2014 Tanana Inventory Pilot: Project Overview Hans Andersen, Robert Pattison & Ken Winterberger Resource Monitoring & Assessment/Forest Inventory & Analysis Pacific Northwest Research Station USDA Forest Service Seattle, WA & Anchorage, AK handersen@fs.fed.us rrpattison@fs.fed.us kwinterberger@fs.fed.us Ross Nelson, Bruce Cook & Doug Morton NASA-Goddard Space Flight Center Biospheric Sciences Laboratory Greenbelt, MD ross.f.nelson@nasa.gov bruce.cook@nasa.gov douglas.morton@nasa.gov USDA Forest Service PNW Research Station Forest Inventory and Analysis Forest Inventory and Analysis (FIA) a very brief history and description Forest Inventory and Analysis (FIA) started in the U.S. in 1929 Initially aimed at reporting to the U.S. Congress the volume and area of productive timberland It has evolved dramatically. Currently aimed at measuring and monitoring all forest land throughout the United States (including Hawaii), Puerto Rico, and the Pacific Trust Islands Made up of a Systematic Sample across all lands in two phases Phase 1 – remotely sensed Phase 2 – ground USDA Forest Service PNW Research Station Forest Inventory and Analysis Phase 1 and Phase 2 – a very brief description of the grid in Alaska • • • • • • • • USDA Forest Service PNW Research Station BLM – 38% forest 62% nonforest DOD – 89% forest 11% nonforest FWS – 41% forest 59% nonforest NPS – 21% forest 79% nonforest TVSF – 100% forest 0% nonforest FS – 51% forest 49% nonforest Major Land Managers – 37% forest 69% nonforest The Shebang – 38% forest 62% nonforest Forest Inventory and Analysis What types of information do we need from a FIA inventory in interior Alaska? How much forest land is there, and how is it changing? What is the biomass and timber resource for local communities? Species, size class, volume, productivity Are the forests of interior Alaska a net carbon source or sink? Climate change, fire, drought, insects Growth, mortality, land cover, soils What is the composition of plants and the quality of wildlife habitat? Wildlife potential, subsistence for local communities USDA Forest Service PNW Research Station Forest Inventory and Analysis Interior AK inventory units 2014 Tanana Inventory Pilot project Pilot/Proof of Concept project will be undertaken in 2014 in the Tanana valley of interior Alaska Project Objectives: Extend FIA “footprint” in interior AK from BNZ/CPCRW experimental forests to Tanana Valley State Forest (1.8 M ac) and Tetlin NWR (~700K ac) – discrete management units, relatively “accessible” Increase/develop familiarity with interior AK logistics Test modified interior AK plot protocols (soils, lichens, etc.) Evaluate utility of airborne lidar+hyperspectral remote sensing Forested area: ~113,000,000 acres information Periodic inventory (5 units) ValleyNASA is most “accessible” unit in interior Alaska Build on relationships with cooperatorsTanana (UAF, (ABoVE, CMS) & land management agencies in interior AK (AKDNR, USF&WS, etc.) Funding provided by a variety of sources (PNW, NASA, etc.) USDA Forest Service PNW Research Station Forest Inventory and Analysis 2014 Tanana Inventory Pilot Project: Overview Field component 1/4th intensity regular hexagonal grid (1 plot per 24,000 acres) within TVSF and Tetlin NWR – 99 total plots Plots randomly moved (~ 50 m) to preserve confidentiality of locations Road-, ATV-, and river-accessible plots established by UAF Helicopter-access plots established by PNW-FIA (Anchorage FSL) Modified field protocol (based on ANC pilot, experimental forest projects, USGS input, etc.) Remote Sensing component Cooperation with NASA-Goddard (Bruce Cook, Ross Nelson, Doug Morton) and Michigan State University (Andrew Finley) State-of-the-art airborne remote sensing instrument (G-LiHT – Goddard LiDAR Hyperspectral Thermal) Airborne RS collected in strip sample (9 km spacing b/n strips) over entire Tanana inventory unit (~145K sq. km) and covering every FIA plot Evaluation of several