机载激光雷达亚热带森林乔木层垂直结构分类方法

doi:10.13466/j.cnki.lyzygl.2022.01.013

摘要/Abstract

摘要：

森林垂直结构分类具有重要的生态学和林学意义。以广西为研究区,通过10阶多项式对样地的离散机载激光点云的高度—覆盖度频率分布进行拟合,得到反映冠层物质垂直分布的垂直冠层剖面(伪波),通过伪波提取有效峰、冠层表面高、层下高、林层高与冠层表面高比值等冠层结构参数,建立分类规则,将林分乔木层垂直结构分为6个类型,采用混淆矩阵评估分类精度,并选取一个面积为1 369km²的区域进行制图以检验分类规则的可推广性。结果表明:1)1 147个样地的总体分类精度为93.9%,Kappa系数为0.913;2)单峰、双峰、3峰剖面的分类错误率为6.2%,7.4%和9.1%,杉木林、松树林、桉树林和阔叶林的分类错误率分别为9%,6.4%,2.4%和6.9%,说明林分垂直结构越复杂分类精度越低;3)各个林层的检测精度均高于96%,漏检率均小于4%,误检率均小于10%,表明各个林层都能够得到准确的检测;4)制图区域的分类规则的覆盖率达到99.8%。研究表明,乔木层垂直结构分类方法具有分类精度高、普适性强、可推广性好、空间信息丰富的特点,适用于大区域亚热带森林乔木层垂直结构分类制图。

关键词: 连续冠层垂直剖面, 伪波, 冠层结构参数, 混淆矩阵

Abstract:

The vertical structure classification of forest plays an important role in ecology and forestry. A vertical canopy profile (pseudo-wave) was obtained by fitting the frequency distribution of height and coverage of discrete laser point cloud in Guangxi by using the tenth order polynomial method,which reflected the vertical distribution of canopy material. Canopy structure parameters such as effective peak,stand surface height,sub-storey height,and crown ratio were extracted by pseudo-wave and classification rules were established to divide the vertical structure of stands into six types.Confusion matrix was used to evaluate the classification accuracy,and an area of 1369km² was selected for mapping to test the generalization of classification rules. The results showed that: 1) In the classification results of 1 147 sample plots,the overall classification accuracy was 93.9%,and the Kappa coefficient was 0.913;2) The error rates of single-peak,double-peak and triple-peak were 6.2%,7.4% and 9.1% respectively,while the error rates of Chinese fir forest,pine forest,eucalyptus forest and broadleaved forest were 9%,6.4%,2.4% and 6.9%,respectively,indicating that the more complex the vertical structure of the stand,the lower the accuracy of classification;3) The accuracy of each forest layer was higher than 96%,the omission errors were less than 4%,and the commission errors were less than 10%,indicating that each forest layer can be accurately detected;4) The coverage of classification rules in mapping areas reached 99.8%. In this study,vertical forest classification method with high accuracy,good generalization and rich spatial information is suitable for overstory vertical structure classification mapping of large regional subtropical forest.

Key words: canopy vertical profile, pseudo-wave, canopy structure parameters, confusion matrix

中图分类号:

S771.8

周相贝, 李春干, 余铸, 陈中超, 苏凯. 机载激光雷达亚热带森林乔木层垂直结构分类方法[J]. 林业资源管理, 2022,(1): 106-113.

ZHOU Xiangbei, LI Chungan, YU Zhu, CHEN Zhongchao, SU Kai. Classification of Vertical Forest Structure of Overstory in Subtropical Forests Using Airborne Lidar Data[J]. FOREST RESOURCES WANAGEMENT, 2022,(1): 106-113.

