Forest Assessments with LiDAR: Forest Assessments with LiDAR: from Research to Operational from Research to Operational

ProgramsPrograms

David L. EvansDavid L. EvansDepartment of ForestryDepartment of Forestry

Forest and Wildlife Research CenterForest and Wildlife Research CenterMississippi State UniversityMississippi State University

Aerial Aerial Photography:Photography:

Stand MappingStand Mapping

Manual Manual MeasurementsMeasurements

Forest Remote Sensing: Forest Remote Sensing: Then and NowThen and Now

Satellite Data: Satellite Data:

Type MappingType Mapping

Could Not See Could Not See the Treesthe Trees

LandsatLandsat

HighHigh--resolution digital imagery: resolution digital imagery: can see the trees, can see the trees,

hard to get measurementshard to get measurements

Presentation OutlinePresentation OutlineProfiling LiDAR ResearchProfiling LiDAR Research

Scanning LiDAR Research and Scanning LiDAR Research and

ApplicationsApplications–– Individual tree measurementsIndividual tree measurements

–– Timber inventoryTimber inventory

–– Forest structure analysisForest structure analysis

–– Stand visualizationStand visualization

Profiling LiDAR ConceptProfiling LiDAR Concept

Nelson, Krabill, and MacLean, 1984Nelson, Krabill, and MacLean, 1984

LiDAR ProfileLiDAR Profile

Early Profiling LiDAR FindingsEarly Profiling LiDAR Findings

COV of laser heights indicated canopy COV of laser heights indicated canopy structurestructure–– High COV = open canopy (regeneration High COV = open canopy (regeneration

and mature standsand mature stands–– Low COV = closed canopy (midLow COV = closed canopy (mid--rotation)rotation)Average tree height matched closely Average tree height matched closely with average height of the upper 5with average height of the upper 5--10% 10% of laser measurementsof laser measurements

Scanning LiDARScanning LiDAR◄ Typical multi-return

data acquisition

Courtesy of the Spencer B. Gross, Inc.website Courtesy of the Spencer B. Gross, Inc.website

▼ An AeroScan LiDAR Sensor

HiHi--Res MultiRes Multi--spectral + LiDARspectral + LiDAR

LiDAR Study SitesLiDAR Study Sites

Coastal forests in WA Coastal forests in WA (Douglas(Douglas--fir, western fir, western hemlock, red alder)hemlock, red alder)Interior forest in ID Interior forest in ID (Ponderosa pine, (Ponderosa pine, DouglasDouglas--fir, lodgepole fir, lodgepole pine, grand fir, pine, grand fir, Engelmann spruce, Engelmann spruce, subalpine fir)subalpine fir)Southern coastal plain in Southern coastal plain in MS, TX, GA, LA (loblolly MS, TX, GA, LA (loblolly pine, mixed oak/gum pine, mixed oak/gum hardwoods)hardwoods)RCW habitat study in NC RCW habitat study in NC (longleaf and loblolly (longleaf and loblolly pine)pine)

TXMS GA

WA

ID

LA

NC

Measurements of InterestMeasurements of Interest

Stem densityStem densityTotal heightTotal heightDBHDBHHeight to base of live crown (HBLC)Height to base of live crown (HBLC)Crown diameter Crown diameter Understory densityUnderstory density

Tree Recognition and Tree Recognition and MeasurementMeasurement

Generate canopy and ground surfacesGenerate canopy and ground surfacesVariable radius focal maximum filterVariable radius focal maximum filter–– Based on relative target densityBased on relative target density–– Finds tall objects and gets highest value Finds tall objects and gets highest value

per objectper object–– Output is a GIS of Output is a GIS of ““treetree”” locations and locations and

associated heightsassociated heightsCrown size from canopy returnsCrown size from canopy returns

Surfaces from LiDARSurfaces from LiDARCanopy Terrain

Tree IdentificationTree Identification

Canopy surfaceCanopy surface

Tree locationsTree locations

•• Combine LiDAR and MS data Combine LiDAR and MS data •• 6565--93% location accuracy93% location accuracy•• Determine height Determine height --

(canopy (canopy –– ground) at tree loc.ground) at tree loc.

Tree Locations Used for: Height, Tree Locations Used for: Height, Crown Size, Leaf Area, and VolumeCrown Size, Leaf Area, and Volume

LiDAR scatter plotLiDAR scatter plot

Tree locationsTree locationsCrown depthCrown depth(center of LA)(center of LA)

Total heightTotal height

GroundGround

Tree Height ComparisonTree Height Comparison

0

10

20

30

40

50

0 10 20 30 40 50

Tree height lidar: linTop/linBot ( m)

Tota

l tre

e he

ight

mea

sure

d in

fiel

d (m

)

selectionArea08

selectionArea37

isolated individual trees

Douglas-fir

Timber Inventory Timber Inventory

First tested in Idaho First tested in Idaho 2001 (SE < 12%)2001 (SE < 12%)Also tested in Louisiana Also tested in Louisiana in two studies in two studies (SE 2.7 to < 10%)(SE 2.7 to < 10%)

Timber InventoryTimber InventoryBased on Double SamplingBased on Double Sampling

Relate sparse ground sample to very large remote Relate sparse ground sample to very large remote (LiDAR) sample(LiDAR) sampleLiDAR flown in stripsLiDAR flown in stripsVery precise tree ht. from LiDAR but biased Very precise tree ht. from LiDAR but biased –– use use regression to field tree heights to adjustregression to field tree heights to adjustApply field relationship of DBH to tree ht. (regression) Apply field relationship of DBH to tree ht. (regression) to predict DBH on LiDARto predict DBH on LiDAR--identified treesidentified treesCalculate volumes on LiDAR treesCalculate volumes on LiDAR treesThe double sample adjusts LiDAR estimate of vol. The double sample adjusts LiDAR estimate of vol. based on sparse field sample.based on sparse field sample.See Parker and Evans (2004) W. J. Appl. Forestry for See Parker and Evans (2004) W. J. Appl. Forestry for the gory detailsthe gory details

