基于GF-1数据的二阶段遥感特征优选与森林地上生物量反演模型

doi:10.11707/j.1001-7488.LYKX20240485

Abstract

Abstract:

Objective: In this study, a two-stage remote sensing feature selection method combined with multiple machine learning models was proposed to explore an efficient model for estimating forest aboveground biomass (AGB) using domestic GF-1 satellite data, thereby providing scientific reference for the application of GF-1 satellite data in forestry monitoring. Method: A total of 44 remote sensing feature variables, such as spectral features, texture features, and vegetation indices, were first extracted from the GF-1 data. Then a two-stage feature selection strategy was proposed and then applied for feature optimization. In the first stage, the filtered method (Relief) and embedded method (Lasso, RF) were used for initial screening, and in the second stage, the feature subset was further optimized by the wrapper method (RFE) to enhance the model estimation capability. Finally, 5 machine learning models including the K nearest neighbor (KNN), support vector regression (SVR), random forest (RF), gradient boosting (GBRT), gradient boosted regression tree (GBRT) and extreme gradient boosting tree (XGBoost) models were selected and applied for AGB estimation. They were applied for evaluating the match between the different feature selection methods and models and their effects on forest AGB estimation accuracy. According to the results, the optimal strategy was explored for estimating for forest AGB based on GF-1 data. Result: 1) The Relief-RFE method achieved the best results in AGB inversion using the XGBoost model with R2 = 0.811, RMSE = 8.45 t·hm^?2, rRMSE = 23.39%. 2) The green light band texture feature (mea-b2) was able to capture spatial distribution information of forest surface cover, such as canopy density, tree distribution, and other spatial structural changes of the forest; The spectral characteristics of the blue light band (b1) were sensitive to changes in chlorophyll and leaf pigment concentrations, and was able to characterize vegetation health status and growth stage information. In the two-stage feature selection of various methods, both of the above two features were selected as core features for forest AGB estimation. 3) Feature selection significantly improved model performance, with XGBoost showing a more pronounced improvement compared to KNN, SVR, RF, and GBRT. Conclusion: By comparing the application of different feature selection methods combined with machine learning model in forest AGB estimation, the results have demonstrated the effectiveness of the proposed two-stage hybrid feature selection strategy in forest AGB estimation based on GF-1 data, especially combining it with the XGBoost model, a highly accurate and robust forest AGB estimation strategy can be constructed.

Key words: GF-1 data, forest above-ground biomass (AGB), hybrid feature selection, machine learning, two-stage optimization strategy

CLC Number:

S771.8

Feifei Yang,Wangfei Zhang,Lei Zhao,Han Zhao,Yongjie Ji,Mengjin Wang. Two-Stage Remote Sensing Feature Optimization and GF-1 Data-Supported Forest Above-Ground Biomass Inversion[J]. Scientia Silvae Sinicae, 2025, 61(4): 9-19.

Figures/Tables 8

Table 1

Fig.1

Table 2

First-stage optimization parameters and estimation accuracy of GF-1"

