基于Stacking模型集成算法的莲都区南方红豆杉潜在分布区
|
摘要
研究使用R软件中的CaretEnsemble和Caret程序包,并基于Stacking方法来实现模型集成,研究南方红豆杉Taxus chinensis var.mairei在浙江省丽水市莲都区的潜在分布区,并比较5种单一模型的模拟结果及其与集成模型的差异。结果表明:单一模型中极端梯度上升模型表现最好,其次是随机森林模型、支持向量机模型、朴素贝叶斯模型和分类回归树模型,集成模型模拟结果好于单一模型,其Kappa值达0.80,准确率达0.90。集成模型模拟结果显示:影响南方红豆杉分布的主要环境因子为海拔、归一化植被指数和年平均最少降雨量。南方红豆杉主要适宜生长在浙江省丽水市莲都区的山地丘陵地区,中部盆地及平原地区不适宜南方红豆杉的生长,其在莲都区的潜在分布区面积为5.01万hm~2。构建的集成模型在一定程度上提高了模型精度,使预测效果更优。图1表3参23
To study the potential distribution of Taxus chinensis var.mairei in Liandu District,the Caret and Caretensemble package in R were used to obtain an ensemble model based on the Stacking method.Then simulation results of five single models[the Extreme Gradient Boosting(XGBoost)Model,the Random Fores(RF)Model,the Support Vector Machine(SVM)Model,the Native Bayes(NB)Model,and the Classification and Regression Tree(CART)Model)]and their differences with the ensemble model were compared.Using40 presence-only points and generate the same number of pseudo-absences points for modeling,divide the dataset using 10-fold cross-validation and verify model accuracy using Kappa and overall accuracy.Results showed that XGBoost performed best as a single model followed by RF,SVM,NB,and CART.However,the ensemble model was better than all single models with its Kappa value reaching 0.80 and having an overall accuracy of 0.90.According to simulation results of the ensemble model,the main environmental factors affecting the distribution of T.chinensis var.mairei were altitude,normalized difference vegetation index(NDVI),and average annual minimum rainfall.T.chinensis var.mairei was suitable for growing in the mountainous and hilly areas of Liandu District but not in the Central Basin and plains area with the potential area for distribution in Liandu District being 5.01×10~4hm~2.Overall,the ensemble model used here improved the precision of the model somewhat making the prediction results better.[Ch,1 fig.3 tab.23 ref.]
To study the potential distribution of Taxus chinensis var.mairei in Liandu District,the Caret and Caretensemble package in R were used to obtain an ensemble model based on the Stacking method.Then simulation results of five single models[the Extreme Gradient Boosting(XGBoost)Model,the Random Fores(RF)Model,the Support Vector Machine(SVM)Model,the Native Bayes(NB)Model,and the Classification and Regression Tree(CART)Model)]and their differences with the ensemble model were compared.Using40 presence-only points and generate the same number of pseudo-absences points for modeling,divide the dataset using 10-fold cross-validation and verify model accuracy using Kappa and overall accuracy.Results showed that XGBoost performed best as a single model followed by RF,SVM,NB,and CART.However,the ensemble model was better than all single models with its Kappa value reaching 0.80 and having an overall accuracy of 0.90.According to simulation results of the ensemble model,the main environmental factors affecting the distribution of T.chinensis var.mairei were altitude,normalized difference vegetation index(NDVI),and average annual minimum rainfall.T.chinensis var.mairei was suitable for growing in the mountainous and hilly areas of Liandu District but not in the Central Basin and plains area with the potential area for distribution in Liandu District being 5.01×10~4hm~2.Overall,the ensemble model used here improved the precision of the model somewhat making the prediction results better.[Ch,1 fig.3 tab.23 ref.]
引文
[1]许仲林,彭焕华,彭守璋.物种分布模型的发展及评价方法[J].生态学报, 2015, 35(2):557-567.XU Zhonglin, PENG Huanhua, PENG Shouzhang. The development and evaluation of species distribution models[J].Acta Ecol Sin, 2015, 35(2):557-567.