DoubleDouble--Sample Timber InventorySample Timber Inventory

_ _ _ __ _ __ ____YYlrlr = y + = y + BB (Lvo (Lvo –– lvo)lvo)__YYlrlr is the regression adjusted is the regression adjusted mean volume per acremean volume per acre__y is mean ground volume on y is mean ground volume on Phase 2Phase 2

BB is relationship of ground is relationship of ground volume (y) to LiDAR volume volume (y) to LiDAR volume (lvo) on Phase 2 sample(lvo) on Phase 2 sample______Lvo is mean LiDAR volumeLvo is mean LiDAR volumeon Phase1on Phase1____lvo is mean LiDAR volume lvo is mean LiDAR volume on Phase 2on Phase 2

General ObservationsGeneral Observations

With traditional photogrammetric sampling:With traditional photogrammetric sampling:Use ground volume vs. photo volume.Use ground volume vs. photo volume.

With LiDAR double sampling procedures:With LiDAR double sampling procedures:Use ground volume vs. LiDAR basal area or volume.Use ground volume vs. LiDAR basal area or volume.Tree height strongly related to DBH and LiDAR provides Tree height strongly related to DBH and LiDAR provides consistent height measure.consistent height measure.LiDAR height and DBH biases LiDAR height and DBH biases ““adjustedadjusted”” by regression by regression estimator for basal area or volume.estimator for basal area or volume.Volume differentiation by species presents challenges.Volume differentiation by species presents challenges.Works well for monocultures.Works well for monocultures.

Forest Structure and Forest Structure and Wildlife Habitat AssessmentWildlife Habitat Assessment

CombinedCombinedInformationInformation

CoverCoverTypeType

H. StructureH. Structure

Tree SizeTree Size

V. StructureV. Structure

InputInputVariablesVariables

Habitat Habitat AssessmentsAssessments

HabitatHabitatclassification classification accuracy lowaccuracy lowin part due to in part due to insufficient insufficient information information about forest about forest structurestructure.

Problem:Problem:

ClassificationClassification

SingleSingle--storystoryoror

MultiMulti--storystory

Vertical Structure Classification from TreeVertical Structure Classification from TreeHeight Variation at Landscape ScaleHeight Variation at Landscape Scale

MultiMulti--spectral and LiDAR Inputs to spectral and LiDAR Inputs to Habitat Suitability ModelsHabitat Suitability Models

1. Map stands by type1. Map stands by type

2. Identify snags2. Identify snags

Output from Habitat Model Output from Habitat Model (Pigmy Nuthatch) Aggregated by Stand(Pigmy Nuthatch) Aggregated by Stand

Stand Visualization from LiDARStand Visualization from LiDAR

Visualization toolsVisualization toolsPoint clouds for forest structurePoint clouds for forest structureTree lists to stand graphicsTree lists to stand graphicsA vision of the future?A vision of the future?

Forest VisualizationForest VisualizationPlotPlot StandStand LandscapeLandscape

SVSSVS

LMSLMSEnVisEnVis

SmartForestSmartForestSource Source http://www.vterrain.orghttp://www.vterrain.orghttp://forsys.cfr.washington.eduhttp://forsys.cfr.washington.edu

Examination of Point CloudsExamination of Point Clouds

LiDAR Trees in SVSLiDAR Trees in SVS

Virtual Environments from LiDAR Virtual Environments from LiDAR Parameters of Individual TreesParameters of Individual Trees

Graphic Tree ModelsGraphic Tree Models

Low (pine) 224 triangles

Med (pine) 508 triangles

High (pine) 2,400 triangles

HardwoodBillboards

Experimental MediaExperimental MediaStand Video vs. Graphic SimulationsStand Video vs. Graphic Simulations

Trunk views Canopy views

Video

Graphic

Video

Graphic

Virtual Environment TestsVirtual Environment Tests

Graphic realism (simple to complex)Graphic realism (simple to complex)Interactivity (virtual navigation and stand Interactivity (virtual navigation and stand examinations)examinations)

Graphic Aids in VEGraphic Aids in VE

Height poleHeight pole

CompassCompass

10 x 10 foot grid10 x 10 foot grid

Possible Applications of Possible Applications of Virtual Stand ModelsVirtual Stand Models

Forest inventoryForest inventoryWildlife habitat assessmentWildlife habitat assessmentStand managementStand managementRecreationRecreationEnvironmental impact assessmentsEnvironmental impact assessmentsMilitaryMilitaryTeaching tools (for all the above)Teaching tools (for all the above)

Current ResearchCurrent ResearchLiDAR and hyperspectral/multispectral data LiDAR and hyperspectral/multispectral data for RCW habitat modelingfor RCW habitat modelingLiDAR sampling theory LiDAR sampling theory –– interaction of interaction of LiDAR with targets / target recognitionLiDAR with targets / target recognition

Understory modeling from LiDAR returnsUnderstory modeling from LiDAR returns

Interactive tools for stand treatment and Interactive tools for stand treatment and investigation in research/teachinginvestigation in research/teaching

Growth and yield modeling and visualization Growth and yield modeling and visualization

Thank YouThank You

Thompson Hall, Mississippi State UniversityThompson Hall, Mississippi State University