Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim...

46
Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science NC STATE UNIVERSITY DEPARTMENT OF SOIL SCIENCE Horizontal Resolution and Data Density Effects on LIDAR-based DEM

Transcript of Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim...

Page 1: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

Jim Thompson, Eric Anderson, & Rob AustinNC State University, Department of Soil Science

Jim Thompson, Eric Anderson, & Rob AustinNC State University, Department of Soil Science

NC STATE UNIVERSITY DEPARTMENT OF SOIL SCIENCE

Horizontal Resolution and Data Density Effects on

LIDAR-based DEM

Horizontal Resolution and Data Density Effects on

LIDAR-based DEM

Page 2: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

OutlineOutline

Digital elevation models (DEM)Digital elevation models (DEM)– Terrain attributes calculated from DEMTerrain attributes calculated from DEM

– Horizontal resolution effectsHorizontal resolution effects

– Data density issuesData density issues

LIDAR Data Density vs. DEM QualityLIDAR Data Density vs. DEM Quality– MethodsMethods

– ResultsResults

– Conclusions Conclusions

Page 3: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

OutlineOutline

Digital elevation models (DEM)Digital elevation models (DEM)– Terrain attributes calculated from DEMTerrain attributes calculated from DEM

– Horizontal resolution effectsHorizontal resolution effects

– Data density issuesData density issues

LIDAR Data Density vs. DEM QualityLIDAR Data Density vs. DEM Quality– MethodsMethods

– ResultsResults

– Conclusions Conclusions

Page 4: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

ElevationElevation

400 m 422 m

Page 5: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Terrain AttributesTerrain Attributes

Slope gradientSlope gradient

Slope aspectSlope aspect

Upslope contributing areaUpslope contributing area

Slope curvatureSlope curvature– Profile curvatureProfile curvature

– Contour curvatureContour curvature

– Total curvatureTotal curvature

Page 6: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Slope GradientSlope Gradient

0% 17%

Page 7: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Upslope Contributing AreaUpslope Contributing Area

Page 8: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Profile CurvatureProfile Curvature

Page 9: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Topsoil ThicknessTopsoil Thickness

0 m 2 m

Page 10: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

DEM ResolutionDEM Resolution

0% 17%

0% 15%

10 m DEM10 m DEM

30 m DEM30 m DEM

Page 11: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Horizontal Resolution EffectsHorizontal Resolution Effects

Decreasing the horizontal resolution produces Decreasing the horizontal resolution produces (Thompson and Bell, 2001):(Thompson and Bell, 2001):

– lower slope gradients on steeper slopeslower slope gradients on steeper slopes

– steeper slope gradients on flatter slopessteeper slope gradients on flatter slopes

– narrower ranges in curvaturesnarrower ranges in curvatures

– larger catchment areas in upper landscape positions larger catchment areas in upper landscape positions

– lower catchment areas in lower landscape positionslower catchment areas in lower landscape positions

and leads to the loss of small-scale featuresand leads to the loss of small-scale features

Page 12: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

DEM Horizontal ResolutionDEM Horizontal Resolution

Page 13: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

DEM ResolutionDEM Resolution

0% 17%

0% 15%

10 m DEM10 m DEM

30 m DEM30 m DEM

Page 14: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

LIDAR-Based DEMLIDAR-Based DEM

Interpolation of spot elevation dataInterpolation of spot elevation data

Vertical precision of 0.1 mVertical precision of 0.1 m

Horizontal resolutionHorizontal resolution– Best possible for any given applicationBest possible for any given application

NC Floodplain Mapping Program LIDAR DataNC Floodplain Mapping Program LIDAR Data– Bare earth mass dataBare earth mass data

– 20 ft (~6 m) gridded DEM20 ft (~6 m) gridded DEM

– 50 ft (~15 m) gridded DEM50 ft (~15 m) gridded DEM

Page 15: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

LIDAR Data DensityLIDAR Data Density

4000-35,000 laser pulses per second4000-35,000 laser pulses per second

2-7 returns collected for each laser pulse 2-7 returns collected for each laser pulse

