Selecting the optimal sample size accurately is important for improving the measurement precision of soil physicochemical properties. In this study, we examined the within-site heterogeneity of soil physical and chemical properties in an evergreen broad-leaved forest and a Chinese fir (Cunninghamia lanceolata) plantation in subtropical China. We applied both traditional and bootstrapping technique to determine the optimal sample size, the number of observation (n), for predicting soil nutrient content with an allowable error of 10% at a 95% confidence level for both forest types. The results showed that soil pH and C/N had low variation, while the soil nutrients had a moderate variation in both forests. The coefficients of variation (CV) of total element concentration in the evergreen broad-leaved forest, such as total carbon and total nitrogen, were significantly larger than those in the Chinese fir plantation, while the CVs of available nutrients were similar between two forests. The evaluation error of all soil properties in two forests showed a stable trend after a sharp decline with the increase of sample size. The mean values also changed from violent fluctuation to stable, and extravagant sample could not effectively improve the accuracy of mean value measurement. The sample sizes of soil water content, total carbon, total nitrogen, nitrate (NO3--N), total potassium, total calcium, and total magnesium in the evergreen broad-leaved forest were signifi-cantly larger than those in the Chinese fir plantation, while the sample size for ammonium (NH4+-N) showed opposite pattern, and the others had no significant difference. In addition, the sample sizes of different soil physicochemical properties varied significantly in the same forest. The available elements, such as NH4+-N and NO3--N, needed a larger sample size than total elements in both forests. Accordingly, we proposed that the heterogeneity and sample size of soil physicochemical properties should be considered before soil survey. An adequate sample size may be varied according to the purpose of sample and the site-specific variation of soil characters.
准确获取最佳取样量有助于提高土壤理化性质的测量精度.采用经典统计学方法,研究了亚热带常绿阔叶林和杉木人工林2个林型土壤理化性质的异质性,运用传统法和自助法研究在90%精度下评估土壤养分含量的最佳取样量.结果表明: 2个林型中土壤pH值和C/N表现为弱变异,其他养分指标均为中等程度变异;常绿阔叶林土壤全C、全N等全量土壤指标的变异系数显著大于杉木人工林,而速效指标差异不显著;2个林型中所有土壤指标的评估误差均随取样量的增加呈现出先急剧曲线下降后稳定的趋势,均值也由剧烈波动转向平稳,并且过多的取样并不能有效提高均值测定精度;常绿阔叶林土壤含水率、全C、全N、NO3--N、全K、全Ca、全Mg的取样量都显著大于杉木人工林,杉木林土壤NH4+-N的取样量显著大于阔叶林,而其他指标二者间差异不显著;同一林型各个指标间取样量差异更加明显,NH4+-N、NO3--N等速效指标的取样量显著大于全C、全N等全量指标.因此,建议在土壤调查或监测过程中充分考虑土壤理化性质的异质性,最佳取样量必须依据于取样的目的和土壤指标的变异信息.
Keywords: bootstrapping; evergreen broad-leaved forest; optimal sample size; sampling design; soil heterogeneity.