陈瀚阅1,,,
张黎明1,
聂祥琴1,
邢世和1,
范协裕1
1.福建农林大学资源与环境学院/福建省土壤生态系统健康与调控重点实验室 福州 350002
2.福建农林大学公共管理学院 福州 350002
基金项目: 国家自然科学基金项目41971050
福建省科技计划项目2017N5006
农业农村部农技推广项目KLD19H01A
中央引导地方科技发展专项2018L3013
福建农林大学科技创新专项基金项目KFA18106A
详细信息
作者简介:任必武, 主要进行土壤属性制图研究。E-mail: rbw126@126.com
通讯作者:陈瀚阅, 主要研究方向为定量遥感和土壤地理。E-mail: Chenhanyue.420@163.com
中图分类号:S15计量
文章访问数:210
HTML全文浏览量:31
PDF下载量:32
被引次数:0
出版历程
收稿日期:2020-11-22
录用日期:2021-01-20
网络出版日期:2021-06-22
刊出日期:2021-06-01
Comparison of machine learning for predicting and mapping soil organic carbon in cultivated land in a subtropical complex geomorphic region
REN Biwu1, 2,,CHEN Hanyue1,,,
ZHANG Liming1,
NIE Xiangqin1,
XING Shihe1,
FAN Xieyu1
1. College of Resources and Environment, Fujian Agriculture and Forestry University/Key Laboratory of Soil Ecosystem Health and Regulation of Fujian Province, Fuzhou 350002, China
2. School of Public Management, Fujian Agriculture and Forestry University, Fuzhou 350002, China
Funds: the National Natural Science Foundation of China41971050
the Science and Technology Planning Project of Fujian Province2017N5006
the Agricultural Technology Extension Project of the Ministry of Agriculture and Rural Affairs of the People's Republic of ChinaKLD19H01A
the Central Committee Guides Local Science and Technology Development Projects of China2018L3013
the Special Fund for Science and Technology Innovation of Fujian Agriculture and Forestry UniversityKFA18106A
More Information
Corresponding author:CHEN Hanyue, E-mail: Chenhanyue.420@163.com
摘要
HTML全文
图
参考文献
相关文章
施引文献
资源附件
访问统计
摘要
摘要:耕地土壤有机碳（SOC）是土壤质量的重要指标，也是生态系统健康的重要表征。当前机器学习（Machine Learning，ML）用于SOC数字制图日益热门，但不同算法在高空间分辨率SOC数字制图中的对比研究尚有欠缺。本研究以福建省东北部复杂地形地貌区为例，采用10 m空间分辨率Sentinel-2影像数据，选取地形、气候、遥感植被变量为驱动因子，重点分析当前常用的机器学习算法——支持向量机（Support Vector Machine，SVM）、随机森林（Random Forest，RF）在SOC预测中的差异，并与传统普通克里格模型（Ordinary Kriging，OK）进行比较。结果表明：基于地形、遥感植被因子和气候因子构建的RF模型表现最佳（RMSE=2.004，r=0.897），其精度优于OK模型（RMSE=4.571，r=0.623），而SVM模型预测精度相对最低（RMSE=5.190，r=0.431）；3种模型预测SOC空间分布趋势总体相似，表现为西高东低、北高南低，其中RF模型呈现的空间分异信息更加精细；最优模型反演得到耕地土壤有机碳平均含量为15.33 g·kg-1；RF模型和SVM模型变量重要性程度表明：高程和降水是影响复杂地貌区SOC空间分布的重要变量，而遥感植被因子重要性程度低于高程。
关键词:复杂地貌区/
耕地土壤有机碳/
机器学习算法/
普通克里格/
数字土壤制图
Abstract:Soil organic carbon (SOC) is a key indicator of soil quality and ecosystem health. At present, machine learning (ML) models for predicting soil properties based on environmental variables are increasingly popular; however, the performance of different ML algorithms in predicting and mapping SOC, especially at high spatial resolutions, have not been compared. This study aimed to develop, evaluate, and compare the performance of Support Vector Machine (SVM), Random Forest (RF), and Ordinary Kriging (OK) models for predicting and mapping the SOC contents in the northeast of Fujian Province. Remote sensing vegetation indices were derived from Sentinel-2 image data with a spatial resolution of 10 m. These vegetation indices, along with selected terrain and climate factors, were adopted as environmental variables to map SOC using the SVM and RF models. The results showed that the performance of the RF model (RMSE[root-mean-square error]=2.004, r=0.897) was better than that of the OK model (RMSE=4.571, r=0.623) and explained most of the SOC spatial heterogeneity. The SVM model had the poorest prediction accuracy (RMSE=5.190, r=0.431). SOC mapped from the three models had similar spatial patterns, with an increasing SOC gradient from east to west and from south to north of the study area. SOC in the farmlands predicted with the RF model varied in the range of 15.33±4.07 g·kg-1. Elevation and rainfall were the most important variables for the RF and SVM models, respectively, whereas the remote sensing vegetation indices were less important than elevation.