[2]李国庆,刘长成,刘玉国,等.物种分布模型理论研究进展[J].生态学报, 2013, 33(16):4827-4835.LI Guoqing, LIU Changcheng, LIU Yuguo, et al. Advances in theoretical issues of species distribution models[J]. Acta Ecol Sin, 2013, 33(16):4827-4835.
[3]张雷,王琳琳,张旭东,等.随机森林算法基本思想及其在生态学中的应用:以云南松分布模拟为例[J].生态学报, 2014, 34(3):650-659.ZHANG Lei, WANG Linlin, ZHANG Xudong, et al. The basic principle of random forest and its applications in ecology:a case study of Pinus yunnanensis[J]. Acta Ecol Sin, 2014, 34(3):650-659.
[4]左闻韵,劳逆,耿玉英,等.预测物种潜在分布区:比较SVM与GARP[J].植物生态学报, 2007, 31(4):711-719.ZUO Wenyun, LAO Ni, GENG Yuying, et al. Predicting species, potential distribution:SVM compared with GARP[J]. Chin J Plant Ecol, 2007, 31(4):711-719.
[5]吴建国,周巧富.气候变化对6种荒漠动物分布的潜在影响[J].中国沙漠, 2011, 31(2):464-475.WU Jianguo, ZHOU Qiaofu. Potential effect of climate change on distribution of 6 desert animals in China[J]. J Desert Res, 2011, 31(2):464-475.
[6]李丽鹤,刘会玉,林振山,等.基于MAXENT和ZONATION的加拿大一枝黄花入侵重点监控区确定[J].生态学报, 2017, 37(9):3124-3132.LI Lihe, LIU Huiyu, LIN Zhenshan, et al. Identifying priority areas for monitoring the invasion of Solidago canadensis based on MAXENT and ZONATION[J]. Acta Ecol Sin, 2017, 37(9):3124-3132.
[7]ARA'UJO M B, NEW M. Ensemble forecasting of species distributions[J]. Trends Ecol Evol, 2007, 22(1):42-47.
[8]周志华.机器学习[M].北京:清华大学出版社, 2016.
[9]WOLPERT D H. Stacked Generalization[M]. New York:Springer, 2017:241-259.
[10]BREIMAN L. Random forests[J]. Mach Learning, 2001, 45(1):5-32.
[11]VAYSSIéRES M P, PLANT R E, ALLEN-DIAZ B H. Classification trees:an alternative non-parametric approach for predicting species distribution[J]. J Veg Sci, 2000, 11(5):679-694.
[12]WU Y. Statistical Learning Theory[J]. Ann Inst Stat Math, 2003, 55(2):371-389.
[13]钱永兰,吕厚荃,张艳红.基于ANUSPLIN软件的逐日气象要素插值方法应用与评估[J].气象与环境学报,2010, 26(2):7-15.QIAN Yonglan, L譈Houquan, ZHANG Yanhong, et al. Application and assessment of spatial interpolation method on daily meteorogical elements based on ANUSPLIN software[J]. J Meteorol Environ, 2010, 26(2):7-15.
[14]张雷.气候变化对中国主要造林树种/自然植被地理分布的影响预估及不确定性分析[D].北京:中国林业科学研究院, 2011.ZHANG Lei. Projectd Effects of Climate Change on Tree Species/Natural Vegetation Geographical Distribution in China and Uncertainty Analysis[D]. Beijing:Chinese Academy of Forestry, 2011.
[15]李艳芳,王钰,李济洪.几种交叉验证检验的可重复性[J].太原师范学院学报(自然科学版), 2013, 12(4):46-49.LI Yanfang, WANG Yu, LI Jihong. The replicabilityof several cross-validated tests[J]. J Taiyuan Norm Univ Nat Sci Ed, 2013, 12(4):46-49.
[16]COHEN J. A coefficient of agreement for nominal scales[J]. Educ Psychol Meas, 1960, 20(1):37-46.