模型 Model	方法 Method	R²	RMSE/ (t·hm^?2)	rRMSE （%）	特征 Feature
KNN	Relief	0.209	14.85	41.10	b3、DVI、b2、b4、b1、PVI、var_b4、var_b3、con_b4、con_b3、var_b2、con_b2、mea_b3、mea_b4、mea_b2、var_b1、con_b1、ent_b1
	RF	0.320	13.68	37.86	mea_b1、cor_b1、mea_b2、cor_b2、mea_b3、con_b4、NDVI、SR、EVI、b1、b2、b3、b4
	Lasso	0.232	15.13	41.88	var_b1、con_b1、ent_b1、Cor_b1、mea_b2、dis_b2、ent_b2、var_b3、con_b3、mea_b4、var_b4、con_b4、dis_b4、ent_b4、Cor_b4、PVI、DVI、EVI、b1、b2、b3、b4
SVR	Relief	0.262	11.87	32.85	b3、DVI、b2、b4、b1、PVI、var_b4、var_b3、con_b4、con_b3、var_b2、con_b2、mea_b3、mea_b4、mea_b2、var_b1、con_b1、ent_b1
	RF	0.321	10.89	30.14	mea_b1、cor_b1、mea_b2、cor_b2、mea_b3、hom_b4、NDVI、SAVI、SR、 EVI、MSAVI、b1、b2、b3、b4
	Lasso	0.386	9.82	27.18	var_b1、con_b1、ent_b1、Cor_b1、mea_b2、dis_b2、ent_b2、var_b3、con_b3、mea_b4、var_b4、con_b4、dis_b4、ent_b4、Cor_b4、PVI、DVI、EVI、b1、b2、b3、b4
RF	Relief	0.220	18.80	52.03	b3、DVI、b2、b4、b1、PVI、var_b4、var_b3、con_b4、con_b3、var_b2、con_b2、mea_b3、mea_b4、mea_b2、var_b1、con_b1、ent_b1
	RF	0.404	14.07	38.94	b3、DVI、b2、b4、b1、PVI、var_b4、var_b3、con_b4、con_b3、var_b2、con_b2、mea_b3、mea_b4、mea_b2、var_b1、con_b1、ent_b1
	Lasso	0.267	15.50	42.90	b3、DVI、b2、b4、b1、PVI、var_b4、var_b3、con_b4、con_b3、var_b2、con_b2、mea_b3、mea_b4、mea_b2、var_b1、con_b1、ent_b1
GBRT	Relief	0.135	15.21	42.10	mea_b1、mea_b2、cor_b2 、mea_b3、sec_b3、con_b4、NDVI、SR、 VARI、EVI、b1、b2、b3、b4
	RF	0.119	15.60	43.18	mea_b1、cor_b1、mea_b2、cor_b2、mea_b3、mea_b4、con_b4、NDVI、SR、EVI、b1、b2、b3
	Lasso	0.132	15.23	42.15	mea_b1、cor_b1、mea_b2、cor_b2、mea_b3、con_b4、SR、VARI、EVI、 MSAVI、b1、b2、b3、b4
XGBoost	Relief	0.296	10.81	29.92	var_b1、con_b1、ent_b1、Cor_b1、mea_b2、dis_b2、ent_b2、var_b3、con_b3、mea_b4、var_b4、con_b4、dis_b4、ent_b4、Cor_b4、PVI、DVI、EVI、b1、b2、b3、b4
	RF	0.416	9.90	27.40	var_b1、con_b1、ent_b1、Cor_b1、mea_b2、dis_b2、ent_b2、var_b3、con_b3、mea_b4、var_b4、con_b4、dis_b4、ent_b4、Cor_b4、PVI、DVI、EVI、b1、b2、b3、b4
	Lasso	0.444	9.90	27.40	var_b1、con_b1、ent_b1、Cor_b1、mea_b2、dis_b2、ent_b2、var_b3、con_b3、mea_b4、var_b4、con_b4、dis_b4、ent_b4、Cor_b4、PVI、DVI、EVI、b1、b2、b3、b4

