ASSESSING DECIDUOUS FOREST STRUCTURE IN CONNECTICUT USING IMAGERY AND LIDAR

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: COMPLETE
• Funding Source: MCINTIRE-STENNIS
• Reporting Frequency: Annual
• Accession No.: 0211809
• Project Start Date: Oct 1, 2007
• Project End Date: Sep 30, 2010
• Program Code: [(N/A)]- (N/A)

Recipient Organization
UNIV OF CONNECTICUT
438 WHITNEY RD EXTENSION UNIT 1133
STORRS,CT 06269

Performing Department
Natural Resources & the Environment

Non Technical Summary
Maturing and late successional forest ecosystems are scarce in Connecticut yet highly valued for their structural and concomitant biological diversity. Because a forest becomes more structurally complex with age, the surface of the canopy also becomes `rougher.' We plan to identify the location and extent of mature and late successional forests by analyzing the texture of the forest canopy from aerial photos, combined with Light Detection And Ranging (LiDAR) data of topography.

Animal Health Component

30%

Research Effort Categories

Basic

10%

Applied

30%

Developmental

60%

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
123	0620	1070	50%
123	0620	2020	40%
123	0620	2060	10%

Knowledge Area
123 - Management and Sustainability of Forest Resources;

Subject Of Investigation
0620 - Broadleaf forests of the North;

Field Of Science
2020 - Engineering; 1070 - Ecology; 2060 - Geography;

Keywords

forest structure

lidar

remote sensing

canopy texture

forest canopy

forest developmental stage

Goals / Objectives
1. identify one or more potential study areas within the state of Connecticut for which high resolution multispectral digital aerial remote sensing and airborne LiDAR data are available, and in which representative forest types are located, 2. for the selected study area(s), compile aerial remote sensing imagery and LiDAR data with urban and other unsuitable areas excluded from further examination, to improve our ability to identify correctly mature canopy structure, 3. design and execute an algorithm that would combine imagery and LiDAR data to identify a range of canopy textures and their locations on the landscape, and 4. validate the forest ecosystems identified as mature and old growth by field assessment consisting of identification of species, aging dominant trees by taking tree cores and estimating the amounts of coarse woody debris.

Project Methods
By using existing remote sensing imagery and LiDAR data in the possession of the Department of Natural Resources Management and Engineering (NRME) and the Center for Land Use Education and Research (CLEAR), data costs for this project will be kept low or nil. During the first year, the graduate student will compile imagery for select, predominantly forested areas. LiDAR data will be examined to confirm canopy heights are appropriate for mature and late successional growth forests. The first summer will be spent establishing at least 30-40 (GPS located) control points in the midst of known canopy structure which will enable us to establish the spectral signatures of four texture/structural classes. For each frequency, summing values on all possible traveling directions on a one hectare window will yield an accumulated radial spectrum, which quantifies canopy coarseness. The four textural classes will approximately correspond to phases of stand development (Oliver 1980); stand initiation, stem exclusion, understory reinitiation (mature) and finally old-growth or late successional. We expect fewer reference points to be located in the stand initiation and late successional development phases as, in Connecticut, these stands are fewer in number. The second year will focus on the design and execution of an algorithm that will combine imagery and LiDAR data at established reference points to identify the spectral signature of the four canopy textures. LiDAR data will aid in the correction of topographic shadows that could skew the spectral signature from the FFT. The second summer will also include validation of the algorithm by field assessment. We will apply the texture analysis technique to forest areas considered a conservation priority to locate late successional structure therein. After applying the algorithm to select forests, areas identified as mature and late successional will be investigated by field assessment. The success rate of correctly identified structures will allow us to asses the usefulness of our technique. Field validation will consist of identification and measurement (height and diameter) of dominant tree species, and taking tree cores of dominant trees for aging, and assessment of woody debris. Aging of dominant trees will be done in our tree ring lab after mounting and preparation of core samples. At least 10 dominant trees will be sampled from each site. An important component of late successional forests (Leverett 1996) decay class and amounts of coarse woody debris will be measured using fixed radius plots. The third year of the project will incorporate any refinements to the algorithm and begin applying the model to other areas of the state for which high resolution multispectral aerial imagery and LiDAR data are available. Results would be shared with state government and groups interested in forest conservation.

