Abstract:

Soil organic matter (SOM) is extremely important for maintaining soil fertility and linking carbon cycle within terrestrial ecosystem. In order to quantify the spatial patterns of SOM content in Hailun County of China, ordinary kriging (OK), universal kriging (UK), cokriging (CK), regression kriging (RK), and secondary data derived from soil total nitrogen (TN), cation exchangeable capacity (CEC) and digital elevation model (DEM) were employed to map the spatial distribution patterns of SOM with different sampling densities. The results of SOM spatial structures identified by different sampling densities indicated that the increase of soil sampling densities may not benefit the identification of SOM continuous component and optimal design of sampling locations was even more important than improving sampling densities. With respect to spatial predictive performance, CK with TN as covariable always had the highest accuracy, and more than 50% of the SOM variation can be explained by CK method. The local detail of SOM variability derived from CK method also outperformed OK, UK, and RK, and the predictive accuracy of RK method may not be improved when compared with univariate OK method. The soil sampling density is not the exclusive factor affecting the spatial predictive accuracy of SOM, and secondary data (ancillary information) significantly correlated to SOM as well as denser coverage of the study area would be more helpful for improving the spatial predictive performance of SOM with limited number of soil samples.

Key words: soil organic matter, ordinary kriging, universal kriging, cokriging, regression kriging