| 摘要: |
| 土壤养分是土壤肥力的重要指标，了解森林土壤养分垂直分布状况可为林业土壤科学施肥提供依据。研究对云浮市森林土壤构建从上到下分5 层的人工神经网络模型，通过对候选的模型输入参数进行筛选获得最优组合，并生产空间分布图。结果表明，所建模型对土壤养分预测能力不同，其中碱解氮、全钾较好，有机质、速效钾、全磷一般，速效磷、全氮相对较差。筛选出的最优模型输入组合显示，地形位置指数、坡度、泥沙输移比、坡向对土壤养分的预测能力最强，水流长度、潜在太阳辐射次之，水流流向、土壤地形因素、垂直坡位相对较差，且同一参数对不同土壤养分指标的预测能力不同；最优模型生产的土壤养分图更详细的反应土壤养分的空间分布。 |
| 关键词: ANN模型 最优组合 土层 输入参数 土壤养分 |
| DOI： |
| 分类号: |
| 基金项目:1.国家自然科学,基于模型预测土壤属性的大面积高分辨率森林生态立地类型的分类研究项目(31500385)。2.广东省林业科技计划项目,广东省梅州市森林土壤调查(2019-07)。,国家自然科学基金 |
|
| Establishment of A Model for Predicting Spatial Distribution of Multilayer Forest Soil Nutrients in Yunfu |
|
SUN Dongxiao1, YANG Qi1, ZHAO Zhengyong1, DING Xiaogang2, LI Yingying1, Zhu Hangyong
|
|
1.Guangxi University,Nanning;2.Guangdong Academy of Forestry Sciences,Guangzhou
|
| Abstract: |
| Soil nutrient is an important index of soil fertility. Understand the vertical distribution of forest soil nutrient scan provide the theoretical basis for guiding the soil fertilization scientifically. In this study, an artificial neural network prediction model of five soil layers from top to bottom was constructed. The optimal combination of parameters was obtained by selecting the candidate model inputs, and the spatial distribution map was produced. The results showed that the prediction performance of the model to individual soil nutrient was different. It was good for Total Potassium (TK) and Available Nitrogen (AN), general for Soil Organic Matter(SOM), Available Potassium (AK) and Total Phosphorous (TP), and relatively poor for Available Phosphorus (AP) and Total Nitrogen (TN). The optimal combination of model inputs showed that Topographic Position Index（TPI）, Slope, Sediment delivery ratio (SDR) and Aspect had the strongest prediction ability for soil nutrients;Flow Length (FL) and Potential Solar Radiation (PSR) were the second; Flow Direction (FD), Soil Topographical
Factors (STF) and Vertical Slope Position (VSP) were relatively poor; and each parameter’s performance forpredicting specific soil nutrient was different. The soil nutrient maps predicted by the optimal model could reflect the spatial distribution of soil nutrients in more details. |
| Key words: ANN model optimal combination soil layer input parameter soil nutrient |
