A progressive simulation method to estimate the stocks of surface soil organic carbon and total nitrogen in an estuarine archipelago

Simulating the soil organic carbon (SOC) and total nitrogen (TN) stocks in estuarine areas is important for the development of digital soil mapping and the regional inventory of carbon and nitrogen, and simulation accuracy is the key issue. In this study, a progressive simulation method was proposed to further improve accuracy through the adequate utilization of the correlations among soil parameters and their relations with the comprehensive land surface factor system. Three soil parameters, namely, SOC, TN, and bulk density (BD), were simulated using the method for estimating SOC and TN stocks. The simulations were progressively conducted in three levels (Levels 1-3) by using partial least squares regression (PLSR) and cokriging. Each of the parameters was simulated separately in Level 1, one parameter was simulated based on each of the two other parameters in Level 2, and one parameter was simulated based on the two other parameters in Level 3. A typical estuarine archipelago in southern China was selected to demonstrate the performance of the method in surface soil (0-20 cm). Results showed the effectiveness of the progressive simulation in improving accuracy when the PLSR was used. The simulation results in Level 3 had the highest accuracy among the results in all levels, increasing the accuracies for SOC, TN, and BD by 54.35%, 57.24%, and 5.61%, respectively, compared with those in Level 1 and achieving high accuracies compared with those in similar studies. The SOC and TN stocks amounted to 270.66 and 26.96 Gg, respectively, and the mean values of SOC and TN densities were 2.65 and 0.26 kg/m(2), respectively, which are lower than global values and those for China. The SOC and TN densities exhibited distinct spatial heterogeneities, which were fundamentally driven by different types of island exploitations and large-scale sea reclamation activities.

Automated summary

The progressive simulation method proposed in this study, using different regression techniques to simulate SOC, TN, and BD parameters at three levels, has been shown to improve accuracy in surface soil in estuarine areas. Results indicate that the highest accuracy was achieved in the simulation results at Level 3.