[17]翟天庆,李欣海.用组合模型综合比较的方法分析气候变化对朱鹮潜在生境的影响[J].生态学报, 2012, 32(8):2361-2370.ZHAI Tianqing, LI Xinhai. Climate change induced potential range shift of the crested ibis based on ensemble models[J]. Acta Ecol Sin, 2012, 32(8):2361-2370.
[18]张雷,刘世荣,孙鹏森,等.气候变化对物种分布影响模拟中的不确定性组分分割与制图:以油松为例[J].生态学报, 2011, 31(19):5749-5761.ZHANG Lei, LIU Shirong, SUN Pengsen, et al. Partitioning and mapping the sources of variations in the ensemble forecasting of species distribution under climate change:a case study of Pinus tabulaeformis[J]. Acta Ecol Sin,2011, 31(19):5749-5761.
[19]MONSERU R A, LEEMANSB R. Comparing global vegetation maps with the Kappa statistic[J]. Ecol Modelling,1992, 62(4):275-293.
[20]PEARSON R G, DAWSON T P. Predicting the impacts of climate change on the distribution of species:are bioclimate envelope models useful?[J]. Global Ecol Biogeogr, 2010, 12(5):361-371.
[21]张殷波,高晨虹,秦浩.山西翅果油树的适生区预测及其对气候变化的响应[J].应用生态学报. 2018, 29(4):1156-1162.ZHANG Yinbo, GAO Chenhong, QIN Hao. Prediction of suitable distribution of Elaeagnus mollis in Shanxi Province,China and its response to climate change[J]. Chin J Appl Ecol, 2018, 29(4):1156-1162.
[22]HERNANDEZ P A, GRAHAM C H, MASTER L L, et al. The effect of sample size and species characteristics on performance of different species distribution modeling methods[J]. Ecography, 2006, 29(5):773-785.
[23]ENGLER R, GUISAN A, RECHSTEINER L. An improved approach for predicting the distribution of rare and endangered species from occurrence and pseudo-absence data[J]. J Appl Ecol, 2004, 41(2):263-274.
[2]李国庆,刘长成,刘玉国,等.物种分布模型理论研究进展[J].生态学报, 2013, 33(16):4827-4835.LI Guoqing, LIU Changcheng, LIU Yuguo, et al. Advances in theoretical issues of species distribution models[J]. Acta Ecol Sin, 2013, 33(16):4827-4835.
[3]张雷,王琳琳,张旭东,等.随机森林算法基本思想及其在生态学中的应用:以云南松分布模拟为例[J].生态学报, 2014, 34(3):650-659.ZHANG Lei, WANG Linlin, ZHANG Xudong, et al. The basic principle of random forest and its applications in ecology:a case study of Pinus yunnanensis[J]. Acta Ecol Sin, 2014, 34(3):650-659.
[4]左闻韵,劳逆,耿玉英,等.预测物种潜在分布区:比较SVM与GARP[J].植物生态学报, 2007, 31(4):711-719.ZUO Wenyun, LAO Ni, GENG Yuying, et al. Predicting species, potential distribution:SVM compared with GARP[J]. Chin J Plant Ecol, 2007, 31(4):711-719.
[5]吴建国,周巧富.气候变化对6种荒漠动物分布的潜在影响[J].中国沙漠, 2011, 31(2):464-475.WU Jianguo, ZHOU Qiaofu. Potential effect of climate change on distribution of 6 desert animals in China[J]. J Desert Res, 2011, 31(2):464-475.
[6]李丽鹤,刘会玉,林振山,等.基于MAXENT和ZONATION的加拿大一枝黄花入侵重点监控区确定[J].生态学报, 2017, 37(9):3124-3132.LI Lihe, LIU Huiyu, LIN Zhenshan, et al. Identifying priority areas for monitoring the invasion of Solidago canadensis based on MAXENT and ZONATION[J]. Acta Ecol Sin, 2017, 37(9):3124-3132.
[7]ARA'UJO M B, NEW M. Ensemble forecasting of species distributions[J]. Trends Ecol Evol, 2007, 22(1):42-47.