图/表 7

表1

图1

图2

图3

表2

基于冠层连续剖面的乔木层垂直结构分类规则

编号	分类规则
A(1)	Null	UT₁
B(1a)	$H UL 1$ ≥hp99/2 and C_INF<0.1	OT₁
B(1b)	$H UL 1$ ≥hp99/2 and C_INF≥0.1	T₁T₃
B(2a)	(hp99/3< $H UL 1$ ≤hp99/2)and C_INF<0.1 and H_LA1/H_CS≥0.65	OT₁T₂
B(2b)	(hp99/3< $H UL 1$ ≤hp99/2)and C_INF<0.1 and H_LA1/H_CS<0.65	OT₁
B(2c)	(hp99/3< $H UL 1$ ≤hp99/2)and C_INF≥0.1 and hp99≥15.0	T₁T₂T₃
B(2d)	(hp99/3< $H UL 1$ ≤hp99/2)and C_INF≥0.1 and hp99<15.0	UT₁T₂
B(3a)	(3.0< $H UL 1$ ≤hp99/3)and hp99≥15.0 and H_LA1/H_CS≥0.80	T₁T₂T₃
B(3b)	(3.0< $H UL 1$ ≤hp99/3)and hp99≥15.0 and(0.65≤H_LA1/H_CS<0.80)and C_INF<0.1	OT₁T₂
B(3c)	(3.0< $H UL 1$ ≤hp99/3)and hp99≥15.0 and(0.65≤H_LA1/H_CS<0.80)and C_INF≥0.1	UT₁T₂
B(3d)	(3.0< $H UL 1$ ≤hp99/3)and hp99≥15.0 and H_LA1/H_CS<0.65	OT₁
B(3e)	(3.0< $H UL 1$ ≤hp99/3)and hp99<15.0 and H_LA1/H_CS≥0.65	OT₁T₂
B(3f)	(3.0< $H UL 1$ ≤hp99/3)and hp99<15.0 and H_LA1/H_CS<0.65 and C_INF<0.1	OT₁
B(3g)	(3.0< $H UL 1$ ≤hp99/3)and hp99<15.0 and H_LA1/H_CS<0.65 and C_INF≥0.1	UT₁
B(4a)	(1.0< $H UL 1$ ≤3.0)and hp99≥15.0 and H_LA1/H_CS≥0.80	T₁T₂T₃
B(4b)	(1.0< $H UL 1$ ≤3.0)and hp99≥15.0 and H_LA1/H_CS<0.80	UT₁T₂
B(4c)	(1.0< $H UL 1$ ≤3.0)and(10.0≤Hp99<15.0)and H_LA1/H_CS≥0.8	UT₁T₂
B(4d)	(1.0< $H UL 1$ ≤3.0)and(10.0≤hp99<15.0)and H_LA1/H_CS<0.8	UT₁
B(4e)	(1.0≤ $H UL 1$ <3.0)and hp99<10.0	UT₁
B(5a)	$H UL 1$ <1.0 and H_CS<10.0	UT₁
B(5b)	$H UL 1$ <1.0 and H_CS≥10.0	UT₁T₂
C(1a)	H_PEAK1≥2/3 hp99 and H_PEAK2≥hp99/3 and H_UL2≥3.0 and(H_LA1+H_LA2)/H_CS<0.65	OT₁
C(1b)	H_PEAK1≥2/3hp99 and H_PEAK2≥hp99/3 and H_UL2≥3.0 and(H_LA1+H_LA2)/H_CS≥0.65	OT₁T₂
C(1c)	H_PEAK1≥2/3hp99 and H_PEAK2≥hp99/3 and H_UL2<3.0 and hp99≥15.0	T₁T₂T₃
C(1d)	H_PEAK1≥2/3hp99 and H_PEAK2≥hp99/3 and H_UL2<3.0 and hp99<15.0	UT₁T₂
C(2a)	H_PEAK1≥2/3hp99 and(3.0<H_PEAK2≤hp99/3)and hp99≥15.0 and C_MID<0.1	T₁T₃
