Assessing forest canopy closure in a geospatial medium to address management concerns for Tropical IslandsSouth East Asia by Rama Chandra Prasad, K.S.Rajan
in Environmental Monitoring and Assessment DOI 10.1007/s10661-008-0717-4
Report No: IIIT/TR/2009/80
Centre for Spatial Informatics International Institute of Information Technology Hyderabad - 500 032, INDIA April 2009
Environ Monit Assess DOI 10.1007/s10661-008-0717-4
Assessing forest canopy closure in a geospatial medium to address management concerns for tropical islands—Southeast Asia P. Rama Chandra Prasad · Nidhi Nagabhatla · C. S. Reddy · Stutee Gupta · K. S. Rajan · S. H. Raza · C. B. S. Dutt
Received: 9 July 2008 / Accepted: 23 December 2008 © Springer Science + Business Media B.V. 2009
Abstract The present study outlines an approach to classify forest density and to estimate canopy closure of the forest of the Andaman and Nicobar archipelago. The vector layers generated for the study area using satellite data was validated with the field knowledge of the surveyed ground control points. The methodology adopted in this present analysis is three-tiered. First, the density stratification into five zones using visual
P. R. C. Prasad (B) · K. S. Rajan Laboratory for Spatial Informatics, International Institute of Information Technology, Gachibowli, Hyderabad 500 032, India e-mail:
[email protected] N. Nagabhatla World Fish Centre, Batu Muang, Penang, Malaysia C. S. Reddy Forestry and Ecology Division, National Remote Sensing Agency, Department of Space, Balanagar, Hyderabad, India S. Gupta Forest Research Institute, Dehradun, India S. H. Raza Aurora’s Scientific Technological & Research Academy, Hyderabad, India C. B. S. Dutt Indian Space Research Organization, Department of Space, Anthariksh Bhavan, Bangalore, India
interpretation for the complete archipelago. In the second step, we identified two island groups from the Andaman to investigate and compare the forest strata density. The third and final step involved more of a localised phytosociological module that focused on the North Andaman Islands. The results based on the analysis of the high-resolution satellite data show that more than 75% of the mangroves are under high- to very high-density canopy class. The framework developed would serve as a significant measure to forest health and evaluate management concerns whilst addressing issues such as gap identification, conservation prioritisation and disaster management—principally to the post-tsunami assessment and analysis. Keywords Tropical · Islands · Forest · Canopy density · South Asia · Geospatial · Management
Introduction Tropical rainforests are bestowed with a wide variety of species richness and diversity patterns (Alwyn and Calaway 1987; Jacobs 1988). Apart from these characteristics, they are recognised as dense forests due to the high density of vegetation formed by the clumped distribution of individuals and also enormously tangled undergrowth of different herbs, shrubs, lianas and climbers.
Environ Monit Assess
Detailed and accurate maps of these forests— condition and structure—are needed for assessment of the flora and faunal biodiversity as well as for sustainable ecosystem management (Blodgett et al. 2000). The conventional way of ground monitoring for density estimation can be tedious and time-consuming, whilst the use of geographical information system (GIS) and remote sensing as a platform for estimating the density of these forests may speed up the process and provide for a more efficient option (Blodgett et al. 2000). Additionally, the remote sensing data also facilitate spatial delineation of vegetation density maps through various techniques using satellite imagery in conjunction with GIS and phytosociological ground data (Chauhan 2004; Prasad 2006). In addition, these data techniques can be combined for spatial prediction and modelling of the vegetation’s biophysical properties (Muukkonen and Heiskanen 2005). Forest canopy density is one of the important parameters in the planning and implementation of forest rehabilitation programs (Rikimaru et al. 2002). It has been suggested that canopy density is an essential parameter to assess and analyse the factors affecting forest growth, its regeneration and to keep a check on management initiatives in gap area plantations and regeneration status (Chauhan 2004). Further, as discussed by Blodgett et al. (2000), the multi-spectral earth observation data with appropriate ground measurements can suitably extrapolate across a large geographic region, and this has significant advantages for forest management, especially in areas where forests are in remote locations or are inaccessible. Also, the spatial techniques to measure the canopy closure along with canopy height profiling has been illustrated from different regions using a wide array of spatial and spectral resolutions (Roy et al. 1990, 1996; Rikimaru and Miyatake 1997; Katul and Albertson 1998; Harding et al. 2001; Marshall et al. 2002; Singh et al. 2003). Background Advancement in geospatial technologies provides a medium to evaluate forest cover in the inaccessible and remote isles. This can also enable a multiscalar assessment of the associated ecosystem
