Stream flow, ground water and climate change 
179 
Fig. 2. (a) Reference ET over Nee Soon freshwater swamp forest catchment in Dec 2012 
(Source: MOD 16); (b) Leaf Area Index across Nee Soon freshwater swamp forest in Dec 2012 
(Source: GLASS-MODIS). 
broadleaf canopies (Watson, 1947). LAI can be determined directly through sample 
measuring and indirectly such as hemispherical photography. This study acquires the 
LAI information from the Global and Surface Satellite (GLASS)-MODIS LAI dataset, 
a global LAI product released by the Center for Global Change Data Processing 
and Analysis (CGCDPA) of Beijing Normal University (Liang & Xiao, 2012). The 
GLASS-MODIS LAI dataset is retrieved using the general regression neural networks 
(GRNNs) trained with the MODIS and CYCLOPES LAI products as well as the 
reprocessed MODIS reflectance products (Xiao et al., 2013). Samples of the GLASS- 
MODIS LAI over the Nee Soon catchment are plotted in Fig. 2 (b). Typical LAI values 
range from 0 to 7, implying areas from no vegetation to dense canopy coverage. 
Simulation results 
Fig. 3 compares the simulated with the observed water depths at Upper and Mid stream 
gauges, whereas Fig. 4 shows the comparison between the simulated and the observed 
groundwater tables at stations DP4 and DP9. The numerical model simulates the water 
depth reasonably well, with root mean square error (RMSE) respectively being 0.11 
m and 0.17 m for these two stations. Both DP4 and DP9 are located upstream, where 
the water table varies within 1 m below the ground surface. The numerical simulation 
successfully captures the rising and falling trends within the series of observations, 
producing insignificant model errors (RMSE respectively being 0.08 m and 0.11 m). 
The model errors are mainly caused by the uncertainty in the reference level arising 
from the smoothing effect of the 20 m grid. 
