基于LSSVM-CA模型的复杂土地利用变化模拟——以鄂州市为例

doi:10.18402/resci.2016.08.11

摘要/Abstract

摘要：

本文探索了最小二乘支持向量机（Least Squares Support Vector Machine ,LSSVM）获取元胞转换规则的可行性,并应用于复杂土地利用变化模拟预测。以湖北省鄂州市为研究区,以1991-2004年土地利用变化数据作为模型训练数据,运用改进的ROC分析方法对比分析了LSSVM和逻辑回归方法获取的元胞转换规则,在此基础上运用LSSVM-CA模型模拟了研究区2013年的土地利用情景,并对2020年和2030年土地利用情景进行预测。研究结果表明：①LSSVM对数量较大、变化过程较复杂土地利用类型的空间分布模拟效果更好,如耕地、建设用地、养殖水体和其他用地;②与2013年实际土地利用情景相比,LSSVM-CA模拟结果总体精度为0.80,Kappa系数为0.73,处于较高一致性水平,优于逻辑回归-CA模型结果;③未来,鄂州市主城区、城西新区、“葛华新城”、“红莲湖新城”以及南部的花湖开发区、沼山镇、太和镇建设用地需求较大,将占用大量耕地,东部和南部低丘岗地区的耕地将大量转变为林地。研究结论为LSSVM方法可用于获取元胞转换规则进行复杂土地利用变化模拟,并能取得较好的效果,模拟结果可为研究区土地规划、耕地和生态环境保护等提供决策参考。

关键词: 元胞自动机, 最小二乘支持向量机, 土地利用, 鄂州市

Abstract:

Here we validate the feasibility of gaining transformation rules for cellular automata modeling using Least Squares Support Vector Machine （LSSVM）methods,and apply it to the simulation and prediction of complex land use change. We took land use change information from 1991 to 2004 for Ezhou,Hubei as training data. With the application of LSSVM and Logistic Regression methods,we obtained two sets of transformation rules respectively. Comparison and analysis were made between them with the help of an enhanced ROC method. We then simulated land use scenarios for 2013 using a united LSSVM-CA model and predictions to 2020 and 2030. The results show that the LSSVM method has a better output than the Logistic Regression method in simulating the spatial distribution of land use types that have a large area and a complex change process,such as cultivated land,construction land,aquaculture land and other land. Compared with the actual land use scenarios in 2013,the precision of simulation output derived from the LSSVM-CA model is 0.80. The Kappa coefficient is 0.73,which is at a high level of consistency and better than the simulation results gained from Logistic-CA model. In the future,some districts will have a large demand of construction land and a large number of cultivated land will be occupied. Farmland in the eastern and southern low-hilly area will transform into forest land in a big way. We conclude that the LSSVM method can be used to obtain transformation rules in a cellular automata model and can gain a good result in simulating land use changes. The simulation results provide meaningful decision-making reference points for the study area in land planning,farmland protection and ecological environmental protection.

Key words: cellular automata, least squares support vector machine, land use, Ezhou City

杨建新, 龚健, 李江风. 基于LSSVM-CA模型的复杂土地利用变化模拟——以鄂州市为例[J]. 资源科学, 2016, 38(8): 1525-1537.

YANG Jianxin,GONG Jian,LI Jiangfeng. Complex land use changes simulation in Ezhou City using cellular automata based on least squares support vector machine[J]. Resources Science, 2016, 38(8): 1525-1537.