Table 2

Fig.2

Fig.3

Table 3

Fig.4

Fig.5

References 0

	丁家祺, 黄文丽, 刘迎春, 等. 基于机器学习和多源数据的湘西北森林地上生物量估测. 林业科学, 2021, 57 (10): 36- 48.
	Ding J Q, Huang W L, Liu Y C, et al. Estimation of forest aboveground biomass in northwest Hunan Province based on machine learning and multi-source data. Scientia Silvae Sinicae, 2021, 57 (10): 36- 48.
	黄从德, 张　健, 杨万勤, 等. 四川省及重庆地区森林植被碳储量动态. 生态学报, 2008, 28 (3): 966- 975. doi: 10.3321/j.issn:1000-0933.2008.03.008
	Huang C D, Zhang J, Yang W Q, et al. Dynamics on forest carbon stock in Sichuan Province and Chongqing City. Acta Ecologica Sinica, 2008, 28 (3): 966- 975. doi: 10.3321/j.issn:1000-0933.2008.03.008
	黄晓娟. 2018. 面向特征选择的Relief算法研究. 苏州: 苏州大学.
	Huang X J. 2018. Research on relief algorithm for feature selection. Suzhou: Soochow University. ［in Chinese］
	菅永峰, 韩泽民, 黄光体, 等. 基于高分辨率遥感影像的北亚热带森林生物量反演. 生态学报, 2021, 41 (6): 2161- 2169.
	Jian Y F, Han Z M, Huang G T, et al. Estimation of forest biomass using high spatial resolution remote sensing imagery in north subtropical forests. Acta Ecologica Sinica, 2021, 41 (6): 2161- 2169.
	吕梓晴, 段爱国. 不同产区杉木生物量与碳储量模型. 林业科学, 2024, 60 (2): 1- 11. doi: 10.11707/j.1001-7488.LYKX20220526
	Lü Z Q, Duan A G. Biomass and carbon storage model of Cunninghamia lanceolata in different production areas. Scientia Silvae Sinicae, 2024, 60 (2): 1- 11. doi: 10.11707/j.1001-7488.LYKX20220526
	蒙诗栎, 庞　勇, 张钟军, 等. WorldView-2纹理的森林地上生物量反演. 遥感学报, 2017, 21 (5): 812- 824.
	Meng S (L /Y), Pang Y, Zhang Z J, et al. Estimation of aboveground biomass in a temperate forest using texture information from WorldView-2. National Remote Sensing Bulletin, 2017, 21 (5): 812- 824.
	潘婧靓. 2020. 联合GF: 3 PolSAR数据和Landsat: 8 OLI数据的森林地上生物量估测方法研究. 哈尔滨: 东北林业大学.
	Pan J L. 2020. Research on forest aboveground biomass estimation methods using combined GF-3 PoISAR data and Landsat-8 OLI data. Harbin: Northeast Forestry University. ［in Chinese］
	武小军, 周文心, 董永新. 一种改进的嵌入式特征选择算法及应用. 同济大学学报(自然科学版), 2022, 50 (2): 153- 159.
	Wu X J, Zhou W X, Dong Y X. A novel embedded feature selection algorithm and its application. Journal of Tongji University (Natural Science), 2022, 50 (2): 153- 159.
	于晓辉. 2019. 森林生物量遥感估测模型构建中的特征选择方法对比研究. 杭州: 浙江农林大学.
	Yu X H. 2019. Comparative study of feature selection methods in the construction of remote sensing models for forest biomass estimation . Hangzhou: Zhejiang A&F University. ［in Chinese］
	张　志, 田　昕, 陈尔学, 等. 森林地上生物量估测方法研究综述. 北京林业大学学报, 2011, 33 (5): 144- 150.
	Zhang Z, Tian X, Chen E X, et al. Review of methods on estimating forest above ground biomass. Journal of Beijing Forestry University, 2011, 33 (5): 144- 150.
	周俊宏, 王子芝, 廖声熙, 等. 基于GF-1影像的普达措国家公园森林地上生物量遥感估算. 农业工程学报, 2021, 37 (4): 216- 223.
	Zhou J H, Wang Z Z, Liao S X, et al. Remote sensing estimation of forest aboveground biomass in Potatso National Park using GF-1 images. Transactions of the Chinese Society of Agricultural Engineering, 2021, 37 (4): 216- 223.
	朱教君, 王高峰, 张怀清, 等. 关于“气候智慧林业” 研究的思考. 林业科学, 2024, 60 (7): 1- 7.
	Zhu J J, Wang G F, Zhang H Q, et al. On the research of climate-smart forestry. Scientia Silvae Sinicae, 2024, 60 (7): 1- 7.
	Afreixo V, Cabral J, Macedo P. 2023. Comparison of feature selection methods in regression modeling: a simulation study. Computational Science and Its Applications-ICCSA 2023 Workshops, 150−159.
	Altman N. An introduction to kernel and nearest-neighbor nonparametric regression. The American Statistician, 1992, 46, 175- 185. doi: 10.1080/00031305.1992.10475879
	Bolón-Canedo V, Sánchez-Maroño N, Alonso-Betanzos A. Feature selection for high-dimensional data. Progress in Artificial Intelligence, 2016, 5, 65- 75.
	Breiman L. Random forests. Machine Learning, 2001, 45, 5- 32. doi: 10.1023/A:1010933404324
	Brovkina O, Novotny J, Cienciala E, et al. Mapping forest aboveground biomass using airborne hyperspectral and LiDAR data in the mountainous conditions of Central Europe. Ecological Engineering, 2017, 100, 219- 230. doi: 10.1016/j.ecoleng.2016.12.004
	Chen T Q, Guestrin C, Chen T Q, et al. 2016. XGBoost. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, California, USA, 785−794.
	Dong H, Xu X, Wang L, et al. Gaofen-3 PolSAR image classification via XGBoost and polarimetric spatial information. Sensors, 2018, 18 (2): 611. doi: 10.3390/s18020611
	Dube T, Mutanga O. Evaluating the utility of the medium-spatial resolution Landsat 8 multispectral sensor in quantifying aboveground biomass in uMgeni catchment, South Africa. ISPRS Journal of Photogrammetry and Remote Sensing, 2015, 101, 36- 46. doi: 10.1016/j.isprsjprs.2014.11.001