Key words:Complex landform/
Soil organic carbon of cultivated land/
Machine learning/
Ordinary kriging/
Digital soil mapping
HTML全文
图1研究区地理位置、采样点分布及不同采样点土壤有机碳(SOC)含量范围
Figure1.Location, distribution of soil sampling sites and soil organic carbon content (SOC) ranges of sampling sites of the study areas
下载: 全尺寸图片幻灯片
图2随机森林(RF, a)、支持向量机(SVM, b)和传统普通克里格模型(OK, c)预测土壤有机碳含量(SOC)模型表现和不同模型预测土壤有机碳范围均方根误差(RMSE, d)
Figure2.Performance of soil organic carbon content (SOC) prediction of Random Forest (RF, a), Support Vector Machine (SVM, b) and Ordinary Kriging (OK, c) models and RMSEs in different soil organic carbon ranges predicted by different models (d)
下载: 全尺寸图片幻灯片
图3随机森林(RF)和支持向量机(SVM)模型因子重要性对比
Figure3.Comparison of importance of environmental variables in Random Forest (RF) and Support Vector Machine (SVM) models
下载: 全尺寸图片幻灯片
图4基于随机森林(RF, Ⅰ)、支持向量机(SVM, Ⅱ)和传统普通克里格模型(OK, Ⅲ)的福建省土壤有机碳空间分布图
Figure4.Spatial distribution maps of soil organic carbon of Fujian Province estimated by Random Forest (RF, Ⅰ), Support Vector Machine (SVM, Ⅱ) and Ordinary Kriging (OK, Ⅲ) models
下载: 全尺寸图片幻灯片
图5基于随机森林(RF)、支持向量机(SVM)和传统普通克里格模型(OK)预测的土壤有机碳(SOC)含量与代表环境因子的关系
Figure5.Relationships between representative environmental factors and soil organic carbon (SOC) contents predicted by Random Forest (RF), Support Vector Machine (SVM) and Ordinary Kriging (OK) models
下载: 全尺寸图片幻灯片表1影响土壤有机碳的环境变量
Table1.Environmental variables affecting soil organic carbon
| 变量类别 Variable category | 指标 Index |
| 植被指数 Vegetation index (VI) | 比值植被指数Ratio vegetation index (RVI) |
| 归一化植被指数 Normalized difference vegetation index (NDVI) | |
| 气候因子 Climate factor (CF) | 年最高气温Max annual temperature (Maxt) |
| 年最低气温Min annual temperature (Mint) | |
| 年降水量Mean annual rainfall (Rainfall) | |
| 地形因子 Topographical factor (TF) | 数字高程模型Digital evaluation model (DEM) |
| 地形湿度指数Topographic wetness index (TWI) | |
| 坡度Slope | |
| 坡向Aspect | |
| 平面曲率Plan | |
| 剖面曲率Profile | |
| 地形起伏度Relief of topography (Rel) |
下载: 导出CSV表2不同模型土壤有机碳(SOC)预测值与实测值对比
Table2.Comparison of predicted by different models and measured soil organic carbon (SOC) values
| 模型 Model | 最小值 Min (g?kg?1) | 平均值 Mean (g?kg?1) | 中值 Median (g?kg?1) | 最大值 Max (g?kg?1) | 标准偏差 SD (g?kg?1) | 变异系数 CV (%) |
| SOC | 1.874 | 15.343 | 14.849 | 35.269 | 5.843 | 38.08 |
| 传统普通克里格Ordinary Kriging | 3.428 | 15.339 | 16.059 | 26.428 | 3.531 | 23.02 |
| 支持向量机Support Vector Machine | 8.827 | 14.964 | 15.020 | 23.512 | 2.269 | 15.16 |
| 随机森林Random Forest | 3.569 | 15.333 | 15.110 | 29.907 | 4.065 | 26.51 |
下载: 导出CSV参考文献
| [1] | 李玲, 张少凯, 吴克宁, 等. 基于土壤系统分类的河南省土壤有机质时空变异[J]. 土壤学报, 2015, 52(5): 979-990 https://www.cnki.com.cn/Article/CJFDTOTAL-TRXB201505003.htm LI L, ZHANG S K, WU K N, et al. Analysis on spatio-temporal variability of soil organic matter in Henan Province based on soil taxonomy[J]. Acta Pedologica Sinica, 2015, 52(5): 979-990 https://www.cnki.com.cn/Article/CJFDTOTAL-TRXB201505003.htm |
