Conference Agenda

Overview and details of the sessions and sub-session of this conference. Please select a date or session to show only sub-sessions at that day or location. Please select a single sub-session for detailed view (with abstracts and downloads if available).

Please note that all times are shown in CEST. The current conference time is: 13th Dec 2021, 09:45:09am CET

 
 
Session Overview
Session
Dr5 S.5.1: ECOSYSTEMS
Time:
Wednesday, 21/July/2021:
10:45am - 12:05pm

Session Chair: Dr. Andy Zmuda
Session Chair: Prof. Yong Pang
Workshop: Dragon 5

ID. 59257 Data Fusion 4 Forests Assessement
ID. 59307 3D Forests from POLSAR Data
ID. 59358 China-ESA Forest Observation
ID. 59313 Grassland Degredation by RS


Show help for 'Increase or decrease the abstract text size'
Presentations
10:45am - 11:05am
Accepted
ID: 260 / Dr5 S.5.1: 1
Oral Presentation for Dragon 5
Ecosystem: 59257 - Mapping Forest Parameters and Forest Damage For Sustainable Forest Management From Data Fusion of Satellite Data

Mapping Forest Parameters and Forest Damage for Sustainable Forest Management From Data Fusion of Satellite Data

Xiaoli Zhang1, Johan E.S. Fransson2, Ning Zhang3, Langning Huo2, Karin Öhman2, Tiecheng Huang1, Eva Lindberg2, Yueting Wang1, Henrik J. Persson2, Guoqi Chai1, Niwen Li1, Long Chen1

1Beijing Forestry University, China; 2Swedish University of Agricultural Sciences, Department of Forest Resource Management, SWEDEN; 3Chinese Academy of Agriculture Sciences, China

ID.59257: Mapping forest parameters and forest damage for sustainable forest management from data fusion of satellite data

PI Europe: Dr. Johan Fransson, Swedish University of Agricultural Sciences, Department of Forest Resource Management, SWEDEN

PI China: Prof. Xiaoli Zhang, Beijing Forestry University, CHINA

Forests play a critical role in the Earth’s ecosystem and have a strong impact on the environment. Under the threat of global climate change, remote sensing techniques provide information for a better understanding of the forest ecosystems, early detection of forest diseases, and both rapid and continuous monitoring of forest disasters. This project concerns the topic Ecosystems and spans the subtopics Collaborative estimation of forest quality parameters and Forest and grassland disaster monitoring. The aim is to study and explore remote sensing techniques in forest applications, especially on data fusion of satellite images, laser scanning, and hyperspectral drone images. The research contents are mainly tree species classification, forest parameter estimation and forest insect damage detection.

1. Work carried out during the current year:

(1) Satellite images:

We applied for and acquired ESA, Copernicus Sentinel and Chinese EO data, such as RADARSAT-2, SPOT, WorldView, Sentinel and Gaofen series, combined with hyperspectral data of UAV and LiDAR data to carry out relevant research. The satellite images acquired for the study area in China are:

RADARSAT-2, two scenes altogether. One scene data covers Fushun County, one scene covers Qingyuan County.

  • SPOT, two images altogether. One scene data covers Wangyedian, one scene covers Gaofeng.
  • Sentinel-1, time-series images from 2019 to 2020, covering Wangyedian.
  • Sentinel-2, time-series cloud-free images from 2019 to 2020, covering Wangyedian.
  • Gaofen-2, two images acquired 2020, that cover Gaofeng.

The satellite images acquired for the study area in Sweden are:

  • Sentinel-2, time-series cloud-free images from 2018 to 2021, covering Remningstorp in Sweden.
  • RADARSAT-2, one image acquired 2020 and a second to be acquired during the summer 2021, that cover Remningstorp in Sweden.
  • Pleiades, one image (per 29 Apr 2021), of Remningstorp, Sweden.