parameters (forest canopy density, habitat diversity, etc.). The satellite remote sensing is best suited for analysis of canopy closures, as elucidated by Roy et al. (1994) whilst modelling the biophysical spectral response for forest density stratification in evergreen forests of South Andaman division and dry deciduous forests over central India. The study was based on the assumption that any alteration in forest canopy is reflected in the crown structure. With inspiration from the above statement, this study illustrates the density analysis and the associated attributes for the tropical Andaman and Nicobar (AN) archipelago, the islands characterised by notable ecological complexity, using a three-tier approach. The biodiversity of these islands is noteworthy and the detail assessment poses a challenge. Past extraction practices, overexploitation of natural resources, and population pressure has seriously affected the ecosystem stability and forest density. Currently, these forests are prone to anthropogenic disturbances due to the high immigration rates and the introduction of various exotic animals like elephants (Rauf 2004). This paper, whilst addressing the need to understand the canopy closure or density of forest communities which are critical for a management planning process, also describes the use of the spatial tools viz., a combination of earth observation data along with the GIS tools that can pertinently contribute to the process. Towards this end, the authors have mapped the forest canopy density as part of the first tier for the complete archipelago, whilst the second tier takes up an in-depth analysis of two islands in the Andaman group—North and Baratang Islands (BI) and the final tier relates the phytodiversity and the density stratification for the North Andaman Islands (NAI). It is envisioned that the derived thematic outputs will provide a synoptic understanding of the extent of open and closed forests in the island landscape. The detailed analysis of the two islands will depict the scenario of forest density individually, for each forest type, thereby providing information on forest stock which can enable selective logging operations by the forest department. In addition, the other outputs may assist in the identification of illegal logging sites and the extent of forest fragmentation.
Environ Monit Assess
Study area and objectives The Andaman and Nicobar Islands (ANI) spreads over the geographic location of 6◦ 5 to 14◦ 45 N lat. and 92◦ to 94◦ E long and vertically stretch over 800 km in the Bay of Bengal with a total area of 8,249 km2 . Nicobar Islands are a group of 28 islands (total area, 1,841 km2 ) separated from the Andaman group (6,408 km2 ) by a 10◦ channel. ANI are one among the nation’s richest biodiversity regions and contains a great assemblage of endemic plant species coupled with high species diversity and density (Roy et al. 2005; Prasad et al. 2007a; Nagabhatla and Roy 2007). The second tier of the study focuses on NAI (12◦ 95 N and 92◦ 86 E) and BI (12◦ 18 N and 92◦ 32 E) of AN archipelago; NAI constitutes a continuous land mass (1,458 km2 ) surrounded by few small islands like Paget, Point, Smith, Landfall, Interview and Narcondam, whilst the BI is more of scattered (small/medium-sized islands separated by straits) landscape with 28 islands, among which the chief are Baratang, Evergreen, Colebrook, Spike, Havelock, Peel, Wilson, Henry Lawrence, John Lawrence, Outram and Neil. The islands have been selected with the purpose of studying the variance in density and diversity in two varied landscapes. Predominantly five major vegetation types, viz. evergreen, semi-evergreen, moist deciduous, mangrove and littoral, flourish in these islands. It has been observed that the forests of these islands have undergone significant changes during the last few decades due to large-scale in-migration, extensive logging, land conversion and forest degradation (details are in Anonymous 2003; Nagabhatla et al. 2006; Prasad et al. 2009). To understand and evaluate the status of the forests before the tsunami struck these regions, the current work analyses the canopy closure. Such an analysis along with the relevant data can prove as a significant baseline to further study the impacts of tsunami in the post-disaster years. The canopy closure analysis in this work has adopted a three-tier approach which will provide insights into the regional, inter-island and local characterisations of the ANI archipelago. In the first level, the entire ANI is analysed using the visual classification approach to stratify the region
into different forest density zones. At the second level, the goal is to understand the variation in the forest canopy closures in a continuous land mass (NAI) vs. scattered islands (BI). These two tiers of the geospatial assignment encapsulate the statistics to decipher the dissimilarity and overlaps in the forest communities of the ANI in general and the two zones in detail both in terms of composition and health. In the third tier, an effort has been made for in-depth analysis of the forest density patterns vis-a-vis phytodiversity for the NAI.
