水保所知识产出(1956---)知识库

针对区域畦灌随机模拟模型中直接蒙特卡洛模拟方法存在的抽样重复性及收敛性较差等问题，借助仅考虑
土壤特性空间随机性的区域畦灌随机模拟模型，研究拉丁超立方模拟、修正拉丁超立方模拟与直接蒙特卡罗模拟
在抽样精确性、区域畦灌模拟计算收敛性、准确性、稳定性和计算效率等方面的差异。结果表明: 修正拉丁超立方
模拟和拉丁超立方模拟能够很好地表征土壤干容重的区域分布特征，抽样精确性较高。修正拉丁超立方模拟收敛
速度最快，能够显著改善抽样质量，模拟准确性与分布形态优于拉丁超立方模拟与直接蒙特卡洛模拟。修正拉丁
超立方模拟具有比拉丁超立方模拟与直接蒙特卡洛模拟更好的稳定性。基于修正拉丁超立方的区域畦灌随机模
拟模型计算效率比拉丁超立方抽样和直接蒙特卡洛模拟分别提高0. 23 倍和1. 80 倍，说明修正拉丁超立方模拟能
够解决原有模拟方法计算效率较低的问题，有助于缩短计算时间，降低计算成本。

The uncertainty of border irrigation parameters will directly affect the simulation results of
border irrigation flow model for surface irrigation management at regional scale． Latin hypercube sampling
( LHS) ，updated Latin hypercube sampling ( ULHS) and simple Monte Carlo sampling ( SMCS) are
studied in sampling convergence，stability and simulation efficiency based on regional scale border
irrigation simulation method and soil properties spatial randomness． In terms of the sampling accuracy，
ULHS and LHS can satisfactorily represent the statistical characteristics of soil bulk density． In the
simulation convergence aspect，ULHS has the faster convergence than LHS and SMCS，which indicates
that ULHS can significantly improve the sampling quality and reduce the sampling frequency． In the
simulation accuracy，simulation times of ULHS are less than those of LHS and SMCS，simulation
accuracy and distribution pattern are better than those of LHS and SMCS． Stability simulation of ULHS is
better than those of LHS and SMCS． Computational efficiency of ULHS is increased by 0. 23-fold and
1. 8-fold than those of LHS and SMCS，respectively． Convergence rate of ULHS is faster than those of LHS
and SMCS． In addition，ULHS can improve computational efficiency and sampling stability． ULHS can help to
improve the simulation performance of regional scale border irrigation stochastic simulation model．