| [2] | FROGBROOK Z L, OLIVER M A. Comparing the spatial predictions of soil organic matter determined by two laboratory methods[J]. Soil Use and Management, 2006, 17(4): 235-244 doi: 10.1111/j.1475-2743.2001.tb00033.x |
| [3] | 鲁如坤. 土壤农业化学分析方法[M]. 北京: 中国农业科技出版社, 2000 LU R K. Analysis Methods of Soil and Agricultural Chemistry[M]. Chinese Agricultural Science and Technology Press, 2000 |
| [4] | FORKUOR G, HOUNKPATIN O K L, WELP G, et al. High resolution mapping of soil properties using remote sensing variables in south-western Burkina Faso: a comparison of machine learning and multiple linear regression models[J]. PLoS One, 2017, 12(1): e0170478 doi: 10.1371/journal.pone.0170478 |
| [5] | POKHREL R M, KUWANO J, TACHIBANA S. A kriging method of interpolation used to map liquefaction potential over alluvial ground[J]. Engineering Geology, 2013, 152(1): 26-37 doi: 10.1016/j.enggeo.2012.10.003 |
| [6] | DAI F Q, ZHOU Q G, LV Z, et al. Spatial prediction of soil organic matter content integrating artificial neural network and ordinary kriging in Tibetan Plateau[J]. Ecological Indicators, 2014, 45: 184-194 doi: 10.1016/j.ecolind.2014.04.003 |
| [7] | 徐尚平, 陶澍, 曹军. 内蒙古土壤pH值、粘粒和有机质含量的空间结构特征[J]. 土壤通报, 2001, 32(4): 145-148 doi: 10.3321/j.issn:0564-3945.2001.04.001 XU S P, TAO S, CAO J. Spatial structure pattern of soil pH, clay and organic matter contents in the Inner Mongolia area[J]. Chinese Journal of Soil Science, 2001, 32(4): 145-148 doi: 10.3321/j.issn:0564-3945.2001.04.001 |
| [8] | LARK R M. Soil-landform relationships at within-field scales: an investigation using continuous classification[J]. Geoderma, 1999, 92(3/4): 141-165 http://www.sciencedirect.com/science/article/pii/S0016706199000282 |
| [9] | 王茵茵, 齐雁冰, 陈洋, 等. 基于多分辨率遥感数据与随机森林算法的土壤有机质预测研究[J]. 土壤学报, 2016, 53(2): 342-354 https://www.cnki.com.cn/Article/CJFDTOTAL-TRXB201602007.htm WANG Y Y, QI Y B, CHEN Y, et al. Prediction of soil organic matter based on multi-resolution remote sensing data and random forest algorithm[J]. Acta Pedologica Sinica, 2016, 53(2): 342-354 https://www.cnki.com.cn/Article/CJFDTOTAL-TRXB201602007.htm |
| [10] | 丁世飞, 齐丙娟, 谭红艳. 支持向量机理论与算法研究综述[J]. 电子科技大学学报, 2011, 40(1): 2-10 https://www.cnki.com.cn/Article/CJFDTOTAL-DKDX201101003.htm DING S F, QI B J, TAN H Y. An overview on theory and algorithm of support vector machines[J]. Journal of University of Electronic Science and Technology of China, 2011, 40(1): 2-10 https://www.cnki.com.cn/Article/CJFDTOTAL-DKDX201101003.htm |
| [11] | WIESMEIER M, BARTHOLD F, BLANK B, et al. Digital mapping of soil organic matter stocks using Random Forest modeling in a semi-arid steppe ecosystem[J]. Plant and Soil, 2011, 340(1/2): 7-24 doi: 10.1007/s11104-010-0425-z |
