论文部分内容阅读
以云浮市云城区和云安区森林土壤为研究对象,应用传统统计学和地统计学方法结合GIS技术分析其土壤养分的空间变异性,预测土壤养分空间分布。地统计学主要从空间插值模型和人工神经网络模型的角度对土壤养分的空间变异性进行诠释,并利用平均绝对误差(MAE)和均方根误差(RMSE)两个指标以及模型预测点与实测点间的相关系数作为判断模型好坏的标准。研究结果认为插值模型中泛克里格插值法在样点密度较小时显示出明显优势,而BP-ANN在模型的稳定性和推广性表现尤为突出,最后对泛克里格插值模型和BP-ANN下的有机碳、全氮、全磷、全钾4种养分空间预测分布特征进行描述。
Taking Yuncheng District of Yunfu City and Yun’an District as the research object, the spatial variability of soil nutrients was analyzed by using traditional statistics and geo-statistics methods combined with GIS technology to predict the spatial distribution of soil nutrients. Geostatistics mainly interprets the spatial variability of soil nutrients from spatial interpolation model and artificial neural network model, and makes use of the two indexes of average absolute error (MAE) and root mean square error (RMSE) The correlation coefficient between points as a good judge of the quality of the standard. The results show that interpolation Kriging interpolation method shows a clear advantage when the sample density is small, but BP-ANN is particularly prominent in the stability and generalization of the model. Finally, ANN was used to describe the spatial distribution of four nutrients in organic carbon, total nitrogen, total phosphorus and total potassium.