GISdevelopment.net ---> AARS ---> ACRS 1995 ---> Forest/Vegetation Mapping

Evaluation of Asian Elephant Habitat

Li Zhixi Li Hongga Lu Feng
Institute of Remote sensing & GIS, Southwest Forestry College
Kunming , 650224, P.R.China
E-mail: HUSSIN@ITC.NL, DEGIER@ITC.NL


Abstract
In this paper, in view of complicated topography and ecological environment in Xishuangbanna, we offer an approach to quantitatively analyse the spatial features of Asian elephant habitat, in which forest sampling, interpretation of TM data and aerophotos, as well as GIS techniques are mutual complementary. The results demonstrate the potential for integrating RS , GIS and sampling into accurate spatial habitat assessment.

Introduction
The United Nations Conference on Environment and Development (UNCED) which took place Rio de Janeiro, Brazil, 3-14 June 1992, represented the threshold of a fundamental change in the attitudes towards environment and sustainable development, signed the 21 Century Agenda and Biodiversity Convention, Which put forward that the protection of biodiversity is important and imperative to safeguard the basic heritage of mankind, the quality of the earth environment, for future generations. Asian elephant (Elephas maximus Linnaeus) belongs to large proboscidean herbivorous mammal, Which survives only In Asia. Due to stern reliance on suitable environment, its distributions becomes more and more narrow. According to the historical records, Asian elephants scattered over China. In Shang Dynasty (1000 B.C.), elephants even reached the range of the Yellow River, but in Epoch of Division Between North and South (500 A.D.) shrank back to the range of the Yangtze River. Closely following destruction of forest, so far, elephants mainly survive in Xishuangabanna primitive tropic forest, and become rarer. Remote Sensing and Geographic Information System, as part of the information technology, take techniques into evaluating wild animal habitat such as using Landsat Data to interpret aquatic animal habitat in the eastern South Dakota (R.G. Best 1978)[1], using Landsat digital data to analyse Cygnets habitat along the coast of British Columbia (Y.Jim,Leo,1987) using GIS and digital elevation model to analyse the Chequered skipper and Curlew habit patches in Scotland (R. Asp in all, 1992)[2], using GIS to manage natural reserves in Thailand (Youngynt Trisurat, 1992)[3] and using Landsat TM to Identify the Amblyomma variegatum habitats in Guadeloop (M. hugh Jones et al. 1992)[4] etc. In addition, we applied Landsat to evaluate the Giant Panda habitat[5]

2. Study Area
The Xishuangbanna Dai Autonomous Prefecture is situated in the south front of yunnan Province ,a transition from Himalayas to Malaysia, N 210 08 - 22o 35, E 99o 53 - 101o 50. The special situation of low center surrounding high mountains shapes a largescale tropic climate. Because of the distinctive position, topography as well as climate, tropic rain forest with massive flora and fauna scatter on the lowland. In Xishuangbanna, from 1965 to 1988, forest coverage decreases from 46. 46% to 33. 72%, with an average of 0.55% annually [6]. With the lessening of living spaces, shortage of foods, degrading of habitat qualities as well as the illicit hunting for elephants, elephants are confronted with the danger of extinction. So it is imperative to monitor and asses the habitat, furthermore, to put forward the measures to protect elephants. In this study, the study area lies in Mengyang Nature Reserve, the largest one of the five reserves in Xishuang - banna, between the XiaoHei stream and the Mekong River. Its total area is 118, 41ha and the elevation varies from 600 meters to 1600 meters. The forest vegetation types mainly comprise seasonal rain forest, mountain fain forest monsoon forest and evergreen broad - leaf forest, etc.

3.Method
In this study , we took ecological and biological factors as the guide, aerophotos and satellate images as the media, GIS as an analysis tool to establish a GIS of assessing and monitoring Asian Elephant habitat. The framework includes the habitat analysis, interpretation and assessment.

3.1 The habit analysis

3.1 I Field sampling survey

According to ecological and biological features, we divide the habit factors into aliment, water, shelter and living space. In view of the interpretation, we select the following factors to survey: forest vegetation type, crown degree, distance, we add human activity intensity. The field work mainly surveys the number of elephant trails under different factors. In order to ensure the typicalness of samples, we take the systematic sampling. Samples are arranged with 100 meter in length, 2 meter in breadth. 71 samples are set up in order that there are no less than 5 samples in each factor. In field survey, it is necessary to record elephant trails and the relevant ecological, biological and human activity, intensity, besides marking the site.

