| The soil nitrogen content is an important indicator of soil fertility and soil quality, and is also an important source and sink in the global nitrogen cycle. Study on spatial variability of soil total nitrogen can correctly evaluate supplying capacity of the soil nitrogen, guide regional utilization and management of soil resource, which would play a very important role in the regional agricultural production and environmental management. The spatial heterogeneity of soil nitrogen is affected by the natural factors such as soil parent material, topography, biology, time and climate and artificial factors such as fertilization and tillage. Whether the environmental factors as auxiliary variables to guide regional prediction of soil properties and mapping can improve the prediction accuracy is not conclusive. In large regional scales and limited samples, it is rare to compare the spatial distribution of soil properties by using different statistical methods.This study selected Hainan Island as the study area to build models. 29 samples that were consist of one or two samples randomly selected in every county as validation set from 159 samples data. The left samples were used as prediction set to build models for predicting TN content. The study combined the terrain and remote sensing index and other environmental factors as auxiliary variables. The spatial distribution of soil TN was analyzed respectively by ordinary Kriging(OK), multiple linear regression(MLR), regression Kriging(RK) and co-Kriging(CK). Then the remaining 29 samples data as validation data set were used to perform precision. The spatial variation characteristics of soil TN in the 1980 s and 2012 were compared and its causes were discussed. Then the conclusions were given, as follows:(1) Except that the optimal prediction model of the topsoil(0-5cm) TN was CK, the optimal prediction models of other three layers were OK When using few samples to predict the distribution of soil nitrogen in a large region, prediction accuracy of ordinary Kriging was higher than that of Cokriging. Prediction accuracy of CK was better than or equal to prediction accuracy of RK. The prediction accuracy of RK was always lower than that of OK.Compared with Multiple Linear Regression(MLR) and ordinary kriging(OK), Co-kriging(CK) and regression kriging(RK) could eliminate the smoothing effect of some extreme values.(2) Soil TN content of four layers showed a decreasing trend from east to west direction. High value mainly distributed in the northeast and the south. The soil TN content of top soil(0-5 cm and 0-20 cm) respectively was at level 3 and level 4 medium level as a whole. The soil TN content of 20-40 cm and 40-60 cm soil layer was at the lower six levels, which area ratio of 20-40 cm and 40-60 cm soil layer reached 51.5% and 90.9% respectively. TN content of different soil layers showed a decreasing trend with increasing soil depth.(3) Soil TN content of different soil layers and different environmental factors showed the interdependency of varying degrees. The relationship among total nitrogen content of 0-5cm soil layer and slope, NDVI and other terrain factors do not reach significantly associated, but only TN content of 0-5cm soil layer and land use showed an significant correlation. The soil total nitrogen contents of both 20-40 cm and 40-60 cm layers was significantly or extremely significantly associated with NDVI, slope. The soil total nitrogen contents of 0-20 cm soil layer correlated with NDVI significantly.(4) After 30 years, the overall spatial distribution trend of TN in Hainan Island soil is similar, high value remains in the southern mountainous regions and northeastern regions, but TN content decreased. Taking Qiongzhong County for instance, the study disscussed the main reason of TN spatial and temporal variation. The main reason is artificial unreasonable reclamation of the mountain with high slope. |
PDF Full Text Request