FOREST HEALTH RX: GEOSPATIAL RAPID FOREST HEALTH ASSESSMENT IN HETEROGENEOUS FORESTS

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: COMPLETE
• Funding Source: MCINTIRE-STENNIS
• Reporting Frequency: Annual
• Accession No.: 0222974
• Project Start Date: Aug 1, 2010
• Project End Date: Sep 30, 2011
• Program Code: [(N/A)]- (N/A)

Recipient Organization
UNIVERSITY OF WASHINGTON
4333 BROOKLYN AVE NE
SEATTLE,WA 98195

Performing Department
Ecosystem Sciences

Non Technical Summary
Technological advancements in hyperspatial remote sensing are explored to effectively unify LiDAR and hyperspectral imagery to assess the susceptibility of heterogeneous forests in urban and arid ecosystems to summer drought and insect stress, respectively. Methodologies developed at the UW Remote Sensing and Geospatial Analysis Laboratory in the last three years are integrated to demonstrate new potential to the remote sensing and forestry disciplines. The techniques piloted here aid in prescribing sustainable management practices for healthy forest ecosystems under pressure from climate and anthropogenic change. Moreover, the project facilitates the training of a workforce (graduate student and workshop participants) capable of utilizing state-of-the-art remote sensing tools to address current and future forest management issues while facing ever-growing social and economic needs.

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
123	0612	2020	100%

Knowledge Area
123 - Management and Sustainability of Forest Resources;

Subject Of Investigation
0612 - Conifer forests of the West;

Field Of Science
2020 - Engineering;

Keywords

hyperspatial remote sensing

lidar

hyperspectral imagery

heterogeneous forests

forest health assessment

climate change

anthropogenic change

forest management

Goals / Objectives
Human mediated changes through fire suppression, livestock grazing, logging and land use change have led to a growing concern over the integrity of Western forest ecosystems (Dellasala et al. 2004; Keane et al. 2008). In particular, there is a fear that fire suppression has resulted in stands that are overly dense and outside their historical range of variability (Hessburg et al. 2005). In urban areas one of the most common stresses to forest canopies is lack of water, which can be exasperated by the 'heat island effect' (Cregg and Dix 2001), changes in climate and further degradation by insects. The proposed development, calibration, application and demonstration of a rapid forest health assessment will focus on two forest health conditions: drought susceptibility in the urban areas and Mountain Pine Beetle susceptibility in semi-arid Eastern Washington. Both forest types are very heterogeneous and thus pose a challenge to the traditional precision forestry remote sensing approaches (Moskal et al. 2009). This project is built on the following hypothesis: In order to understand and thus sustainably manage, forest health and ecosystem services, a monitoring approach needs to link, harmonize and integrate a multi-scale and multi-resolution array of remote sensing platforms capable of assessing the structural and spectral conditions. The main goal of the proposed research is to develop such an approach. To achieve this goal the following research objectives will be pursued: (1) Develop a dynamic framework for unifying LiDAR and hyperspectral data; (2) Calibrate the framework through validation with existing field and other datasets; (3) Apply the framework to monitoring of forest health and ecosystem services in heterogeneous landscapes; and, (4) Demonstrate the framework approach at a workshop hosted by the PFC.

