机构地区:[1]中国热带农业科学院橡胶研究所,海南儋州571737 [2]中国热带农业科学院科技信息研究所,海南儋州571737 [3]海南农垦科学院,海口570206
出 处:《南方农业学报》2015年第10期1839-1848,共10页Journal of Southern Agriculture
基 金:现代农业产业技术体系建设专项项目(CARS-34-GW6)
摘 要:【目的】将土壤属性和连续型环境变量相结合,研究橡胶园管理分区,为大尺度范围内橡胶园管理提供参考。【方法】以土壤属性(p H、有机质、全氮、有效磷和速效钾)和连续型环境变量(高程、坡度、平均降雨量、平均气温和归一化植被指数)为数据源,在对土壤属性进行空间变异分析及对所有变量完成主成分分析的基础上,利用K-均值聚类法对橡胶园进行管理分区。【结果】土壤p H为弱变异性,有效磷属于强变异性,其他土壤属性为中等变异性;连续型环境变量中,高程、平均降雨量和平均气温属于低变异性,坡度和归一化植被指数为中等变异性。半方差分析中,除土壤有机质符合球状模型外,其余土壤属性均符合指数模型,同时除土壤有效磷具有中等空间自相关性,其余土壤属性具有强烈的空间自相关性。分别以土壤属性、环境变量及土壤属性与环境变量的组合为输入变量进行K-均值聚类,均得到4个分区,单以土壤属性或环境变量划分管理分区显著降低分区内所用变量自身属性的变异系数,而对其他变量属性变异系数降低有限,但将土壤属性和环境变量结合起来进行管理分区可同时大幅度提高不同分区内土壤属性和环境变量的均一性。【结论】综合利用土壤属性和环境变量是在划分管理区时的最佳选择,当土壤属性缺乏或不易获取时,也可利用易于获取的环境变量进行管理分区。[ Objective ]In order to provide reference for improve management efficiency of large-scale rubber planta- tions, the present experiment were conducted to investigate management zones of rubber plantations based on soil proper- ties and environmental variables. [Method]Based on data about soil properties viz., pH, organic matter(OM), total ni- trogen (TN), available phosphorus (AP), available potassium (AK) and environmental variables viz., parent materials (Pa), elevation (ELE), slope (SLO), aspect (ASP), mean precipitation (PRE), mean temperature (TMEAN), normalized difference vegetation index (NDVI), the spatial variation of soil properties and principal components of all variables were analysed. The study area of rubber plantations were partitioned into different management zones by K-means cluster- ing method. [ResultlThe soil pH showed weak spatial variability, AP showed strong spatial variability, other soil prop- erties showed moderate spatial variability. Among environmental variables, ELE, monthly PRE and monthly TMEAN showed weak spatial variability. SLO and NDVI showed moderate spatial variability. In addition, the semivariance analysis indicated that all soil properties were fitted with exponential model, except for OM (fitted with spherical model), and AP showed moderate spatial autocorrelation, other soil properties showed strong spatial autocorrelation. Using soil properties, environmental variables and combination of both as input variables, the study area was partitioned into four management zones. Mean coefficient of variation(C.V. ) of soil properties and environmental variables after delineating management zones were calculated and compared with the original datasets, the results showed that using soil properties or environ- mental variables alone only markedly reduced C.V. of soil properties or environmental variables, while using both of these two datasets significantly cut down C.V. of soil properties and environmental variables simultane
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...