Source: STATE UNIV OF NEW YORK submitted to NRP
CHARACTERIZATION OF MONTANE FOREST ECOSYSTEMS USING ADVANCED REMOTE SENSING TECHNOLOGY
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
MCINTIRE-STENNIS
Reporting Frequency
Annual
Accession No.
0222204
Grant No.
(N/A)
Cumulative Award Amt.
(N/A)
Proposal No.
(N/A)
Multistate No.
(N/A)
Project Start Date
Aug 15, 2010
Project End Date
Sep 30, 2012
Grant Year
(N/A)
Program Code
[(N/A)]- (N/A)
Recipient Organization
STATE UNIV OF NEW YORK
(N/A)
SYRACUSE,NY 13210
Performing Department
Environmental Resources Engineering
Non Technical Summary
While Adirondack forests are currently protected by state policy, they may not be resilient to climate and other drivers of changes because of past management and ongoing stressors such as acid deposition. Nonetheless, research on the relationships between drivers of changes and forest ecosystem characteristics and health in the Adirondacks is very limited due to lack of high quality, regionally extensive forest inventory data, especially in the nearly three million acres of state-owned Forest Preserve. Remote sensing has been utilized for effective and efficient forest management, but advanced remote sensing applications over montane areas such as the Adirondacks are still in their infancy. This project will focus on a portion of the Adirondacks and utilize multi-sensor remote sensing data to characterize the montane forest ecosystems. The goal of this research is to fill a knowledge gap about the Adirondack Forest Preserve using advanced remote sensing technology and data fusion techniques, thus improve forest management and modeling, and conservation planning in the Adirondacks. This research will 1) investigate advanced remote sensing methodologies through the synergistic use of hyperspectral imagery and high posting density LiDAR data for a portion of the Adirondacks to identify the spatial distribution of local forest communities and their biophysical/structural parameters, and 2) quantify changes in the spatial distribution of local forest communities over the past 25 years using multispectral satellite imagery. Artificial intelligence (i.e., decision/regression trees and neural networks) will be used for classification and biophysical parameter estimation. Typical statistical approaches such as the maximum likelihood method for classification and regression analysis for biophysical parameter estimation will also be used. This research will provide methods to identify forest composition and its changes over time, and biophysical and structural parameters across vast areas of unmanaged Forest Preserve lands, where knowledge is limited and resources for forest inventory are lacking. These methods can be applied to other similar montane areas particularly in the Northern Forest region. This research will also provide the Adirondacks with the accurate spatial forest baseline data, with an advanced technological framework for regular updating and validation over time. Successful results from this research will provide a basis for future long-term research for monitoring the Adirondacks and surrounding Northern forest ecosystems under changing environments.
Animal Health Component
(N/A)
Research Effort Categories
Basic
(N/A)
Applied
(N/A)
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
1230613107020%
1230613202010%
1230613206010%
1237210107020%
1237210202020%
1237210206020%
Knowledge Area
123 - Management and Sustainability of Forest Resources;

Subject Of Investigation
7210 - Remote sensing equipment and technology; 0613 - Mixed conifer-broadleaf forests;

