Number of soil profiles needed to give a reliable overall estimate of soil organic carbon storage using profile carbon density data

Abstract

Soil profile data are the basis for estimating soil organic carbon (SOC) storage and any changes in SOC storage, which are of great significance in terms of global climate change. Estimates based on various profile data have been made for SOC storage in China. Modeling studies have given contrasting results on changes in SOC storage in Chinese croplands. A certain number of measured soil profile data are needed to validate the modeled results. In the present study, we examined the relationships between sample size, population variance and detection limit using the central limit theorem and the statistical properties of the normal distribution. Based on the profile dataset from the Second National Soil Survey in China, we calculated that to derive a reliable estimate of the overall mean SOC density for all the soils of China, a sample size of 4,000 soil profiles is needed. In this case, a reliable estimate is defined as having a 95% confidence interval and allowing a ±5% detection limit of SOC. The necessary sample size for cropland soils is 1,250. Our results indicate that approximately 100 samples only are needed to validate a modeled SOC loss of 20–30% in cropland soils in China. By aggregating the soil profiles in the dataset into soil orders and calculating the variance of each soil order, we show that the sample sizes in the dataset are insufficient to give reliable estimates on the carbon densities of most soil orders; thus, we conclude that there is considerable uncertainty in the SOC distribution maps resulting from the Second National Soil Survey.