Biodivers Conserv (2009) 18:1001–1018 DOI 10.1007/s10531-008-9490-5 ORIGINAL PAPER

Spatial and temporal deforestation dynamics in protected and unprotected dry forests: a case study from Myanmar (Burma) Melissa Songer Æ Myint Aung Æ Briony Senior Æ Ruth DeFries Æ Peter Leimgruber

Received: 4 January 2008 / Accepted: 18 September 2008 / Published online: 15 October 2008 Ó Springer Science+Business Media B.V. 2008

Abstract Tropical dry forests are more threatened, less protected and especially susceptible to deforestation. However, most deforestation research focuses on tropical rain forests. We analyzed spatial and temporal changes in land cover from 1972 through 2005 at Chatthin Wildlife Sanctuary (CWS), a tropical dry forest in Myanmar (Burma). CWS is one of the largest protected patches of tropical dry forest in Southeast Asia and supports over half the remaining wild population of the endangered Eld’s deer. Between 1973 and 2005, 62% of forest was lost at an annual rate of 1.86% in the area, while forest loss inside CWS was only 16% (0.45% annually). Based on trends found during our study period, dry forests outside CWS would not persist beyond 2019, while forests inside CWS would persist for at least another 100 years. Analysis of temporal deforestation patterns indicates the highest rate of loss occurred between 1992 and 2001. Conversion to agriculture, shifting agriculture, and flooding from a hydro-electric development were the main deforestation drivers. Fragmentation was also severe, halving the area of suitable Eld’s deer habitat between 1973 and 2001, and increasing its isolation. CWS protection efforts were effective in reducing deforestation rates, although deforestation effects extended up to 2 km into the sanctuary. Establishing new protected areas for dry forests and finding ways to mitigate human impacts on existing forests are both needed to protect remaining dry forests and the species they support. Keywords Tropical dry forests
Biodiversity
Change detection
Deforestation rates
Fragmentation
Protected areas
Forest dynamics
Land use change
Remote sensing
Landsat
ASTER

M. Songer (&)
B. Senior
P. Leimgruber Smithsonian Institution, National Zoological Park, Conservation Ecology Center, 1500 Remount Road, Front Royal, VA 22630, USA e-mail: [email protected] Myint Aung SICAS Office, 30 5/A Thirimingalar Avenue/Kabar Aye, Kamayut Township, Yangon, Myanmar R. DeFries Department of Ecology, Evolution, and Environmental Biology, Columbia University, 10th Floor Schermerhorn Ext., 1200 Amsterdam Avenue, New York, NY 10027, USA

123

1002

Abbreviations ASTER CWS GIS GPS ISODATA Landsat ETM? Landsat MSS Landsat TM

Biodivers Conserv (2009) 18:1001–1018

Advanced Spaceborne Thermal Emission and Reflection Radiometer Chatthin Wildlife Sanctuary Geographic Information Systems Global Positioning System Iterative Self-Organizing Data Analysis Techniques Landsat Enhanced Thematic Mapper Landsat Multispectral Scanner Landsat Thematic Mapper

Introduction Tropical dry forests are more threatened, less protected and especially susceptible to deforestation in comparison to tropical rain forests (Bunyavejchewin 1982; Murphy and Lugo 1986; Janzen 1988; Bullock et al. 1995; Maass 1995). However, most deforestation research and protection focuses on tropical rain forests (Sa´nchez-Azofeifa et al. 2005), likely due to their well-publicized species richness (Myers 1984; Gentry 1984; Sutton et al. 1984; Raven 1988). In Myanmar (Burma) tropical dry forests experienced some of the highest deforestation rates of any forest types during the 1990s (Leimgruber et al. 2005), yet only 4% of these forests are legally protected as opposed to 8% of other forest types (Songer 2006). Protection of tropical dry forests poses problems less frequently encountered in tropical rain forest protection. Dry forests are more often associated with high human population densities (Murphy and Lugo 1986), are more vulnerable to extraction pressures (Bullock et al. 1995; Maass 1995; Miles et al. 2006), and are usually places that have been exposed to human habitation and use for centuries (Janzen 1988; Stott 1990). Tropical dry forests also are frequently on arable soils preferred for agricultural use, particularly rice and sugar cane, and have better biological and climatic conditions for humans compared to other tropical ecosystems (Ewel 1999). As a result[80% of Myanmar’s population resides in the central dry zone where they can benefit from the climate and soil of dry forest ecosystems. Myanmar has more remaining forest cover than most countries on the Southeast Asian mainland (Leimgruber et al. 2005) and possesses some of the last strongholds for tropical dry forests in Southeast Asia. Chatthin Wildlife Sanctuary (CWS) is the best example of protected dry forests in the country and supports well over half of the world’s remaining population of the endangered Eld’s deer (Cervus eldi). Despite their importance, little is known about the status of dry forests in and around CWS, or about their rate of decline. In addition, there are no quantitative or qualitative assessments of how effectively CWS preserves dry forests. There is much debate about the best way to measure protected area effectiveness (Parrish et al. 2003; Hockings et al. 2004; Chape et al. 2005; Mas 2005). Biodiversity protection is the primary purpose for establishing protected areas, and their effectiveness may depend to a large extent on their capacity to retain the biodiversity of an area as well as the original land cover supporting this biodiversity (Bruner et al. 2001; Liu et al. 2001). Past assessments of protected area effectiveness often relied on combinations of interview survey data, information on staffing and financing from government agencies, and expert assessments (Rao et al. 2002; Ervin 2003; Goodman 2003; Myint Aung et al. 2004; Struhsaker et al. 2005; Myint Aung 2007). More recently, several studies use the relationship between