25,000 points per square mile25,000 points per square mile

ABUNDANCE OF DATAABUNDANCE OF DATA

Page 16: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

LIDAR Data DensityLIDAR Data Density

Page 17: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

LIDAR Data DensityLIDAR Data Density

Page 18: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Data ReductionData Reduction

LIDAR tileLIDAR tile LIDAR tileLIDAR tile

50% Test 50% Test 50% Test 50% Test 50% Validation 50% Validation 50% Validation 50% Validation

25% Test25% Test25% Test25% Test

10% Test10% Test10% Test10% Test

5% Test5% Test 5% Test5% Test

1% Test1% Test 1% Test1% Test

Page 19: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Data ReductionData Reduction

50 25 10 5 1

% of Original Density

0

10

20

30

40

RM

SE

(cm

)

IDW

OK

50 25 10 5 1

% of Original Density

0

10

20

30

40

RM

SE

(cm

)

IDW

OK

}}10 cm10 cm

IDW vs. OKIDW vs. OK

Page 20: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

ObjectivesObjectives

Produce a series of DEM at different horizontal Produce a series of DEM at different horizontal resolutions at multiple LIDAR data densitiesresolutions at multiple LIDAR data densities

Compare each of these DEM to a DEM Compare each of these DEM to a DEM produced from 100% of the LIDAR dataproduced from 100% of the LIDAR data

Determine the optimum LIDAR point density Determine the optimum LIDAR point density suitable for producing a DEM at a given suitable for producing a DEM at a given horizontal resolutionhorizontal resolution

Evaluate the effects of LIDAR data density on Evaluate the effects of LIDAR data density on the production of DEM at different resolutionsthe production of DEM at different resolutions

Page 21: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

OutlineOutline

Digital elevation models (DEM)Digital elevation models (DEM)– Terrain attributes calculated from DEMTerrain attributes calculated from DEM

– Horizontal resolution effectsHorizontal resolution effects

– Data density issuesData density issues

LIDAR Data Density vs. DEM QualityLIDAR Data Density vs. DEM Quality– MethodsMethods

– ResultsResults

– Conclusions Conclusions

Page 22: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Study SiteStudy Site

Page 23: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Study SiteStudy Site

Hofmann ForestHofmann Forest– 32,500 ha forest ecosystem32,500 ha forest ecosystem

– Lower Coastal PlainLower Coastal Plain

– >90% forested>90% forested

– 12 to 20 m above MSL12 to 20 m above MSL

LIDAR DataLIDAR Data– NC Floodplain Mapping ProgramNC Floodplain Mapping Program

– 61 LIDAR tiles61 LIDAR tiles

– 9,000,000 LIDAR points9,000,000 LIDAR points

Page 24: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Data ReductionData Reduction

100%100%

50%50%

25%25%

10%10%

5%5%

1%1%

Page 25: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

DEM5DEM55050DEM5DEM55050DEM10DEM105050

DEM10DEM105050

Data Reduction and ComparisonData Reduction and Comparison

Data Data ReductionReduction

Data Data ReductionReduction

DEM30DEM305050DEM30DEM305050

Compare Compare ElevationsElevations

Compare Compare ElevationsElevations

Compare Compare ElevationsElevations

Compare Compare ElevationsElevations

Compare Compare ElevationsElevations

Compare Compare ElevationsElevations

DEM5DEM5100100DEM5DEM5100100DEM10DEM10100100

DEM10DEM10100100DEM30DEM30100100DEM30DEM30100100

Generate DEM Generate DEM (ANUDEM)(ANUDEM)

Generate DEM Generate DEM (ANUDEM)(ANUDEM)

100% LIDAR100% LIDARData SetData Set

100% LIDAR100% LIDARData SetData Set

Generate DEM Generate DEM (ANUDEM)(ANUDEM)

Generate DEM Generate DEM (ANUDEM)(ANUDEM)