Materials and method Satellite data Indian Remote Sensing Satellite (IRS) 1C/1D Linear Imaging Self Scanner-III (LISS III; 23.5-m resolution), Panchromatic (PAN; 5.8-m resolution) and merged LISS-PAN (multi-spectral 5.8 m) datasets were used. Wherever IRS-1C LISSIII cloud-free data were not available, LandsatThematic Mapper (TM) data were schematically referred. The datasets used in this study are listed in Table 1 with relevant spectral and spatial resolution. Further, the reconnaissance field survey was undertaken to collect ground truth information about the patterns of vegetation in the area— traverses along all roads and major drainage channels, hill tops, creeks and sandy beaches. To compliment the analysis and create a compilation of the existing knowledge base, literature survey was carried out and also interaction with forest department and local institutions was established.
Processing Visual interpretation approach was effectively utilised in this study, which optimised full potential of different band combinations and image enhancement in the multi-spectral data. The image enlargement techniques in the PAN data added to the feature delineation. Visual interpretation technique was opted in this study to capture the variability in the complex tropical formations. In
Environ Monit Assess Table 1 Tabulation of multi-spectral and multi-sensor image data used for delineating the forest canopy for the Andaman and Nicobar Islands (Anonymous 2003) Union territory
Satellite
Path
Row
Date of acquisition
Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Andaman Islands Nicobar Islands Nicobar Islands Nicobar Islands Nicobar Islands Nicobar Islands Nicobar Islands Nicobar Islands Nicobar Islands Andaman and Nicobar Islands
IRS IC–LISS III IRS IC–LISS III IRS IC–LISS III IRS IC–LISS III IRS IC–LISS III IRS IC–LISS III IRS IC–LISS III Landsat TM Landsat TM IRS IC–PAN IRS IC–PAN IRS IC–PAN IRS IC–PAN IRS IC–PAN IRS IC–PAN IRS IC–PAN IRS IC–LISS III IRS IC–LISS III IRS IC–LISS III IRS IC–LISS III Landsat TM Landsat TM IRS IC–PAN IRS IC–PAN IRS IC–PAN and IRS IC–LISS/PAN merged
114 114 114 115 115 115 115 134 134 114 114 114 115 115 115 115 115 116 116 117 132 133 115 115
64 65 66 63 64 65 66 51 52 64 65 66 63 64 65 66 67 68 69 69 55 54 67 68
Mar. 2000 Feb. 2000 and Mar. 2000 Feb. 2000 Jan. 2000 Jan. 2000 and Mar. 2000 Jan. 2000 Jan. 1999 Mar. 2001 Feb. 1997 Feb. 1998 Feb. 1998 Feb. 1998 Feb. 1998 Feb. 1998 Feb. 1998 Feb. 1998 Feb. 1999 Dec. 1999 Aug. 1999 Mar. 1998 Mar. 2000 Feb. 2001 Feb. 1998 Feb. 1998 Mar. 1998, Sept. 1999, April 2001, Feb. 2002
the case of the automated classification methods, the intermixing of closely related forest communities was relatively high, resulting in low classification accuracy due to limited spectral bandwidth (80–100 nm) available with IRS 1C/1D satellite data (Prasad et al. 2007b). Hence, visual techniques were used coupled with intense ground verification. The field knowledge base along with the ancillary data, viz. road, settlement, division boundaries, supported the delineation process. However, we do recognise that the possibilities of upscaling this approach can be complex. The canopy density was primarily mapped using grey tones of PAN data, whilst fused composite image (LISS-III/PAN merged) shored up the refinement and verification process. The method applied for merging the LISS-III and PAN was the high pass filter resolution merge algorithm embedded in the ERDAS digital image processing software
that pan-sharpens the high-resolution PAN imagery with lower resolution multi-spectral data. The technique extracts edge information of highresolution data, thereby adding it pixel by pixel to low-resolution data (Rokhmatuloh et al. 2003), and exhibits minimal distortions in the spectral characteristics of data compared to other image fusion techniques (Chavez et al. 1991). At the onset of the process, the LISS III data were geometrically rectified with reference to Survey of India toposheet, and the PAN data were rectified using LISS III by image-to-image rectification process. A detailed vegetation (forest and non-forest) map with various land use/land cover features along with predominant forest types was prepared using LISS III data by onscreen visual interpretation technique for ANI (Anonymous 2003), NAI (Prasad et al. 2007b) and BI (Nagabhatla and Roy 2007). The forest
Environ Monit Assess Table 2 Levels of density classification in Andaman and Nicobar Islands Level
% Canopy cover
Category
Density class
D5 D4 D3 D2 D1
> 80 60–80 40–60 20–40 < 20
Very high High Medium Low Very low/Degraded