123

Biodivers Conserv (2009) 18:1001–1018

1003

habitat size and biodiversity as an indicator (Brooks et al. 1997; Pimm and Raven 2000) and have shown this pattern in relationship to protected areas (Gurd et al. 2001; Newmark 1987, 1996). Beyond protected area size, another important factor affecting biodiversity may be the degree of difference between the land cover of the protected area and that of the surrounding landscape matrix (Hansen and Rotella 2001; Carroll and Miquelle 2004; Mas 2005). Retaining original land cover is particularly important when the land cover outside the protected areas is experiencing rapid degradation and land cover changes. Such land cover based assessments are especially useful when the target is an ecosystem type, such as dry forest or dry deciduous forest, rather than a specific set of species. Satellite remote sensing provides a useful tool for measuring a park’s effectiveness in retaining original land cover and assessing the surrounding landscape by providing accurate and current information about land cover changes in even the most remote areas (Zheng et al. 1997; Foody and Cutler 2003; Linkie and Smith 2004; Sa´nchez-Azofeifa et al. 2005; Mas 2005; Trigg 2006). However, frequently these analyses rely on only a few satellite images, collected at one or two dates in the recent past. These assessments fail to capture inter-annual changes in land cover. We used five satellite images of CWS and surrounding areas—spanning 32 years—to conduct a case study of the spatial and temporal deforestation and fragmentation dynamics in Southeast Asian dry forests. Based on our analyses we addressed three questions: 1. What are the spatial and temporal patterns in deforestation and fragmentation? 2. Based on these patterns what are the major drivers of this land cover change? 3. How effective is CWS in preserving dry forests and Eld’s deer habitat in the area?

Methods Study area CWS is located at the northern edge of Myanmar’s central dry zone (95°240 E–9.95°400 E, 23°300 N–23°420 N; Fig. 1a). The monsoonal climate has three seasons, including a rainy season (*0.4 m rainfall annually), a cool-dry season, and a hot-dry season (Myint Aung et al. 2004). Crops are grown August through January. During the hot-dry season (February through May) the area burns, due to both human and natural causes. The dry deciduous forests, known locally as ‘‘Indaing’’, are dominated by the dipterocarp species Dipterocarpus tuberculatus; grassland and evergreen forest patches are intermittently found throughout. Myanmar’s dry deciduous forests are found in a horseshoe-shaped area, wedged between the country’s hill region and its central dry zone (Fig. 1b). The dry zone had substantial dry deciduous forest cover (Kurz 1877; Stamp 1925), but it rapidly disappeared as human populations and agriculture expanded, and what remains now is restricted to the fringes of the ecosystem’s environmental envelope (Koy et al. 2005; Leimgruber et al. 2005). Our study area (CWS) is one of the largest existing patches of dry deciduous forest areas in Myanmar. No forest remains to the south of CWS and to the north dry deciduous forest transitions into evergreen and mixed-deciduous as elevation increases. However, to the east and west, CWS is part of a landscape mosaic with interspersed patches of agriculture and dry deciduous forest (Fig. 1b). Using a Global Positioning System (GPS), CWS staff recorded geographic positions for each boundary pillar. Based on these positions and knowledge of the park warden, we

123

1004

Biodivers Conserv (2009) 18:1001–1018

Fig. 1 a Chatthin Wildlife Sanctuary (CWS), Thapanseik Reservoir and surrounding villages (marked with triangles; background based on ASTER image from 2005; color display: red = band 3N, green = band 2, blue = band 1); b location of CWS in Myanmar’s dry deciduous forest areas (white areas on map)

digitized the sanctuary boundary in a Geographic Information System (GIS). The resulting map showed a demarcated area of 362 km2 for CWS, though the official area is reported as 320 km2. CWS is one of Myanmar’s oldest protected areas, with a complex environmental history (Myint Aung et al. 2004). Under British rule the area was created as a fuel reserve in 1919. In 1941 it was converted into a sanctuary, to conserve Eld’s deer and its dry deciduous forest habitat. Three villages inside the sanctuary have been ‘‘grandfathered’’ into the reserve. There are also 34 villages within 10 km of the sanctuary, totaling *4,000 households and [25,000 people. Villagers in the area are primarily subsistence farmers who depend on the sanctuary’s forest to supplement their harvest (Songer 2006). For a more detailed description of CWS and its history, see Myint Aung et al. (2004) and Allendorf et al. (2006). Satellite data Availability of satellite data for the study area is limited; we had to include imagery from different sensors with varying characteristics. Browsing different satellite image archives we selected all freely available mid-resolution satellite imagery that exist for the area and fulfilled our selection criteria: (a) images had to be acquired between the end of the rainy season (October) and the beginning of the dry season (February); and (b) images had to have cloud cover of \5%. We found five images from three different sensors including Landsat Multispectral Scanner (MSS; 19 Nov 1973; 57 m resolution), Landsat Thematic Mapper (TM; 23 Jan 1989 & 25 Dec 1992; 30 m), Landsat Enhanced Thematic Mapper (ETM?; 23 Oct 2001; 30 m), and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER; 3 Jan 2005; 15 m) imagery. The sum of cloud and cloud shadow in

