Time series high-resolution leaf area index estimation and change monitoring in the Saihanba area

被引：0

作者：

Zhou H. ^{[1
]}

Zhang G. ^{[1
,2
]}

Wang C. ^{[1
,2
]}

Wang J. ^{[1
]}

Cheng S. ^{[3
]}

Xue H. ^{[2
]}

Wan H. ^{[4
]}

Zhang L. ^{[3
]}

机构：

[1] State Key Laboratory of Remote Sensing Science, Beijing Engineering Research Center for Global Land Remote Sensing Products, Faculty of Geographical Science, BNU, Beijing

[2] School of Surveying & Land Information Engineering, Henan Polytechnic University, Jiaozuo

[3] Hebei saihanba Mechanic forest farm, Chengde

[4] Satellite Environment Application. Center of Ministry of Ecology and Environmental, Beijing

来源：

National Remote Sensing Bulletin | 2021年 / 25卷 / 04期

基金：

中国国家自然科学基金;

关键词：

Change detection; Deep learning method; Leaf area index; The Ensemble Kalman Filter algorithm; Time series high resolution;

D O I：

10.11834/jrs.20219447

中图分类号：

学科分类号：

摘要：

A 30 m-spatial-resolution LAI time series estimation method was proposed on the basis of the ensemble Kalman filter (EnKF). Time series LAI of 2000-2018 was produced in the Saihanba area, and vegetation change monitoring was applied. The detected disturbance was consistent with climate condition and field management. Time series LAI is critical for vegetation growth monitoring, surface process simulation, and global change research. Saihanba is an important ecological environment protection area in China, and long-term monitoring of this area is significant for forest management and development. In this study, MODIS LAI products and Landsat surface reflectance data were used to generate time series high-resolution LAI datasets from 2004 to 2018 in Saihanba by using EnKF. Vegetation changes were then monitored on the basis of the generated LAI time series with the Prophet model. First, the multistep Savitzky-Golay filtering algorithm was used to smooth the MODIS LAI data, and the upper envelope of time series LAI was generated. A dynamic model was constructed in accordance with the trend of LAI upper envelope to provide a short-range forecast of LAI. Then, the ground measured LAI data and the corresponding Landsat reflectance data were used to train a Back Propagation (BP) neural network. The high-resolution LAI data from the BP model were used to update the dynamic model in real time to generate high-resolution time series LAI data based on the EnKF method. Lastly, the time series LAI data were used as the input of the Prophet deep learning model to obtain the LAI time series prediction values of a certain year. The correlation coefficient and root-mean-square error distribution maps could be obtained from the comparison of the prediction results with the LAI of the current year. A Support Vector Machine (SVM) method was used to classify the disturbed and normal pixels. The EnKF algorithm can generate continuous high-resolution LAI data, and the estimation results are consistent with the field LAI values with R2 of 0.9498 and RMSE of 0.1577. At the regional scale, the estimation LAI maps have high consistency with the Landsat reference LAI maps, the R2 is higher than 0.87, and the RMSE is less than 0.61. The Prophet and SVM models detected that the vegetation in Saihanba was severely disturbed in 2009, 2010, 2013, 2014, and 2015, mainly due to the low annual rainfall and deforestation. The detection results are consistent with the local precipitation and logging data. The algorithm proposed in this paper can be used for time series high-spatial-resolution LAI data inversion on a large scale, and the inversion results can be used for vegetation change detection. This work has important reference significance for the planning and management of Saihanba and even the national forest area. © 2021, Science Press. All right reserved.

引用

页码：1000 / 1012

页数：12

共 42 条

[1] Baret F, Hagolle O, Geiger B, Bicheron P, Miras B, Huc M, Berthelot B, Nino F, Weiss M, Samain O, Roujean J L, Leroy M., LAI, fAPAR and fCover CYCLOPES global products derived from VEGETATION: Part 1: principles of the algorithm, Remote Sensing of Environment, 110, 3, pp. 275-286, (2007)
[2] Chander G, Helder D L, Markham B L, Dewald J D, Kaita E, Thome K J, Micijevic E, Ruggles T A., Landsat-5 TM reflective-band absolute radiometric calibration, IEEE Transactions on Geoscience and Remote Sensing, 42, 12, pp. 2747-2760, (2004)
[3] Chapelle O, Vapnik V, Bousquet O, Mukherjee S., Choosing multiple parameters for support vector machines, Machine Learning, 46, 1, pp. 131-159, (2002)
[4] Chen J M, Black T A., Defining leaf area index for non-flat leaves, Plant, Cell and Environment, 15, 4, pp. 421-429, (1992)
[5] Cohen W B, Yang Z Q, Stehman S V, Schroeder T A, Bell D M, Masek J G, Huang C Q, Meigs G W., Forest disturbance across the conterminous United States from 1985-2012: the emerging dominance of forest decline, Forest Ecology and Management, 360, pp. 242-252, (2016)
[6] Delegido J, Verrelst J, Alonso L, Moreno J., Evaluation of sentinel-2 red-edge bands for empirical estimation of Green LAI and chlorophyll content, Sensors, 11, 7, pp. 7063-7081, (2011)
[7] Evensen G., The Ensemble Kalman Filter: theoretical formulation and practical implementation, Ocean Dynamics, 53, 4, pp. 343-367, (2003)
[8] Fang H L, Liang S L, Townshend J R, Dickinson R E., Spatially and temporally continuous LAI data sets based on an integrated filtering method: examples from North America, Remote Sensing of Environment, 112, 1, pp. 75-93, (2008)
[9] Fu L Z, Qu Y H, Wang J D., Bias analysis and validation method of the MODIS LAI product, Journal of Remote Sensing, 21, 2, pp. 206-217, (2017)
[10] He T, Liang S L, Wang D D, Cao Y F, Gao F, Yu Y Y, Feng M., Evaluating land surface albedo estimation from Landsat MSS, TM, ETM+, and OLI data based on the unified direct estimation approach, Remote Sensing of Environment, 204, pp. 181-196, (2018)

← 1 2 3 4 5 →