Project Methods
Two remotely sensed data types will be necessary for the project: high point density discreet LiDAR with at least 4 returns per 8 points in a m2 and hyperspectral imagery with a spatial resolution of 5 meters or finer in the range of 400-2500 nanometers with band increments of no less than 10 nanometers. The dynamic framework for unifying LiDAR and hyperspectral remotely sensed data (PHASE I) will combine methods by utilizing the work of two of Moskal's former graduate students: Kato et al. (2009) and Erdody and Moskal (2009). Second, satellite hyperspectral datasets have already been acquired in 2008 by PFC and will be used as an auxiliary dataset. An alternate 2008 LiDAR dataset also exists for an area in the Washington Park Arboretum (Richardson et al. 2009) and can be used as a substitute. The remotely sensed data will be acquired and analyzed for two pilot sites. First, for an urban area within the King Country boundaries that captures a gradient from high to low density. The extent of this area is constrained by flight limitations and acquisition costs. We will have access to appropriate field data provided by others. The second pilot site will be in the Coleville National Forest, for which a 2008 LiDAR dataset is already available, thus a larger hyperspectral dataset will be acquired. Both LiDAR and field plot data (60 sites) were acquired in 2008. Furthermore, 2009 and 2010 aerial flights identifying Mountain Pine Beetle infested areas will be utilized to determine susceptibility to attack using the spectral and structural information from the remotely sensed data. The calibration (PHASE II) between LiDAR and field data and auxiliary data sets will rely on all of the above mentioned data. In PHASE III the above datasets will be utilized to assess the susceptibility to drought (urban) and Mountain Pine Beetle (Coleville). The metrics derived from the LiDAR and imagery will follow a technique already applied to LiDAR and non-hyperspectral imagery in Canada by Barter et al. (2010), but the authors state the results could be improved with an increase in spectral information such as that of hyperspectral data. The Canadian study was unable to address the issue of susceptibility due to the lack of aerial flights delineating past attacks, thus, this work would further the research beyond the attack stage (red) to the susceptibility stage (green). In PHASE IV of the project the pilot projects will be utilized as technology exchange through demonstration tools during a one day workshop hosted at the University of Washington by the Precision Forestry Cooperative. The two pilot projects, one in an urban area and one in an arid eastern Washington forest will demonstrate that precision forestry tools including hyperspatial remote sensing with LiDAR and hyperspectral imagery are relevant, interchangeable and applicable in both urban and non-urban forests.

Progress 08/01/10 to 09/30/11

Outputs
OUTPUTS: This project leveraged partnerships between the UW Precision Forestry Cooperative, public (USDA Forest Service), private (Watershed Sciences, Inc.) and non-profit (Cascade Land Conservancy) agencies to develop, calibrate, apply and demonstrate a rapid forest health assessment tool. We completed the following milestones: 1. Applied and made necessary adjustments of developed methodology to an urban study site (South Seattle). a. Applied the framework developed and tested for Arboretum (semi-urban study site) on the datasets acquired for South Seattle Zip Code (urban study site). b. Adjusted the segmentation and classification algorithms to incorporate the urban structure of the environment (the biggest challenge was to discern individual trees from man-made objects). 2. Compiled a hyperspectral signatures (126 bands) dataset incorporating all the trees used in the study. 3. Analyzed the hyperspectral data for comparison (band range and width) of two hyperspectral sensors: HyMap and CASI, for the purposes of distinguishing various tree species and tree health conditions within the species. PARTICIPANTS: L.Monika Moskal, Assistant Professor, University of Washington School of Forest Resources. Miles Logsdon, Reseach Assistant Professor, University of Washington School of Oceanography. Dylan G. Fischer, Faculty, Evergreen State College and the Evergreen Ecological Observation Network. Alexandra Kazakova, graduate student, University of Washington School of Forest Resources. TARGET AUDIENCES: Target audiences include the remote sensing research community, forest managers, educators, students, and the general public. PROJECT MODIFICATIONS: Not relevant to this project.

Impacts
List of deliverables: 1) Georeferenced (using geometric look up values) Hymap imagery (5 flightlines). 2) Individual tree segments for the full extent of the study areas (shapefile). 3) OBIA classification and segmentation algorithm used for the creation of individual tree crown segments. 4) Segment centers (representing the geometrical center of each tree) (shapefile). 5) Apex points of each tree (the tallest point of the tree) (shapefile). 6) Data table that contains all the sampled trees and their spectral values in 125 bands. 7) A presentation containing graphs of various tree species signatures, band separations, tree health signatures (healthy vs unhealthy), and spectral variability within and between species.

Publications

• Kazakova, A. N., Moskal, L.M., and Styers, D.M. 2011. Hyperspectral remote sensing of urban tree species, 2011 AAG Annual Meeting, Seattle, WA, AAG Paper Session:5219 Advancements in Hyperspectral Remote Sensing. Abstract available online at: http://meridian.aag.org/callforpapers/program/AbstractDetail.cfmAbst ractID=36105, link verified 1/6/2012.