123

Biodivers Conserv (2009) 18:1001–1018

1005

all our imagery amounted to less than 1%; these areas were removed from further analysis. To assure good co-registration, all images were georeferenced to an orthorectified Landsat image from 1989 based on ground control points that were identifiable in both the 1989 image and the image being registered. The orthorectified image was obtained from NASA’s Geocover (Tucker et al. 2004). Deforestation analysis was restricted to the 3,897 km2 area covered by the ASTER image, which includes CWS, Thapanseik Dam and the immediate surroundings (Fig. 1a). Measuring deforestation We performed unsupervised classifications on each image separately, using clustering (ISODATA algorithm; ERDAS 1997), cluster labeling, and cluster busting (Jensen et al. 1987; Rutchey and Vilcheck 1994) as needed. Clusters were interpreted as forest ([30% canopy cover), non-forest, or water and then recoded. To assess classification accuracy of these maps, we used 245 ground-truthing points collected in the area during 2001 (Koy et al. 2005) and compared them to the 2001 classification. CWS staff conducted vegetation surveys at the points, using a GPS unit to obtain locations. Survey points were selected randomly within a homogeneous area that was not near an edge of the vegetation type. Based on these vegetation surveys we designated each point as forest or non-forest and cross-tabulated these designations with the categories in the 2001 map. Out of 245 survey points, 157 were located in forest and 88 in non-forest. The 2001 classification had an overall accuracy of 92.6%. Ground-truthing points and aerial photos are not available for other years, but the 2001 analysis is an indicator of our overall ability to classify forest cover correctly for the study area. After classifications were completed, Landsat MSS and the ASTER classifications were resampled to a cell size of 30 m to allow for integration with the Landsat TM and ETM? classifications. After resampling, all images were smoothed using a 3 9 3 cell majority filter to reduce noise in the data. Using GIS to overlay the classifications we determined deforestation between time periods and assessed the spatial and temporal dynamics of these changes. We produced deforestation maps for each period and one for overall forest loss from 1972 to 2005. Maps incorporated the following categories: (1) non-change classes or categories that remain under the same land cover type for both time steps being compared, including: (a) forest—all areas with [30% canopy cover; (b) water—lakes, reservoirs, rivers and streams; and (c) non-forest—all other areas; (2) change classes, including: (a) deforestation—non-forest or water areas that were classified forest in the previous time step; (b) regrowth—forested areas that were non-forest in the previous time step. During the 1992–2001 analysis *15% of the area was impacted by flooding from a hydroelectric dam. We masked out this area and re-calculated deforestation rates during this period to compare rates with and without the impact of the flood. Differences in spatial resolution in our satellite imagery could potentially impact deforestation estimates. We tried to minimize these biases and also conducted a control study to assess the size of this potential bias. To minimize resolution biases in deforestation rates, we used 30 m as the base resolution for all analyses. This is justified because it allowed us to retain the greatest amount of detail—keeping three out of five images at their native resolution—while not biasing the results significantly towards the finer or coarser resolution data sets. Our control study showed only small biases in annual deforestation rates as the result of resolution differences (0.03–0.22%). For the control study we resampled the 1989 Landsat TM image to 60 m, then classified it and calculated change from the 1973 Landsat MSS

123

1006

Biodivers Conserv (2009) 18:1001–1018

classification. Similarly, we degraded the resolution of the 2005 ASTER image to 30 m, classified its land cover, and compared it to the 2001 Landsat ETM? classification. Land cover change index To explore spatial deforestation dynamics, we developed a land cover change index ranging from 0 to 4, indicating the number of changes that occurred between the acquisition dates of the five satellite images. Based on this data set we analyzed spatial and temporal patterns inside and outside the sanctuary. By combining the land cover change index map with land cover maps for 1973 and 2005, we identified driving forces behind various patterns of land cover change, specifically deforestation. The assumptions for this analysis were as follows: (a)

Areas forested in 1973 and experiencing one change to non-forest are considered permanently converted to agricultural areas or associated village areas. (b) All forested areas experiencing more than one change (excluding water change) are considered to be shifting cultivation. (c) Non-forest areas that changed to forest at least once between 1973 and 2005 are considered to be shifting cultivation. (d) All forest and non-forest areas that experienced a single change or multiple changes to water during any time period were considered to be flooded or floodplains. Protected area effectiveness We assessed CWS’ effectiveness at retaining original land cover by comparing inter-period and overall rates of net deforestation found inside and outside CWS. Prior to 2005 there were no tree planting or other restoration efforts going on inside CWS, so all reforestation detected within the area was due to natural regeneration. To assess the patterns of forest loss inside CWS we calculated deforestation rates within 1 km wide buffers reaching from the boundary to the center of the sanctuary. Fragmentation and potential impact on Eld’s deer We used the FRAGSTATS (McGarigal et al. 2002) to measure changes in fragmentation levels, specifically number of patches (n), forest area (a), mean patch size (mps), mean nearest neighbor distance (mnn), and edge density (ed). Forest patches were defined as forest area C15 km2, the minimum area required for a single Eld’s deer home range (Myint Aung et al. 2001). Edge density was calculated as the total length of all edge perimeters divided by the summed area of all patches. Nearest neighbor for each patch is the single shortest distance to another patch.

Results Temporal deforestation dynamics and forest losses Declines in forest cover in the study area were high between 1973 and 2005 (Table 1). In 1973, 2,738 km2 ([70%) of the landscape was forested but within 32 years forest cover had declined to 1,092 km2 (28%). Declines in percent forest cover were more dramatic

123

Biodivers Conserv (2009) 18:1001–1018 Table 1 Forest area inside Chatthin Wildlife Sanctuary (CWS) and the surrounding landscape between 1973 and 2005

1007

Year

Inside (km2)

Outside (km2)

Total (km2)

1973

333

2,405

2,738

1989

310

1,734

2,044

1992

296

1,624

1,920

2001

281

720

1,001

2005

286

806

1,092

outside than inside CWS (Fig. 2). If these trends continue unchanged, all forest outside CWS will be lost by 2019. Protection inside CWS slowed declines and CWS forests may persist until at least 2165 under these trends. Deforestation rates inside and outside CWS explain these patterns and illustrate the temporal dynamics of land cover change (Table 2; Figs. 3, 4). About 1,688 km2 (62%) of forests were lost in the study area since 1973 and the total annual deforestation rate of 1.86% is well above the global average of 0.2%. Based on this rate, the area is losing nearly 20% of its forest cover every 10 years. However, even this high deforestation rate is probably an underestimate influenced by relatively low deforestation rates between 1973 and 1989. Inter-period deforestation rates were generally much higher and show a continuous increase from 1.57% in 1973 to 5.26% in 2001. The peak inter-period deforestation rate occurred between 1992 and 2001 (Table 2), when a substantial forest area was flooded by a hydroelectric development (Fig. 3c). Flooding increased already high annual deforestation rates from 4.12 to 5.26%. In anticipation of the flooding, large forest areas southwest of CWS were logged by 2001. Some of these areas never flooded and resulting regrowth strongly influenced annual net deforestation rates from 2001 to 2005 (-2.25%; yellow areas in Fig. 3d). Deforestation was most severe outside CWS, with annual rates ranging from 1.73 to 6.11% (Table 2; Fig. 3). Forest losses were also high inside CWS, with net annual deforestation rates about twice the global average. However, the difference between inside and outside deforestation rates is considerable (Table 2). Temporal patterns in