3.1.2 Quantitative analysis
The number of elephants trails under various situation of eight factors is referred to table 1. The table 1 is only single-factor analysis. In view of interaction among factors, we take multi-factor composite analysis, i.e., multivariate regression to quantitatively analyse habitat factors. The formula is following:

Y: b0 +b1X1+b2X2 +b3X3 +b4X4 +b5X5+b6X6 +b7X7+b8X8

where, Y is the frequency of elephants activity (unit is trails per ha ). b0 is regression constant.b1, b2, b3, b4, b5,b6,b7, b8 are regression coefficients.

X1,X2,X3,X4,X5,X6, X7,X8, represent correlative factors (see table 1) ,respectively.

Some factors in table 1 are qualitative, which cannot be directly inputed to the equation, in addition , the relation between different conditions of each factor and elephant trails is often nonlinear. In order to ensure the regression precise, we transform data from qualitative to quantitative.

Table 1.
Factor (xi) Conditions (xij) Trails(yiha) relation
Vegetation(x1) Ever green broad-leaf (x1-1)
Bamboo forest (x1-2)
Shrub (x1-3)
Dry farmland (x1-4)
1120
1380
300
230
2
1
3
4
Crown density (x2) Median 0.2 - 0.5 D (x2-2)
Median 0.2 - 0.5 (x2-2)
Dense >0.5 (x2-3)
620
620
990
2
2
1
Distance from water (x3) Near <100m (x3-1)
Median100m - 500m (x3-2)
Far >500m (x3-3)
1270
780
440
1
2
3
Slope (x4) <10°- (x4-1)
10 - 20°- (x4-2)
21 - 30°- (x4-3)
31 - 40°- (x4-4)
>41°- (x4-5)
440
650
1550
950
760
5
4
1
2
3
Aspect (x5) North - east (x5-1)
West to east (x5-2)
South to west (x5-3)
1350
760
270
1
2
3
Location (x6) High (6-1)
Median (x6-2)
Low (x6-3)
640
1050
830
3
2
1
Elevation (x7) <800m (x7-1)
800 - 1000m (x7-2)
>1000m (x7-3)
660
800
970
3
2
1
Human activity (x8) Force (x8-1)
Median (x8-2)
Weak (x8-3)
220
530
1540
3
2
1

In the transformation, the minimum trails of each factor under different conditions are defined as 1 1.0, and the others are proportional vested , e. g. the vegetation type (see table 2). I

Table 2.
conditions trails /ha numerical value
evergreen broad-leaf forest 1120 4.9
bamboo forest 1380 6.1
shrub 300 1.3
dry farmland 230 1.0

The computed coefficients are as table 3.

The precision is evaluated as following :
Lyy (the total square error of a sampling)=64889577
U (the regression sum of squares)=35931762
Q(the residual sum of squares )=28957815
F(8, 62)(the test of significance of regression equation) =9. 62
FO. 01(8,62)=2.80 F > FO. 01 R=O. 76

Table 3.
factor
vegetation
crown
water
slope
aspect
location
elevation
human activity
coefficient bj
156.40
-196.90
101.44
142.80
157.95
12. 17
115.29
132.99
constant bo
- 1705.03












The test of significance of each regression coefficient is in table 4.
Table 4.
factor F priority
vegetation 4.973 4
crown density 1.504 5
water 5.731 3
slope 0.800 6
aspect 10.519 2
location 0.009 8
elevation 0.364 7
human activity 12.819 1

the study, we get the precision of 82% under the reliability of 95%.

3. 2 Interpretation of the habitat
Among factors, topographic factors, i.e. slope, aspect, location, elevation and distance from water, can be directly obtained from the topographic map. The human activity intensity, e.g. density of population, can be obtained from the statistics of late villages situation. And we gain vegetation type and crown density from interpret action of images. Although vegetation only take the forth place, the first place human activity intensity, the second aspect are directly or indirectly reflected by vegetation. And we regard the vegetation as the foundation of the habitat. In the interpretation, vegetation is considered as the main object.