| [12] | SREENIVAS K, DADHWAL V K, KUMAR S, et al. Digital mapping of soil organic and inorganic carbon status in India[J]. Geoderma, 2016, 269: 160-173 doi: 10.1016/j.geoderma.2016.02.002 |
| [13] | EMADI M, TAGHIZADEH-MEHRJARDI R, CHERATI A, et al. Predicting and mapping of soil organic carbon using machine learning algorithms in northern Iran[J]. Remote Sensing, 2020, 12(14): 2234 doi: 10.3390/rs12142234 |
| [14] | WERE K, BUI D T, DICK ? B, et al. A comparative assessment of support vector regression, artificial neural networks, and random forests for predicting and mapping soil organic carbon stocks across an Afromontane landscape[J]. Ecological Indicators, 2015, 52: 394-403 doi: 10.1016/j.ecolind.2014.12.028 |
| [15] | SIEWERT M B. High-resolution digital mapping of soil organic carbon in permafrost terrain using machine learning: a case study in a sub-Arctic peatland environment[J]. Biogeosciences, 2018 http://smartsearch.nstl.gov.cn/paper_detail.html?id=58087218617a7963360da722381f4fed |
| [16] | 杨煜岑, 杨联安, 任丽, 等. 基于随机森林的农耕区土壤有机质空间分布预测[J]. 浙江农业学报, 2018, 30(7): 1211-1217 doi: 10.3969/j.issn.1004-1524.2018.07.15 YANG Y C, YANG L A, REN L, et al. Prediction for spatial distribution of soil organic matter based on random forest model in cultivated area[J]. Acta Agriculturae Zhejiangensis, 2018, 30(7): 1211-1217 doi: 10.3969/j.issn.1004-1524.2018.07.15 |
| [17] | 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(8): 141-146 https://www.cnki.com.cn/Article/CJFDTOTAL-GHZH201808021.htm REN L, YANG L A, WANG H, et al. Spatial prediction of soil organic matter in apple region based on random forest[J]. Journal of Arid Land Resources and Environment, 2018, 32(8): 141-146 https://www.cnki.com.cn/Article/CJFDTOTAL-GHZH201808021.htm |
| [18] | JOHNSON R W. An introduction to the bootstrap[J]. Teaching Statistics, 2001, 23(2): 49-54 doi: 10.1111/1467-9639.00050 |
| [19] | BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32 doi: 10.1023/A:1010933404324 |
| [20] | Boser B E. A training algorithm for optimal margin classifiers[J]. Proceedings of the Fifth Annual Workshop on Computational Learning Theory, 2008, 5: 144-152 http://portal.acm.org/citation.cfm?id=130401 |
| [21] | 李永娜. 基于支持向量机的回归预测综述[J]. 信息通信, 2014, 27(11): 32-33 doi: 10.3969/j.issn.1673-1131.2014.11.017 LI Y N. A review of regression prediction based on support vector machines[J]. Information & Communications, 2014, 27(11): 32-33 doi: 10.3969/j.issn.1673-1131.2014.11.017 |
| [22] | 张偲敏, 汪艳, 郭天太, 等. 支持向量回归机的参数择优算法[J]. 中国科技信息, 2015, (Z3): 26-27 https://www.cnki.com.cn/Article/CJFDTOTAL-XXJK2015Z3013.htm ZHANG S M, WANG Y, GUO T T, et al. Parameter selection algorithm of support vector regression machine[J]. China Science and Technology Information, 2015, (Z3): 26-27 https://www.cnki.com.cn/Article/CJFDTOTAL-XXJK2015Z3013.htm |