Very dense, undisturbed intact patches Forest extracted once, now well protected and managed Average forest subjected to disturbance Disturbed and extracted Degraded (natural or anthropogenic processes)
community layer was extracted from the above thematic output and was used as base map to derive density zones for each of the forest types. The density typology described in Table 2 shows the zones based on percentage of crown cover, each at an interval of 20%. The primary forest types were classified into five density levels by visually delineating the PAN data based on spectral reflectance and characteristic of canopy cover. The delineated canopy density layer was refined using the merged IRS LISS-PAN data. For delineating the density zones, the map was enhanced to 1:30,000, compared to 1:50,000 used in the forest cover mapping, as the close cover of the crown caused intermixing and made it difficult to extract the accurate information. Field inventory Remote-sensing-based forest-type strata in conjunction with topography and climate provided a spatial framework for ground sampling. Ground control points were distributed based on proportion of the forest classes and were used to collect, verify and validate the overall approach. For NAI, phytosociological data (60 plots of 0.1 ha) representing different density classes were collected and a field inventory was prepared. About 15 plots (32 × 32 m) were laid out in each density class (D1–D4), whilst D5 class was not sampled due to difficulty in accessing the area. Within each plot, measurements of girth at breast height were made for all trees having a girth of more than 30 cm at breast height along with its height. Data were analysed for stem density, basal area, species richness (the number of species recorded in sampled area), diversity using Shannon diversity index (Shannon and Weiner 1963) and important value index (IVI)—a sum of relative density and relative basal area (Blair and Brunett 1976).
Results and discussion Vegetation-type-wise density map The ground information, topographic maps and the digital elevation map supported the interpretation of the earth observation data for density stratification. The density zones also describe the status of forest resources in the islands. The classified output (vegetation map) derived from the LISS III data show five dominant types, viz. evergreen, semi evergreen, moist deciduous, littoral and mangrove forest, along with other classes like degraded forest agriculture, barren, mudflats, etc. (details of the classification can be refereed from Anonymous 2003; Prasad et al. 2007b; Nagabhatla and Roy 2007). The PAN data being monochromatic add information for density zoning taking into account the variation in tone (grey. . . black) and texture (smooth. . . speckle) formed from the sparse and clumped nature of stems on the real ground. Andaman and Nicobar Islands The density stratification reflected that nearly 37% (1,738 km2 ) of the forest area in Andaman Islands falls under high density (Table 3; Fig. 1). Semi-evergreen, moist deciduous and littoral forests occur mostly as medium density zones. They are subjected to high anthropogenic and recreational pressure, and this high human influence is attributed to their peripheral placement in the landscape. Very low-density areas have been classified under degraded forests. The terrain and accessibility conditions in the Nicobar Islands have kept the area comparatively free from disturbance, and therefore, high density (86%, 1,219 km2 ) is maintained in general. These islands have dense and intact forests on
Environ Monit Assess Table 3 Area statistics showing the forest canopy density in major vegetation communities of Andaman and Nicobar Islands (derived from Anonymous 2003) Forest classes Andaman Tropical evergreen Tropical semi evergreen Tropical moist deciduous Mangroves Littoral forest Nicobar Tropical evergreen Mixed evergreen Moist deciduous Lowland swamp Mangroves Littoral forest
% Area covered under each forest class Very high High
Medium
Low
8.96 14.56 10.69 58.41 0.27
36.23 34.95 40.25 37.23 31.70
29.42 38.46 37.46 2.80 46.08
5.39 12.03 11.61 1.57 21.95
– – – – 100 –
94.65 38.80 8.68 89.38 – 32.13
4.16 49.77 73.87 7.94 – 52.01
1.18 11.42 17.45 2.68 – 15.86
< 20% (very low) density class was mapped as degraded forest (Refer Anonymous 2003)
