基于遥感影像和决策树算法的土壤制图
|
摘要
传统土壤信息获取方法已经无法完全满足当前各领域对土壤数据的需求,如何结合新的技术提高土壤普查效率,获取高精度土壤图成为了现阶段的研究重点。本研究综合利用高分二号遥感影像提取的遥感光谱指数以及DEM数据提取的地形因子,通过决策树算法进行数据挖掘,获取各土壤类型的土壤—环境规则,然后利用SoLIM结合土壤—环境规则进行推理制图,获得研究区的土壤类型分布图。结果表明,预测土壤图总体精度为88%,高于传统土壤图的精度72%,且在三种不同的采样方式(均匀采样、横截面采样和主观采样)下土壤预测精度分别为89%、88%、86%,均高于传统土壤图。这说明,预测土壤图比传统土壤图更能反映土壤类型空间差异,且预测土壤图在表达土壤类型整体空间分布信息的同时也可捕捉到土壤类型与地貌类型的耦合关系。
Traditional soil information acquisition methods can not fully meet the new needs of various related fields.Therefore, how to combine new technology to improve soil survey efficiency and obtain high-precision soil maps has become the focus of research at this stage. Spectral indices of GF-2 satellite images and topographic data captured by DEM were used for data mining in combination with decision tree algorithm. The soil-environment rules under different soil types obtained by decision tree algorithm, which were input into SoLIM software to infer the distribution of prediction soil map. The overall accuracy of the predicted soil map was 88%, higher than that of the original soil map of 72%. And the accuracy of predicted soil map were 89%, 88%, 86% by using the methods of uniform sampling,random sampling, section sampling, respectively. The predicted soil map could more accurately reflect the spatial diversity of soil types than the original one, and predicted soil map can express the spatial distribution information of soil types and can also capture the coupling relationship between soil type and geomorphic type.
Traditional soil information acquisition methods can not fully meet the new needs of various related fields.Therefore, how to combine new technology to improve soil survey efficiency and obtain high-precision soil maps has become the focus of research at this stage. Spectral indices of GF-2 satellite images and topographic data captured by DEM were used for data mining in combination with decision tree algorithm. The soil-environment rules under different soil types obtained by decision tree algorithm, which were input into SoLIM software to infer the distribution of prediction soil map. The overall accuracy of the predicted soil map was 88%, higher than that of the original soil map of 72%. And the accuracy of predicted soil map were 89%, 88%, 86% by using the methods of uniform sampling,random sampling, section sampling, respectively. The predicted soil map could more accurately reflect the spatial diversity of soil types than the original one, and predicted soil map can express the spatial distribution information of soil types and can also capture the coupling relationship between soil type and geomorphic type.
引文
[1]THWAITES RN,SLATER BK.Soil-landscape resource assessment for plantations--a conceptual framework towards an explicit multi-scale approach[J].Forest Ecology&Management,2000,138(1-3):123-138.
[2]韩宗伟,黄魏,罗云,等.基于路网的土壤采样布局优化——模拟退火神经网络算法[J].应用生态学报,2015,26(03):891-900.
[3]HUDSON B D.The Soil Survey as Paradigm-based Science[J].Soil Science Society of America Journal,1992,56(3):836-841.
[4]孙福军,雷秋良,刘颖,等.数字土壤制图技术研究进展与展望[J].土壤通报,2011,42(06):1502-1507.
[5]周斌,王繁.基于决策树模型的土壤性质空间推断[J].土壤通报,2004(04):385-390.
[6]任丽,杨联安,王辉,等.基于随机森林的苹果区土壤有机质空间预测[J].干旱区资源与环境,2018,32(08):141-146.
[7]谢军,秦承志,肖桂荣,等.模糊聚类方法在南方红壤小流域土壤属性制图中的应用——以长汀朱溪河小流域为例[J].中国水土保持科学,2015,13(05):132-139.
[8]翟天林,金贵,邓祥征,等.基于多源遥感影像融合的武汉市土地利用分类方法研究[J].长江流域资源与环境,2016,25(10):1594-1602.
[9]刘峰,朱阿兴,李宝林,等.利用陆面反馈动态模式来识别土壤类型的空间差异[J].土壤通报,2009,40(03):501-508.
[10]LIU F,GENG X,ZHU A X,et al.Soil texture mapping over low relief areas using land surface feedback dynamic patterns extracted from MODIS[J].Geoderma,2012,s 171-172(2):44-52.
[11]LIU F,GENG X,ZHU A X,et al.Soil polygon disaggregation through similarity-based prediction with legacy pedons[J].Journal of Arid Land,2016,8(5):760-772.
[12]MAHMOUDABADI E,KARIMI A,HAGHNIA G H,et al.Digital soil mapping using remote sensing indices,terrain attributes,and vegetation features in the rangelands of northeastern Iran[J].Environmental Monitoring&Assessment,2017,189(10):500.
[13]郭燕,程永政,王来刚,等.利用高光谱和GF-1模拟多光谱进行土壤有机质预测和制图研究[J].土壤通报,2016,47(03):537-542.
[14]TAYLOR SA,ASHCROFT GL.Physical edaphology:The physics of irrigated and nonirrigated soils[J].Journal of Environmental Quality,1974,3(2):iv.
[15]DE B S,WIELEMAKER W G,MOLENAAR M.Formalisation of soil-landscape knowledge through interactive hierarchical disaggregation[J].Geoderma,1999,91(1):151-172.
[16]ENDRE D,ERIKA M,BAUMGARDNER M F,et al.Use of combined digital elevation model and satellite radiometric data for regional soil mapping[J].Geoderma,2000,97(3-4):367-391.
