Spatial variation in drivers of karst rocky desertification based on geographically weighted regression model
XUErqi Key Laboratory of Land Surface Pattern and Simulation,Institute of Geographic Sciences and Natural Resources Research,Chinese Academy of Sciences,Beijing 100101,China 收稿日期:2017-08-23 修回日期:2017-09-14 网络出版日期:2017-10-20 版权声明:2017《资源科学》编辑部《资源科学》编辑部 基金资助:国家自然科学基金项目(41601095)国家重点基础研究发展计划项目(973 计划)(2015CB452702) 作者简介: -->作者简介:许尔琪,男,广东汕头人,博士,主要从事土地利用及空间格局、生态环境效应研究。E-mail:xueq@igsnrr.ac.cn
关键词:石漠化;影响因子;地理加权回归模型;空间分异;黔桂喀斯特山地 Abstract Analysis and identification of key drivers of karst rocky desertification (KRD)can contribute to effective management and restoration. Ignoring heterogeneity leads to statistical bias and influences the specificity of desertification control planning. Taking Guizhou and Guangxi Karst Mountainous Regions as the study area,this paper chose twelve factors,including socioeconomic,spatial distance,topography,climate,soil,lithology and land use,as drivers of KRD. Geographically Weighted Regression modeling (GWR)and embedding spatial factors to the traditional ordinary linear regression (OLS)model were used to analyze the spatial distribution of influence on KRD. An obvious spatial agglomeration of KRD in the study area according to Moran’s I at the significant level of 99% was found. Coefficients of determination (R2)of GWR were much higher for OLS (0.508 vs. 0.156),indicating a much better fit for the GWR model. Coefficients of GWR between twelve drivers and an obvious spatial distribution of value magnitudes,negative or positive effects and combined types. The specifics of the karst background create a fragile and vulnerable environment that is susceptible to human activities. Meanwhile,intense human activities lead to a sharp change in KRD status,which included the predatory sabotage for casing the severe KRD and the KRD restoration projects for reversing KRD to no KRD. The regression coefficients of the twelve drivers and their linear combined characteristics showed different spatial distributions based on GWR modeling. Using the GWR model revealed the spatial discrimination of the effects of driving forces on KRD and identified key drivers of KRD at the local area,revealing the spatial distribution of the joint effect of different driving forces helping provide a scientific reference to differential KRD control at small watershed scales.
本文以位于贵州和广西的黔桂喀斯特山地为研究区[29],该区覆盖峰林、峰丛和洼地等主要喀斯特地貌类型,面积为21.41万km2,约位于22º8'54"N-28º12'27"N,104º18'27"E-110º20'40"E之间(图1)。地势为西北高,东南低,山地多,平地少。该区属亚热带季风湿润气候,年均降雨约为800~1900mm,年均温度为9℃~23℃。作为中国石漠化分布的最严重区域,各类型石漠化皆有分布,是石漠化与影响因子关系研究的典型区。 显示原图|下载原图ZIP|生成PPT 图1黔桂喀斯特山地范围 -->Figure 1Boundary of Guizhou and Guangxi Karst Mountainous Region -->
