自组织神经网络在流域土壤水分含量空间模式识别中的应用

Application of Self-Organizing Neural Networks in Identification of Soil Moisture Spatial Patterns

由于土壤水分含量受众多因素的影响,空间变异性很大,给土壤水分含量空间分布的研究带来了很大的困难。空间模式识别是处理土壤水分含量空间数据的方法之一,能够分析得到土壤水分含量空间数据的聚类结果。基于自组织特征映射和自适应共振理论的自组织神经网络模型在空间数据模式识别中得到广泛应用,针对澳大利亚tar-rawarra试验流域土壤水分含量的观测数据,应用自组织神经网络,建立动态土壤水分含量的空间模式识别模型,并用半变异函数对识别结果进行检验,实例研究表明该方法是一种行之有效的方法。

It is very difficult to study soil moisture spatial distribution because of its great spatial variability affected by many factors. The spatial pattern identification is one of the ways to deal with the soil moisture spatial data, which can obtain clustering of soil moisture spatial data. The self-organizing neural network models find wide uses in identification spatial soil moisture pattern, which based on the self-organizing reflection and self-adaptive resonance theory. A dynamic soil moisture spatial identification model is developed against at the soil moisture observation data in the Tarrawarra Catchment in Australia using self-organizing neural network. The semi-variance function is used to verify the identification result, and the results of verification indicate that the method is a valid one.