Field Of Science
2020 - Engineering; 1070 - Ecology; 2060 - Geography;
Goals / Objectives
This project will utilize multi-sensor (multispectral, hyperspectral and LiDAR) remote sensing data to characterize Adirondack forest ecosystems. Our goal is to fill a knowledge gap about the Adirondack Forest Preserve using advanced remote sensing technology and data fusion techniques, thus improve forest management and modeling, and conservation planning in the Adirondacks. The objectives of this research are to (1) investigate the individual and synergistic use of hyperspectral and LiDAR remote sensing data to estimate the biophysical/structural parameters (i.e., LAI, biomass, DBH, canopy height) of local forest communities in a portion of the Adirondacks, (2) identify the spatial distribution of local forest communities using both hyperspectral and high density LiDAR data, and (3) quantify the spatial distribution changes of forest communities over the past 25 years using multispectral satellite imagery. While the first two objectives are to accurately characterize the current status of local forest communities, the third one is to identify spatial changes of local forest communities over time based on the current status. This research will provide novel multi-sensor data fusion methods to accurately characterize the health and dynamics of the Adirondack forest communities. These methods can be applied to other similar montane areas in the Northern Forest region. This research will provide the Adirondacks with the accurate spatial forest baseline data, with an advanced technological framework for regular updating and validation over time. This work is also intended to considerably improve policy decisions and management strategies under changing environments in the Adirondacks. Using the accurate and timely remote sensing-based monitoring and assessment results, decision makers or managers can determine effective responses and actions. This research will lay the groundwork for a long-term research program into the effects of climate and other drivers of changes on forest health, dynamics and biodiversity in order to promote sustainable forest management in the Adirondacks and the Northern Forest region by properly utilizing advanced remote sensing technology.
Project Methods
This research will focus on Huntington Wildlife Forest (HWF), located in the central Adirondacks. HWF is a 15,000 acre research forest with comprehensive and long-term forest inventory data (over 300 Continuous Forest Inventory (CFI) plots sampled since 1950) and containing one of the longest-term climate and acidic deposition monitoring sites in the Adirondacks. HWF offers a range of forest types and conditions with various disturbance histories. HWF is an ideal site for a remote sensing study because of the availability of forest inventory and environmental monitoring data on-site, and the presence of the ESF-operated Adirondack Ecological Center (AEC) and associated research and support facilities. A satellite remote sensor, the Hyperion imaging spectrometer, will be used to collect hyperspectral imagery over HWF. Hyperion is one of the sensors onboard the NASA Earth Observing-1 spacecraft. It collects data at 220 bands from 400 - 2,500 nm at a nominal ground sampling distance of 30 x 30 m. High posting density (> 4 points per square meter) LiDAR data will be collected using a Leica ALS50 LiDAR system by Kucera International, Inc. Both satellite Hyperion and airborne LiDAR data will be collected over HWF approximately the same week during the leaf-on season of 2011 to reduce any errors caused by the discrepancy of data collection dates. Prior to and during the remote sensing data collection, in situ ground measurement data will be collected to calibrate and validate remote sensing products. In situ data include the forest species identity, the biophysical (LAI, biomass) and structural (DBH, canopy height) parameters of the local forest communities, and spectral reflectance using a handheld spectroradiometer. Artificial intelligence (i.e., decision/regression trees and neural networks) will be used to map forest species and estimate biophysical parameters through data fusion (LiDAR and hyperspectral data). Typical statistical approaches such as the maximum likelihood method for classification and regression analysis for biophysical parameter estimation will also be used. In order to identify spatial distribution changes of local forest communities over the past 25 years, multi-temporal Landsat satellite TM data will be investigated. About three Landsat TM data pair (i.e., leaf-on and leaf off seasons) collected around anniversary dates with minimum cloud cover at 8-year interval. The CFI data will be used as ground reference data for old Landsat data analysis, and the same classification logic (i.e., artificial intelligence) will be applied to map the forest communities over time. As some forest species are sensitive to topographic characteristics (i.e., elevation and aspect) ancillary data such as elevation will be incorporated. Landsat data classification over time will focus on two types of forests (i.e., northern-hardwood forest and boreal forest) for a more reliable and robust investigation. The multi-temporal analysis will use the artificial intelligence techniques as well as the conventional approaches mentioned above.