Fig. 2 Declines in percent forest area inside and outside Chatthin Wildlife Sanctuary (CWS) between 1973 and 2005

123

1008 Table 2 Deforestation rates inside and outside Chatthin Wildlife Sanctuary (CWS) between 1973 and 2005

Biodivers Conserv (2009) 18:1001–1018

Period

Annual deforestation rate (%) Inside

Outside

Total

1973–1989

0.42

1.73

1.57

1989–1992

1.44

2.10

2.00

1992–2001

0.59

6.11

5.26

2001–2005

-0.44

-2.95

-2.25

1973–2005

0.45

2.06

1.86

Fig. 3 Temporal and spatial deforestation dynamics. Land cover change analyses for each time period. a 1973–1989; b 1989–1992; c 1992–2001; d 2001–2005. Forest, Non-forest, and Water categories represent areas that did not change during the study period. Original forest cover from 1973 is represented by green and purple in (a). CWS = Chatthin Wildlife Sanctuary

123

Biodivers Conserv (2009) 18:1001–1018

1009

Fig. 4 Overall land cover change between 1973 and 2005 inside and outside Chatthin Wildlife Sanctuary (CWS)

deforestation rates were variable inside the reserve; in contrast to the area outside the reserve, the greatest forest destruction occurred between 1989 and 1992 when deforestation rates briefly went up to about 70% of outside rates, but declined afterwards. Spatial deforestation dynamics and land cover change Analysis of landscape patterns revealed substantially increased fragmentation and patchiness of dry deciduous forests in the study area from 1973 to 2005 (Fig. 5). Overall the amount of habitat area within patches large enough to cover the average home range of an Eld’s deer (i.e. area C15 km2) decreased by two-thirds from 2,658 to 883 km2 (Fig. 5). At the same time, mean patch size decreased, and number of patches as well as edge density increased. This indicates substantial declines in forest area and connectivity associated with severe landscape fragmentation. Additionally, connectivity among patches declined markedly during the study period from a mean nearest neighbor distance of 0.13–1.73 km. Although there was no net deforestation between 2001 and 2005 (gross deforestation was balanced out by regrowth) this did not substantially change fragmentation patterns. (Fig. 5d, e). Based on the analysis of the land cover change index, most of the area (62%) experienced at least one change in land cover between 1973 and 2005 (Table 3; Fig. 6a). The majority of these changes were unidirectional; i.e., a change occurred only once and was permanent. Relatively little change occurred inside CWS; almost three-quarters of the sanctuary remained unchanged over time. Conversion of forest to agriculture was the most important driving force for deforestation in the study area (Table 4; Fig. 6b). Over one-third of all land experienced this change once and subsequently never returned to forest. However, shifting cultivation and flooding also affected large portions of the land. Most of the flooding occurred between

123

1010

Biodivers Conserv (2009) 18:1001–1018

Fig. 5 Fragmentation of Eld’s deer habitat between 1973 and 2005 in and outside Chatthin Wildlife Sanctuary (CWS). Abbreviations: n = number of patches C15 km2; a = total area of all patches C15 km2; mps = mean patch size for all patches C15 km2; ed = edge density; mnn = mean nearest neighbor distance

123

Biodivers Conserv (2009) 18:1001–1018

1011

Table 3 Land cover change index describing the number of land cover changes inside Chatthin Wildlife Sanctuary (CWS) and outside its boundaries between 1973 and 2005 Change indexa

Area (km2)

Percent

Inside

Outside

Total

Inside

Outside

Total

0

271

1,243

1,514

74

35

38

1

48

1,603

1,650

13

46

43

2

30

450

480

8

13

12

3

13

184

196

4

5

5

4

2

26

28

1

1

1

a

The land cover change index was calculated by summing the changes occurring for each pixel during each time step for a maximum of four changes

Fig. 6 Land cover change index and land use changes that are driving deforestation; a Change index representing number of land cover changes occurring at each pixel throughout the study; b Land use change associated with deforestation. Land use change categories: (A) agricultural conversion = areas characterized by unidirectional change from forest to non-forest; (B) shifting cultivation = forest or non-forest areas changing cover more than once; (C) flooded/floodplain = areas with one or more changes to water

1992 and 2001 when Thapanseik dam was built and its flooding affected *420 km2 of forest and cropland. Only 21% of the entire study area remained under forest cover and was never affected by a land cover change that we could detect (Table 4; Fig. 6b). Patterns of land use changes are different inside and outside CWS, with shifting cultivation more prevalent inside CWS (12% vs. 9% for agricultural conversion). Almost all agricultural conversion inside CWS is limited to boundary areas or in proximity to villages grandfathered into the sanctuary (Fig. 6b). The most severe forest losses inside CWS occurred within 2 km of the sanctuary boundary (Fig. 7). These impacts declined sharply moving inward and all areas beyond 4 km from the boundary experienced deforestation rates below the countrywide rate of 0.3% (Leimgruber et al. 2005).

