Cheometric Analysis
363
0
200
400
600
800
1000
1200
1400
1600
1800
2000
0
200
400
600
800
1000
1200
1400
1600
1800
2000
DOC Measured
DOC predicted in cross−validation
Estuary
WTP
River
Figure 10.10. PLS prediction of DOC from fluorescence in the Horsens catchment. A single twolatent-variable model incorporating data from all sites produces poor predictions for the estuarine
samples.
Table 10.2. PLS prediction of DOC concentration from fluorescence intensities
in the Horsens catchment, using separate models for the river and wastewater
samples (n = 302) and estuary samples (n = 211)
River and wastewater
Estuary
nLV
RMSE cv
R
2 cv
RMSE cv
R
2 cv
1
120.59
0.80
46.51
0.24
2
104.11
0.85
36.57
0.52
3
98.25
0.87
34.51
0.58
4
93.07
0.88
35.53
0.58
5
89.07
0.89
33.21
0.63
6
90.32
0.89
32.49
0.64
7
90.17
0.89
31.48
0.66
Note: The “best model,” representing a trade-off between minimum number of latent variables (nLV), minimum RMSE cv and maximum R
2 cv is indicated in each case in bold.
363
0
200
400
600
800
1000
1200
1400
1600
1800
2000
0
200
400
600
800
1000
1200
1400
1600
1800
2000
DOC Measured
DOC predicted in cross−validation
Estuary
WTP
River
Figure 10.10. PLS prediction of DOC from fluorescence in the Horsens catchment. A single twolatent-variable model incorporating data from all sites produces poor predictions for the estuarine
samples.
Table 10.2. PLS prediction of DOC concentration from fluorescence intensities
in the Horsens catchment, using separate models for the river and wastewater
samples (n = 302) and estuary samples (n = 211)
River and wastewater
Estuary
nLV
RMSE cv
R
2 cv
RMSE cv
R
2 cv
1
120.59
0.80
46.51
0.24
2
104.11
0.85
36.57
0.52
3
98.25
0.87
34.51
0.58
4
93.07
0.88
35.53
0.58
5
89.07
0.89
33.21
0.63
6
90.32
0.89
32.49
0.64
7
90.17
0.89
31.48
0.66
Note: The “best model,” representing a trade-off between minimum number of latent variables (nLV), minimum RMSE cv and maximum R
2 cv is indicated in each case in bold.