50% LIDAR50% LIDARData SetData Set

50% LIDAR50% LIDARData SetData Set25% LIDAR25% LIDAR

Data SetData Set

25% LIDAR25% LIDARData SetData Set DEM30DEM302525

DEM30DEM302525 DEM10DEM102525DEM10DEM102525 DEM5DEM52525

DEM5DEM5252510% LIDAR10% LIDARData SetData Set

10% LIDAR10% LIDARData SetData Set DEM30DEM301010

DEM30DEM301010 DEM10DEM101010DEM10DEM101010 DEM5DEM51010

DEM5DEM510105% LIDAR5% LIDARData SetData Set

5% LIDAR5% LIDARData SetData Set DEM30DEM3055

DEM30DEM3055 DEM10DEM1055DEM10DEM1055 DEM5DEM555

DEM5DEM5551% LIDAR1% LIDARData SetData Set

1% LIDAR1% LIDARData SetData Set DEM30DEM3011

DEM30DEM3011 DEM10DEM1011DEM10DEM1011 DEM5DEM511

DEM5DEM511

Page 26: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

DEM ComparisonsDEM Comparisons

Evaluate the effects of data density at each Evaluate the effects of data density at each horizontal resolutionhorizontal resolution

Cell-by-cell comparisonCell-by-cell comparison– Assumed DEM created from the complete LIDAR data set Assumed DEM created from the complete LIDAR data set

were the “best” DEMwere the “best” DEM

– Comparisons always made back to DEM created using the Comparisons always made back to DEM created using the total original data set total original data set

Elevation values compared using paired t-testsElevation values compared using paired t-tests

Differences regressed against data densityDifferences regressed against data density

Page 27: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Data Reduction and Point DensityData Reduction and Point Density

30 m 10 m 5 m

Data reduction Designation Points/cell Designation Points/cell Designation Points/cell

100 DEM30100 26.16 DEM10100 2.90 DEM5100 0.72

50 DEM3050 13.08 DEM1050 1.45 DEM550 0.36

25 DEM3025 6.54 DEM1025 0.73 DEM525 0.18

10 DEM3010 2.62 DEM1010 0.29 DEM510 0.07

5 DEM305 1.31 DEM105 0.15 DEM55 0.04

1 DEM301 0.26 DEM101 0.03 DEM51 0.01

Page 28: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Elevation DifferencesElevation Differences

30 m 10 m 5 m

Data reduction Difference (m) P value Difference (m) P value Difference (m) P value

50 -0.01 0.4379 -0.01 0.7680 -0.01 0.1824

25 -0.02 0.1095 -0.02 0.0552 -0.03 0.0145

10 -0.02 0.1830 -0.03 0.0117 -0.04 0.0007

5 -0.03 0.0441 -0.04 0.0086 -0.06 0.0006

1 -0.06 0.0013 -0.04 0.0047 -0.07 0.0001

Page 29: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Elevation DifferencesElevation Differences

All reduced density DEM overestimated All reduced density DEM overestimated elevations for all horizontal resolutionselevations for all horizontal resolutions

As data density decreased, the elevation As data density decreased, the elevation difference increaseddifference increased

As mean points per grid cell decreased, the As mean points per grid cell decreased, the elevation difference increasedelevation difference increased

Page 30: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Elevation DifferencesElevation Differences

0.00 0.10 0.20 0.30 0.40 0.50

Mean points per grid cell

-0.08

-0.06

-0.05

-0.03

-0.02

0.00

Mea

n d

iffe

ren

ce (

m)

5 m5 m

0.00 0.50 1.00 1.50 2.00

Mean points per grid cell

-0.08

-0.06

-0.05

-0.03

-0.02

0.00M

ean

dif

fere

nce

(m

)

10 m10 m

0 2 4 6 8 10 12 14

Mean points per grid cell

-0.08

-0.06

-0.05

-0.03

-0.02

0.00

Mea

n d

iffe

ren

ce (

m)

30 m30 m

Page 31: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Data DensityData Density