Progress 08/15/10 to 09/30/12

Outputs
OUTPUTS: Full waveform lidar data and products generated during the project were distributed to communities of interest via the NYView website or in person. In this project, we used multitemporal Landsat TM satellite images and full waveform lidar data to characterize the montane forest in the Huntington Wildlife Forest in the Adirondacks. We investigated the forest change over the past 25 years using the Landsat images and quantified the aboveground biomass and carbon stocks using the airborne lidar data. We found that remote sensing-based approaches are very efficient for characterizing montane forests and can be used to provide reliable quantitative results such as forest species maps, and biomass and carbon stock maps. The data and products were distributed to the public via the NYView website, and the research findings were published (will be published) in the peer-reviewed journal articles. PARTICIPANTS: Colin Gleason, MS student at ESF, now PhD student at UCLA He tested and developed some of the methodologies used in this project and tested for other regions using airborne lidar data. Manqi Li, MS student at ESF She performed core data analyses for this project. Most of the research findings from this project is included in her thesis. TARGET AUDIENCES: Not relevant to this project. PROJECT MODIFICATIONS: Not relevant to this project.

Impacts
Machine learning methodologies used in this project were implemented as a standalone software tool with support from another research project. The tool is "Advanced Remote Sensing Information Extractor," which is currently in the final stage and will be distributed to the public soon. Full waveform lidar data collected over the HWF study site and associated products are also distributed to the communities of interest. The data and products will be helpful not only to geospatial scientists and forest managers or ecologists, but also to hodrologists and biologists, who are especially interested in Adirondack forests.

Publications

  • Gleason, C., Im, J. (2012). A fusion approach for tree crown delineation from LiDAR data, Photogrammetric Engineering & Remote Sensing, 78(7): 679-692
  • Li, M., Im, J., Liu, T., Quackenbush, L.J. (2013). Forest biomass and carbon stock quantification using full waveform LiDAR data in montane forests, ISPRS Journal of Photogrammetry and Remote Sensing, in review.
  • Li, M., Im, J., Beier, C. (2013). Machine learning approaches for forest classification and change analysis using multi-temporal Landsat TM images over Huntington Wildlife Forest, Journal of Applied Remote Sensing, in re-review.
  • Gleason, C., Im, J. (2012). Forest biomass estimation from airborne LiDAR data using machine learning approaches, Remote Sensing of Environment, 125: 80-91.


Progress 10/01/10 to 09/30/11

Outputs
OUTPUTS: We classified forest type (deciduous and coniferous) over Huntington Wildlife Forest using multi-temporal Landsat TM images collected in 1991, 2001, and 2011. These collection dates are correspondent to those of the CFI plot data survey. Three forest type maps were generated and are ready for distribution. Full waveform LiDAR remote sensing data was collected early in September 2011 by Kucera International, Inc. We received the post-processed LiDAR points generated from discrete return data. We processed them to generate several surface layers including 1m and 2m Digital Terrain Models (DTMs) and 0.5m and 1m Digital Surface Models (DSMs). These products are ready for distribution through the NYView website. The waveform signals are still under processing by the vendor and the final waveform LiDAR products will be available in March 2012. PARTICIPANTS: Manqi Li (graduate student) - research assistant; Colin Gleason (graduate student) - funded from GA and worked closely with the project; Zhenyu Lu (graduate student) - funded from GA and participated in lidar data processing and algorithm development TARGET AUDIENCES: Nothing significant to report during this reporting period. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.

Impacts
A novel tree crown delineation method (COTH) using a synthesis of genetic algorithm optimized object recognition, statistical prediction interval maxima filtering, and hill climbing to associate crown objects to treetops was developed through this project and successfully tested using LiDAR data over Heiberg Memorial Forest. The delineated tree crowns provided model estimates of individual tree biomass and the effects of delineation accuracy on biomass modeling was investigated using machine learning methods. These tree crown delineation and biomass estimation methods have not been tested over Huntington Wildlife Forest (i.e., study area of this project) because the waveform LiDAR data have not been fully processed. The multi-temporal forest type classification provided supportive results about upslope shift in the Northeastern hardwood - boreal forest ecotone in Huntington Wildlife Forest.

