Methodology to Forecast Road Surface Temperature with Principal Components Analysis and Partial Least-Square Regression: Application to an Urban Configuration
Table 3
Results of PLS models with optimum number of thermal fingerprints.
PLS configuration
Date
Measured RST (°C)
RST forecast (°C)
Bias
Standard deviation
RST <10°C (2 factors used out of 6 of the PLS model using 8 samples)
2012-12-11
0.4
0.4
0.0
1.9
2013-10-24
12.1
10.7
−1.5
1.7
RST <10°C (6 factors used out of 6 of the PLS model using 8 samples)
2012-12-11
0.4
0.1
−0.3
1.9
2013-10-24
12.1
11.3
−0.8
1.6
RST <10°C (2 factors used out of 4 of the PLS model using 6 samples)
2012-12-11
0.4
0.4
0.0
2.2
2013-03-22
0.8
0.1
−0.7
2.4
RST <10°C (4 factors used out of 4 of the PLS model using 6 samples)
2012-12-11
0.4
−0.1
−0.5
2.4
2013-03-22
0.8
−0.2
−1.0
2.7
Where bias , with , RSTmeasured, respectively, modelled and measured road surface temperature.