3.6. Evaluating the indicators for loading assessment
After analyzing the spatial correlations of the soil parameters
with diffuse N loading at two depths, the partial least squares (PLS)
regression method was applied to seek important soil variables
affecting diffuse N loading. Because diffuse N loading originated
mainly in the paddy rice and upland areas, the PLS was focused on
these two types of sub-basins. The interaction analysis proved that
the relationship observed in the surface was much stronger than
the subsurface, so the PLS analysis was only applied in the top layer.
The relationship assessment standard was to the plot from w*c of
one regression dimension versus another. The plot showed that the
X-variables (PLSR weight for the 1st component, percentage)
presented the information of the dominant factor. The Y-variables
(PLSR weight for the 2nd component, percentage) modeled well
the second leading factor. The R2 value for
first two factors of the
top surface in the PLS analysis of paddy rice and upland were 0.514
and 0.460, respectively. The contributions of the eight parameters
to the two latent vectors for the two types of land use are shown in
Fig. 7.