(2) Survey data:

  • One field investigation of forest parameters was conducted in Nanning, Guangxi, China, and the parameters including diameter at breast height (>5cm), tree height, under branch height, environmental parameters and the coordinates of four corners were obtained.
  • For forest pests and diseases, a field survey of forests damaged by pine wood nematodes was carried out in Fushun, Liaoning, China. The spectral information of the damaged trees and healthy trees in different susceptible stages were collected.
  • The forest information of the field sample plots in the study area Remningstorp were updated and stand-level information about forest managements were noticed. We have previously used the same field sample plots for analysis of tree species from Sentinel-1 and Sentinel-2 data.
  • A controlled experiment is conducting at Remningstorp in 2021. Pheromone dispensers was set in 24 plots, expecting the bark beetles attacking around 180 trees. The process of infestations will are recorded during April 2021 to June 2021.

(3) Technical progress:

  • In the section on tree species classification, we proposed two deep learning models, an improved prototypical networks (IPrNET) and a new prototypical networks combined with attention mechanism CPAM-P-NET model, using UAVs hyperspectral data, and obtained good classification results for eight major tree species in southern China.
  • For forest parameters such as height, biomass, we combined the ZY-3 stereo image and DEM to automatically extract a high resolution spatially continuous tree height product. In addition, the acquired tree height product combined with Sentinel-2 data were used to achieve the forest above-ground biomass distribution map.
  • In the section on forest biotic disturbance detection, pine wood nematode (Bursaphelenchus xylophilus) and Dendrolimus tabulaeformis Tsai et Liu (D. tabulaeformis) and the European spruce bark beetle (Ips typographus [L.]) three kinds of forest insect damage were studied.

For D. tabulaeformis, we proposed a spectral-spatial classification framework combining UAV-based hyperspectral images and digital images to achieve more detailed classification and automatically extract damaged tree crowns. For Bursaphelenchus xylophilus, we analyzed the spectral characteristics of three tree species (Pinus tabulaeformis, Pinus koraiensis and Pinus tabulaeformis) in Weihai, Shandong Province and Fushun, Liaoning Province during different susceptible stages, and explored the spectral response characteristics of plants under stress. For spruce bark beetles, we have been developing algorithms to detect infestations using Sentinel-2 images in 2018 and 2019.

2.Plan for next year

(1) Data acquisition:

We would like to apply for TanDEM data covering Wangyedian and Weihai, as well as WorldView-3 data covering Wangyedian, Anhui and Fushun.

We have ordered one WorldView-2 image and one WorldView-3 image covering Remningstorp in May 2021. We will also acquire one more RADARSAT-2 image in September, and one or two RADARSAT-2 images in the winter 2021-2022.

(2) The research content:

  • For tree species classification, we will acquire WorldView-3 and Sentinel-2 images for repeated experiments in southern China, and further analyze tree species classification based on deep learning model. And in the study area Remningstorp, the planned analysis from WorldView-3 images will be done for individual trees thanks to the higher resolution of the images.
  • In the section on forest parameters, we will combine the Sentinel-1 SAR data for biomass estimation, and combine the satellite images and LiDAR data for tree crown extraction.
  • We will use the RADARSAT-2 images for developing change detection methods of forest biomass.
  • In the section on forest insect damage detection, early detection of Bursaphelenchus xylophilus based on spectral characteristics of different stages combined with remote sensing data will be studied. And for spruce bark beetles, we will acquire WorldView-3 images and field data from a controlled experiment in 2021. Early detection using WorldView-3 images will be studied.

(3) Cooperation plan:

Our two teams share similar interests in forest remote sensing, conduct similar projects to communicate and collaborate more in scientific research. Annual seminars are planned in 2021 to 2024. The two team will collaborate in joint studies and publications. Niwen Li, a PhD student in BFU, will have a research exchange to SLU from August 2021 to July 2022. And more exchange of PhD students is under consideration. The scientific communication will be conducted every month by the researchers.

