forest Research- environmental-quality in Hainan Province, P.R. of China. Revue Canadienne De Recherche Forestiere, 30,1913–1921. Ecological Engineering, 4,223–229. Zhang, R. J., Xing, F. W., Ng, S. C., Ye, Y. S., Wang, F. G., & Chen, Zhou,J.,Wei,F.,Li,M.,Zhang,J.F.,Wang,D.,&Pan, H. Q. (2007). The alien plants and environment evaluation of R. L. (2005). Hainan black-crested gibbon is headed Yinggeling Moutain, Hainan, China. Ecology and for extinction. International Journal of Primatology, Management, 16, 906–911 (in Chinese with English abstract). 26,453–465.