C(2b)	H_PEAK1≥2/3hp99 and(3.0<H_PEAK2≤hp99/3)and hp99≥15.0 and C_MID≥0.1	T₁T₂T₃
C(2c)	H_PEAK1≥2/3hp99 and(3.0<H_PEAK2≤hp99/3)and hp99<15.0	UT₁T₂
C(3a)	H_PEAK1≥2/3hp99 and(1.0≤H_PEAK2<3.0)and hp99≥15.0 and H_LA1/H_CS<0.65 and H_LA2/H_CS<0.35	OT₁
C(3b)	H_PEAK1≥2/3hp99 and(1.0≤H_PEAK2<3.0)and hp99≥15.0 and H_LA1/H_CS≥0.65 and H_LA2/H_CS<0.35	OT₁T₂
C(3c)	H_PEAK1≥2/3hp99 and(1.0≤H_PEAK2<3.0)and hp99≥15.0 and H_LA1/H_CS<0.65 and H_LA2/H_CS≥0.35 and(H_LA1+H_LA2)/H_CS≥0.80	T₁T₂T₃
C(3d)	H_PEAK1≥2/3hp99 and(1.0≤H_PEAK2<3.0)and hp99≥15.0 and H_LA1/H_CS<0.65 and H_LA2/H_CS≥0.35 and(H_LA1+H_LA2)/H_CS<0.80	T₁T₃
C(3e)	H_PEAK1≥2/3hp99 and(1.0≤H_PEAK2<3.0)and hp99<15.0 and H_LA1/H_CS≥0.65	UT₁T₂
C(3f)	H_PEAK1≥2/3hp99 and(1.0≤H_PEAK2<3.0)and hp99<15.0 and H_LA1/H_CS<0.65 and H_UL2<hp99/3	UT₁
C(3g)	H_PEAK1≥2/3hp99 and(1.0≤H_PEAK2<3.0)and hp99<15.0 and H_LA1/H_CS<0.65 and H_UL2≥hp99/3	OT₁
C(4a)	(hp99/3≤H_PEAK1<2/3hp99)and hp99≥15.0 and(H_LA1+H_LA2)/H_CS≥0.8	T₁T₂T₃
C(4b)	(hp99/3≤H_PEAK1<2/3hp99)and hp99≥15.0 and(H_LA1+H_LA2)/H_CS<0.8 and H_UL2<hp99/3	UT₁T₂
C(4c)	(hp99/3≤H_PEAK1<2/3hp99)and hp99≥15.0 and(H_LA1+H_LA2)/H_CS<0.8 and H_UL2≥hp99/3	OT₁T₂
C(4d)	(hp99/3≤H_PEAK1<2/3hp99)and hp99<15.0 and(H_LA1+H_LA2)/H_CS<0.65 and H_UL2<hp99/3	UT₁
C(4e)	(hp99/3≤H_PEAK1<2/3hp99)and hp99<15.0 and(H_LA1+H_LA2)/H_CS<0.65 and H_UL2≥hp99/3	OT₁
C(4f)	(hp99/3≤H_PEAK1<2/3hp99)and hp99<15.0 and(H_LA1+H_LA2)/H_CS≥0.65 and H_UL2<hp99/3	UT₁T₂
C(4g)	(hp99/3≤H_PEAK1<2/3hp99)and hp99<15.0 and(H_LA1+H_LA2)/H_CS≥0.65 and H_UL2≥hp99/3	OT₁T₂
D(1a)	H_PEAK3≥3.0 and(H_LA1+H_LA2+H_LA3)/H_CS≥0.80	T₁T₂T₃
D(1b)	H_PEAK3≥3.0 and(H_LA1+H_LA2+H_LA3)/H_CS<0.80	OT₁T₂
D(2a)	H_PEAK3<3.0 and(H_LA1+H_LA2+H_LA3)/H_CS≥0.80	T₁T₂T₃
D(2b)	H_PEAK3<3.0 and(H_LA1+H_LA2+H_LA3)/H_CS<0.80	OT₁T₂