2.2 研究方法与数据来源
2.2.1 石漠化分级 因石漠化分级标准较多,尚无统一定论,本文采用国家林业局官方发布的最新一期石漠化空间分布数据(2011年)[30],并对部分结果进行人工修正,该数据主要依据基岩裸露、植被覆盖和土壤覆盖程度进行石漠化分级[31](表1),将石漠化强度等级分为无石漠化、潜在石漠化、轻度石漠化、中度石漠化、强度石漠化和极强度石漠化等6级,分别用1~6表示。 2.2.2 影响因子选取 对比以往研究对石漠化影响因子贡献程度的结果[14,17-20],考虑数据的可获得性,本文共选取12个影响因子(图2),包括社会经济、空间距离、地形、气候、土壤和土地利用等因素,其数据来源及处理过程见表2(见第1978页)。以1km×1km栅格单元对石漠化和影响因子进行空间重采样,共计21.41万个单元,进行后续分析。 显示原图|下载原图ZIP|生成PPT 图2黔桂喀斯特山地石漠化影响因子空间差异 -->Figure 2Driving forces of karst rocky desertification in Guizhou and Guangxi Karst Mountainous Region -->
Table 2 表2 表2黔桂喀斯特山地石漠化影响因子及其数据来源、处理过程 Table 2Driving forces of karst rocky desertification and data sources and processing in Guizhou and Guangxi Karst Mountainous Region
黔桂喀斯特山地2011年在各石漠化等级均有所分布(图3),其中,无石漠化和潜在石漠化面积最大,分别为13.59万km2和3.35万km2,各占总面积的63.48%和15.66%;其余类型石漠化面积较低,重度、中度和轻度石漠化分列三到五位,比例分别为8.23%、6.73%和4.84%,极重度石漠化面积最少,比重仅为1.06%。Moran’s I指数为0.56,标准化统计量Z(I)值为514.97,大于正态函数在99%显著水平时的临界值2.58,表明石漠化存在着显著空间集聚现象。 显示原图|下载原图ZIP|生成PPT 图32011年黔桂喀斯特山地石漠化空间分布 -->Figure 3Spatial distribution of karst rocky desertification in Guizhou and Guangxi Karst Mountainous Region in 2011 -->
GWR模型与OLS模型相比,其拟合效果显著提高(表4),GWR的拟合优度R2(0.508)明显高于OLS中的R2(0.156),GWR的AIC(42 551.165)远小于比OLS的AIC(564 901.811)。同时,GWR模型的标准化残差值的范围在[-3.37,3.65],其中约99.76%的范围在[-2.58,2.58],表明标准化残差值在99%的显著性水平下是随机分布。 Table 4 表4 表4GWR 模型参数估计及检验结果 Table 4Parameter estimation and test results of the GWR model
模型参数
OLS模型
GWR模型
Bandwidth
-
6 928.466
Residual Squares
-
9 823.062
Effective Number
-
1 678.221
Sigma
-
0.779
AIC
564 901.811
42 551.165
R2
0.156
0.554
Adjusted R2
0.156
0.508
新窗口打开 GWR模型对自变量在每个空间位置的回归系数进行局部统计,分别得到回归系数的统计特征(表5)以及直方图累积分布(图4)。人口密度(x1)、到主要公路的距离(x4)、海拔(x5)、坡度(x6)、年均气温(x8)、植被指数(x9)、土壤砂粘百分比(x10)、碳酸盐比例分级(x11)和土地利用分级(x12)等因素对石漠化影响程度的波动区间较小,GDP(x2)、到主要铁路的距离(x3)、年均降水(x7)等因素对石漠化的影响程度波动区间较大。因子回归系数均呈现出正负变动幅度,主要集中在[-1,1]范围内,围绕在0上下浮动。除人口密度、GDP、海拔、坡度和植被指数等5个因子有一定偏态分布,其余因子呈较明显正态分布。GDP、植被指数、碳酸盐岩百分比等因子回归系数小于0的比例较大,其他因素与石漠化的回归系数大于0的比重较大。 Table 5 表5 表5黔桂喀斯特山地石漠化影响因子的GWR模型回归系数的描述性统计 Table 5Statistical characteristics of GWR regression coefficients for driving forces of karst rocky desertification in Guizhou and Guangxi Karst Mountainous Region
影响因子
平均值
标准差
最大值
最小值
下四分位值
中位值
上四分位值
x1
-0.064
0.191
1.230
-1.734
-0.148
0.033
0.089
x2
-0.018
0.596
14.950
-12.900
-0.078
-0.012
0.018
x3
0.022
1.235
14.126
-18.324
-0.489
0.001
0.534
x4
0.024
0.197
1.049
-1.112
-0.068
0.001
0.115
x5
0.179
0.585
5.539
-2.585
-0.020
0.112
0.557
x6
0.082
0.106