statistical estimation/inference and modeling/mapping approaches (Bayesian hierarchical, etc.) USDA Forest Service PNW Research Station Forest Inventory and Analysis 2014 Tanana pilot: Field plots Tanana Valley State Forest: ~ 72 forested plots (37 heli-access, 35 ground-access) Tetlin NWR: ~27 forested plots (25 heli-access, 2 ground access) Total: ~99 forested plots USDA Forest Service PNW Research Station Forest Inventory and Analysis 2014 Tanana Inventory Pilot: Plot Measurements Key Features: Current tree protocols – with additional microplot Current P2 understory vegetation protocols Current P2 down woody materials protocols Ground layer sampling (composition and carbon content) Soil sampling (current thaw depth and belowground carbon) High-accuracy GPS plot positions USDA Forest Service PNW Research Station Forest Inventory and Analysis 2014 Tanana Pilot: Tree and Vegetation Measurements Current Phase 2 tree measurement protocols used Additional microplot on each subplot Purpose of second microplot is to sample small-diameter trees 24 foot radius subplot (1/24th ac), trees and snags >= 5 inches Canopy cover in layer by growth habit Current Phase 2 vegetation protocols used Veg. collected on 24-foot radius subplots Species and abundance (cover) for four most abundant species per growth habit, per subplot Structure – recorded as cover by growth habit by layer on each subplot USDA Forest Service PNW Research Station 6.8 foot radius microplots (1/300th ac) seedlings, saplings Forest Inventory and Analysis 2014 Tanana Pilot: Down woody material (DWM) DWM is a valuable indicator of: Quality and status of wildlife habitat Structural complexity of forests Fuel loading and potential fire behavior Amount of carbon stored in forests Site productivity / nutrient cycling Harvest residues 24 ft P2 Down woody material protocol: Two 24-foot transects per subplot (8 per plot) Measure diameters of dead wood (stems/branches) that transects cross USDA Forest Service PNW Research Station Forest Inventory and Analysis 2014 Tanana Pilot: Ground layer sampling Developed by Sarah Jovan (PNWRMA Corvallis FSL), Rob Smith (OSU) and others Objectives: Estimate biomass, C and N content among terrestrial bryophytes and lichens Estimate the landscape-level cover and biomass of important functional groups Cryptograms sampled at 32 microquads (8 per subplot) 20 cm × 50 cm Daubenmire plots 13 nonvascular groups Percent cover and depth USDA Forest Service PNW Research Station Forest Inventory and Analysis 2014 Tanana Pilot: Soil sampling protocol Current thaw depth Depth of Litter/live moss Organic soil layers Mineral soil layer 30 ft Collect samples for further analysis in lab (C, bulk density, etc.) Thaw depth and soil USDA Forest Service PNW Research Station Forest Inventory and Analysis 2014 Tanana Pilot: High-accuracy GPS Accurate plot locations are critical for matching high-resolution remote sensing and field data • In 2-phase sampling designs, error in plot locations directly influences the precision of parameter estimates Dual-frequency GPS+GLONASS receivers can acquire coordinates with < 1 m error in all boreal forest conditions Coordinates obtained at subplot level (enables comparison at ind tree level) USDA Forest Service PNW Research Station Acquiring accurate FIA plot locations using survey-grade GPS receiver on Kenai Peninsula (August, 2008) Courtesy: Ray Koleser Forest Inventory and Analysis Field logistics & schedule Cooperation with UAF Helicopter-access plots established by PNW-FIA crews Ground-access plots established by UAF field crews (via road, trail, river) 3–4 person field crews Bell 206 Long Ranger (still TBD) Schedule Training in Anchorage in May Manley Hot Springs/Fairbanks: July 3–8 Delta Junction: July 15–18 Tok: July 19–22 Northway (camp): July 28–August 5 USDA Forest Service PNW Research Station Forest Inventory and Analysis 2014 Tanana Inventory Pilot project Field component 1/4th intensity regular hexagonal grid (1 plot per 24000 acres) – approx. 