3.2. 1 Image processing[6]
Data available for analysis are following: the black and white panchromatic aerophotos with scale of 1: 20, 000 taken in March 1989; Landsat TM data respectively recorded on 2-Feb 1988and 28-May 1992; Spot XS data recorded on 16 Feb 1988, etc. Firstly, we carry out the image processing e. g. contrast enhancement, color synthesis. The correlation coefficient matrix of TM bands in Mengyang , illuminates that minor correlation are respectively between TM2 and TM4'TM3 and TM4, TM4 and TM7. Only considering correlation, TM346 is the best, and TM234 takes second place. But, when we take further steps to consider functions of each band, the synthesis of TM234 is prior to others, for its susceptibility to vegetation type and partly eliminating hill shadows. Secondly, we use maximum likelihood for image classification. By virtue of actual situation in Mengyang nature reserve, we set up training units of 11 types, including forest, bamboo ,tea plantation, farmland, rubber woods, meadow, shrub, water and the various tansitions , e. g. from forest to farmland, from farmland to forest, from meadow to shrub. The results possess 79% precision with the reliability of 95 %. When 11 types are composed to 6 types, the precision increases to 86 %.

3.2.2 Interpretation
The objection of interpretation training is to establish marks and to set up units for computer automatic classification. In addition, the ecological law of vegetation and the vertical vegetation distribution which are formed by difference of heat and precipitation, are also applied in the interpretation. By cartological choise, boundary lines of status quo are demarcated and transformed to topographical maps one by one to form the base maps .

3. 3 Habitat Assessment

3. 3. 1 Software and Hardware

The habitat is assessed with ARC/INFO version 7.03 on SUN SPARC LX workstation, Calcomp 9500 digitizer and Calcomp 3036 plotter.

3. 3. 2 The Database Design
Based on factors analysis and interpretation, topography, hydrology, forest vegetation, density of population, road, village and administative boundary are used as layers to establish GIS.The organization of layers sees table 5. In the design, we use the special dictionary[7]to establish Mengyang nature reserve GIS.

Table 5.
layers geographic data attribution data
topography line contour
hydrology line river type, name
vegetation polygon vegetation type
road line road, path, trail
village administative point name, population ,income etc.
bounds line county, village, reserve etc.

3. 3. 3 The Database Establishment
The data are generally divided into two parts, geographic data and associated attributes. To establish geographic database, we capture geographic features with Calcomp 9500 digitizer (1000lpi) .Well-distributed 17 control points in Mengyang reserve are selected to ensure the accuracy. Meanwhile, the input error of points is under 0.2mm on maps (scale 1: 50,000). The graphics are transformed and compiled to for the topology. Based on geographic database , the associated attributes such as vegetation, elevation, hydrology, road, landuse, annotion, text, etc. , are defined and inputed by users, besides the automatically produced items such as area, length , perimeter. Once the input of attributes is completed , we can identify, operate and analyse the relation between geographic data and attribution and the relation among layers .

3. 3. 4 Analysis
According to units divided for management, and using 8 factors (see table 1) , through regression equation, we estimate trails. We take the second unit as a example, vegetation(x1) is bamboo , vested to 6.1 (see table 2); according to this reason, crown density(X2) is 1.9;distance from water (X3) is 400 meters, 5. 4; slope(X4) between 20.-30. 2.6; aspect (X5) is west, 2.9 ; location (X6) is median, 4.9; elevation(X7) is less than 800 meters, 3.0; human activity intensity(X8) is weak, 7. 0. According to table 3, the result is following: .

Y = -1705.03 + 156.40*6.1-196.90*1.9 + 101.44*5.4 + 142.80*2.6 + 157.95*2.9 + 12.17*4.9 + 115.29*3.0+ 132.99*7.0 = 1766 trails/ha.

In the light of trails per ha and management levels of status quo, we classify Asian elephant habitat (see table 6) .

Table 6.
trails / ha class
< 600 bad
601 -1200 moderate
1201 -1800 good
> 1801 excellent

At above example, the unit belongs to good class. According to the method, by evaluating every Unit, we obtain digital theme map of Asian Elephant habitat in Mengyang.

4. Results and Discussions

4. 1 Results

4. 1. 1 Firstly, we establish an assessment system of Asian elephant habitat for protecting elephants in Mengyang. Meanwhile, we Ouput the forest vegetation map ( see map 1) and Asian Elephant assessment map ( see map 2) , obtain present spatial information of elephant habitat , provide the foundament of protection and management work .