表2

表3

图4

参考文献 24

[1]	Zimblea D A, Evansb D L, Carlson G C, et al. Characterizing vertical forest structure using small-footprint airborne LiDAR[J]. Remote Sensing of Environment, 2003,87(2-3):171-182. doi: 10.1016/S0034-4257(03)00139-1
[2]	Pommerening A, Sánchez Meador A J. Tamm review:Tree interactions between myth and reality[J]. Forest Ecology and Management, 2018,424:164-176. doi: 10.1016/j.foreco.2018.04.051
[3]	惠刚盈, 赵中华, 陈明辉. 描述森林结构的重要变量[J]. 温带林业研究, 2020,3(1):14-20.
[4]	方红亮. 森林垂直结构参数实测与遥感研究进展:以叶面积指数和聚集指数为例[J]. 科学通报, 2021,66(24):3141-3153.
[5]	Bourdier T, Cordonnier T, Kunstler G, et al. Tree size inequality reduces forest productivity:an analysis combining inventory data for ten european species and a light competition model[J]. Plos One, 2016,11(3):e0151852. doi: 10.1371/journal.pone.0151852
[6]	Krishna M P, Mohan M. Litter decomposition in forest ecosystems:a review[J]. Energy,Ecology and Environment, 2017,2(4):236-249. doi: 10.1007/s40974-017-0064-9
[7]	楼一恺, 范忆, 戴其林, 等. 天目山常绿落叶阔叶林群落垂直结构与群落整体物种多样性的关系[J]. 生态学报, 2021,41(21):8568-8577.
[8]	Hall S A, Burke I C, Box D O, Kaufmann M R, Stoker J M. Estimating stand structure using discrete-return lidar:an example from low density,fire prone ponderosa pine forests[J]. Forest Ecology and Management, 2005,208:189-209. doi: 10.1016/j.foreco.2004.12.001
[9]	Baker P J, Wilson J S. A quantitative technique for the identification of canopy stratification in tropical and temperate forests[J]. Forest Ecology and Management, 2000,127:77-86. doi: 10.1016/S0378-1127(99)00118-8
[10]	Bouvier M, Durrieu S, Fournier R A, et al. Generalizing predictive models of forest inventory attributes using an area-based approach with airborne LiDAR data[J]. Remote Sensing of Environment, 2015,156:322-334. doi: 10.1016/j.rse.2014.10.004
[11]	李增元, 刘清旺, 庞勇. 激光雷达森林参数反演研究进展[J]. 遥感学报, 2016,20(5):1138-1150.
[12]	刘浩, 张峥男, 曹林. 机载激光雷达森林垂直结构剖面参数的沿海平原人工林林分特征反演[J]. 遥感学报, 2018,22(5):872-888.
[13]	Morsdorf F, Mårell A, Koetz B, et al. Discrimination of vegetation strata in a multi-layered Mediterranean forest ecosystem using height and intensity information derived from airborne laser scanning[J]. Remote Sensing of Environment, 2010,114(7):1403-1415. doi: 10.1016/j.rse.2010.01.023
[14]	Leiterer R, Torabzadeh H, Furrer R, et al. Towards automated characterization of canopy layering in mixed temperate forests using airborne laser scanning[J]. Forests, 2015,6(11):4146-4167. doi: 10.3390/f6114146
[15]	Latifi H, Heurich M, Hartig F, et al. Estimating over- and understorey canopy density of temperate mixed stands by airborne LiDAR data[J]. Forestry, 2016,89(1):69-81. doi: 10.1093/forestry/cpv032
[16]	Kwon S, Jung H, Baek W, et al. Classification of forest vertical structure in south korea from aerial orthophoto and lidar data using an artificial neural network[J]. Applied Sciences, 2017,7(10):1046. doi: 10.3390/app7101046
[17]	Adnan S, Maltamo M, Coomes D A, et al. A simple approach to forest structure classification using airborne laser scanning that can be adopted across bioregions[J]. Forest Ecology and Management, 2019,433:111-121. doi: 10.1016/j.foreco.2018.10.057
[18]	Xu Zhaoshang, Zheng Guang, Moskal L M. Stratifying forest overstory for improving effective LAI estimation based on aerial imagery and discrete laser scanning data[J]. Remote Sensing, 2020,12(13):2126. doi: 10.3390/rs12132126
[19]	Jarron L R, Coops N C, MacKenzie W H, et al. Detection of sub-canopy forest structure using airborne LiDAR[J]. Remote Sensing of Environment, 2020,244:111770. doi: 10.1016/j.rse.2020.111770
[20]	骆期邦, 曾伟生, 贺东北 .等. 林业数表模型——理论、方法与实践[M]. 长沙: 湖南科学技术出版社, 2001.
[21]	Parker G G, Brown N J. Forest canopy stratification-is it useful?[J]. American Naturalist, 2000,155(4):473-484. doi: 10.1086/303340
[22]	Neto S E N, Paula A D, Tagliaferre C, et al. Performance assessment of methodologies for vertical stratification in native forest[J]. Ciênc.Florest, 2018,28(4):1583-1591.
[23]	Morsdorf F, Kötz B, Meier E, et al. Estimation of LAI and fractional cover from small footprint airborne laser scanning data based on gap fraction[J]. Remote Sensing of Environment, 2006,104(1):50-61. doi: 10.1016/j.rse.2006.04.019
[24]	Muss J D, Mladenoff D J, Townsend P A, et al. A pseudo-waveform technique to assess forest structure using discrete lidar data[J]. Remote Sensing of Environment, 2011,115(3):824-835. doi: 10.1016/j.rse.2010.11.008

森林类型	样地数/个	林分高		林分直径		林分蓄积量
森林类型	样地数/个	均值/m	变动系数/%	均值/cm	变动系数/%	均值/(m³/hm²)	变动系数/%
松树林	265	13.77	27.32	19.04	28.71	190.35	47.76
杉木林	233	10.80	28.51	12.17	28.90	191.57	47.77
桉树林	289	15.94	21.73	11.16	22.82	138.27	47.28
阔叶林	360	8.79	45.71	11.41	52.53	81.94	89.07

类型	真实值						总数	用户精度/%
类型	(1)UT₁	(2)OT₁	(3)UT₁T₂	(4)OT₁T₂	(5)T₁T₃	(6)T₁T₂ T₃	总数	用户精度/%
(1)UT₁	190	0	1	0	0	0	191	99.5
(2)OT₁	5	534	0	2	0	2	543	98.3
(3)UT₁T₂	9	0	67	6	1	1	84	79.8
(4)OT₁T₂	17	5	6	164	1	0	193	85.0
(5)T₁T₃	0	3	0	0	21	1	25	84.0
(6)T₁T₂ T₃	1	1	5	1	2	101	111	91.0
总数	222	543	79	173	25	105	1147
生产者精度/%	85.6	98.3	84.8	94.8	84.0	96.2