the western coast, as settlements and communication are restricted to the eastern side. The salt-tolerant mangrove ecosystem found mainly in tropical and subtropical intertidal regions which show very high density reflect the low or almost no disturbance to them. NAI versus BI Andaman and Nicobar Islands The canopy density map shown in Fig. 2 exhibits major portion of NAI (506 km2 ) under dense class (with 60–80% canopy cover), thus suggesting high intact condition of the forest. The very low-density zone covers an area of 75.3 km2 , indicating low forest fragmentation rate. These areas were also regarded as degraded forests and mostly occur near the habitations or settlement areas (like Diglipur, Ramnagar, Radhanagar, Kalipur, Durga nagar, etc.) or along the road sides. The accessibility to these areas allows for higher human lopping practices (Table 4). The D4 density class (represented by major portion of the study area) occurs primarily in mangrove (79.7%), followed by littoral (43.9%) and evergreen (37.1%). Both the mangrove and evergreen forest types display high proportion of dense area with undisturbed conditions existing at various spatial levels, i.e. evergreen on high altitude areas and mangroves near inaccessible coastal areas. The low density of canopy cover observed in moist deciduous is due to topographic
gradients and easy approachability that allows frequent anthropogenic disturbances. An overall observation of the density map obtained depicts the domination of D3 (40–60%) and D4 (60–80%) density zones in the NAI forests. Baratang Islands The density pattern of the forest formations in BI show that the vegetation communities here have been under serious anthropogenic influence. Combining the major vegetation types with the density classification shows that most of the region has canopy cover of 60– 80% (Table 4). The factors viz. terrain, accessibility and non-proximity to a disturbance source have helped to preserve higher densities in remote areas and uninhabited islands like John Lawrence, Henry Lawrence, Outram and North Passage. It is observed that about 60% of evergreen forests and majority of mangrove forests (77.16%) are under high- to very high-density level. The semievergreen and deciduous forests are categorised as medium to moderately dense. The overall picture of canopy closure in the region reflects that Havelock and Neil Islands have been exposed to large-scale extraction and increase in secondary formations. The Ritchie’s group of islands shows more or less homogenous vegetation with less human interference. During the past years, the forests have been exposed to heavy extraction activities resulting in low density in Baratang and Havelock Islands. Of
Environ Monit Assess Fig. 1 Density stratification for major vegetation communities in Andaman and Nicobar Islands, also highlighting the areas for in-depth analysis in Tier2 (Anonymous 2003)
the total, about 32% area is inhabited with open to medium dense forest. It is observed that low to medium dense forests are concentrated around habitations and along the road. The general trend shows that the inhabited islands depict an average density between 40% and 60%, whereas the uninhabited islands have comparatively larger area under high-density forest cover (see Fig. 2). The islands covered under the Rani Jhansi Marine
National Park (John Lawrence, Henry Lawrence and Outram) show comparatively high canopy density due to the protected status of the specified islands. The NAI zone falls in the average forest density between 40% and 80% slotted into two categories of D3 and D4, whilst the BI falls more between 40% and 60% in the main Baratang stretch. The other islands in the group, of which only few are
Environ Monit Assess Landfall Island Outram Island
Henry Lawrence Island Paget Island Baratang Island
Peel Island John Lawrence Island
Havelock Island
Sound Island
Neil Island
Interview Island
North Andaman Islands
Baratang Islands
Fig. 2 Forest canopy closure analysis for North Andaman and Baratang Islands (Anonymous 2003)
inhabited, sum up the average density cover of D4 zone. The reason being that most of these are designated under reserve forest, tribal reserve and/or wildlife sanctuary, although the inaccessibility and its isolation adds to the cause. The made-up pattern and the placement of the islands in the archipelago also explains the variation, as NAI is a continuous land mass connected to the other islands of the Andaman group by Andaman Trunk Road (ATR), thus facilitating people to settle around in contrast to the less chances of settlement penetration in the internal dense forest regions. Therefore, the human disturbances mainly occur along the wayside. Whilst in the case of BI, the small scattered islands (such as Neil, Havelock, etc.) provides a probability to settle along the accessible coastlines in addition to the settlements along the ATR in the main stretch of Baratang.