	Fassnacht F E, Hartig F, Latifi H, et al. Importance of sample size, data type and prediction method for remote sensing-based estimations of aboveground forest biomass. Remote Sensing of Environment, 2014, 154, 102- 114. doi: 10.1016/j.rse.2014.07.028
	Freeman E A, Moisen G G, Coulston J W, et al. Random forests and stochastic gradient boosting for predicting tree canopy cover: comparing tuning processes and model performance. Canadian Journal of Forest Research, 2016, 46 (3): 323- 339. doi: 10.1139/cjfr-2014-0562
	Friedman J H. Greedy function approximation: a gradient boosting machine. Annals of Statistics, 2001, 29 (5): 1189- 1232. doi: 10.1214/aos/1013203450
	Gitelson A A, Kaufman Y J, Stark R, et al. Novel algorithms for remote estimation of vegetation fraction. Remote Sensing of Environment, 2002, 80 (1): 76- 87.
	Guyon I, Elisseeff A. An introduction to variable and feature selection. Journal of Machine Learning Research, 2003, 3, 1157- 1182.
	Guyon I, Weston J, Barnhill S, et al. Gene selection for cancer classification using support vector machines. Machine Learning, 2002, 46 (1): 389- 422.
	Huang H J, Wu D S, Fang L M, et al. Comparison of multiple machine learning models for estimating the forest growing stock in large-scale forests using multi-source data. Forests, 2022, 13 (9): 1471. doi: 10.3390/f13091471
	Huete A, Didan K, Miura T, et al. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sensing of Environment, 2002, 83 (1/2): 195- 213.
	Huete A R. A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment, 1988, 25 (3): 295- 309. doi: 10.1016/0034-4257(88)90106-X
	Jin Q W, Fan X T, Liu J, et al. Estimating tropical cyclone intensity in the South China Sea using the XGBoost model and FengYun satellite images. Atmosphere, 2020, 11 (4): 423. doi: 10.3390/atmos11040423
	Jordan C F. Derivation of leaf-area index from quality of light on the forest floor. Ecology, 1969, 50 (4): 663- 666. doi: 10.2307/1936256
	Kira K, Rendell L A. 1992. A practical approach to feature selection. Machine Learning Proceedings 1992, Morgan Kaufmann, 249−256.
	Kohavi R, John G H. Wrappers for feature subset selection. Artificial Intelligence, 1997, 97 (1/2): 273- 324.
	Konstantinavičienė J, Vitunskienė V. Definition and classification of potential of forest wood biomass in terms of sustainable development: a review. Sustainability, 2023, 15 (12): 9311. doi: 10.3390/su15129311
	Li X H. Using “random forest” for classification and regression. Chinese Journal of Applied Entomology, 2013, 50 (4): 1190- 1197.
	Liu H, Setiono R. 1996. A probabilistic approach to feature selection — a filter solution. International Conference on Machine Learning, 1−9.
	Qi J, Chehbouni A, Huete A R, et al. A modified soil adjusted vegetation index. Remote Sensing of Environment, 1994, 48 (2): 119- 126. doi: 10.1016/0034-4257(94)90134-1
	Richardson A J, Wiegand C L. Distinguishing vegetation from soil background information. Photogrammetric Engineering and Remote Sensing, 1977, 43 (12): 1541- 1552.
	Rouse J W, Haas R H, Schell J A, et al. Monitoring vegetation systems in the great plains with ERTS. NASA Special Publication, 1974, 351 (1): 309.
	Tibshirani R. Regression shrinkage and selection via the lasso: a retrospective. Journal of the Royal Statistical Society Series B: Statistical Methodology, 2011, 73 (3): 273- 282. doi: 10.1111/j.1467-9868.2011.00771.x
	Vabalas A, Gowen E, Poliakoff E, et al. Machine learning algorithm validation with a limited sample size. PLoS One, 2019, 14 (11): e0224365. doi: 10.1371/journal.pone.0224365
	Verma R K, Sharma L K, Lele N. AVIRIS-NG hyperspectral data for biomass modeling: from ground plot selection to forest species recognition. Journal of Applied Remote Sensing, 2023, 17 (1): 014522- 014522.
	Zhang H, Song T Q, Wang K L, et al. Biomass and carbon storage in an age-sequence of Cyclobalanopsis glauca plantations in southwest China. Ecological Engineering, 2014, 73, 184- 191. doi: 10.1016/j.ecoleng.2014.09.008
	Zhang Y L, Wang N, Wang Y L, et al. A new strategy for improving the accuracy of forest aboveground biomass estimates in an alpine region based on multi-source remote sensing. GIScience & Remote Sensing, 2023, 60 (1): 2163574.
	Zhu T T. Analysis on the applicability of the random forest. Journal of Physics: Conference Series, 2020, 1607 (1): 012123. doi: 10.1088/1742-6596/1607/1/012123