[17]ODEH I O A,MCBRATNEY A B.Using AVHRR images for spatial prediction of clay content in the lower Namoi Valley of eastern Australia[J].Geoderma,2000,97(3):237-254.
[18]陈波,胡玉福,喻攀,等.基于纹理和地形辅助的山区土地利用信息提取研究[J].地理与地理信息科学,2017,33(01):1-8.
[19]黄魏,许伟,汪善勤,等.基于不确定性模型的土壤——环境关系知识获取方法的研究[J].土壤学报,2018,55(01):54-63.
[20]庞素琳,巩吉璋.C5.0分类算法及在银行个人信用评级中的应用[J].系统工程理论与实践,2009,29(12):94-104.
[21]薛薇,陈欢歌.Clementine数据挖掘方法及应用[M].电子工业出版社,2010.
[22]郭立力,赵春江.十折交叉检验的支持向量机参数优化算法[J].计算机工程与应用,2009,45(08):55-57.
[23]朱阿兴,李宝林,杨琳,等.基于GIS、模糊逻辑和专家知识的土壤制图及其在中国应用前景[J].土壤学报,2005(05):142-149.
[24]ZHU A X.A similarity model for representing soil spatial information[J].Geoderma,1997,77(2-4):217-242.
[25]李超,文天晟,张凤荣,等.半干旱沙区土类/亚类的遥感调查制图方法[J].农业工程学报,2018,34(06):189-196.
[2]韩宗伟,黄魏,罗云,等.基于路网的土壤采样布局优化——模拟退火神经网络算法[J].应用生态学报,2015,26(03):891-900.
[3]HUDSON B D.The Soil Survey as Paradigm-based Science[J].Soil Science Society of America Journal,1992,56(3):836-841.
[4]孙福军,雷秋良,刘颖,等.数字土壤制图技术研究进展与展望[J].土壤通报,2011,42(06):1502-1507.
[5]周斌,王繁.基于决策树模型的土壤性质空间推断[J].土壤通报,2004(04):385-390.
[6]任丽,杨联安,王辉,等.基于随机森林的苹果区土壤有机质空间预测[J].干旱区资源与环境,2018,32(08):141-146.
[7]谢军,秦承志,肖桂荣,等.模糊聚类方法在南方红壤小流域土壤属性制图中的应用——以长汀朱溪河小流域为例[J].中国水土保持科学,2015,13(05):132-139.
[8]翟天林,金贵,邓祥征,等.基于多源遥感影像融合的武汉市土地利用分类方法研究[J].长江流域资源与环境,2016,25(10):1594-1602.
[9]刘峰,朱阿兴,李宝林,等.利用陆面反馈动态模式来识别土壤类型的空间差异[J].土壤通报,2009,40(03):501-508.
[10]LIU F,GENG X,ZHU A X,et al.Soil texture mapping over low relief areas using land surface feedback dynamic patterns extracted from MODIS[J].Geoderma,2012,s 171-172(2):44-52.
[11]LIU F,GENG X,ZHU A X,et al.Soil polygon disaggregation through similarity-based prediction with legacy pedons[J].Journal of Arid Land,2016,8(5):760-772.
[12]MAHMOUDABADI E,KARIMI A,HAGHNIA G H,et al.Digital soil mapping using remote sensing indices,terrain attributes,and vegetation features in the rangelands of northeastern Iran[J].Environmental Monitoring&Assessment,2017,189(10):500.
[13]郭燕,程永政,王来刚,等.利用高光谱和GF-1模拟多光谱进行土壤有机质预测和制图研究[J].土壤通报,2016,47(03):537-542.
[14]TAYLOR SA,ASHCROFT GL.Physical edaphology:The physics of irrigated and nonirrigated soils[J].Journal of Environmental Quality,1974,3(2):iv.
[15]DE B S,WIELEMAKER W G,MOLENAAR M.Formalisation of soil-landscape knowledge through interactive hierarchical disaggregation[J].Geoderma,1999,91(1):151-172.
[16]ENDRE D,ERIKA M,BAUMGARDNER M F,et al.Use of combined digital elevation model and satellite radiometric data for regional soil mapping[J].Geoderma,2000,97(3-4):367-391.
[17]ODEH I O A,MCBRATNEY A B.Using AVHRR images for spatial prediction of clay content in the lower Namoi Valley of eastern Australia[J].Geoderma,2000,97(3):237-254.
[18]陈波,胡玉福,喻攀,等.基于纹理和地形辅助的山区土地利用信息提取研究[J].地理与地理信息科学,2017,33(01):1-8.
[19]黄魏,许伟,汪善勤,等.基于不确定性模型的土壤——环境关系知识获取方法的研究[J].土壤学报,2018,55(01):54-63.
[20]庞素琳,巩吉璋.C5.0分类算法及在银行个人信用评级中的应用[J].系统工程理论与实践,2009,29(12):94-104.
[21]薛薇,陈欢歌.Clementine数据挖掘方法及应用[M].电子工业出版社,2010.
[22]郭立力,赵春江.十折交叉检验的支持向量机参数优化算法[J].计算机工程与应用,2009,45(08):55-57.
[23]朱阿兴,李宝林,杨琳,等.基于GIS、模糊逻辑和专家知识的土壤制图及其在中国应用前景[J].土壤学报,2005(05):142-149.
[24]ZHU A X.A similarity model for representing soil spatial information[J].Geoderma,1997,77(2-4):217-242.
[25]李超,文天晟,张凤荣,等.半干旱沙区土类/亚类的遥感调查制图方法[J].农业工程学报,2018,34(06):189-196.