htly lower than 
> level of kappa 
that DOMAIN, 
ar research and 
d to extend the 
watersheds and 
ablished by the 
d. The result 
patterns of JETs 
By comparing 
watersheds, the 
in of these two 
nsequently, the 
bles performed 
veen these two 
es of predictive 
es of the two 
The results 
1 from SPOT-5 
' for a widely 
ons of spectral 
OT-5 imagery. 
tral information 
ata and LIDAR 
lopment so that 
[AIN. 
     
  
  
  
  
  
  
  
  
  
  
  
Study Area “Tong-Feng” Samples *Kuan-Dau" Samples 
Statistics — Elevation Slope Aspect Elevation Slope Aspect Elevation Slope Aspect 
m 06 06 IP Vl E fy eB Tye oy 0 py ei TRV 
Mean 1314 34 — 5 24 1712 22 = 7 24 1259 27 — 7 30 
Mode 1239 37 127 6 22 1227 22 247 7-22 1076 22 7 7 21 
Max. 2418 89 361 8 119 2027 46 358 8 51 1559 38 355 8 51 
Min. 454 0 0 + 0 1122 2 2 2.20 1076 16 7 4 21 
VI: Vegetation Index. ~~ TP: Terrain position. 
Table 1. The statics of environmental factors with study area and the samples of each watershed. 
SD1 SD2 SD3 
Method and Dataset 
OA (96) kappa OA (96) kappa OA (%) kappa 
Training 81 0.44 72 0.22 76 0.44 
DA Test 88 0.49 73 0.23 79 0.49 
Average 85 0.47 73 0.23 78 0.47 
Training 96 0.79 97 0.81 92 0.76 
DT Test 93 0.64 96 0.79 91 0.69 
Average 95 0.72 97 0.80 92 0.73 
Training 90 0.62 94 0.68 90 0.70 
MAXENT Test 91 0.63 93 0.66 89 0.66 
Average 91 0.63 94 0.67 90 0.68 
Training 84 0.49 88 0.56 79 0.50 
ML Test 86 0.52 88 0.51 82 0.53 
Average 85 0.51 88 0.54 81 0.52 
Training 97 0.86 98 0.91 95 0.85 
DOMAIN Test 96 0.79 97 0.82 94 0.80 
Average 97 0.83 98 0.87 95 0.83 
  
OA: Overall Accuracy. SDI: “Tong-Feng” samples. SD2: “Kuan-Dau” samples. — SD3: *Merged samples of two watersheds". 
Table 2. The model evaluation result of each method in the three SDs. 
  
SD1: Using KDS as IS 
SD2: Using TFS as IS 
SD-3: Using TFS as IS SD3: Using KDS as IS 
  
  
Method 
OA (96) kappa OA (%) kappa OA (%) kappa OA (90) kappa 
DA 75 0.32 77 0.51 80 0.43 78 0.28 
DT 77 0.13 70 -0.00 92 0.63 93 0.64 
MAXENT 74 0.08 74 0.22 91 0.61 91 0.53 
ML 73 0.21 67 0.07 81 0.43 81 0.37 
DOMAIN 78 0.18 72 0.10 98 0.91 97 0.82 
  
  
OA: Overall Accuracy. SDI: “Tong-Feng” model. SD2: “Kuan-Dau” model. SD3: “Merged samples of two watersheds” model. 
TFS: Test dataset of Tong-Feng Watershed. KDS: Test dataset in Kuan-Dau watershed. IS: Independent Samples. 
Table 3. The evaluation results of extrapolation ability. 
6. REFERENCE 
Atkinson, P. M. and P. Lewis, 2000. Geostatistical 
classification for remote sensing: an introduction. Computers 
& Geosciences, 26, pp.361-371. 
Bourg, N. A., W. J. Mcshea and D. E. Gill, 2005. Putting a 
CART before search: successful habitat prediction for a rare 
forest herb. Ecology, 86 (10), pp. 2793-2804. 
Carpenter, G, A. N. Gillison and J. Winter, 1993. Domain: a 
flexible modelling procedure for mapping potential distribution 
of plants and animal. Biodiversity and Conservation, 2, pp. 
667-680. 
Congalton, R. G, 1991. A review of assessing the accuracy of 
classification of remotely sensed data. Remote Sensing of 
Environment, 37, 99, pp. 35-46 
De'ath, G. and K. E. Fabricius, 2000. Classification and 
regression trees: a powerful yet simple technique for ecological 
data analysis. Ecology, 81 (11), pp. 3178-3192. 
Felicísimo. A. M. and A. Gómez-Mufioz, 2004. GIS and 
predictive modelling: a comparison of methods applied to 
forestal management and decision-making, Proceedings of GIS 
Research. University of East Anglia, UK, pp. 143-144. 
Guisan, A. and Zimmermann, N. E., 2000. Predictive habitat 
distribution models in ecology. Ecological Modeling, 135, pp.