论文部分内容阅读
针对传统的时空克里格算法的精度受到时空变异函数的影响,而时空变异函数理论模型的选择常受主观因素影响和理论半变异函数局限,没有普适性的建模方法;加之参数较多估计困难,致使插值精度不高的问题,该文提出一种普适性的基于广义回归神经网络自适应时空克里格插值变异函数拟合方法,在此基础上建立了广义回归神经网络与时空克里格结合的新颖时空混合插值算法。通过与传统插值方法在民勤县地下水埋深插值中的比较研究表明,该时空混合插值算法的插值精度显著提高,并且是一个普适性的插值法。
The accuracy of the traditional space-time Kriging algorithm is affected by the spatiotemporal variation function. However, the selection of the theoretical model of the spatio-temporal variation function is often affected by the subjective factors and the theoretical semi-variogram, and there is no universal modeling method. It is difficult to estimate the accuracy of the interpolation, resulting in the problem of low interpolation accuracy. In this paper, a universal fitting method based on generalized regression neural network adaptive spatiotemporal Kriging interpolation variation function is proposed. Based on this, a generalized regression neural network Kriging’s Novel Spatial-temporal Hybrid Interpolation Algorithm. The comparison with traditional interpolation method in groundwater depth interpolation in Minqin County shows that the interpolation accuracy of the spatiotemporal hybrid interpolation algorithm is significantly improved and is a universal interpolation method.