图/表 11

图1

图2

图3

图4

表1

图5

表2

图6

表3

表4

图7

参考文献 33

[1]	唐华俊,吴文斌,杨鹏,等. 土地利用/土地覆被变化(LUCC)模型研究进展[J]. 地理学报,2009,64(4):456-468.
	[Tang H J,Wu W B,Yang P,et al.Recent progresses of land use and land cover change (LUCC)models[J]. Acta Geographica Sinica,2009,64(4):456-468.]
[2]	张丁轩,付梅臣,陶金,等. 基于Clue-S模型的矿业城市土地利用变化情景模拟[J]. 农业工程学报,2013,(12):246-256.
	[Zhang D X,Fu M C,Tao J,et al.Scenario simulation of land use change in mining city based on LUE-S model[J]. Transactions of the Chinese Society of Agricultural Engineering,2013,(12):246-256.]
[3]	Xu Y,Mcnamara P,Wu Y,et al.An econometric analysis of changes in arable land utilization using multinomial logit model in Pinggu District,Beijing,China[J]. J Environ Manage,2013,128(20):324-334.
[4]	Mozumder C,Tripathi N K.Geospatial scenario based modelling of urban and agricultural intrusions in Ramsar wetland Deepor Beel in Northeast India using a multi-layer perceptron neural network[J]. International Journal of Applied Earth Observation and Geoinformation,2014,32(1):92-104.
[5]	Rasmussen L V,Rasmussen K,Reenberg A,et al.A system dynamics approach to land use changes in agro-pastoral systems on the desert margins of Sahel[J]. Agricultural Systems,2012,107(1):56-64.
[6]	Ralha C G,Abreu C G, Coelho C G C, et al. A multi-agent model system for land-use change simulation[J]. Environmental Modelling & Software,2013,42(4):30-46.
[7]	Zhang P,Liu Y,Pan Y,et al.Land use pattern optimization based on clue-s and swat models for agricultural non-point source pollution control[J]. Mathematical and Computer Modelling,2013,58(3-4):588-595.
[8]	Mitsova D,Shuster W,Wang X.A cellular automata model of land cover change to integrate urban growth with open space conservation[J]. Landscape and Urban Planning,2011,99(2):141-153.
[9]	高志强,易维. 基于CLUE-S和Dinamica EGO模型的土地利用变化及驱动力分析[J]. 农业工程学报,2012,28(16):208-216.
	[Gao Z Q,Yi W.Land use change in China and analysis of its driving forces using CLUE-S and Dinamica EGO model[J]. Transactions of the Chinese Society of Agricultural Engineering,2012,28(16):208-216.]
[10]	Arsanjani J J,Helbich M,Kainz W,et al.Integration of logistic regression,Markov chain and cellular automata models to simulate urban expansion[J]. International Journal of Applied Earth Observation and Geoinformation,2013,21(4):265-275.
[11]	Verstegen J A,Karssenberg D,Hilst F V D, et al. Identifying a land use change cellular automaton by Bayesian data assimilation[J]. Environmental Modelling & Software,2014,53(1):121-136.
[12]	Rienow A,Goetzke R.Supporting sleuth-enhancing a cellular automaton with support vector machines for urban growth modeling[J]. Computers,Environment and Urban Systems,2015,49:66-81.
[13]	Guan Q L W,Clarke K C. An artificial-neural-network-based,constrained CA model for simulating urban growth[J]. Carto-graphy & Geographic Information Science,2005,32(4):369-380.
[14]	张亦汉,黎夏,刘小平,等. 基于数据同化的元胞自动机[J]. 遥感学报,2011,15(3):475-491.
	[Zhang Y H,Li X,Liu X P,et al.CA model based on data assimilation[J]. Journal of Remote Sensing,2011,15(3):475-491.]
[15]	张亦汉,黎夏,刘小平,等. 耦合遥感观测和元胞自动机的城市扩张模拟[J]. 遥感学报,2013,17(4):872-886.
	[Zhang Y H,Li X,Liu X P,et al.Urban expansion simulation by coupling remote sensing observations and cellular automata[J]. Journal of Remote Sensing,2013,17(4):872-886.]
[16]	黎夏. 协同空间模拟与优化及其在快速城市化地区的应用[J]. 地球信息科学学报,2013,15(3):321-327.
	[Li X.Collaborative spatio-simulation and optimization and its application in fast growing regions[J]. Journal of Geo-Information Science,2013,15(3):321-327.]
[17]	周竹,李小昱,李培武,等. 基于GA-LSSVM和近红外傅里叶变换的霉变板栗识别[J]. 农业工程学报,2011,27(3):331-335.
	[Zhou Z,Li X Y,Li P W,et al.Near-infrared spectral detection of moldy chestnut based on GA-LSSVM and FFT[J]. Transactions of the Chinese Society of Agricultural Engineering,2011,27(3):331-335.]
[18]	Mitra V,Wang C,Banerjee S.Text classification:A least square support vector machine approach[J]. Applied Soft Computing Journal,2007,7(3):908-914.
[19]	Ismail S,Shabri A,Samsudin R.A hybrid model of self-organi-zing maps (SOM)and least square support vector machine (LSSVM)for time-series forecasting[J]. Expert Systems with Appli-cations,2011,38(8):10574-10578.
[20]	Xu H,Guohua C.An intelligent fault identification method of rolling bearings based on LSSVM optimized by improved PSO[J]. Mechanical Systems and Signal Processing,2013,35(1-2):167-175.
[21]	曹雪,罗平,李满春,等. 基于扩展CA模型的土地利用变化时空模拟研究-以深圳市为例[J]. 资源科学,2011,33(1):127-133.
	[Cao Z,Luo P,Li M C,et al.Spatio-temporal simulation of land use change based on an extended CA model:A case study of Shenzhen City,China[J]. Resources Science,2011,33(1):127-133.]
[22]	Wang H,Hu D.Comparison of SVM and LS-SVM for Regression[R]. Beijing:2005 International Conference on Neural Networks and Brain,2005.
[23]	Kirkpatrick S,Gelatt J C D,Vecchi M P. Optimization by simulated annealing[J]. Science (New York,N.Y.),1983,220(4598):671-680.
[24]	Xavier-De-Souza S,Suykens J A K,Vandewalle J,et al. Coupled simulated annealing[J]. IEEE Transactions On Systems,Man,and Cybernetics,Part B (Cybernetics),2010,34(2):320-335.
[25]	王霞,王占岐,金贵,等. 基于核函数支持向量回归机的耕地面积预测[J]. 农业工程学报,2014,30(4):204-211.
	[Wang X,Wang Z Q,Jin G,et al.Land reserve prediction using different kernel based support vector regression[J]. Transactions of the Chinese Society of Agricultural Engineering,2014,30(4):204-211.]
[26]	Huang Y,Mccoll W F.An Improved Simplex Method for Function Minimization[C]. Beijing:IEEE International Conference,1996.
[27]	Wang P C,Shoup T E.Parameter sensitivity study of the Nelder-Mead Simplex Method[J]. Advances in Engineering Software,2011,42(7):529-533.
[28]	黎夏,叶嘉安,廖其芳. 利用案例推理(CBR)方法对雷达图像进行土地利用分类[J]. 遥感学报,2004,8(3):246-253.
	[Li X,Ye J A,Liao Q F. Case-Based Reasoning (CBR)for land use classification using radar images[J]. Journal of Remote Sensing,2004,8(3):246-253.]
[29]	Madevska-Bogdanova A,Nikolik D,Curfs L.Probabilistic SVM outputs for pattern recognition using analytical geometry[J]. Neurocomputing,2004,62(1):293-303.
[30]	黎夏,叶嘉安. 基于神经网络的元胞自动机及模拟复杂土地利用系统[J]. 地理研究,2005,24(1):19-27.
	[Li X,Ye J A.Cellular automata for simulating complex land use systems using neural networks[J]. Geographical Research,2005,24(1):19-27.]
[31]	Sangermano F,Eastman J R,Zhu H.Similarity weighted instance-based learning for the generation of transition potentials in land use change modeling[J]. Transactions in GIS,2010,14(5):569-580.
[32]	Pontius R G,Parmentier B.Recommendations for using the relative operating characteristic (ROC)[J]. Landscape Ecology,2014,29(3):367-382.
[33]	肖明,吴季秋,陈秋波,等. 基于CA-Markov模型的昌化江流域土地利用动态变化[J]. 农业工程学报,2012,28(10):231-238.
	[Xiao M,Wu J Q,Chen Q B,et al.Dynamic change of land use in Changhua downstream watershed based on CA-Markov model[J]. Transactions of the Chinese Society of Agricultural Engineering,2012,28(10):231-238.]