Zhang-Mapping Forest Parameters and Forest Damage for Sustainable Forest Management-260Oral5.pdf


11:05am - 11:25am
Accepted
ID: 324 / Dr5 S.5.1: 2
Oral Presentation for Dragon 5
Ecosystem: 59307 - 3-D Characterization and Temporal Analysis of Forests and Vegetated Areas Using Time-Series of Polarimetric SAR Data and Tomographic Processing

Assessment Of The Performance of Polarimetric And Tomographic SAR Configurations For The Characterization of Tropical Forests

Laurent Ferro Famil1,2, Ludovic Villard2, Thuy Le Toan2, Eric Pottier1, Erxue Chen3, Zengyuan Li3

1University of Rennes 1, France; 2Cesbio, France; 3IFRIT, Chines Academy of Forestry, Beijing, China

The European Space Agency (ESA) will launch in 2023 the BIOMASS spaceborne mission, based on a SAR device operating over several polarizations at P band, and whose main objective is the characterization of dense forests. One of the original features of this mission concerns its ability to provide 3-D SAR information through polarimetric and tomographic SAR (PolTomoSAR) data processing. The whole mission lifetime will be split into two phases: during the first phase (15 months) stacks of 7 coherent images will be acquired over all forested areas, whereas during the second phase (4 years), data will be measured in dual-baseline Polarimetric SAR Interferometry (PolInSAR) configuration, with 3 images every 7 months. This contribution proposes to compare the performance of the different observations modes of the future BIOMASS mission for the characterization of tropical forests, and to evaluate the gain in performance conferred by a synergistic combination of the two phases of the sensor’s life and by the use external information, such Lidar acquisitions etc. In particular, it provides indicators of variability for different typical descriptors of the SAR response of a forest, and tests several scenari for the fusion of multi-source information using SAR images acquired at P band in the frame of the TropiSAR campaign and auxiliary data.

Ferro Famil-Assessment Of The Performance of Polarimetric And Tomographic SAR Configurations-324Oral5.pdf


11:25am - 11:45am
Accepted
ID: 331 / Dr5 S.5.1: 3
Oral Presentation for Dragon 5
Ecosystem: 59358 - CEFO: China-Esa Forest Observation

1st Year Progress of CEFO Project (China-ESA Forest Observation)

Yong Pang1, Wen Jia1, Jacqueline Rosette2, Juan Suárez3, Zeng-yuan Li1, Xiao-jun Liang1, Tao Yu1, Shi-li Meng1, Ming Yan1

1Chinese Academy of Forestry, China, People's Republic of; 2Department of Geography, Swansea University, Swansea SA2 8PP, UK; 3Forest Research, Northern Research Station, Roslin, Midlothian EH25 9SY, Scotland, UK

This joint project combines the use of field and airborne remote sensing data to validate and calibrate innovative new satellite sensors from CNAS, ESA and NASA for forest inventory, assessment and monitoring applications in China and the UK.

The Pu’er airborne remote sensing experiment was conducted in December 2020, which covered a subtropical forest area near the city of Pu’er in the Yunnan province of China. The multi-sensor airborne remote sensing data covered 900 km2 area were acquired using two Chinese Academy of Forestry’s Airborne Observation Systems (CAF-AOS). The LiDAR point cloud has an average footprint density of 2.3 pts/m2. The hyperspectral image has a spatial resolution of 2 m and 125 bands with the spectral range spanning from 400 to 990 nm. The spatial resolution of CCD image is 0.5 m. The spatial resolution of the middle-wave infrared image and long-wave infrared image is 1.87 m and 0.85 m, respectively. Moreover, 100 forest plots, and other ground measurements such as land-use/land-cover (LULC) ground truth points, Leaf Area Index (LAI), soil moisture, and chlorophyll content were conducted during that time. These high quality airborne and ground observed data provide a great support for the validation of high resolution satellite remote sensing products.

Field data, ground based structural surveys and UAV remote sensing campaigns took place at the Aberfoyle Research Forest, Loch Lomond and Trossachs National Park, Scotland, during October – November 2019. More field data collection is planned in this area for the summer of 2021. The majority of this site comprises conifer plantations with a predominance of Sitka spruce (60%), Scots pine, Japanese larch and Norway spruce. Additionally, there are some stands of native broadleaf species. Monitoring plots for yield assessment and windthrow risk modelling were surveyed using a handheld laser scanner, enabling the beneath canopy structure and density to be recorded. Sites were overflown with UAV RGB sensors, enabling a 3D reconstruction of the canopy using structure from motion (SfM) analysis. The Research Forest was additionally flown using airborne LiDAR during April 2021, updating a sequence of previous data captured during 2002, 2006, and 2012. All the areas covered in the field and by airborne methods constitute a set of fiducial sites for the constant calibration and validation of satellite imagery. The main goals of the Scottish experiments are the creation of superior and more intensive forest inventory products, improved growth models and a more accurate production forecast system.

