华北农田土壤水分运动参数的Bayes估计

为了估计具有垂向空间异质性的田间尺度土壤水分运动参数,提出利用原位多点土壤含水率观测数据反演多层土壤持水和导水特征参数的Bayes方法。该方法基于Hydrus-1D模拟模型建立Bayes推断模型,采用Markov Chain Monte Carlo(MCMC)方法求解Bayes联合概率分布,得到参数的后验边缘分布。将该方法应用于华北平原一典型农田土壤剖面,参数反演结果反映了空间变异性,与室内试验值具有较好的一致性,相应的土壤水分运动模拟具有较高精度,从而验证了该方法的有效性。

Bayesian estimate of the soil hydraulic parameters in agricultural land in North China

The vadose zone is vertically heterogeneous with variable soil hydraulic parameters in the field.A Bayesian model was developed to estimate the multilayer parameters of the soil water retention and hydraulic conductivity functions from in situ measurements of soil water content at several depths.The inference of the Bayesian model was based on simulations of the soil moisture variation using a Hydrus-1D model.A Markov Chain Monte Carlo(MCMC) method was used to sample the Bayesian joint posterior probability function to get the marginal posterior distributions of the soil hydraulic parameters.The Bayesian model was used to analyze a soil profile in agricultural land in North China.The calibrated soil hydraulic parameters show the heterogeneity and are consistent with that obtained in laboratory tests.Reasonably high accuracy simulations of the soil moisture variation were possible with the calibrated parameters.Thus,the results illustrated the validity of this method.