| [23] | 马冉. 流域尺度土壤特性空间分布及影响因素研究——以三峡库区草堂河流域为例[D]. 重庆: 西南大学, 2019 MA R. Spatial distribution and influential factors of soil properties at A watershed scale—A case study on Caotang River Basin in the Three Gorges Reservoir Area[D]. Chongqing: Southwest University, 2019 |
| [24] | GARCIA-PAUSAS J, CASALS P, CAMARERO L, et al. Soil organic carbon storage in mountain grasslands of the Pyrenees: effects of climate and topography[J]. Biogeochemistry, 2007, 82(3): 279-289 doi: 10.1007/s10533-007-9071-9 |
| [25] | MZA B, SA A, AJ C, et al. Digital mapping of soil properties using multiple machine learning in a semi-arid region, central Iran[J]. Geoderma, 2019, 338: 445-452 doi: 10.1016/j.geoderma.2018.09.006 |
| [26] | 张厚喜, 林丛, 程浩, 等. 武夷山不同海拔梯度毛竹林土壤有机碳特征及影响因素[J]. 土壤, 2019, 51(4): 821-828 https://www.cnki.com.cn/Article/CJFDTOTAL-TURA201904025.htm ZHANG H X, LIN C, CHENG H, et al. Variation of soil organic carbon content of moso bamboo forest along altitudinal gradient in Wuyi Mountain in China[J]. Soils, 2019, 51(4): 821-828 https://www.cnki.com.cn/Article/CJFDTOTAL-TURA201904025.htm |
| [27] | 钟兆全. 闽北毛竹林土壤有机碳含量特征及其影响因素[J]. 福建林业科技, 2017, 44(2): 36-42 https://www.cnki.com.cn/Article/CJFDTOTAL-FJLK201702008.htm ZHONG Z Q. Characteristics of soil organic carbon content and its influencing factors of Phyllostachys edulis forest in north of Fujian Province[J]. Journal of Fujian Forestry Science and Technology, 2017, 44(2): 36-42 https://www.cnki.com.cn/Article/CJFDTOTAL-FJLK201702008.htm |
| [28] | GUO P T, LI M F, LUO W, et al. Digital mapping of soil organic matter for rubber plantation at regional scale: an application of random forest plus residuals kriging approach[J]. Geoderma, 2015, 237/238: 49-59 http://smartsearch.nstl.gov.cn/paper_detail.html?id=a2653895c12bd4ce93ed2a95a1b12815 |
| [29] | 齐雁冰, 王茵茵, 陈洋, 等. 基于遥感与随机森林算法的陕西省土壤有机质空间预测[J]. 自然资源学报, 2017, 32(6): 1074-1086 https://www.cnki.com.cn/Article/CJFDTOTAL-ZRZX201706016.htm QI Y B, WANG Y Y, CHEN Y, et al. Soil organic matter prediction based on remote sensing data and random forest model in Shaanxi Province[J]. Journal of Natural Resources, 2017, 32(6): 1074-1086 https://www.cnki.com.cn/Article/CJFDTOTAL-ZRZX201706016.htm |
| [30] | 卢宏亮, 赵明松, 刘斌寅, 等. 基于随机森林模型的安徽省土壤属性空间分布预测[J]. 土壤, 2019, 51(3): 602-608 https://www.cnki.com.cn/Article/CJFDTOTAL-TURA201903025.htm LU H L, ZHAO M S, LIU B Y, et al. Spatial prediction of soil properties based on random forest model in Anhui Province[J]. Soils, 2019, 51(3): 602-608 https://www.cnki.com.cn/Article/CJFDTOTAL-TURA201903025.htm |
| [31] | BARITZ R, SEUFERT G, MONTANARELLA L, et al. Carbon concentrations and stocks in forest soils of Europe[J]. Forest Ecology and Management, 2010, 260(3): 262-277 doi: 10.1016/j.foreco.2010.03.025 |