123

1012

Biodivers Conserv (2009) 18:1001–1018

Table 4 Land use changes associated with deforestation between 1973 and 2005 Land use/land use changea

Area (km2)

Percent

Agricultural conversion

1,367

35

Shifting agriculture

482

12

Flooded/floodplain

506

13

Forest

816

21

Agriculture

686

18

11

\1

Water

a Land use change categories are defined as: (a) agricultural conversion = areas characterized by a unidirectional change from forest to non-forest; (b) shifting cultivation = areas starting as forest or non-forest and changing between forest or non-forest more than once; (c) flooded/floodplain = areas with at least one change and classified as water at least once

Fig. 7 Declines in annual deforestation rates with increasing distance from the boundary of Chatthin Wildlife Sanctuary (CWS)

Discussion Dry forest losses and driving forces Deforestation is rapidly changing the landscape around CWS. Our findings support other studies showing that Myanmar’s dry forests are among the most threatened and least protected forest ecosystems in the country (Koy et al. 2005; Leimgruber et al. 2005) and is also representative of dry forest declines in other countries in the region (DeFries et al. 2005). Deforestation patterns show that forest declines primarily resulted from agricultural conversion and subsistence uses rather than large-scale logging. As rural populations are expanding, pioneering farmers are changing forests to agriculture. Some of this is the result of government policies that promote expansion of agriculture and subsistence farming in these formerly remote areas (Eberhardt 2003; McShea et al. 1999). In 1987–1988 the

123

Biodivers Conserv (2009) 18:1001–1018

1013

Myanmar Agricultural Enterprise encouraged planting sugar cane in dry forest areas (Myint Aung et al. 2004). Sugar cane is 2–3 times more valuable than typical crops and can be exported to China, India, and Thailand. Increased cane production has additional indirect effects on forest because it requires a great deal of fuelwood for processing. Shifting cultivation is the second major force driving deforestation. Shifting cultivation has a long tradition in Myanmar and other countries in the region (Menzies 1995, cited by Eberhardt 2003; Eberhardt 2003). While areas used for shifting cultivation arguably can still reforest, trends lean towards permanent conversion into agriculture for several reasons. First, increasing populations and food demands have led to a shortening of rotation cycles beyond sustainability, causing transitions to permanent agriculture. Second, expansion of land registration policies increasingly move land under the control of large agrobusinesses, further increasing the transition rate (Eberhardt 2003). Hydroelectric development is another major deforestation driver and probably has increasingly affected forests throughout Myanmar and Southeast Asia. Between 1988 and 2002 at least 23 hydroelectric dams and 129 irrigation dams were constructed in Myanmar (Ministry of Information 2002). The negative environmental consequences of these dam developments have been pointed out in a region-wide analysis (Hirsch 1995, cited by Eberhardt 2003), but little information currently exists about how dam developments are affecting forests in Myanmar, and in particular dry forests. Studies of impacts of hydroelectric development in Thailand showed native small mammal species declined in just 5 years, with several local extinctions caused by fragmentation of forests on small islands in the new reservoir (Lynam 1997). Similar studies in Venezuela show that fragmentation caused by hydroelectric development resulted in ecological imbalances through the disappearance of sensitive species and trophic guilds, such as predators and parasites, and the loss of key species and overabundance of others, ultimately leading toward biodiversity loss and a more simplified ecosystem (Terborgh et al. 1997). In addition to immediate forest losses from flooding, there are cascading impacts as villagers are forced to relocate, rebuild homes, and acquire new fields. Flooding of Thapanseik dam forced the relocation of 52 villages; 5 moved adjacent to the southwestern CWS boundary. Many households had few economic assets and relied even more heavily on forest resources for fodder and other products (Myint Aung et al. 2004). In anticipation of the flood a 165 km2 reserve forest located just south of CWS was logged—an area that had connectivity to CWS and maintained a population of Eld’s deer before it was eliminated. Landscape dynamics Chatthin Wildlife Sanctuary undergoes two phases of forest change during the study period. During the first phase the area transitions from a nearly contiguous forest area into increasingly smaller patches as the area is fragmented by conversion to agriculture. By 2001 nearly all forest patches surrounding CWS have been removed and during the last inter-period we see dynamics due to regrowth and continued subsistence use along the boundaries. To understand the patterns of the landscape, we need to evaluate the spatial and temporal dynamics, i.e., where a forest disappears and regrows is as important how much is disappearing over time. Although there seems to be a positive forest trend during the last time period, we can see this is patchy regrowth, with a few patches returning near the flooded area (Fig. 3d). These forests were cut in anticipation of the flood, yet did not flood, resulting in regrowth. It is likely that many of these areas will be degraded or converted in the near future.