30 m 10 m 5 m

Data reduction Points/ha P value P value P value

50 145 0.4379 0.7680 0.1824

25 73 0.1095 0.0552 0.0145

10 29 0.1830 0.0117 0.0007

5 15 0.0441 0.0086 0.0006

1 3 0.0013 0.0047 0.0001

Page 32: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Data Density RequirementsData Density Requirements

55 1010 1515 2020 2525 3030

DEM horizontal resolution (m)DEM horizontal resolution (m)

00

3030

6060

9090

120120

150150

LID

AR

(p

oin

ts h

aL

IDA

R (

po

ints

ha-1-1

)) Excessive data densityExcessive data density

Threshold data densityThreshold data density

Insufficient dataInsufficient data densitydensity

Page 33: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

ConclusionsConclusions

Target resolution of DEM determines the level Target resolution of DEM determines the level of acceptable LIDAR data reductionof acceptable LIDAR data reduction

Higher density LIDAR data needed to model Higher density LIDAR data needed to model higher resolution topographic featureshigher resolution topographic features

Results are a function of landscape Results are a function of landscape morphology and land usemorphology and land use

Page 34: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Page 35: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,
Page 36: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,
Page 37: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

DEM PrecisionDEM Precision

0% 17%

0% 20%

0.1 m precision0.1 m precision

1.0 m precision1.0 m precision

Page 38: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Vertical Precision EffectsVertical Precision Effects

Decreasing the vertical precision of the DEM Decreasing the vertical precision of the DEM produces (Thompson and Bell, 2001): produces (Thompson and Bell, 2001):

– more flat cells (slope gradient = 0 and curvature = 0) more flat cells (slope gradient = 0 and curvature = 0)

– more steep cells more steep cells

– a wider distribution in curvaturesa wider distribution in curvatures

– discretization of elevation values, which leads to discretization of elevation values, which leads to discretization of other terrain attributes and reduced discretization of other terrain attributes and reduced similarity of data setssimilarity of data sets

Page 39: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

DEM Vertical PrecisionDEM Vertical Precision

Page 40: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

DEM PrecisionDEM Precision

0% 17%

0% 20%

0.1 m precision0.1 m precision

1.0 m precision1.0 m precision

Page 41: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Study SiteStudy Site

Page 42: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

0.00 0.10 0.20 0.30 0.40 0.50

Mean points per grid cell

-0.08

-0.06

-0.05

-0.03

-0.02

0.00

Me

an

dif

fere

nc

e (

m)

0.00 0.50 1.00 1.50 2.00

Mean points per grid cell

-0.08

-0.06

-0.05

-0.03

-0.02

0.00

Me

an

dif

fere

nc

e (

m)

0 2 4 6 8 10 12 14

Mean points per grid cell

-0.08

-0.06

-0.05

-0.03

-0.02

0.00

Me

an

dif

fere

nc

e (

m)

Page 43: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

0.00 0.10 0.20 0.30 0.40 0.50

Mean points per grid cell

-0.08

-0.06

-0.05

-0.03

-0.02

0.00

Mean difference (m)

0.00 0.50 1.00 1.50 2.00

Mean points per grid cell

-0.08

-0.06

-0.05

-0.03

-0.02

0.00

Mean di

fferenc

e (m)

0 2 4 6 8 10 12 14

Mean points per grid cell

-0.08

-0.06

-0.05

-0.03

-0.02

0.00

Me

an

d

iffe

re

nc

e (m

)

Page 44: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

LIDAR Data DensityLIDAR Data Density

Collected as point filesCollected as point files

Linear interpolation techniques are needed for Linear interpolation techniques are needed for improved use of LIDAR (Lloyd and Atkinson, improved use of LIDAR (Lloyd and Atkinson, 2002)2002)

Page 45: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Profile CurvatureProfile Curvature

Page 46: Jim Thompson, Eric Anderson, & Rob Austin NC State University, Department of Soil Science Jim Thompson, Eric Anderson, & Rob Austin NC State University,

NC STATE UNIVERSITY

Data ReductionData Reduction

100%

1%