0.835
-0.351
0.007
0.058
0.135
x7
-0.028
1.310
22.305
-18.087
-0.781
0.007
0.817
x8
0.040
0.565
3.670
-6.051
-0.187
0.002
0.233
x9
-0.136
0.134
0.349
-0.850
-0.098
-0.053
0.011
x10
0.008
0.163
1.105
-1.962
-0.062
0.003
0.082
x11
-0.098
0.193
1.417
-1.727
-0.183
-0.049
0.002
x12
0.010
0.080
0.527
-0.471
-0.027
0.001
0.044
新窗口打开 显示原图|下载原图ZIP|生成PPT 图4黔桂喀斯特山地石漠化影响因子回归系数统计 -->Figure 4Histogram statistics of regression coefficients of GWR for different driving forces in Guizhou and Guangxi Karst Mountainous Region -->
3.3 石漠化影响因子的空间分异规律
利用GWR模型,分析石漠化影响因子回归系数的空间分布(图5),结果表明,每一因子回归系数呈现不同的正负和数值的空间变化,且各因子系数的空间分布各不相同。一方面,不合理人类活动可加剧石漠化,有效保护的行为和工程则可治理石漠化,这使得量化人类活动强弱的驱动因子与石漠化在不同空间位置可能正负效应和影响大小的差异。另一方面,因子多以联合作用影响石漠化的分布,由于某个关键因子在局部的影响程度以及因子间的相关性,使得其他影响因子在回归方程中系数发生变化。因此,影响因子与石漠化关系,在空间上呈现出影响程度、正负效应和因子组合的显著差异。 显示原图|下载原图ZIP|生成PPT 图5黔桂喀斯特山地石漠化影响因子回归系数 -->Figure 5Spatial distribution of GWR regression coefficients for driving forces of karst rocky desertification in Guizhou and Guangxi Karst Mountainous Region -->
黔桂喀斯特山地石漠化Moran’s I指数为0.56,标准化统计量Z(I)值为514.97,大于正态函数在99%显著水平的数值,存在明显的空间集聚分布。GWR模型分析加入空间要素,克服了经典回归模型统计条件假设的缺陷,其R2(0.508)明显比OLS模型的R2(0.156)高,拟合效果显著提高。 GWR计算的各石漠化影响因子回归系数在正负区间皆有分布,主要集中在[-1,1]范围内,呈现不同空间分布:人口呈“中间低南北高”,GDP为“西低中东高”,植被指数和岩性为“西低东高”,海拔和坡度则是“东南高西北低”,土地利用分级呈“桂中高、其余低”,到主要铁路的距离、到主要公路的距离、年均降水、年均气温和土壤质地对石漠化的影响空间分布较为零散。 喀斯特地区地形破碎,岩性、土壤和植被等类型多样,加之复杂人类活动,对石漠化分布产生显著影响,体现出单因子的正负效应、多种因子的组合类型和贡献程度的空间明显差异,如社会经济发展对石漠化的正负效应,人口与海拔、坡度的组合关系,降雨与植被对石漠化的交互作用等。人类活动叠加在喀斯特特殊的岩性、土壤和植被构成上,更容易发生石漠化,同时,在局部范围,高强度破坏和资源掠夺可导致严重石漠化发生,高投入石漠化治理措施又能够有效治理石漠化。 因此,石漠化治理工程应考虑不同影响因子组合对石漠化的影响及其空间差异。应用GWR模型,分析各影响因子对石漠化影响程度和正负效应,揭示影响因子在不同空间位置的组合,辨析其中影响石漠化的关键影响因子,从而为差别化小流域治理提供科学依据,以有效改善和控制石漠化。 The authors have declared that no competing interests exist.
[Li YB,Luo GJ,Bai XY,et al.The correl-ations among arable land,settlement and karst rocky desertifi-cationcases study based on typical peak-cluster depression [J]. ,2014,34(9):2195-2207.] [本文引用: 1]
[Yuan DX.Global view on karst rock desertification and integrating control measures and experiences of China [J]. ,2008,25(9):19-25.] [本文引用: 2]
[10]
Wang SJ,Li RL,Sun CX,et al.How types of carbonate rock assemblages constrain the distribution of karst rocky desertified land in Guizhou Province,PR China:phenomena and mechani-sms [J]. ,2004,15(2):123-131. [本文引用: 2]
[Su WC.Cont-rolling model for rocky desertification of karst mountainous region and its preventing strategy in Southwest,China [J]. ,2002,16(2):29-32.]
[13]
YanX,Cai YL.Multi-scale anthropogenic driving forces of karst rocky desertification in Southwest China [J]. ,2015,26(2):193-200.