| [32] | MEIER I C, LEUSCHNER C. Variation of soil and biomass carbon pools in beech forests across a precipitation gradient[J]. Global Change Biology, 2010, 16(3): 1035-1045 doi: 10.1111/j.1365-2486.2009.02074.x |
| [33] | WIESMEIER M, PRIETZEL J, BARTHOLD F, et al. Storage and drivers of organic carbon in forest soils of southeast Germany (Bavaria)—Implications for carbon sequestration[J]. Forest Ecology and Management, 2013, 295: 162-172 doi: 10.1016/j.foreco.2013.01.025 |
| [34] | SHI Y, BAUMANN F, MA Y, et al. Organic and inorganic carbon in the topsoil of the Mongolian and Tibetan grasslands: pattern, control and implications[J]. Biogeosciences, 2012, 9(6): 2287-2299 doi: 10.5194/bg-9-2287-2012 |
| [35] | 杨帆, 徐洋, 崔勇, 等. 近30年中国农田耕层土壤有机质含量变化[J]. 土壤学报, 2017, 54(5): 1047-1056 https://www.cnki.com.cn/Article/CJFDTOTAL-TRXB201705001.htm YANG F, XU Y, CUI Y, et al. Variation of soil organic matter content in croplands of China over the last three decades[J]. Acta Pedologica Sinica, 2017, 54(5): 1047-1056 https://www.cnki.com.cn/Article/CJFDTOTAL-TRXB201705001.htm |
| [36] | MCBRATNEY A B, ODEH I O A, BISHOP T F A, et al. An overview of pedometric techniques for use in soil survey[J]. Geoderma, 2000, 97(3/4): 293-327 http://www.sciencedirect.com/science/article/pii/S0016706100000434 |
| [37] | 尹萍. 气候变化及人类活动对中国土壤有机碳储量的影响[J]. 农业与技术, 2012, 32(9): 190 https://www.cnki.com.cn/Article/CJFDTOTAL-NYYS201209152.htm YIN P. Effects of climate change and human activities on soil organic carbon reserves in China[J]. Agriculture and Technology, 2012, 32(9): 190 https://www.cnki.com.cn/Article/CJFDTOTAL-NYYS201209152.htm |
| [38] | 田慎重, 宁堂原, 王瑜, 等. 不同耕作方式和秸秆还田对麦田土壤有机碳含量的影响[J]. 应用生态学报, 2010, 21(2): 373-378 https://www.cnki.com.cn/Article/CJFDTOTAL-YYSB201002019.htm TIAN S C, NING T Y, WANG Y, et al. Effects of different tillage methods and straw-returning on soil organic carbon content in a winter wheat field[J]. Chinese Journal of Applied Ecology, 2010, 21(2): 373-378 https://www.cnki.com.cn/Article/CJFDTOTAL-YYSB201002019.htm |
| [39] | BOU KHEIR R, GREVE M H, B?CHER P K, et al. Predictive mapping of soil organic carbon in wet cultivated lands using classification-tree based models: The case study of Denmark[J]. Journal of Environmental Management, 2010, 91(5): 1150-1160 http://www.cabdirect.org/abstracts/20103130760.html |
| [40] | WALIA N K, KALRA P, MEHROTRA D. Prediction of carbon stock available in forest using naive Bayes approach[C]//2016 Second International Conference on Computational Intelligence & Communication Technology (CICT). Ghaziabad, India. IEEE, 2016: 275-279 |
| [41] | LI Q Q, YUE T X, WANG C Q, et al. Spatially distributed modeling of soil organic matter across China: an application of artificial neural network approach[J]. CATENA, 2013, 104: 210-218 http://www.sciencedirect.com/science/article/pii/s0341816212002494 |
