基于地理探测器的大熊猫生境适宜度评价模型及验证
作者简介:廖颖(1990-),女,硕士生,研究方向为遥感与生态环境评价。E-mail: lysky.cool@163.com
收稿日期: 2015-09-08
要求修回日期: 2015-12-09
网络出版日期: 2016-06-10
基金资助
国家国际科技合作专项(S2013GR0477)
Suitability Assessment and Validation of Giant Panda Habitat Based on Geographical Detector
Received date: 2015-09-08
Request revised date: 2015-12-09
Online published: 2016-06-10
Copyright
动物生境适宜度评价对于野生动物生境保护十分重要。基于物种活动点来建模的生态位模型是目前应用最广泛的动物生境评价方法,但该方法不能直接表达生境适宜度与环境因子间具有生态学意义的数量关系。本文以雅安地区为例,提出一种新的大熊猫(Ailuropoda melanoleuca)生境适宜度评价方法,选取海拔、坡度、坡向、地形指数、距水源距离、植被类型、主食竹及距公路距离8个环境因子,引入地理探测器,在分别基于MAXENT模型和层次分析法(The Analytic Hierarchy Process,AHP)所构建生境适宜度模型的基础上,通过4个地理探测器(风险探测器、因子探测器、生态探测器和交互作用探测器)探寻大熊猫生境与各环境因子间的关系以及环境因子对大熊猫生境的影响机理,并将其预测结果与单一MAXENT模型和AHP法进行对比。结果表明:(1)AHP、AHP-Geogdetector、MAXENT和MAXENT-Geogdetector模型总体评价精度分别为85.6%、86.5%、91.3%和94.2%,kappa系数分别为0.699、0.718、0.821和0.882,AUC值分别为0.902、0.928、0.949和0.966,模型所预测的适宜和较适宜区与实际分布区重叠比分别为63.66%、61.30%、76.70%和90.10%,说明AHP-Geogdetector和MAXENT-Geogdetector模型精度均比相应的单一模型有所提高,且MAXENT-Geogdetector模型精度最高;(2)基于地理探测器的大熊猫生境适宜度评价模型能以“生境适宜度和环境因子间具有生态学意义的数量关系”的形式直接体现环境因子对动物生境利用的生态学作用,具有较好的生态学可解释性。因此,用地理探测器进行大熊猫生境适宜度评价具有较好的可行性。
廖颖 , 王心源 , 周俊明 . 基于地理探测器的大熊猫生境适宜度评价模型及验证[J]. 地球信息科学学报, 2016 , 18(6) : 767 -778 . DOI: 10.3724/SP.J.1047.2016.00767
The suitability assessment of wildlife habitat is very important for wildlife management and protection. Niche model is the most commonly used presence-only based habitat suitability model, which cannot explicitly express the quantitative relationship between the suitability of wildlife habitat and the environmental factors, and would be insufficient to express the ecological effects of environmental factors on wildlife habitat use. In this study, a new giant panda (Ailuropoda melanoleuca) habitat assessment method based on geographical detector (Geogdetector) is proposed. A total of 8 environmental factors were selected for the suitability assessment of giant panda habitat, including the elevation, slope, aspect, topographic position index, distance from drainage system, vegetation type, staple food sources of bamboo, and distance from human settlements. Based on the initial habitat suitability index (HSI) input data calculated by Analytic Hierarchy Process (AHP) and MAXENT model respectively, we used four geographical detectors (the risk detector, factor detector, ecological detector, and interaction detector) to assess the relationship between the suitability of giant panda habitat and their environmental risk factors. Results show that the suitability assessment of giant panda habitat based on the geographical detector has relatively favorable precision and feasibility. (1) The MAXENT-Geogdetector model has higher level of performance on accuracy than the other three methods. The overall accuracy of the prediction results based on AHP, AHP-Geogdetector, MAXENT and MAXENT-Geogdetector are 85.6%、86.5%、91.3% and 94.2% respectively. The kappa coefficients are 0.699, 0.718, 0.821 and 0.882 respectively. The AUC values are 0.902, 0.928, 0.949 and 0.966 respectively. And the overlap ratios of the predicted distribution area to the actual distribution area are 63.66%, 61.30%, 76.70% and 90.10% respectively. (2) The proposed Geogdetector-based method well captures the ecological effects of environmental factors on the wildlife habitat use by indicating the quantitative relationship between the suitability of wildlife habitat and the environmental factors.
