1 / 30

Geospatial Datasets for Forest Inventory Using Laser Altimetry

This study aims to develop geospatial data products for estimating forest inventory variables using small footprint laser altimetry. The research involves quantifying biophysical variables at individual tree scale and generating geospatial mapping products at stand/patch scale.

munozr
Télécharger la présentation

Geospatial Datasets for Forest Inventory Using Laser Altimetry

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. The Development of Geospatial Datasets for Estimating Forest Inventory Variables Using Small Footprint Laser Altimetry. Eric Rowell1, Lee Vierling2, Wayne Sheppard3, Carl Seielstad1, and Lloyd Queen1 1National Center for Landscape Fire Analysis, University of Montana, Missoula, Mt 59801 2College of Natural Resources, University of Idaho, Moscow, ID 83844 3U.S. Forest Service, Rocky Mountain Research Station, 240 West Prospect, Ft. Collins, CO 80526

  2. I n t r o d u c t i o n Background • Pilot study: • Research conducted to quantify forest biophysical variables in the Black Hills of South Dakota • Development of geospatial data products from these variables for use by managers and scientists • Individual tree scale • Broader stand/patch scale • Fire behavior modeling

  3. I n t r o d u c t i o n What’s the advantage of laser altimetry? • Near continuous sampling of large areas: • Generating large volumes of height data • Representative of information regarding the canopy, understory, and ground. • The ability to distinguish unique biophysical variables • (e.g. Stem ID, DBH, and Crown Width) • These biophysical variables can be used to drive the derivation of stand level attributes and indices • (e.g. SDI)

  4. I n t r o d u c t i o n • Primary variable derived from laser altimetry data (drives the whole system). • Usually underestimates maximum tree height in coniferous forests Tree Height Crown Width • Tertiary variable derived from DBH Crown Base • Derived from the Canopy Height Model (CHM) • Secondary variable that is derived from tree height Diameter (DBH) Rationale I: Biophysical Variables

  5. I n t r o d u c t i o n Rationale II: Geospatial Data Products • What scale do these data products need to be? • Individual Trees • Stands (e.g. structure, age class, condition) • Patches (areas of equal ecological quality) • Landscape (a sum of patches or stands) • Which biophysical variables are important to quantify? • For silvics (e.g. tree height, DBH, stem count) • For canopy fuels (e.g. tree height, crown width, HTCB, crown shape, PCC) • For forest ecology (e.g. structure (encompasses silvics and fuels), understory) • From the biophysical variables, what additional variables or indices can be generated? • For silvics: SDI, stem volume, sapwood area • For canopy fuels: CBD, crown weight

  6. I n t r o d u c t i o n Research Objectives I Forest Biophysical Variable Inventory: • Estimation of biophysical variables from laser altimetry at an individual tree scale. • Height, DBH, crown width, and height to crown base • Stem count • Discussion of data segmentation techniques to refine biophysical variable estimation • Validation of estimated biophysical variables from laser altimetry by comparison with coincident plot scale field data.

  7. I n t r o d u c t i o n Research Objectives II Generation of geospatial mapping products: • Aggregate plot scale biophysical variables to a stand/patch scale. • Additional data products and indices for silviculture and fire behavior modeling.

  8. M e t h o d s Laser Altimetry Data Acquisition System: Leica Geosystems ALS40 Airborne Laser Scanner Acquired: August 16, 2002 Nominal Post Spacing: 1.5 m System Specifications: • GPS/IMU correction • High accuracy 15cm vertical, 20-25 cm horizontal • Up to 3 returns per pulse • A return at a minimum every 5 meters or 2 nanoseconds • Variable pulse rate of 15,000- 25,000 pulses per second • Variable scan rate of ±15hz to 20hz • ± 30 centimeter foot print dependent on altitude • Reflectance Intensity

  9. M e t h o d s Data Segmentation • Separate ground returns from vegetation returns (two steps) • Virtual Deforestation Algorithm; open source (Haugerud, 2001) • Terrascan software; closed source (TerraSolid, Helsinki, Finland) • Adapted Variable Sized Window Algorithm • Determines individual tree stems from Canopy Height Model (CHM) • Uses a local maximum filter window using a window size determined from an allometric model for crown width from height • Data is smoothed to create a convex hull shape for easier ID of individual trees • End product is a feature class that has stem locations and height associated with each stem

  10. Data Interpolation Issues • Using a smoothing array to better estimate stem locations leads to further underestimation of height • Heights from the unaltered canopy height model are tagged to the stem locations

  11. M e t h o d s Adapted Variable Widow Analysis Process

  12. M e t h o d s Ikonos false color infrared image August 2001 overlayed on lidar baldearth DEM Study Area: Black Hills Experimental Forest Plots: SDSMT n = 25 RMRS n = 41 Total Number of Trees Sampled: SDSMT n = 2111 RMRS n = 2885 Data Collected for each tree: Height DBH Crown Width HTCB

  13. M e t h o d s Data Product Flow Chart

  14. M e t h o d s Formulas for Forest Biophysical Variables • Height is the native data product • Tends to underestimate the maximum height • Needs to be corrected for this underestimation • Height to Crown Base is derived from the native height data • Canopy Height Model (CHM) • DBH and Crown width are allometrically derived

  15. M e t h o d s Observed Basal Area (observed) Predicted Canopy Cover (laser) Basal Area Class BAC 1 mSD = 7.0 mDBH = 36.9cm mBA = 5.3 m2 BAC 2 mSD = 29.0 mDBH = 29.9cm mBA = 12.1 m2 BAC 3 mSD = 43.5 mDBH = 30.4cm mBA = 20.5 m2 BAC 4 mSD = 96.7 mDBH = 21.9cm mBA = 29.3 m2 BAC 5 mSD = 156.7 mDBH = 16.4cm mBA = 43.1 m2 Refinement of DBH Estimation • Because different thinning treatments and stocking density a single model proves inadequate to estimate DBH • To correct for these conditions we apply a classification to stratify the height data. • The data is stratified as a function of field observed basal area (BA) and laser derived percent crown cover (PCC)

  16. M e t h o d s Basal Area Class DBH from Height Linear Model R2 1 y = 1.9763x - 0.6047 0.786 2 y = 1.7796x + 1.0648 0.791 3 y = 1.6768x - 0.5704 0.755 4 y = 1.5921x - 3.6192 0.595 5 y = 0.3937x + 8.0436 0.612 DBH Models by Basal Area Classification • Specific models are applied to each BA class to better estimate DBH from laser estimated tree height.

  17. R e s u l t s R e s e a r c h O b j e c t i v e I Biophysical Variable Estimation • Comparison of observed and predicted:

  18. R e s u l t s R e s u l t s R e s e a r c h O b j e c t i v e I Comparison of Stem Counts • Stem locations were most accurately predicted in low density stands. • Sensitivity analysis of stands where stem density is greater than 200 stems/900m2 show the variable sized window underestimates stems by a factor of two. • The key to accurate stem identification is smoothing the data to create a convex hull.

  19. Forest Stand Model

  20. R e s u l t s R e s u l t s R e s e a r c h O b j e c t i v e I I Stand Density Index • SDI is a summation index used to assess levels of growing stock • The index is a relative density measure that is a function of tree DBH and stem density per hectare • For Even Aged ponderosa Pine • DBH*(TPH/25)1.6

  21. R e s u l t s R e s e a r c h O b j e c t i v e I I Patch Analysis Results • Using the basal area classification we can separate patches of trees from laser estimated PCC • We can then populate the PCC feature class with mean height, DBH,HTCB, and crown width values

  22. R e s u l t s R e s e a r c h O b j e c t i v e I I Mapping Product • Patch scale mapping products are of better scale for forest managers making decisions for thinning prescriptions • The values associated with each class can be used in fire behavior modeling (FARSITE) and forest growth models (FVS)

  23. Conclusions • We successfully predicted individual tree biophysical variables: • Height, DBH, crown width, and HTCB • laser altimetry data were segmented to refine DBH estimates • Laser altimetry data were aggregated to stand scale: • By development of biophysically based indices (SDI) • And data stratification through comparisons of field observed (BA) and laser predicted variables (PCC)

  24. Phase II: Fuels Mapping • Our next phase of research • Surface fuels mapping • Transpose research conducted in the Tenderfoot Experimental Forest by Seielstad and Queen to the Black Hills Experimental Forest • Canopy fuels mapping • Transpose research conducted in the Black Hills Experimental Forest by Rowell to the Tenderfoot Experimental Forest

  25. Surface Fuels • Used ground/near surface roughness to explain presence or absence of coarse woody debris • Obstacle density: # of near-ground returns normalized by all points (ground + fuel bed) • Standard deviation and Kurtosis of the ground and near-ground height distributions.

  26. Results of Surface Fuel Estimated • Woody Debris Estimates: • Obstacle density, STD, and Kurtosis of the ground height distribution can be used to estimate total dead fuel volume and 1000 hr fuels • Roughness metrics did not prove effective for identifying fine fuels • Roughness metrics predicted fuel volumes within ± 25 percent of field estimated fuel volumes • Variability within the laser estimated fuel volumes was significantly less than field estimated suggesting that the volume of laser estimates may be better characterizing the fuel bed

  27. Canopy Fuels • Crown bulk density: • Predicted from laser estimated live crown weight (Brown, 1978) EXP( 0.2680 + 2.0740(LN DBH)) • Crown volume estimated from tree crown width, crown base height, and modeled tree crown shape

  28. Initial findings • Live crown weight correlated well between observed and predicted (R2 =.78) • Crown width and CBH relationships areequally well correlated • Crown shape is the unknown: • Do we model a cone, ellipsoid, or paraboloid • We will then apply the model on an individual tree scale to produce canopy bulk density estimates across the landscape

  29. Acknowledgements • My field crew Meghan Calhoon, Shane Hansen, and Beth Hansen. • We would like to thank Horizons, Inc. for the acquisition of the laser altimetry data and the support of ER during this study. • Thanks also to the USDA Rocky Mountain Research Laboratory for the contribution of field data collection and logistical support. • This work was partially supported by NASA EPSCoR grant NCC5-588 and the Upper Midwest Aerospace Consortium.

More Related