变化类型	耕地-养殖水面	耕地-建设用地	耕地-林地	其他用地-林地	水体-养殖水面	耕地-其他土地	…
γ	6.304	3.469	4.100	5.000	4.002	5.826	…
σ²	0.025	0.216	0.015	0.021	0.050	0.003	…

		耕地	林地	建设用地	水体	养殖水面	其他用地
尤拉Q系数	LSSVM	0.63	0.74	0.85	0.54	0.91	0.50
尤拉Q系数	逻辑回归方法	-0.21	0.78	-0.04	0.62	0.34	-0.13

景观格局指数		实际情景	模拟情景	指数差值	取值范围
形状特性	平均斑块分维数（FRAC_MN）	1.06	1.05	0.01	[1,2]
	平均斑块相关外接圆指数（CIRCLE_MN）	0.53	0.47	0.06	[0,1]
	平均斑块聚集指数（CONTIG_MN）	0.40	0.44	0.04	[0,1]
	周长面积分形维数（PAFRAC）	1.39	1.25	0.14	[1,2]
聚散特性	蔓延度指数（CONTAG）	42.32	48.40	6.08	[1,100]
	散布于并列指数（IJI）	74.76	75.43	0.67	[1,100]
	斑块结合度指数（COHESION）	99.12	99.11	0.01	[1,100]
	聚集度指数（AI）	84.71	91.49	6.78	[1,100]

		实际面积
		耕地	林地	建设用地	水体	养殖水面	其他用地	精度	总精度
模拟面积	耕地	53 225.28	1 392.75	2 641.14	326.16	2 581.65	1 501.65	86	80
	林地	1 493.91	11 465.01	523.71	54.27	99.63	903.42	79
	建设用地	2 666.61	610.56	22 301.82	253.62	1 004.31	659.25	81
	水体	278.64	39.06	328.50	2 190.24	807.57	133.56	58
	养殖水面	2 669.40	67.14	996.39	823.41	16 728.39	336.33	77
	其他用地	1 344.06	961.47	711.45	119.52	394.02	3 821.22	52