Fig.1 Location of Ya′an prefecture图1 雅安研究区地理位置示意图 |
Tab.1 Redefined interaction relationships in a coordinate system表1 因子交互关系的坐标轴表达 |
图示 | 表达式 | 交互关系 |
---|---|---|
PD(x∩ y)<Min(PD(x),PD(y)) | 非线性减弱 | |
Min(PD(x), PD(y))<PD(x∩ y)< Max(PD(x),PD(y)) | 单方减弱 | |
PD(x∩ y)>Max(PD(x),PD(y)) | 相互增强 | |
PD(x∩ y)=PD(x)+PD(y) | 相互独立 | |
PD(x∩ y)>PD(x)+PD(y) | 非线性增强 |
Fig.3 The flowchart of assessing the habitat suitability for giant panda based on the geographical detector图3 基于地理探测器的大熊猫生境适宜度评价流程图 |
Tab.2 Assessment criteria for the affecting factors on giant panda's habitat in Ya'an prefecture表2 雅安地区大熊猫生境适宜度影响因子评价准则 |
目标层 | 因素层 | 指标层 | 适宜 | 较适宜 | 不适宜 |
---|---|---|---|---|---|
雅安地区大熊猫生境适宜性评A | 地形因素B1 | 海拔C1/m | 1800~3200 | 1200~1800,3200~3800 | <1200,>3800 |
坡度C2/° | 0~30 | 30~40 | >40 | ||
坡向C3 | 南、东南、西南 | 东、东北、西 | 北、西北 | ||
地形指数C4 | -5~5 | -15~ -5,5~15 | >15,<-15 | ||
距水源距离C5/m | <2000 | 2000~4000 | >4000 | ||
生物因素B2 | 植被类型C6 | 针叶林、针阔混交林 | 阔叶林、灌丛及草甸 | 其他 | |
主食竹C7 | 冷箭竹、短锥玉山竹、拐棍竹 | 其他竹类 | 无竹区 | ||
干扰因素B3 | 距公路距离C8/m | >1000 | 500~1000 | <500 | |
适宜度赋值 | 5 | 3 | 1 |
3.2.2 样本值与采样点 |
Tab.3 Statistically significant difference of the average habitat suitability index between three primary bamboo zones表3 大熊猫主食竹不同类别分区生境适宜度的统计显著性差异 |
主食竹显著差异 | 不适宜 | 较适宜 | 适宜 |
---|---|---|---|
不适宜 | |||
较适宜 | Y | ||
适宜 | Y | Y |
注:Y表示2个主食竹类别分区对大熊猫生境的影响具有显著性差异(置信度为95%) |
Tab.4 Statistically significant difference of the impacts for different risk factors on habitat suitability index表4 不同环境因子对生境适宜度影响的统计显著性差异 |
显著差异 | 海拔 | 坡度 | 坡向 | 地形指数 | 距水源距离 | 植被类型 | 主食竹 | 距道路距离 |
---|---|---|---|---|---|---|---|---|
海拔 | ||||||||
坡度 | N | |||||||
坡向 | Y | Y | ||||||
地形指数 | Y | Y | N | |||||
距水源距离 | Y | Y | N | N | ||||
植被类型 | N | N | Y | Y | Y | |||
主食竹 | N | N | Y | Y | Y | N | ||
距道路距离 | Y | Y | N | N | N | Y | Y |
Tab.5 Interactive impact of each paired factors on the habitat of giant panda表5 2种环境因子(PD值)对大熊猫生境影响的交互作用 |
0.2745 | 0.1887 | 0.0026 | 0.0582 | 0.0126 | 0.2705 | 0.6936 | 0.0315 | ||
---|---|---|---|---|---|---|---|---|---|
海拔 | 坡度 | 坡向 | 地形指数 | 距水源距离 | 植被类型 | 主食竹 | 距道路距离 | ||
0.2745 | 海拔 | ||||||||
0.1887 | 坡度 | 0.362 | |||||||
0.0026 | 坡向 | 0.292 | 0.191 | ||||||
0.0582 | 地形指数 | 0.291 | 0.238 | 0.067 | |||||
0.0126 | 距水源距离 | 0.279 | 0.200 | 0.013 | 0.068 | ||||
0.2705 | 植被类型 | 0.432 | 0.380 | 0.281 | 0.323 | 0.275 | |||
0.6936 | 主食竹 | 0.719 | 0.721 | 0.695 | 0.707 | 0.694 | 0.757 | ||
0.0315 | 距道路距离 | 0.365 | 0.228 | 0.037 | 0.096 | 0.048 | 0.320 | 0.709 |
Fig.4 Distribution of giant pandas' habitats in Ya′an prefecture图4 雅安地区大熊猫生境分布 |
Tab.6 Classification accuracy of the four models表6 模型分类精度 |
AHP | AHP-Geogdetector | |||||||
---|---|---|---|---|---|---|---|---|
实际类别 | 预测类别 | 实际类别 | 预测类别 | |||||
活动栅格 | 非活动栅格 | 精度/(%) | 活动栅格 | 非活动栅格 | 精度/(%) | |||
活动栅格 | 32 | 14 | 69.6 | 活动栅格 | 32 | 14 | 69.6 | |
非活动栅格 | 1 | 57 | 98.3 | 非活动栅格 | 0 | 58 | 100 | |
85.6 | 86.5 | |||||||
MAXENT | MAXENT-Geogdetector | |||||||
实际类别 | 预测类别 | 实际类别 | 预测类别 | |||||
活动栅格 | 非活动栅格 | 精度/(%) | 活动栅格 | 非活动栅格 | 精度/(%) | |||
活动栅格 | 37 | 9 | 80.4 | 活动栅格 | 41 | 5 | 89.1 | |
非活动栅格 | 0 | 58 | 100 | 非活动栅格 | 1 | 57 | 98.3 | |
91.3 | 94.2 |
The authors have declared that no competing interests exist.
[1] |
|
[2] |
[
|
[3] |
|
[4] |
|
[5] |
[
|
[6] |
|
[7] |
|
[8] |
[
|
[9] |
[
|
[10] |
|
[11] |
[
|
[12] |
|
[13] |
[
|
[14] |
|
[15] |
|
[16] |
|
[17] |
[
|
[18] |
[
|
[19] |
雅安市林业局.熊猫家源雅安大熊猫现状令人振奋——第四次大熊猫调查结果出炉[N/OL].雅安市人民政府官网[2015-3-31]. [2015-4-16]..
[ Ya 'an Forestry Bureau. The status of giant panda in Ya’an are giving us exciting insights-the fourth national survey results on giant panda in China[N/OL]. ]
|
[20] |
[
|
[21] |
[
|
[22] |
|
[23] |
[
|
[24] |
[
|
[25] |
|
[26] |
[
|
[27] |
|
[28] |
国家林业局.全国第三次大熊猫调查报告[M].北京:科学出版社,2006.
[
|
[29] |
[
|
[30] |
[
|
[31] |
[
|
/
〈 |
|
〉 |