Dragonflies of Nee Soon swamp forest 
137 
Fig. 8. Principal components analysis (PCA) results showing the loading of the 23 environmental 
variables to the first two PC axes. Square (blue): sites in group 1; Triangle (purple): sites in 
group 2, round (black): sites in group 3. 
(FE), riparian canopy cover (RCA), distance to open area (OA), and stream water with 
high value of ORP, but negatively correlated to water depth, water with high pH and 
stream with steep bank shape. PC 2 explained 17% of the total environmental variation, 
and was positively correlated with water DO, stream velocity and stream order, but 
negatively correlated with water temperature (Tern) and high levels of leaf litter (LL) 
in the substrate. PC 3 explained 11% of the total environmental variation, and was 
positively correlated with existence of pools in water channel and stream substrate 
with high amount of silt, but negatively correlated with stream substrate with high 
amount of sand and stream bank with overhanging tree root. PC 4 explained 9% of the 
total environmental variation, and was positively correlated with riparian vegetation 
heterogeneity, in-stream macrophytes, woody debris and negatively correlated with 
silt substrate. 
These four axes were input as independent variables into an Ordinary Least 
Square (OLS) model to test the significance of the link between habitat characteristics 
and diversity of the odonate community, represented by Shannon-Weaver Index (H’) 
and Species Richness (R). Spatial autocorrelation and homoscedasticity of the residuals 
were investigated with Durbin-Watson test and Breusch-Pagan test respectively. 
OLS results (Table 6, Fig. 10-12) show that only PC 1 and 3 are significantly 
correlated with overall odonate diversity in Nee Soon freshwater swamp forest. PC 
1 was negatively correlated with both Species Diversity Index and with Species 
Richness, while PC3 was positively correlated with only Species Diversity index H’. Its 
correlation with Species Richness was rejected due to high spatial autocorrelation. In 
summary, distance from forest edge, canopy cover of the riparian vegetation, distance 
from nearby open area, shallow water depth, water with high ORP reading, water with 