4. 1.2 Through analysis, we get areas and rates of different classes (see table 7). The main features of four habitat classes are as follows :

Table 7.
class frequency area (ha) rate %
excellent 101 2930 3.5
good 171 31674 28.0
moderate 225 44097 39.0
bad 181 33302 29.5

The excellent weak human activity impact ,elevation between 1000 meters and 1300 meters ,slope from 20°¾ to 30°¾ aspect east-north, lower location e.g. valleys or at the foot of hills, range within 100 meters from water ,bamboo forest with no more crown rate. These are best , and are generally located in core parts of reserve.

The good: little impact of human , aspect north or east, west, distance from water within 500 meters, ever green broad-leaf forest or bamboo forest.

The moderate: the normal impact of human, aspect east, west, distance from water within from 500 to 700 meters, shrub or broad-leaf ever green forest.

The bad: force impact of human, aspect west-south, distance from water more than 700 meter , shrub, meadow, farmland or bare land. These are bad for elephant survival, and generally out of the core.

4. 1.3 According to habitat features, in view of actual situation, correlative measures are put for- ward for protecting Asian elephant habitat. In excellent regions , it is urgent to strictly control and reduce villagers activity, emigrate the inner villages, prohibit the activities such as hunting , cutting, plucking bamboo shoots and destroying wild banana , etc. The inner water i. e. pools , streams are periodly casted the salt and iodine to board elephants. In moderate, bad regions, it is important to inhibit farm cultivation of slashing and burning, advocate agroforestry, set up eco- logical villages , reforest vegetation cover, enhance soil fertilization and water soil retention, in- crease villagers income and etc. Above all, it is imperative to reinforce education and promulgation of the law, punish illegal hunters for wild animals. Meanwhile, the electric fences and other defensive fences are set up to prevent elephants from destroying crops.

4.2 Discussions
4. 2. 1 In this study, considering the reliability of factors, we mainly lay on physical factors. The human factor is limited in the density of population, farm cultivation and roads. On the other hand, social factor has intricate and important influence on habitat, which includes many complicated aspects such as the masses interference and protection etc. For example, illegal hunters killed sixte~,'1 elephants, injured three, from May to June 1994. On the contrary, villagers pre- serve wild animals .Peasant in Mengla nature reserve protected an abondoned elephant infant , and sent it to reserve station. In addition, national factor and religion factor have influence on elephants.

4.2.2 In this study, Forest Sampling, Remote Sensing and Geographic Information System are efficiently integrated. According to Remote Sensing and Permanent Plot Techniques for World Forest Monitoring Conference (IUFO.RO , Thailand, 1992)(8] and International Guidelines for Forest Monitoring[9](IUFRO,1994),the integration of FS, RS and GIS is the developing tendency for natural resources monitoring. So , the paper may be regarded as a test of integration of FS , RS and GIS.

Reference
  • R. G. Best. 1983,Remote Sensing Approaches for Wildlife Management. Proceeding of the RNRF Symposium on the Application of Remote Sensing to Resource Management, Seattle , Washington ,P55-96.
  • Richard Aspinall ,1992. Spatial Pattern Analysis Tools in GIS for Environmental Planning. Proceedings of the IUFRO Centennial Meeting ,Berlin, P 223 -233.
  • Youngyut Trisurat ,1992. Application of GIS in Protected Area Management. Remote Sensing and Permanent Plot Techniques for World Forest Monitoring. Proceeding of the IUFRO S4. 02. 05 Pattaya , Thailand. P 215- 225.
  • M. Hugh-Jones ,N Barre, 1992. Landsat-TM Identification of Amblyomma Variegatum (Acari: Ixodidae) Habitats in Guadeloup. Remote Sensing of Environment. Vol,40 No, 1 P43-54
  • Li Zhixi 1990, Investigation for the Giant Panda habitat Using Remote Sensing, Remote. Sensing of Environment, PRC. Vo15, No 2,P 94-101.
  • Li Zhixi et al. 1993, The tropical Forest Vegetation Mapping Using Multisensor Information. Remote Sensing of Environment, PRC. Vol 8, No3 , P 180 -189.
  • ESRI ,SWFC ,1994. Database Design. Xjshuangbanna Agroforestry development Project.
  • R. H. Cydelund Songkram. Risto Paivinen Thammmincha, 1992. Remote Sensing and Permanent Plot Techniques for World Forest Monitoring. Proceeding of the IUFRO S4. 02. 05 ,Pattaya , Thailand.
  • Risto Paivinen, 1994. International Guidelines for Forest Monitoring. IUFRO world series Vol ,5 P 6-34.