Progress 10/01/07 to 09/30/10

Outputs
OUTPUTS: On-the-ground surveys of forest stands to quantify forest structure can be expensive and time consuming. This project addresses whether the use of remotely sensed data and algorithms such as one dimensional discrete Fourier transforms and grey level co-occurrence matrices, could provide an efficient means of assessing different forest structure. This approach is enabled by using canopy roughness as a surrogate of forest structure. During the summer of 2008 the graduate student Nick McIntosh collected the field data needed to validate the image analysis. Three study sites with unique canopy structure were established across Connecticut. Sites were located in Nehantec State Park in Lyme, the Natchaug State Park in Chaplin and the Yale Meyers Forest in Eastford. At each site, the graduate student Nick McIntosh led field crews to establish two linear transects consisting of four 25 meter radius plots, spaced 100 meters apart. Each plot center was GPS located for subsequent alignment with aerial imagery. Field data collected at each plot were: tree density, stem and crown diameters, tree heights and species. Tree increment cores were taken for a subset of trees in each plot to determine forest age. Cores were prepared (mounted, sanded) and counted in the Rudnicki lab. With aid and instruction from PI's Nick compiled, processed and analyzed the aerial imagery. Image data used was from the 2008 NAIP (National Agriculture Imagery Program) and was provided by the CLEAR (Center for Landuse Education and Research) at the University of Connecticut. The Imagery was cropped into 64X64 meter squares and with GPS locations established at each field plot center, Nick was able to co-locate the aerial imagery. Images were first analyzed with principle components analysis to increase the contrast of the pixels within each image. Correlation analysis was then conducted on each image to determine the optimal sampling scheme for the Fourier analysis. The (optimized) Fourier analysis was then applied to each image and spectra were averaged for each site to determine representative spectral signatures for each forest structure/texture. This process led to the realization that another sampling site, representing a very young and smooth canopy texture, that was added in the spring of 2009. Nick McIntosh was tutored in the use of Mathematica, Ecognition, Erdas Imagine, ArcGIS to conduct spectral analysis of the imagery. He has been tutored in technical writing to be able to prepare products based on this project. Nick has completed compiling, processing and analyzing the aerial imagery. Through his guided and independent reading of the primary literature he has acquired the ability to contextualize research findings to address the concerns of a resource manager. The PI's Rudnicki, Meyer and Civco all contributed to Nick's instruction and mentoring in the execution of the project tasks including the preparation of products in the form of a presentation at the 2009 Connecticut Conference on Natural Resources and a written master's thesis. Nick McIntosh also gave a public seminar at the University of Connecticut of this research project in June of 2010. PARTICIPANTS: The project PI Mark Rudnicki has provided overall project coordination and planning with specific guidance in establishment of study plots and collection of field validation data. Co-PI Thomas Meyer has worked closely with the PI's in project planning and grad student guidance. Specifically Dr. Meyer has provided guidance/training for the student to efficiently conduct image analysis (Fourier transforms and correlations). Co-PI Daniel Civco has also worked closely with the other PI's to plan and guide the project and student, supplying insight into traditional methods of image analysis. Dr. Civco has also supported the project via supplying the necessary imagery and support staff from CLEAR (center for landuse education and research), namely James Hurd and Jason Parent. Graduate student Nick McIntosh has dramatically improved his image analysis skills through this project and learned not only how novel applications of accepted techniques can yield new knowledge but is learning how the process of science and inquiry works. TARGET AUDIENCES: Maintaining diversity of successional stages across the landscape is critical to maintain ecosystem function and diversity across the Connecticut landscape. As much of New England is severely lacking in the late successional growth end of the spectrum state forest and wildlife managers as well as NGO's such as the Nature Conservancy are very interested to locate areas that may be refugia for biodiversity within the state. PROJECT MODIFICATIONS: Not relevant to this project.

Impacts
Nick McIntosh has become well trained in the use of Mathematica, Ecognition, Erdas Imagine, ArcGIS and their application to research analysis. The student has learned statistical analysis and has examined the image processing outputs to determine their significance. He has learned technical writing skills and persisted through several difficult revisions to reach a final professional product. Nick McIntosh learned how to deliver a scientific presentation on his preliminary findings at the regional conference "Connecticut Conference on Natural Resources" on March 9th 2009. This conference is aimed at mangers, scientists and policy makers, many of whom will find this research interesting as it may be applied to help prioritize conservation efforts in the state and possibly throughout the eastern deciduous forest. Nick demonstrated an enormous improvement in his ability to present scientific research during the (June 2010) public seminar associated with the defense of his masters research project.

Publications

• Nicholas A. McIntosh (2010) Assessing the Accuracy of Predicting Connecticut Forest Attributes by One-Dimensional Discrete Fourier Transforms and Grey Level Co-Occurrence Matrices. Master of Science Thesis, University of Connecticut. Storrs, CT. 143pp.