[8]周志华.机器学习[M].北京:清华大学出版社, 2016.
[9]WOLPERT D H. Stacked Generalization[M]. New York:Springer, 2017:241-259.
[10]BREIMAN L. Random forests[J]. Mach Learning, 2001, 45(1):5-32.
[11]VAYSSIéRES M P, PLANT R E, ALLEN-DIAZ B H. Classification trees:an alternative non-parametric approach for predicting species distribution[J]. J Veg Sci, 2000, 11(5):679-694.
[12]WU Y. Statistical Learning Theory[J]. Ann Inst Stat Math, 2003, 55(2):371-389.
[13]钱永兰,吕厚荃,张艳红.基于ANUSPLIN软件的逐日气象要素插值方法应用与评估[J].气象与环境学报,2010, 26(2):7-15.QIAN Yonglan, L譈Houquan, ZHANG Yanhong, et al. Application and assessment of spatial interpolation method on daily meteorogical elements based on ANUSPLIN software[J]. J Meteorol Environ, 2010, 26(2):7-15.
[14]张雷.气候变化对中国主要造林树种/自然植被地理分布的影响预估及不确定性分析[D].北京:中国林业科学研究院, 2011.ZHANG Lei. Projectd Effects of Climate Change on Tree Species/Natural Vegetation Geographical Distribution in China and Uncertainty Analysis[D]. Beijing:Chinese Academy of Forestry, 2011.
[15]李艳芳,王钰,李济洪.几种交叉验证检验的可重复性[J].太原师范学院学报(自然科学版), 2013, 12(4):46-49.LI Yanfang, WANG Yu, LI Jihong. The replicabilityof several cross-validated tests[J]. J Taiyuan Norm Univ Nat Sci Ed, 2013, 12(4):46-49.
[16]COHEN J. A coefficient of agreement for nominal scales[J]. Educ Psychol Meas, 1960, 20(1):37-46.
[17]翟天庆,李欣海.用组合模型综合比较的方法分析气候变化对朱鹮潜在生境的影响[J].生态学报, 2012, 32(8):2361-2370.ZHAI Tianqing, LI Xinhai. Climate change induced potential range shift of the crested ibis based on ensemble models[J]. Acta Ecol Sin, 2012, 32(8):2361-2370.
[18]张雷,刘世荣,孙鹏森,等.气候变化对物种分布影响模拟中的不确定性组分分割与制图:以油松为例[J].生态学报, 2011, 31(19):5749-5761.ZHANG Lei, LIU Shirong, SUN Pengsen, et al. Partitioning and mapping the sources of variations in the ensemble forecasting of species distribution under climate change:a case study of Pinus tabulaeformis[J]. Acta Ecol Sin,2011, 31(19):5749-5761.
[19]MONSERU R A, LEEMANSB R. Comparing global vegetation maps with the Kappa statistic[J]. Ecol Modelling,1992, 62(4):275-293.
[20]PEARSON R G, DAWSON T P. Predicting the impacts of climate change on the distribution of species:are bioclimate envelope models useful?[J]. Global Ecol Biogeogr, 2010, 12(5):361-371.
[21]张殷波,高晨虹,秦浩.山西翅果油树的适生区预测及其对气候变化的响应[J].应用生态学报. 2018, 29(4):1156-1162.ZHANG Yinbo, GAO Chenhong, QIN Hao. Prediction of suitable distribution of Elaeagnus mollis in Shanxi Province,China and its response to climate change[J]. Chin J Appl Ecol, 2018, 29(4):1156-1162.
[22]HERNANDEZ P A, GRAHAM C H, MASTER L L, et al. The effect of sample size and species characteristics on performance of different species distribution modeling methods[J]. Ecography, 2006, 29(5):773-785.
[23]ENGLER R, GUISAN A, RECHSTEINER L. An improved approach for predicting the distribution of rare and endangered species from occurrence and pseudo-absence data[J]. J Appl Ecol, 2004, 41(2):263-274.