植被指数 Vegetation index	植被指数计算公式 Vegetation index calculation equation	文献 References
NDVI	$ \mathrm{NDVI}=\dfrac{\rho_{\text{NIR}}-\rho_{\text{red}}}{\rho_{\text{NIR}}+\rho_{\text{red}}} $	Rouse et al., 1974
SR	$ \mathrm{SR}=\rho_{\text{red}}/\rho_{\text{NIR}} $	Jordan , 1969
VARI	$ \mathrm{VARI}=\dfrac{\rho_{\text{green }}-\rho_{\text{red}}}{\rho_{\text{green}}+\rho_{\text{red}}-\rho_{\text{blue}}^{ }} $	Gitelson et al., 2002
DVI	$ \mathrm{DVI}=\rho_{\text{NIR}}-\rho_{\text{red}} $	Richardson et al., 1977
PVI	$ {\text{PVI}} = {\text{0}}{\text{.939 }} \times {\rho _{{\text{NIR}}}} - {\text{0}}{\text{.344 }} \times {\rho _{{\text{red}}}} + {\text{0}}{\text{.09}} $	Richardson et al., 1977
SAVI	$ \mathrm{SAVI}=\dfrac{1.5\times(\rho_{\text{NIR}}-\rho_{\text{red}})}{\rho_{\text{NIR}}+\rho_{\text{red}}+\text{0}\text{.5}} $	Huete , 1988
EVI	$ \mathrm{EVI}=\dfrac{2.5\times(\rho_{\text{NIR}}-\rho_{\text{red}})}{1+\rho_{\text{NIR}}+6\times\rho_{\text{red}}-7.5\times\rho_{\text{bule}}} $	Huete et al., 2002
MSAVI	$ \mathrm{MSAVI}=\dfrac{2\rho_{\text{NIR}}+1-\sqrt{(2\rho_{\text{NIR}}+1)^2-8(\rho_{\text{NIR}}-\rho_{\text{red}})}}{2} $	Qi et al., 1994