The airborne hyperspectral images at meter-scale spatial resolution, and centimetre resolution from UAV multispectral data, provide the capability to describe the canopy structure and monitor the physiological conditions through some narrow bands spectrum, which has great potential for forest health monitoring, forest type/tree species classification. The reflectance from airborne hyperspectral images was firstly compared with the satellite remote sensing products from Gaofen-1/2/6 and sentinel-2. Then, the forest type/tree species was classified based on the composite Sentinel-2 image with 10/20 m spatial resolution on Google Earth Engine (GEE). An automated training dataset was built based on several published land cover products. Finally, forest type/tree species classification results are generated using random forest classifier. The final forest type/tree species classification results will be evaluated using the LULC ground truth data and plots information for further analysis.

The airborne LiDAR data and UAV SfM data were used to evaluate the performances of forest vertical parameters estimation algorithms using Gaofen-7, GEDI and ICESat-2 data. Both terrain and forest parameters will be analyzed.

Pang-1st Year Progress of CEFO Project-331Oral5.pdf


11:45am - 12:05pm
Accepted
ID: 263 / Dr5 S.5.1: 4
Oral Presentation for Dragon 5
Ecosystem: 59313 - Grassland Degradation Detection and Assessment by RS

Grassland Degradation Detection and Assessment by Remotre Sensing (59313)

Zhihai Gao1,2, Xiaosong Li3, Alan Grainger4, Bin Sun1,2, Yifu Li1,2, Ziyu Yan1,2

1Institute of Forest Resource Information Techniques, Chinese Academy of Forestry, Beijing 100091, China; 2Key Laboratory of Forestry Remote Sensing and Information System, NFGA, Beijing 100094, China; 3Key Laboratory of Digital Earth Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, China; 4School of Geography ,University of Leeds, Leeds LS2 9JT, UK

As grassland is the largest terrestrial ecosystem in China, as well as the sources of many major rivers and key areas of water and soil conservation, it plays an irreplaceable role in ensuring national scale ecological security and promoting ecological civilization construction. However, as an important ecological resource, grassland ecosystem in China has been greatly degrading caused by climate change, overgrazing and other human activities. Therefore, grassland degradation monitoring and assessment have become an extremely urgent work for grassland ecological conservation and restoration of degraded grassland. Some remote sensing mapping methods of grassland types and remote sensing inversion of grassland biomass at regional scale have been put forward in previous studies, but most of them have poor universality and may not meet the needs of grassland dynamic and accurate monitoring.

In Dragon 5 project 59313, we did some scientific studies based on the geomatics methods on remotely sensed data from both European and Chinese side and other geospatial databases. In the first year of Dragon 5, joint research results have been achieved in the following two aspects:

(1) Fully employing the potential of time series sentinel-1 and sentinel-2 of ESA and the observation data of grassland shrubbery sample were used as the data sources. The shrub coverage of Xilingol grassland was estimated by correlation analysis and random forest model methods. It is of great significance for the sustainable utilization and management of grassland and the response analysis of climate change to grasp the information of large-scale grassland shrub accurately.

(2) Soil organic carbon (SOC) and soil total nitrogen (STN) are important indicators of soil quality. On the Google Earth Engine (GEE) cloud computing platform, we choose three machine learning methods: random forest (RF), support vector machine (SVM) and Multi-layer perceptron neural networks (MLP Neural Nets) models. Using Sentinel-2 of ESA, topographic factors and climatic factors carry out 30-meter resolution high-precision mapping of the 0-20cm SOC and STN content of the soil surface of grassland. High-resolution SOC and STN digital maps are of great significance for soil quality assessment and land degradation monitoring.

Gao-Grassland Degradation Detection and Assessment by Remotre Sensing-263Oral5.pdf


 
Contact and Legal Notice · Contact Address:
Privacy Statement · Conference: 2021 Dragon Symposium
Conference Software - ConfTool Pro 2.6.142+TC
© 2001 - 2021 by Dr. H. Weinreich, Hamburg, Germany