2.5. Model accuracy check
This paper test the accuracy of soil PH, SOM, AN, AP, AK, Fe, Cr, Cd,
Zn, Cu and Pb by the constructed panel data model. The results were
shown in Table 2.
As can be seen from Table 2, each soil attribute in the correction set
has a high coefficient R̅ 2, greater than 0.95.
The highest is alkaline hydrolyzed nitrogen (0.998), and the
corresponding root mean square error is 0.67. The lowest value was PH
(0.95), and the corresponding root mean square error was 0.05, both of
which were relatively low, indicating that the inversion model
constructed by the panel data model could simultaneously realize the
inversion of 11 soil properties, have good modeling accuracy.
According to the prediction results of the validation set,R̅v 2 of PH, organic matter,
alkali-hydrolyzed nitrogen, Cd and Cu were lower than the modeling set,
and the rest were increased. In addition to Pb, the root mean square
error was lower than that in the modeling set,R̅v 2 of other soil attributes
were increased. The relative errors are all greater than 2.5, indicating
that the model has good quantitative prediction ability of PH, SOM, AN,
AP, AK, Fe, Cr, Cd, Zn, Cu and Pb.
In order to more clearly display the modeling accuracy of Panel data of
Fixed influence variable coefficient model, the soil attribute content
diagram (Figure 4) and scatter diagram (Figure 5) of the measured value
and the inversion value.
The comparison between Figure 4 and Figure 5 shows that, except for a
few samples differences, the measured values and predicted values of
most samples are concentrated near y = x , i.e. 1:1
line. The correlation coefficients r between measured values and
predicted values all pass the significance test at P = 0.01
level. This indicates that the panel data model with SG-CRspectral transformation as independent variable has good predictive
ability, and can be used to invert multiple soil properties
simultaneously.