Publications

  • Li, M., J. Im, and C. Beier, Forest classification and change analysis using multi-temporal Landsat TM images over Huntington Wildlife Forest in the Adirondacks, Journal of Applied Remote Sensing, to be submitted in February, 2012.
  • Gleason, C. and J. Im, 2012. Forest biomass estimation from airborne LiDAR data using machine learning approaches, Remote Sensing of Environment, in revision.
  • Gleason, C. and J. Im, 2012. A fusion approach for tree crown delineation from LiDAR data, Photogrammetric Engineering & Remote Sensing, in press.
  • Gleason, C. and J. Im, 2011. A review of remote sensing of forest biomass and biofuel: options for small area applications, GIScience and Remote Sensing, 48(2): 141-170.


Progress 08/15/10 to 12/21/10

Outputs
OUTPUTS: Activities: New remote sensing data (i.e., Hyperion hyperspectral imagery and multiple returns LiDAR data) for this project are scheduled to be collected over Huntington Wildlife Forest (HWF) during summer 2011 to investigate forest biophysical and structural characteristics. In 2010, several research activities were conducted to successfully prepare the upcoming remote sensing data collection and subsequent data analysis. Firstly, an extensive literature review on remote sensing of forest biomass was conducted. Secondly, another LiDAR remote sensing mission was conducted over Heiberg Memorial Forest (HMF) near Syracuse, NY in August 2010. Although the site (i.e., HMF) has different characteristics from HWF, remote sensing-based methods that we are developing can be tested with the HMF data. We have been evaluating the HMF LiDAR data (> 8 pts/m2) to identify various forest parameters including tree height, tree crown, LAI, and biomass. We also had field trips to collect reference data from HMF during summer 2010 to calibrate and validate the remote sensing-derived products. This HMF data analysis is expected to be completed by May 2011. The remote sensing methodologies and results from the HMF data analysis will be evaluated using the HWF data in 2011. Finally, we started forest type classification and change analysis using multi-temporal satellite Landsat data (i.e., 1991, 2001) and CFI data of HWF. Two graduate students were involved in these three activities (refer to Participants below). The research findings will be disseminated through presentation at National conferences (American Society for Photogrammetry and Remote Sensing Annual Conference, Milwaukee, WI; Association of American Geographers Annaul Conference, Seattle, WA) in spring 2011. Project results have also been and/or will be submitted for peer review in several journals. PARTICIPANTS: Manqi Li was a new MS student employed as a Research Project Assistant on the project. Colin Gleason was a MS student employed as a Graduate Assistant, but participated in this project to pursue his thesis research focusing on literature review and preliminary analyses of forest biomass using remote sensing for a different forest area. TARGET AUDIENCES: Not relevant to this project. PROJECT MODIFICATIONS: Not relevant to this project.

Impacts
An extensive literature review on remote sensing of forest biomass was conducted and the result was submitted to a journal for peer review (please refer to Publications below). The other research activities are currently under investigation or planned to be conducted in 2011. So, detailed outcomes/impacts will be available in the next annual report. From the ongoing activities mentioned above, we are expecting to submit several papers to major remote sensing and/or forest-related journals. The tentative list (to be submitted in 2011) is: 1. Gleason, C., J. Im, and L.J. Quackenbush, 2011. Determining tree crown dimensions from airborne LiDAR data using a modified hill climbing algorithm and object recognition, Photogrammetric Engineering & Remote Sensing, in prep (February, 2011) 2. Gleason, C. and J. Im, 2011. Quantifying forest biomass for biofuel production from airborne LiDAR data using a transferrable model., Remote Sensing of Environment, in prep (April, 2011) 3. Im, J., M. Li, and M. Dovciak, 2011. Quantification of boreal forest upshift due to climate change, Forest Ecology and Management, in prep (May, 2011) 4. Li, M., J. Im, L.J. Quackenbush, and C. Beier, 2011. Characterization of Huntington Wildlife Forest in the Adirondacks using remote sensing, Remote Sensing of Environment, in prep (December, 2011)

Publications

  • Gleason, C. and J. Im, 2010. A review of remote sensing of forest biomass and biofuel: options for small scale applications. GIScience and Remote Sensing, in review.