123

1014

Biodivers Conserv (2009) 18:1001–1018

Fragmentation, habitat degradation, and increasing human populations have reduced the range of Eld’s deer. Surveys through most of the existing range in 1997 showed evidence of Eld’s deer in 23 townships (McShea et al. 1999). Similar surveys in 2003 found evidence in only 18 townships (Myint Aung, unpublished data). Landscape analysis combined with the 1997 survey data showed that habitat area was the most important variable for predicting presence of Eld’s deer. Koy et al. (2005) found that Eld’s deer population was positively correlated with density of dry forests and density of forests was negatively correlated with human population. As human populations increase so does pressure on the forest. Without improved land use planning and development of more sustainable strategies we can expect continued decline of Eld’ deer populations. Effectiveness of protection at CWS Maintenance of original land cover is essential to conserving biodiversity. Our study demonstrated that CWS is effective in preserving dry forests and Eld’s deer habitat. Our results were similar to those reported from many protected areas throughout the tropics. Bruner et al. (2001) showed that 93 protected areas in tropical forests were successful in slowing forest loss although the surrounding landscapes were largely deforested during the study period. Similar results were presented by a land cover change study of 198 tropical protected areas (DeFries et al. 2005) and a detailed study of deforestation in the Sarapiqui region in Costa Rica (Sa´nchez-Azofeifa et al. 2005). In all of these studies the protected areas were surrounded by rapidly changing landscapes. Although they became increasingly more isolated, they maintained high forest cover within their boundaries, potentially providing strongholds for species and biodiversity conservation. However, maintenance of original land cover alone is not proof of effectiveness as other damaging processes and threats, such as poaching, may take place without affecting the forest size or structure. Myanmar economics and politics, as well as the history of CWS management and protection efforts explain the deforestation patterns observed in our study. Prior to 1988, deforestation in the entire region was negligible and little development or land cover change occurred. Major economic changes on a national level occurred in 1988, including the introduction of market principles in 1988 and the demonetization fo the currency. This resulted in many families losing their savings and being forced to rely more heavily on resource extraction. These economic changes are paralleled with sharply increased deforestation rates in our image analyses for the period from 1989 to 1992. Protection and management efforts were limited during this time period, with few staff members that were headquartered several kilometers from the CWS boundary (Myint Aung et al. 2004). During the 1992–2001 we see deforestation trends outside CWS continue to increase to their highest level (6.11% annually), yet at the same time CWS rates were cut to less than half previous levels (1.44% down to 0.59% annually). This contrast coincides with the initiation of Eld’s deer surveys and other biodiversity inventories by Smithsonian scientists in 1992 (Wemmer et al. 2004). Smithsonian projects not only increased research activities, but also provided support for protection activities (Myint Aung et al. 2004). Rangers began conducting wildlife surveys and patrolling regularly, the CWS boundary was recorded using GPS, and permanent markers were constructed to establish visibility for sanctuary staff and local people. From 1996, staff were based permanently inside the sanctuary and several stations were built. Enforcement focused on stopping extraction, particularly highimpact activities such as hunting and timber removal. In 1999, the warden began community outreach through annual workshops for area village chairman and regular visits to

123

Biodivers Conserv (2009) 18:1001–1018

1015

interior villages to discuss CWS and its purpose. Prior to this villagers had little information about CWS or staff. The combined effects of patrolling, increased staff presence, education, infrastructural improvements, and foreign interest raised the profile of the sanctuary, increased conservation awareness of local people, and mitigated negative impacts on forest by reducing extraction. Infrastructural improvements such as stations helped support patrolling, while boundary markers made it possible to follow through with legal action by allowing staff to show that offenders were aware they were inside CWS. Interviews with villagers and CWS staff indicate that the staff presence and patrolling efforts led villagers to reduce the frequency of trips into CWS and limit activities. For example, most no longer removed large trees for housepoles and hunters scaled back efforts. Patrol staff have broken up hunting rings and enforced punishment through fines and jail sentences. Local villagers were hired as staff, improving relations with villages and information about extractive activities. Improved protection efforts also had clear impacts in slowing biodiversity loss. Interviews to determine historical and present wildlife diversity demonstrate some faunal recovery during the time of improved management efforts (Myint Aung et al. 2004). The large mammals of CWS including gaur, sambar deer, leopards, and tigers, had declined to the point of being critically endangered to extinct by the 1990s. However, medium-sized mammals such as hog deer, wild dog, and muntjac populations that declined through the 1970s and 1980s began to rebound during the 1990s. Similarly, Eld’s deer densities declined between 1983 and 1996 from 8.3 to 4.7 per km2 (McShea et al. 1999); but in 1997 they began to recover and increased to 9.2 per km2 by 2005 (Myint Aung, unpublished data). Protecting remaining dry forests in Myanmar If trends found in this analysis continue, dry forest in this area will be gone by 2019, with the exception of CWS. However, CWS will be a small and severely isolated island in a sea of agriculture with little potential for long-term conservation of dry forest biodiversity or even Eld’s deer. While protection efforts at CWS have slowed deforestation, degradation and poaching on the inside, it cannot maintain the ecosystem in a well connected, intact dry forest landscape. We presented a case study, but the situation for dry forest conservation is likely to be worse throughout the rest of Myanmar and potentially throughout all of Southeast Asia. Between the 1960s and 1980s, Thailand has lost more than half of its original forest cover and little dry forest remains. Most of Myanmar’s remaining dry forests are found along the edges of the hill regions in areas less preferred for agriculture. Yet, as conversion pressures increase, these dry forest areas may decline as well. There is little indication that agricultural conversion can be reversed, even in the case of shifting cultivation, which, when it occurs today, is likely a transitional phase that will eventually move toward permanent conversion. Establishing legal protection is the first important step for maintaining the country’s dry forests and biodiversity, however that is often not sufficient. Government land use policies often do not take environmental concerns into account. For example, the promotion of sugar cane as a crop and the development of Thapanseik dam were government policies that had unintended but severe consequences for CWS. Both increased encroachment in the park. The former by increasing sugar cane yields, the latter by displacing people.

123

1016

Biodivers Conserv (2009) 18:1001–1018

Half of Myanmar’s protected areas have high-impact activities and incompatible components, such as plantations, timber removal, mining, roads, railways, and military encampments, going on inside the boundaries (Rao et al. 2002). These go well beyond the smaller scale extractive activities of local people and only occur with support from government officials; major national level policy changes are required to prevent these activities in the future. Acknowledgments We thank the staff of Chatthin Wildlife Sanctuary for their assistance in collection of field data and Dan Kelly and Melanie Delion for assistance in data management in the laboratory. This project was primarily funded by a grant awarded by NASA to the Conservation Non-Governmental Working Group.