Density stratification vis-a-vis phytosociological analysis: the case of NAI Satellite-derived density classes were sampled for phytosociological data and an attempt was made to analyse and correlate the phytodiversity with the density classes as illustrated in Table 5.
Species richness Species richness was high in D1, decreased in D2 and again increased in D3 and D4 classes. The high species richness observed in D1 is because it is an area with highly disturbed community and large canopy gaps. This is similar to earlier observations that in natural conditions, disturbances either caused naturally or anthropogenically play an ideal role in enhancing the species richness
a For Baratang Island, the very low density class was characterised as degraded forest, and the total area of this class amounted to 2.10 km2 distributed in all forest classes
45.02 40.06 26.99 1.11 15.85 129.03 83.25 79.05 17.82 2.07 40.15 222.34 2.13 1.06 0.13 0 33.64
56.64 65.89 69.56 71.32 43.52
37.76 28.86 25.71 23.51 21.58
14.23 9.35 14.23 0 1.94 39.75
3.47 4.19 4.6 5.17 2.26
– – – – – 2.10
3.74 5.50 8.30 0.00 8.53 11.6 24.8 18.7 0.00 20.1 75.3 12.81 9.88 12.93 10.84 4.91 39.7 44.5 29.2 5.2 11.6 130.2 33.66 50.65 46.35 27.88 5.84 104.3 228.4 104.6 13.4 13.8 464.6 37.11 27.56 25.53 43.90 79.72 115.0 124.3 57.6 21.2 187.9 506.0 12.68 6.41 6.88 17.38 1.00
North Andaman Islands Evergreen 39.3 Semi-evergreen 28.9 Moist deciduous 15.5 Littoral 8.4 Mangroves 2.3 Total density 94.5 Baratang Islands Evergreen 19.91 Semi-evergreen 43.64 Moist deciduous 0.82 Littoral 0.00 Mangroves 43.28 Total density 107.65
Medium density (40–60%) Area (km2 ) Area (%) High density (60–80%) Area (km2 ) Area (%) Very high density (>80%) Area (km2 ) Area (%) Density classes Forest types
Table 4 Aerial statistics of the density classification in North Andaman and Baratang Islands
Low density (20–40%) Area (km2 ) Area (%)
Very low densitya ( 20 m) Canarium euphyllum Bombax insigne Dipterocarpus gracilis Pterocarpus dalbergoides Pterocarpus dalbergoides Amoora wallichi Artocarpus chaplasha Terminalia procera Pisonia excelsa Mitragyna rotundifolia Predominant species (IVI) Pterocarpus dalbergoides Pterocarpus dalbergoides Dipterocarpus gracilis Diospyros oocarpa Celtis wightii Celtis wightii Artocarpus chaplasha Pterygota alata Aglaia andamanica Aglaia oligophylla
D1
Table 6 Tabularised description of top canopy and predominant species in varied density zones of North Andaman Islands
D4
Environ Monit Assess
agriculture lands), whilst the intact high-density zone is concentrated on hilltops and inaccessible areas. In addition, the analysis for NAI reflects that the assessments of phytodiversity in different density classes do not show notable correlation between phytodiversity and density classes. On the contrary, the analysis suggest that forest-typewise density analysis is productive in identifying the general pattern of phytodiversity in different density classes, as species richness and diversity changes with the type of forest. In the present analysis, the variation in the phytodiversity parameters among density classes is the grouping of various forest types within each density class; as a result, the relation between phytodiversity and forest canopy density was not very evident. It is suggested that further investigation should focus on forest-type-wise density–phytodiversity analysis whilst documenting the entire forest irrespective of types. The present analysis supports the study by Dutt et al. (1994) for the island that also promulgated that the closed forests (>40% canopy density) can be considered as ‘conservation zones’ whilst the open forests (