[14]
XuE,ZhangH.Characterization and interaction of driving factors in karst rocky desertification:a case study from Chang-shun,China [J]. ,2014,5(2):1329-1340. [本文引用: 4]
[15]
Bai XY,Wang SJ,Xiong KN.Assessing spatial-temporal evolution processes of karst rocky desertification land:indi-cations for restoration strategies [J]. ,2011,24(1):47-56. [本文引用: 1]
[Wang SJ.The most serious eco-geologically environmental problem in southwestern China-karst rocky desertification [J]. ,2003,22(2):120-126.] [本文引用: 1]
[17]
JiangY,LiL,GrovesC,et al.Relationships between rocky desertification and spatial pattern of land use in typical karst area,Southwest China [J]. ,2009,59(4):881-890. [本文引用: 2]
[18]
Liu YS,Wang JY,Deng XZ.Rocky land desertification and its driving forces in the karst areas of rural Guangxi,Southwest China [J]. ,2008,5(4):350-357. [本文引用: 1]
[19]
Yang QQ,Wang KL,Zhang CH,et al.Spatio-temporal evolu-tion of rocky desertification and its driving forces in karst areas of Northwestern Guangxi,China [J]. ,2011,64(2):383-393. [本文引用: 1]
[20]
LiY,XieJ,LuoG,et al.The evolution of a karst rocky desertification land ecosystem and its driving forces in the Houzhaihe area,China [J]. ,2015,5(10):501-512. [本文引用: 2]
[Zhang XB,Wang SJ,Bai XY,et al.Relationships between the spatial distribution of karst land desertification and geomorphology,lithology,precipitation,and population density in Guizhou province [J]. ,2013,41(1):1-6.] [本文引用: 1]
[25]
Qi XK,Wang KL,Zhang CH.Effectiveness of ecological restoration projects in a karst region of southwest China assessed using vegetation succession mapping [J]. ,2013,54:245-253. [本文引用: 1]
[26]
ZhangM,WangK,LiuH,et al.How ecological restoration alters ecosystem services:an analysis of vegetation carbon sequestration in the karst area of northwest Guangxi,China [J]. ,2015,74(6):5307-5317.
[27]
TongX,WangK,YueY,et al.Quantifying the effectiveness of ecological restoration projects on long-term vegetation dynamics in the karst regions of Southwest China [J]. ,2017,54:105-113.
[Li YB,Luo GJ,Shao JA,et al.The evolving models of rural settlements and population distribution in karst mountains [J]. ,2013,67(12):1666-1674.] [本文引用: 1]
[State Forestry Administr-ation of People 's Republic of China. Bulletin of China 's Karst Rock Desertification [EB/OL].(2012-06-18)[2017-08-23]. .]URL [本文引用: 1]
[31]
Li YB,Shao JA,YangH, et al. The relations between land useand karst rocky desertification in a typical karst area,China [J]. ,57(3):621-627. [本文引用: 1]
[Data Center for Resources and Environmental Sciences,Chinese Academy of Sciences. Population Density Dataset (2010) [EB/OL]. (2014-06-01)[2017-03-10]. .]URL [本文引用: 1]
[Data Center for Resources and Environmental Sciences,Chinese Academy of Sciences. GDP Dataset (2010) [EB/OL]. (2014-06-01)[2017-03-10]. .]URL [本文引用: 1]
[International Scientific & Technical Data Mirror Site,Computer Network Information Center,Chinese Academy of Sciences. SRTM DEM Dataset [EB/OL]. (2017-01-01)[2017-03-10]. .]URL [本文引用: 2]
[Cold and Arid Regions Science Data Center at Lanzhou. Soil Particle-size Distribution Dataset (2011) [EB/OL]. (2013-06-07)[2017-03-10]. .]URL [本文引用: 1]
[Data Center for Resources and Environ-mental Sciences,Chinese Academy of Sciences. Land Use Dataset(2010) [EB/OL]. (2012-02-20)[2017-03-10]. .]URL [本文引用: 1]
[42]
Sawada M. Global Spatial Autocorrelation Indices-Moran's I,Geary's C and the General Cross-Product Statistic [EB/OL].(2009-11-23)[2017-08-23]..URL [本文引用: 1]
[Bai XY,Xiong KN,Li YB,et al.Quantitative research of the space dependence of different grades rocky desertification and the factor of population in karst mountain area [J]. ,2006,24(2):242-248.] [本文引用: 1]
[Qin YR,Lan CY,Shu WS,et al.Problem of karst desertification caused by destru-ction of limestone vegetation in Guangxi,China [J]. ,2007,35(S1):48-51.] [本文引用: 1]
[Deng JF.Main practice and successful mode of karst rocky desertification control in Qianxinan [J]. ,2014,34(1):4-7. ] [本文引用: 1]
[48]
Xu EQ,Zhang HQ,Li M X. Mining spatial information to investigate the evolution of karst rocky desertification and its human driving forces in Changshun,China [J]. ,2013,458-460:419-426. [本文引用: 1]
[Li YB,Wang SJ,Zhou MW,et al.Relationship between karst rocky desertification distribution and slope under different spatial scales [J]. ,2009,16(5):70-77.] [本文引用: 1]
[Sheng MY,LiuY,Xiong KN.Response of soil physical-chemical properties to rocky desertification succession in South China Karst [J]. ,2013,33(19):6303-6313.] [本文引用: 1]