Progress 01/01/09 to 12/31/09

Outputs
OUTPUTS: On-the-ground surveys of forest stands to quantify forest structure can be expensive and time consuming. This project addresses whether the use of remotely sensed data and algorithms, such as one dimensional discrete Fourier transforms and grey level co-occurrence matrices, could provide an efficient means of assessing different forest structure. This approach is enabled by using canopy roughness as a surrogate of forest structure. This year's activities' outputs for the project graduate student (Nick McIntosh) include collection of field data, analysis of imagery and field data. The PI's Rudnicki, Meyer and Civco all contributed to the teaching and mentoring of the graduate student in the execution of the project tasks including the preparation of a project product in the form of a master's thesis. Nick McIntosh was tutored in the use of Mathematica, Ecognition, Erdas Imagine, ArcGIS to conduct spectral analysis of the imagery. He has been tutored in technical writing to be able to prepare products based on this project. Nick has completed compiling, processing and analyzing the aerial imagery. Through his guided and independent reading of the primary literature he has acquired the ability to contextualize research findings to address the concerns of a resource manager. PARTICIPANTS: The project PI Mark Rudnicki has provided overall project coordination and planning. Co-PI Thomas Meyer has provided guidance/training for the student to efficiently conduct image analysis (Fourier transforms and correlations). Co-PI Daniel Civco has guided the student, supplying insight into traditional methods of image analysis. Dr. Civco has also supported the project via supplying the necessary imagery and support staff from CLEAR (center for landuse education and research), namely James Hurd. TARGET AUDIENCES: Maintaining diversity of successional stages across the landscape is critical to maintain ecosystem function and diversity across the Connecticut landscape. As much of New England is severely lacking in the late successional growth end of the spectrum, state forest and wildlife managers as well as NGO's such as the Nature Conservancy are very interested to locate areas that may be refugia for biodiversity within the state. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.

Impacts
Nick McIntosh has become well trained in the use of Mathematica, Ecognition, Erdas Imagine, ArcGIS and their application to research analysis. The student has learned statistical analysis and has examined the image processing outputs to determine their significance. He has learned technical writing skills.

Publications

• No publications reported this period

Progress 01/01/08 to 12/31/08

Outputs
OUTPUTS: The project has been going well and we on track with our stated objectives in terms of both field work and image processing. During the summer of 2008 the graduate student Nick McIntosh collected the field data needed to validate the image analysis. 3 study sites with unique canopy structure were established across Connecticut. Sites were located in Nehantec State Park in Lyme, the Natchaug State Park in Chaplin and the Yale Meyers Forest in Eastford. At each site, the graduate student Nick McIntosh led field crews to establish two linear transects consisting of four 25 meter radius plots, spaced 100 meters apart. Each plot center was GPS located for subsequent alignment with aerial imagery. Field data collected at each plot were: tree density, stem and crown diameters, tree heights and species. Tree increment cores were taken for a subset of trees in each plot to determine forest age. Cores were prepared (mounted, sanded) and counted in the Rudnicki lab. Nick has also made excellent progress in compiling, processing and analyzing the aerial imagery. Image data used was from the 2008 NAIP (National Agriculture Imagery Program) and was provided by the CLEAR (Center for Landuse Education and Research) at the University of Connecticut. The Imagery was cropped into 64X64 meter squares and with GPS locations established at each field plot center, Nick was able to co-locate the aerial imagery. Images were first analyzed with principle components analysis to increase the contrast of the pixels within each image. Correlation analysis was then conducted on each image to determine the optimal sampling scheme for the Fourier analysis. The (optimized) Fourier analysis was then applied to each image and spectra were averaged for each site to determine representative spectral signatures for each forest structure/texture. This process led to the realization that another sampling site, representing a very young and smooth canopy texture, needed to be added and we will be doing so in the spring of 2009. PARTICIPANTS: The project PI Mark Rudnicki has provided overall project coordination and planning with specific guidance in establishment of study plots and collection of field validation data. Co-PI Thomas Meyer has worked closely with the PI's in project planning and grad student guidance. Specifically Dr. Meyer has provided guidance/training for the student to efficiently conduct image analysis (Fourier transforms and correlations). Co-PI Daniel Civco has also worked closely with the other PI's to plan and guide the project and student, supplying insight into traditional methods of image analysis. Dr. Civco has also supported the project via supplying the necessary imagery and support staff from CLEAR (center for landuse education and research), namely James Hurd and Jason Parent. Graduate student Nick McIntosh has dramatically improved his image analysis skills through this project and learned not only how novel applications of accepted techniques can yield new knowledge but is learning how the process of science and inquiry works. TARGET AUDIENCES: Maintaining diversity of successional stages across the landscape is critical for continuing preservation of ecosystem function and diversity across the landscape. As much of New England is severely lacking in the late successional growth end of the spectrum state forest and wildlife managers as well as NGO's such as the Nature Conservancy are very interested to locate areas that can be refugia for biodiversity within the state. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.

Impacts
The Graduate student Nick McIntosh will give an oral presentation on his preliminary findings at the regional conference "Connecticut Conference on Natural Resources" on March 9th 2009. This conference is aimed at mangers, scientists and policy makers, many of whom will find this research interesting as it may be applied to help prioritize conservation efforts in the state and possibly throughout the eastern deciduous forest.

Publications

• No publications reported this period

Progress 01/01/07 to 12/31/07

Outputs
In this first couple of months we have secured a graduate student, Nick McIntosh, to work on the project for partial fulfillment of a master's thesis. With an excellent Geomatics background, Nick has begun gathering the imagery inventory for the state to select the area of focus for the project.

Impacts
We do not have outcomes to report yet.

Publications

• No publications reported this period