Spatial Distribution Simulation Of Soil Nitrogen At Different Scales In The Central Sichuan Hilly Areas

(1) The soil total nitrogen (TN) average contents rised with altitude decreasing at the micro and meso scales. The soil available nitrogen (AN) average contents showed a similar phenomenon in the meso-scale areas. The soil AN in the micro-scale plots showed this order:the high hilly area (73.35±33.80mg kg-1)<low hilly area (77.56±36.00mg kg-1)<medium hilly area (84.93±31.95mg kg-1). At the macro-scale, the soil TN (0.93+0.28g kg-1) and AN (86.18±33.28mg kg-1) content are similar with the medium hilly area at the meso-scale.(2) The spatial variability of soil nitrogen in high, medium hilly areas mainly dued to the combined effect of random and structural variation at the different scales. It was directed by the C0/(C+C0) of soil TN and AN content in most plots are between25%to75%. The spatial variability mainly through randomness factors works in the low hilly areas (C0/(C+C0)>75%), such as land use methods and ecological development. Obtained by Geostatistics method, the spatial distribution pattern transition threshold changed abruptly, and the fragmentation of the distribution showed an increasing trend as the scale up. However, those characteristics would performance differently for the different types of landforms at the same scale.(3) With the scaleing-up, the types and numbers of environment variables which were related significantly with the soil TN and AN increased in the sample plots, areas and region. At the same time, the adjusted coefficient of model for the soil TN and AN would decrease. The F-test confirmed that the fitting models through the environment inversion method was significant level (p<0.01).At the micaro-scale, the geography factors and vegetation indexes were added into the fitting models of soil TN and AN. Most of those models can explain50.00%of the total variance, and the fitting effect is good. At the meso-scale, the DN value (Pixel gray value) was added into the models in the high and medium hilly areas, and three kinds of the variables were all added into the fitting models for the low hilly area. These models can explain4.7%-12.2%of the total variance. At the macro-scale, the models is mainly added with the geography factors and the pixel gray values. The total variance was be explained4.7%with the TN model, and the total variance was be explained11.9%with the AN model. It is be directed by the F-test, the models at three scales were all significant level(p0.01). (4) Obtained by environmental inversion method, the spatial distribution pattern showed obvious scale-effect. At the micro-scale, the distribution patterns combined with the location and altitude of the hills were better than that in the meso and macro-scales. However, sample plots A7and A8got some abnormal values (NoData).The distributed simulations for three scales have shown that the high values were mostly located in the low-lying areas, such as hill valley, river bend, and the low values appeared in some areas with the relatively high altitude, such as the hilltop, ridges in the region. Compared with the Kriging Method, the Environment Inversion Method without the rigorous smooth can display the extreme value in the local areas.(5) Comparison of the two methods, the simulation effect for soil TN and AN showed differences at the micro, meso and macro scales. Even at the same scale, the relative and absolute error obtained by two methods in the different geographical areas displayed differently.At the micro-scale, the Environmental Inversion Method is superior to the Geostatistics for simulation of TN and AN in the high and medium geomorphic plots, while that for TN in low hilly performance opposite. At the meso-scale, the simulation effect for soil TN and AN in the high hilly region performance of ordinary kriging method is better, and the simulation effect of the two methods performed similar in the medium hilly region. In the low hilly region, the simulation effect of TN with Geostatistics method is superior to the environmental inversion method, while the simulation effect of AN performed on the contrary. At the macro-scale, two methods showed similar simulation effect.