landscape position contributed largely to the variability of AN, 22.7% and 11.7%,
respectively.
The adjusted R2 obtained from AP was very low (0.134). Landscape position and slope
aspect explained 8.2% and 5.2% variability of AP, respectively. In general, the relatively
low amount of AP variability explained may be due to the fact that different state factors
are responsible for the accumulation of AP and not all of them were taken into account in
this study. Other factors that were not included, such as the degree of erosion, management
practices and biological activity, may be responsible for the part remaining of AP
variability. Since management practices including tillage and fertilization to influence
on biological activity occurred only in the cropland and orchard, P especially for AP was
expected to be more differently distributed. In addition, the content of AP was largely
dependent on the topsoil erosion. As a result, low amount of AP variability was explained.
The regression models of soil nutrients were used to calculate predicted values for each
land use. A comparison between average predicted and measured values of SOM and soil
nutrients is given by Fig. 2. The average predicted values in TP and AN for each land use
fell with in a 95% confidence interval compared with their corresponding measured
values. Only an average in SOM and TN for fallow land, and in AP for intercropping land,
shrub land and woodland exceeded 95% confidence interval of the average measured
value (Fig. 2). Thus, our regression models were reasonable enough to predict SOM and
soil nutrients in similar loess areas.