References Allendorf TD, Khaing Khaing Swe, Thida Oo, Ye Htut, Myint Aung, Hayek LA, Leimgruber P, Wemmer C (2006) Community attitudes toward three protected areas in upper Myanmar (Burma). Environ Conserv 33(4):344–352. doi:10.1017/S0376892906003389 Myint Aung (2007) Policy and practice in Myanmar’s protected area system. J Environ Manage 84(2):188– 203. doi:10.1016/j.jenvman.2006.05.016 Myint Aung, McShea WJ, Sein Htung, Aung Than, Tin Mya Soe, Monfort S, Wemmer C (2001) Ecology and social organization of a tropical deer (Cervus eldi thamin). J Mammal 82(3):836–847 Myint Aung, Khaing Khaing Swe, Thida Oo, Kyaw Kyaw Moe, Leimgruber P, Allendorf T, Duncan C, Wemmer C (2004) The environmental history of Chatthin Wildlife Sanctuary, a protected area in Myanmar (Burma). J Environ Manage 72:205–216. doi:10.1016/j.jenvman.2004.04.013 Brooks TM, Pimm SL, Collar NJ (1997) Deforestation predicts the number of threatened birds in insular Southeast Asia. Conserv Biol 11(2):382–394. doi:10.1046/j.1523-1739.1997.95493.x Bruner AG, Gullison RE, Rice RE, da Fonseca GAB (2001) Effectiveness of parks in protecting tropical biodiversity. Science 291(5501):125–127. doi:10.1126/science.291.5501.125 Bullock SH, Mooney HA, Medina E (1995) Seasonally dry tropical forests. Cambridge University Press, Cambridge, 450 pp Bunyavejchewin SC (1982) Canopy structure of the dry dipterocarp forest of Thailand. University of Washington, Seattle Carroll C, Miquelle DG (2004) Spatial viability analysis of Amur tiger Panthera tigris altaica in the Russian Far East: the role of protected areas and landscape matrix in population persistence. J Appl Ecol 43(6):1056–1068. doi:10.1111/j.1365-2664.2006.01237.x Chape S, Harrison J, Spalding M, Lysenko I (2005) Measuring the extent and effectiveness of protected areas as an indicator for meeting global biodiversity targets. Philos Trans R Soc B 360:442–455. doi: 10.1098/rstb.2004.1592 DeFries RS, Hansen A, Newton AC, Hansen MC (2005) Increasing isolation of protected areas in tropical forests over the past twenty years. Ecol Appl 15(1):19–26. doi:10.1890/03-5258 Eberhardt K (2003) A review of challenges to sustainable development in the uplands of Myanmar. In: Xu J, Mikesell S (eds) Landscapes of diversity: indigenous knowledge, sustainable livelihoods and resource governance in montane mainland Southeast Asia. Proceedings of the III symposium on MMSEA, Lijiang, August 2002. Yunnan Science and Technology, Kunming, pp 101–111 ERDAS Inc (1997) ERDAS tour guide 8.3. ERDAS Inc., Atlanta Ervin J (2003) Rapid assessment of protected area management effectiveness in four countries. Bioscience 53(9):833–841. doi:10.1641/0006-3568(2003)053[0833:RAOPAM]2.0.CO;2 Ewel JJ (1999) Natural systems as models for the design of sustainable systems of land use. Agrofor Syst 45(1–3):1–21. doi:10.1023/A:1006219721151 Foody GM, Cutler MEJ (2003) Tree biodiversity in protected and logged Bornean tropical rain forests and its measurement by satellite remote sensing. J Biogeogr 30(7):1053–1066 Gentry AH (1984) An overview of neotropical phytogeographic patterns with an emphasis on Amazonia. Missouri Botanical Garden, St. Louis Goodman PS (2003) Assessing management effectiveness and setting priorities in protected areas in KwaZulu-Natal. Bioscience 53(9):843–850. doi:10.1641/0006-3568(2003)053[0843:AMEASP] 2.0.CO;2

