几种土壤粒径分布参数模型在稀疏分级数据中的对比研究

ELEVEN MODELS DESCRIBING SOIL PARTICLE-SIZE DISTRIBUTION FOR SPARSE DATA

采用全国第二次土壤普查的西藏地区1 634个土壤样本的颗粒分析资料以选择最好的参数模型.使用了确定性系数和Akaike资讯规范对模型进行对比研究.结果表明:Skaggs模型的总体表现最好,其次为Weibull和2个Fredlund模型,modified logistic growth,Offset-Nonrenormalized Lognormal,Offset-Renormalized Lognormal,vanGenuchten modified和van Genuchten type等几个模型都较为理想,而Logistic growth和Simple Lognormal模型效果最差.土壤质地对模型的表现存在一定影响.模型的表现随黏粒质量分数的增加而发生变化.

Eleven particle-size distribution （PSD） models were examined in order to determine the best model in representing PSDs of 1634 Tibet soil samples. Coefficient of determination （R^2） was used to measure goodness-of-fit of models and Akaike＇s information criterion （AIC） to compare quality of model fits. The Skaggs model was found to be the best, followed by Weibull and Fredlund model. Performance of modified logistic growth, Offset-Nonrenormalized Lognormal, Offset-Renormalized Lognormal, an Genuchten modified and van Genuchten type model were good, Logistic growth and Simple Lognormal model were the worst. Performance of PSD models was affected by soil texture. Performance of most PSI） models changed with increasing clay content in the soil samples.