模型 Model	一阶段方法 First-stage method	二阶段方法 Two-stage method	R²	RMSE/ (t·hm^?2)	rRMSE （%）	特征 Feature
KNN	Relief	RFE	0.474	11.59	32.08	b3、b2、b4、b1、con_b4、con_b3、mea_b3、mea_b4、mea_b2
	RF		0.384	10.87	30.01	mea_b2、con_b4、EVI、b1、b3、b4
	Lasso		0.417	10.63	29.42	con_b1、cor_b1、mea_b2、con_b3、mea_b4、con_b4、EVI、 b1、b2、b3、b4
SVR	Relief		0.301	11.45	31.69	b3、DVI、b2、b4、b1、con_b4、con_b3、mea_b3、mea_b2
	RF		0.473	9.10	25.19	mea_b2、mea_b3、hom_b4、NDVI、EVI、b1、b4
	Lasso		0.493	10.84	30.00	con_b1、cor_b1、mea_b2、con_b3、mea_b4、con_b4、 EVI、b1、b2、b3、b4
RF	Relief		0.593	8.11	22.45	b3、b2、b4、b1、con_b4、con_b3、mea_b3、mea_b2、con_b1
	RF		0.629	7.29	20.18	b3、b2、b4、b1、con_b4、con_b3、mea_b3、mea_b4、mea_b2
	Lasso		0.571	7.34	20.32	b3、b2、b4、b1、PVI、con_b4、con_b3、mea_b3、mea_b2
GBRT	Relief		0.658	11.51	31.86	mea_b2、mea_b3、con_b4、NDVI、EVI、b1、b4
	RF		0.651	7.54	20.87	mea_b2、mea_b4、con_b4、SR、EVI、b1
	Lasso		0.591	9.12	25.24	mea_b2、con_b4、EVI、MSAVI、b1、b3、b4
XGBoost	Relief		0.811	8.45	23.39	mea_b1、cor_b2、var_b3、con_b3、cor_b3、con_b4、 dis_b4、cor_b4、PVI、DVI、b1
	RF		0.708	6.21	17.19	con_b1、cor_b1、mea_b2、con_b3、con_b4、dis_b4、 EVI、b1、b2、b3、b4
	Lasso		0.781	6.21	17.19	con_b1、cor_b1、mea_b2、con_b3、con_b4、dis_b4、 EVI、b1、b2、b3、b4

[1]	Xiaoning Ge,Xinqiao Xu,Huaiqing Zhang,Jing Zhang,Jie Yang,Zeyu Cui,Rurao Fu,Jinjie Liang,Tianhua Zou,Linlong Wang,Yang Liu. Progress and Reflection on Genotype-Environment Interaction Algorithms in Forest Tree Breeding [J]. Scientia Silvae Sinicae, 2025, 61(3): 1-15.
[2]	Dan Pan,Luyi Luo,Kaiwen Ji,Fanbin Kong. Impact of Rural E-Commerce Development on Intra-Regional Income Gap and Shared Prosperity in Forested Areas [J]. Scientia Silvae Sinicae, 2024, 60(5): 35-50.
[3]	Chenchen Shen,Wenfa Xiao,Jianhua Zhu,Lixiong Zeng,Jizhen Chen,Zhilin Huang. Characterization of Soil Organic Carbon and Key Influencing Factors of Natural Forests in Central China Based on Machine Learning Algorithms [J]. Scientia Silvae Sinicae, 2024, 60(3): 65-77.
[4]	Qiangying Jiao,Zongfu Han,Weiye Wang,Di Liu,Pengxu Pan,Bo Li,Nianci Zhang,Ping Wang,Jinhua Tao,Meng Fan. Driving Factors and Forecasting Model of Lightning-Caused Forest Fires in Daxing’ anling Mountains Based on Multi-Sources Data and Machine Learning Method [J]. Scientia Silvae Sinicae, 2023, 59(6): 74-87.
[5]	Nannan Yang,Yan Bai,Suyi Jiang,Chunmei Yang,Kaihong Xu. Recognition Method of Plate and Wood Based on ALexNet Optimaization [J]. Scientia Silvae Sinicae, 2022, 58(3): 149-158.
[6]	Feng Qi, Chen Erxue, Li Zengyuan, Li Lan, Zhao Lei. Forest Above-Ground Biomass Estimation Method for Rugged Terrain Based on Airborne P-Band PolSAR Data [J]. Scientia Silvae Sinicae, 2016, 52(3): 10-22.

Two-Stage Remote Sensing Feature Optimization and GF-1 Data-Supported Forest Above-Ground Biomass Inversion

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 8

References 0

Related Articles 6

Recommended Articles

Metrics

Comments