123

Biodivers Conserv (2009) 18:1001–1018

1017

Gurd DB, Nudds TD, Rivard DH (2001) Conservation of mammals in Eastern North American wildlife reserves: how small is too small? Conserv Biol 15(5):1355–1363. doi:10.1046/j.1523-1739.2001. 00188.x Hansen AJ, Rotella JJ (2001) Nature reserves and land use: implications of the ‘‘place’’ principle. In: Dale V, Haeuber R (eds) Applying ecological principles to land management. Springer-Verlag, New York Hirsch P (1995) A regional perspective on large dams. As cited in Eberhardt K (2003) A review of challenges to sustainable development in the uplands of Myanmar. In: Xu J, Mikesell S (eds) Landscapes of diversity: indigenous knowledge, sustainable livelihoods and resource governance in montane mainland Southeast Asia. Proceedings of the III symposium on MMSEA, Lijiang, August 2002. Yunnan Science and Technology Press, Kunming, pp 101–111 Hockings M, Stolton S, Dudley N (2004) Management effectiveness: assessing management of protected areas? J Environ Policy Plan 9(2):157–174. doi:10.1080/1523908042000320731 Janzen DH (1988) Tropical dry forests: the most endangered major tropical ecosystem. In: Wilson EO, Peters FM (eds) Biodiversity. National Academy Press, Washington Jensen JR, Ramsey EW, Mackey HE, Christensen E, Sharitz R (1987) Inland wetlands change detection using aircraft MSS data. Photogramm Eng Remote Sensing 9:669–682 Koy K, McShea WJ, Leimgruber P, Haack BN, Myint Aung (2005) Percentage canopy cover-using Landsat imagery to delineate habitat for Myanmar’s endangered Eld’s deer (Cervus eldi). Anim Conserv 8(3):289–296. doi:10.1017/S1367943005002209 Kurz S (1877) Forest flora of British Burma, vols 1 and 2. Office of the Superintendent of Government Printing, Calcutta Leimgruber P, Kelly DS, Steininger MK, Brunner J, Mueller T, Songer M (2005) Forest cover change patterns in Myanmar (Burma) 1990–2000. Environ Conserv 32(4):356–364. doi:10.1017/S037689290 5002493 Linkie M, Smith RJ (2004) Mapping and predicting deforestation patterns in the lowlands of Sumatra. Biodivers Conserv 13(10):1809–1818. doi:10.1023/B:BIOC.0000035867.90891.ea Liu J, Linderman M, Ouyang Z, An L, Yang J, Zheng H (2001) Ecological degradation in protected areas: the case of Wolong Nature Reserve for giant pandas. Science 292(5514):98–101. doi:10.1126/science. 1058104 Lynam AJ (1997) Rapid decline of small mammal diversity in monsoon evergreen forest fragments in Thailand. In: Laurance WF, Bierregaard RO (eds) Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago Maass JM (1995) Conversion of tropical dry forest to pasture and agriculture. In: Bullock SH, Mooney HA, Medina E (eds) Seasonally dry tropical forests. Cambridge University Press, Cambridge Mas J-F (2005) Assessing protected area effectiveness using surrounding (buffer) areas environmentally similar to the target area. Environ Monit Assess 105(1–3):69–80. doi:10.1007/s10661-005-3156-5 McGarigal K, Cushman SA, Neel MC, Ene E (2002) FRAGSTATS: spatial pattern analysis program for categorical maps. Computer software program produced by the authors at the University of Massachusetts, Amherst. Available via: www.umass.edu/landeco/research/fragstats/fragstats.html. Cited 20 Aug 2008 McShea WJ, Leimgruber P, Aung M, Monfort SL, Wemmer C (1999) Range collapse of tropical cervid (Cervus eldi) and the extent of remaining habitat in central Myanmar. Anim Conserv 2(3):173–183. doi:10.1111/j.1469-1795.1999.tb00063.x Menzies NK (1995) The changing dynamics of shifting cultivation practices in upland southwest China. As cited in Eberhardt K (2003) A review of challenges to sustainable development in the uplands of Myanmar. In: Xu J, Mikesell S (eds) Landscapes of diversity: indigenous knowledge, sustainable livelihoods and resource governance in montane mainland Southeast Asia. Proceedings of the III Symposium on MMSEA, Lijiang, August 2002. Yunnan Science and Technology Press, Kunming, pp 101–111 Miles L, Newton AC, DeFries RS, Ravilious C, May I, Blyth S et al (2006) A global overview of the conservation status of tropical dry forests. J Biogeogr 33(3):491–505. doi:10.1111/j.1365-2699.2005. 01424.x Ministry of Information (2002) Myanmar facts and figures. Ministry of Information, University of Myanmar, Yangon Murphy P, Lugo AE (1986) Ecology of tropical dry forests. Annu Rev Ecol Syst 17:67–88. doi:10.1146/ annurev.es.17.110186.000435 Myers N (1984) The primary source: tropical forests and our future. Norton, New York Newmark WD (1987) A land-bridge island perspective on mammalian extinctions in western North Arnerican parks. Nature 325:430–432. doi:10.1038/325430a0

123

1018

Biodivers Conserv (2009) 18:1001–1018

Newmark WD (1996) Insularization of Tanzanian parks and the local extinction of large mammals. Conserv Biol 10:1549–1556. doi:10.1046/j.1523-1739.1996.10061549.x Parrish JD, Braun DP, Unnasch RS (2003) Are we conserving what we say we are? Measuring ecological integrity within protected areas. Bioscience 53:851–860. doi:10.1641/0006-3568(2003)053 [0851:AWCWWS]2.0.CO;2 Pimm SL, Raven P (2000) Extinction by numbers. Nature 403:843–845. doi:10.1038/35002708 Rao M, Rabinowitz A, Saw Tun Khaing (2002) Status review of the protected-area system in Myanmar, with recommendations for conservation planning. Conserv Biol 16(2):360–368. doi:10.1046/j.1523-1739. 2002.00219.x Raven PH (1988) Our diminishing tropical forests. In: Wilson EO, Peters FM (eds) Biodiversity. National Academy Press, Washington Rutchey K, Vilcheck L (1994) Development of an everglades vegetation map using a SPOT image and the global positioning system. Photogramm Eng Remote Sens 60(6):767 Sa´nchez-Azofeifa GA, Quesada M, Rodriguez JP, Nassar JM, Stoner KE, Castillo A et al (2005) Research priorities for neotropical dry forests. Biotropica 37(4):477–485. doi:10.1111/j.1744-7429.2005.00066.x Songer M (2006) Tropical dry forests of Southeast Asia. Dissertation, University of Maryland Stamp LD (1925) The vegetation of Burma from an ecological standpoint. Thacker, Spink and Company, Calcutta Stott P (1990) Stability and stress in the savanna forests of mainland South-East Asia. J Biogeogr 17(4/ 5):373–383. doi:10.2307/2845366 Struhsaker TT, Struhsaker PJ, Siex KS (2005) Conserving Africa’s rain forests: problems in protected areas and possible solutions. Biol Conserv 123(1):45–54. doi:10.1016/j.biocon.2004.10.007 Sutton SL, Whitmore TC, Chadwick AC (eds) (1984) Tropical rain forests: ecology and management. Blackwell, Oxford Terborgh J, Lopez L, Tello J, Yu D, Bruni AR (1997) Transitory states in relaxing ecosystems of land bridge islands. In: Laurance WF, Bierregaard RO (eds) Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago Trigg SN (2006) Utility of landsat 7 satellite data for continued monitoring of forest cover change in protected areas in Southeast Asia. Singap J Trop Geogr 27:49. doi:10.1111/j.1467-9493.2006.00239.x Tucker CJ, Grant DM, Dykstra J (2004) NASA’s global orthorectified landsat data set. Photogramm Eng Remote Sens 70(3):313–323 Wemmer C, Kress WJ, Zug G (2004) Smithsonian collaborations in Myanmar: a compilation of projects and activities between 1994–2003. Smithsonian Institution, Washington Zheng D, Wallin DO, Hao Z (1997) Rates and patterns of landscape change between 1972 and 1988 in the Changbai mountain area of China and North Korea. Landscape Ecol 12(4):241–254. doi:10.1023/ A:1007963324520

123