Using Imagery and LiDAR for cost effective mapping and analysis for timber and biomass inventories

Transcription

1 Using Imagery and LiDAR for cost effective mapping and analysis for timber and biomass inventories Mark Meade: CTO Photo Science Mark Milligan: President LandMark Systems May 2011

2 Presentation Outline Sensor fusion the best of both worlds Imagery LiDAR Goal Better Faster - Cheaper Stand delineation and stratification Summary Questions

3 Geospatial Solutions SENSOR FUSION

4 Digital Imagery High resolution Good for boundaries Texture Stereo Generally only one date With State ortho programs and NAIP now leaf on and leaf off imagery available RGB CIR

5 Other imagery Sources Segments over over over high Summer Spring resolution LANDSAT CIR imagery

6 LiDAR Generates an accurate 3D model of the earth s surface Use an infrared laser and scanning mirror in measurement Sensors have the ability to measure multiple returns from each pulse Also measure the intensity of each of the returns

7 Multiple LIDAR Returns

8 Multiple Returns

9 LiDAR LiDAR is an active sensor Generates infrared light Can be flown day or night LiDAR cannot be flown above clouds during rain with standing water on the ground LiDAR can pass through gaps in some vegetation For bare earth mapping generally flown leaf off

10 LiDAR Point Density Fixed Wing Rotary Wing Mobile Mapping Acquisition Heights 3,000-8,000 AMT AMT Ground based Acquisition Speeds knots knots mph Vertical Accuracy 9-25 cm 3-15 cm 2-10 cm Horizontal Accuracy m cm 3-10 cm Point Density ppsm ppsm 1,000-8,000 ppsm Nominal Post Spacing (NPS): Average distance between adjacent LiDAR points Point Density: Number of LiDAR points per unit area (points per square meter) Root Mean Square Error (RMSE): Statistical value equal to the 68% C. I. Accuracy: 95% confidence interval of the data (vertical = RMSE x 1.96)

11 Cost Considerations Good vegetation mapping often requires high densities Achieving ultra high densities can result in Significant costs Increased data storage Data manipulation issues Acquisition costs can increase to 2x or 4x Data requirements: 1 ppsm point density would equate to 200 Megabytes per mi 2 4 ppsm would be 4X, or 800 Megabytes per mi 2 8 ppsm would be 8X, or 1.6 Gigabytes per mi 2

12 Data Produced LiDAR First return intensity and elevation Last return intensity and elevation Bare earth DEM

13 Data Point Intensity Surface Cloud Model Values

14 Geospatial Solutions GOAL FOR INVENTORY

15 Stratified vs. Stand level Stratified Low cost Statistical estimate for forest Requires strata delineated Shorter duration Produces planning level data Stand Level High cost Stand specific information Requires stands delineated Longer duration Produces operational level data

16 Stand/Strata Delineation Semi-automated when compared to manual PI Better? More consistent Greater detail Faster? Get to an initial delineation faster Scales well in large areas Cheaper? Less time less money Better stratification few plots required

17 Images Courtesy of B. Pliszka Can We Get more Information Tree heights Understory Canopy structure Biomass and carbon

18 Geospatial Solutions STAND DELINEATION AND STRATIFICATION

19 Sub-Stand Analysis Sub-canopy Sub-stand Canopy Elements

20 Classification of height

21 Stand Delineation Semi-automated Automated Image of Oregon Segmentation Stand Forest Delineation

22 Images Courtesy of B. Pliszka Improve Stand Delineation The example below represents a forest stand before the LiDAR based adjustment. High resolution aerial imagery supports the initial delineation of forest stands but sometimes lacks information on canopy height. The same forest stand after the LiDAR based adjustment. The canopy height model (CHM) provides insight into the within-stand height variation not always detected using aerial imagery. Area with height difference

23 Improved Stratification To minimize the number of field plots that are taken need a good stratification of the landscape Semi-automated methods using imagery and LiDAR can be used to rapidly stratify the forest Can then use the stratification to locate field plots and provides area of each stratum

24 Improved Information Remotely sensed measurements of forest structure and fuel loads in the Pinelands of New Jersey, Nicholas Skowronski, Kenneth Clark, Ross Nelson, John Hom, Matt Patterson, RSE 2007 Generally relationships are better in conifers that broadleaves Better with higher point density LiDAR Point spacing m

25 Improved Information: understory Structural characteristics for prescribed fire treatment and unburned areas Attribute Treated Unburned p value Sig. Overstory Cover (%) 50.0± ± NS Height (m) 6.4± ± <0.01 DBH (cm) 12.0± ± <0.001 Understory Sapling cover (%) 0.0± ± <0.001 Seedling cover (%) 16.6± ± <0.01 Shrub cover (%) 36.5± ± <0.001 Sapling height (m) 1.3± ± <0.05 Seedling height (m) 1.0± ± NS Shrub height (m) 0.5± ± NS Shrub biomass (t ha 1) 0.85± ± <0.001 % vegetation cover and height estimated from LIDAR measurements. The recently burned area was the site of a prescribed fire 2 months previously Unburned site has not burned since Differences between normalized percentage of LIDAR returns are significant for 1 2 m and 2 3 m height class bins at p<0.05

26 summary Integration of imagery and LiDAR can produce some valuable information for forest management The LiDAR data is now more common but many forestry applications require high density collections that are more expensive Using lower density (1 ppsm) LiDAR can be used with imagery to delineate stands and classify them into strata that can support inventory both stratified and stand based LiDAR has the potential to provide other information for forest management, like understory, but care should be taken when extrapolating research findings to operational forestry practices

27 Questions Mark Meade: Mark Milligan: Andrew Brenner: