Using lidar for forest assessment and monitoring Hans-Erik Andersen Forest Inventory and Analysis USDA Forest Service PNW Research Station Anchorage, Alaska USDA Forest Service PNW Research Station Forest Inventory and Analysis Cost of helicopter-access plots on Kenai Peninsula, AK: ~ $6000/plot Status of FIA inventory in Alaska Boreal Forests We estimate that Alaska has 17% of U.S. forest land ? Roughly 112 million acres of forest in “interior” Alaska. 9 Vast 9 Mostly low productivity 9 Inaccessible 9 Difficult and expensive to inventory Currently there is no FIA inventory in interior Alaska Remote sensing will necessarily play an integral role in any future inventory USDA Forest Service PNW Research Station Forest Inventory and Analysis Importance of FIA inventory in Alaska Boreal Forests Fore st N PP CO 2 So il t e Fires Insect Snowdamage epth 1955d -2004 1989-2005 Climate Change mp era tur e USDA Forest Service PNW Research Station e r o F es r i st F y p o n Ca a e s a e Dis r e v o C ts c e s n nd I Forest Inventory and Analysis Proposed Periodic Inventory in the Tanana Basin of interior AK Systematic grid for forest inventory in Tanana unit (36 mill ac) 27 million acres of forest (estimated from remote sensing data) 1-in-5 option (~ one plot per 30,000 acres) Fairbanks 1-in-5 option: 1220 plots, about 900 forested USDA Forest Service PNW Research Station Forest Inventory and Analysis What type of information can lidar provide in forest inventory? Lidar is VERY good at measuring certain forest attributes: z z z Overstory tree heights Canopy cover Vertical canopy structure In certain conditions, lidar CAN also provide other information: z z z z Species class (conifer vs. hardwood, esp. in leaf-off conditions) Mortality (esp. in leaf-on conditions) Understory cover (esp. with high density lidar) Surface fuels (esp. with high density lidar) Lidar CANNOT (yet) directly measure other important forest attributes: z z z z z z z Understory individual tree heights Species Stem diameter Age Defect/decay class Lichens/mosses/indicator species/understory vegetation Soils, permafrost, etc. etc. PNW-FIA (Anchorage FSL) is carrying out a pilot project on the Kenai peninsula to investigate the utility of lidar as a tool in the Alaska resource inventory USDA Forest Service PNW Research Station Forest Inventory and Analysis Lidar-derived Bare-Earth Surface Model USDA Forest Service PNW Research Station Forest Inventory and Analysis Lidar-derived Canopy Surface Model USDA Forest Service PNW Research Station Forest Inventory and Analysis Lidar-derived canopy cover An estimate of canopy cover is generated from first-return lidar points: # lidar first returns in canopy (13) Total # lidar first returns (20) USDA Forest Service PNW Research Station = 65% Forest Inventory and Analysis Kenai Lidar Flight Lidar strips & FIA plots Anchorage Acquisition in May, 2004 Covered 120 FIA plots N-S swaths spaced 10 km apart Total linear coverage: approx. 600 km Alaska Total area covered: Kenai Peninsula 19,387 ha (2.3 % of western Kenai lowlands) Total cost: $60K z z Mobilization: 10% Reflights: 10% Total size of raw binary data: 50 GB USDA Forest Service PNW Research Station Forest Inventory and Analysis Kenai Lidar Flight Parameters And System Settings • • • • • • Optech ALTM 3070 Platform: fixed-wing Flying height: 1200 m Flying speed: 250 km/h Scanning swath: ~350 m Laser pulse density: 4 pulses/m2 • Laser pulse rate: 70,000 pulses/second USDA Forest Service PNW Research Station Forest Inventory and Analysis Lidar forest measurement, Kenai Peninsula, AK FIA plot & over Lidar terrain Lidar strips & Lidar canopy Lidar swath Quickbird model surfacesatellite modelFIA plots FIA plotbirch/spruce image, forest on Kenai Peninsula, AK USDA Forest Service PNW Research Station Forest Inventory and Analysis Lidar-based Median individual lidar intensity tree measurement, by species, Kenai KenaiPeninsula, Peninsula,AK AK Field Spc: Spc: Birch Field Ht: 18.8 m FCrownArea:34 m2 Lidar Spc:: Hardwood All DataSpc 50 50 40 Lidar crown Lidar surface Ht: 18.1 m Field–measured treesLidar LCrownArea:31 m segmentation model (> 12.5 cm) 2 40 50 Late May Lidar Acquisition 40 Early May Lidar Acquisition 20 White spruce LCrownArea: LCrownArea: 2 m2 30 Lidar Ht: 11.37 m Lidar individual tree species classification 20 Lidar Spc: Spc: Conifer Birch 30 Median Lidar Intensity 30 FCrownArea: FCrownArea: 3 m2 20 Median Lidar Intensity Field Ht: 11.4 m Median Lidar Intensity Field Spc: Spc: White spruce Brown: Hardwood 94 95 98 375 Species Code 746 747 USDA Forest Service PNW Research Station 10 0 10 0 0 10 Green: Conifer 94 95 98 375 Species Code 746 747 94 95 98 375 Species Code 746 747 Forest Inventory and Analysis Accurate GPS survey of FIA plot locations Accurate plot locations are critical for matching LIDAR & field data Average error of original FIA plot coordinates was 6 m Accurate (<1 m error) plot coordinates acquired over 32 plots on Kenai using surveygrade GPS More plots to be GPSed in summer, 2008 USDA Forest Service PNW Research Station Acquiring GPS location at FIA plot on Kenai Peninsula Forest Inventory and Analysis Methods: Biomass estimation using FIA plot data & lidar strip sample Use FIA plot data from entire Kenai to simulate lidar-derived forest structure metrics at plot level z z z z z Maximum tree/segment height Mean tree/segment height Canopy cover (% plot area > 2 m ht) Tree/segment density Coefficients vary by forest/species type Comparison with lidar metrics on GPSed field plots shows very good match with max ht, more discrepancies in canopy cover Current set of (road-accessible) GPSed plots are not representative of entire western Kenai – need random sample for validation! USDA Forest Service PNW Research Station 25 100 25 20 80 20 15 60 15 10 40 10 Regression relationships using plot-level structural metrics 5 5 20 0 0 z Due to edge effect and recently fallen trees R2 dead = 0.83 Plot-based SQRT(biomass(tons/ha)) canopy cover Field-based maximum ht (m) 15 5Field-based 10 z Apply edge correction to reduce bias in canopy cover Generate simulated lidar canopy height model for each subplot Apply lidar measurement algorithms to simulated canopy height model 0 z Field vs.field lidar biomass Lidar vs. canopy cover Lidar vs. field max ht Plot -level biomass vs. structure metrics Field-based SQRT(Biomass (tons/ha) 0 0 00 20 5 5 5 40 60 80 100 10 10 15 15 20 25 10 15 20 Lidar-based canopy cover Lidar-based maximum(tons/ha)) ht (m) Predicted SQRT(biomass Lidar-based SQRT(Biomass (tons/ha)) Forest Inventory and Analysis LANDFIRE Results: Biomass Estimation on western Kenai using Lidar Strip Sample Lidar-based Biomass Estimates (13 m cells) Black: Low Biomass White: High Biomass USDA Forest Service PNW Research Station Forest Inventory and Analysis Model-assisted Design for Biomass Estimation Lidar strip sample and plot data can be used as components of model-assisted sampling design Approx. design-unbiased regression estimator (Särndal, 1992): ∑ t yˆi sΙ π Ιi + ∑s yk − yˆ k πk Model-based estimate + Adjustment term (based on lidar strips) (corrects for bias using plot data ) Highly efficient if model fits well Still approx. unbiased even if model does not hold Precision further increased w/ post-stratification USDA Forest Service PNW Research Station Forest Inventory and Analysis Results: Assessing Lidar Strip Sampling Efficiency via Simulation Simulation used to assess sampling error for various plot and stripbased estimates Detailed PI-based forest stand classification layer used to generate continuous biomass map over entire western Kenai Precision = RMSE/tot.biomass: Simulated Strip Simulated Strip FIA sampling Sample frame Map Sample Biomass 5 km 15 km 10 km spacing spacing 350-m wide N-S strip samples: 5 km spacing: 0.015 10 km spacing: 0.032 15 km spacing: 0.032 20 km spacing: 0.05 Systematic plot samples: 5 km spacing: 0.033 10 km spacing: 0.084 15 km spacing: 0.126 USDA Forest Service PNW Research Station Forest Inventory and Analysis Future research: Post-stratification of lidar strip sample using LANDFIRE veg class, height, and canopy cover layers USDA Forest Service PNW Research Station Forest Inventory and Analysis Future research: Monitoring growth with multi-temporal lidar Individual tree lidar data, Capitol Forest, WA 2003 1999 1998 3 m height growth USDA Forest Service PNW Research Station Forest Inventory and Analysis Conclusions & Future Directions Lidar has the potential to be a valuable sampling tool for large area resource inventory Lidar may be even more useful as a monitoring tool z Lidar surveys could significantly decrease remeasurement interval z In periodic interior AK inventory, 20 - 25 years between field measurements Plots could be “visited” with lidar every 8 - 10 years Assessment of carbon emitted from fires, etc. Future Research z z z z Assessment of model-assisted sampling design (w/ additional GPSed plots, poststratification w/ LANDFIRE) Forest biomass assessment with spaceborne L-band imaging radar (w/ NASA-JPL & UAF Geophysical Inst.) Assessment of biomass/carbon in NA boreal forests using combination of plot data, airborne profiling lidar & GLAS satellite lidar (w/ Ross Nelson, NASA-Goddard) Investigation of using lidar & NN imputation (w/ Oregon St. Univ.) USDA Forest Service PNW Research Station Forest Inventory and Analysis Questions/Discussion? USDA Forest Service PNW Research Station Forest Inventory and Analysis Future research: Monitoring long-term changes in Interior AK forests with AIRIS Large-Scale Photography and LIDAR 2007 1982aerial CIR aerial photo photo Eutrification of lake USDA Forest Service PNW Research Station Forest Inventory and Analysis