97 plots Plots randomly moved (~ 50 m) to preserve confidentiality of locations Road-, ATV-, and river-accessible plots established by UAF Helicopter-access plots established by PNW-FIA Modified field protocol (based on ANC pilot, experimental forest projects) Remote Sensing component Cooperation with NASA-Goddard scientists (Bruce Cook, Ross Nelson, Doug Morton) State-of-the-art airborne remote sensing instrument (G-LiHT – Goddard LiDAR Hyperspectral Thermal) Airborne RS collected in strip sample (9.2 km spacing b/n strips) over entire Tanana inventory unit (60K sq. miles) & covering every FIA plot USDA Forest Service PNW Research Station Forest Inventory and Analysis 2014 Tanana pilot: Field plots 2014 Tanana pilot: Field plots & G-LiHT flight lines G‐LiHT LiDAR + hyperspectral strip sample: 9.2 km spacing (covers every FIA field plot) 300 meter swath width LidDAR covers 3% of total area More intensive G‐LiHT strip sample collected over FIA plots in Bonanza Creek and Caribou–Poker Creek experimental forests USDA Forest Service PNW Research Station Forest Inventory and Analysis What is G-LiHT? G-LiHT is a portable, airborne imaging system that simultaneously maps the composition, structure, and condition of vegetation using: 1) LiDAR to provide 3D information about the spatial distribution of canopy elements; 2) Imaging spectroscopy to discern species composition and variations in biophysical variables (e.g., photosynthetic pigments, nutrient and water content); and 3) Thermal measurements to quantify surface temperatures and detect heat and moisture stress. Scanning/Profiling LiDAR G‐LiHT System: Synergy and Key Characteristics MEASUREMENT CHARACTERISTICS* Scanning LiDAR Swath width/FOV 387 m (60°) Footprint diameter 10 cm (0.3 mrad) Range precision 5 cm (2 σ) Sampling density at surface 6 pulses m‐2 Max. returns per pulse 8 Imaging spectrometer Swath width/FOV Cross track pixels Spectral range Spectral resolution Bands Irradiance spectrometer Swath width/FOV Spectral range Sample/Band width 310 m (50°) 1,004 400 to 1000 nm 10 nm 402 hemispheric (180°) 350 to 1,100 nm 1.5 and 1.5 nm Thermal camera Swath width/FOV Imaging array size Spectral range 173 m (30°) 384 × 288 8 to 14 μm Spatial resolution of data products Ht. 30 m Temp. 35° C * Flying at 335 m AGL and 110 kt. N 0 m 20° 1 m G-LiHT Platform & Operations Mounted through windshield Piper Cherokee Piper Cherokee Irradiance sensor G-LiHT installed Occupants: 3 max. (pilot, operator, observer) Fuel capacity: 80 gallons avgas Duration: 4 hours w/o reserves Range/Speed: 800 km @ 110 knots Acquisitions: July–August 2014 G-LiHT uninstalled NASA Carbon Monitoring System (CMS) Project R. Nelson (PI); B. Cook, D. Morton (NASA); H. Andersen, R. Pattison (PNW); A. Finley (Michigan State) ⑤ Bayesian joint probability models & maps of carbon and uncertainty ④ Fire and forest carbon stocks and losses ③ Enhanced inventory with tree variables ② Biomass & carbon inventory ① Experimental Design Plots, G‐LiHT and Landsat (146,626 km2) Burn statistics from MODIS and Landsat Tree variables derived from G‐LiHT multi‐sensor data (e.g., species class , size distribution) G‐LiHT LiDAR transects (3780 km2; 3% of total area) FIA field plots (0.006 km2) Data Sources & Upscaling Methodology Courtesy: Bruce Cook NASA-GSFC Example: LiDAR-assisted sampling design for biomass estimation Stage 1: LiDAR strip sample Stage 2: Subsample of field plots & prediction of biomass within strips via regression Plot AGB Total biomass = Estimated mean biomass in LiDAR strips × Total Area (+ correction factor to remove bias in modelassisted designs) AGB = f(LiDAR) LiDAR metrics USDA Forest Service PNW Research Station Forest Inventory and Analysis Example: LiDAR-assisted sampling design for biomass estimation, Kenai Peninsula, Alaska LiDAR acquisition in May, 2004 Covered 120 FIA plots N-S swaths spaced 10 km apart Total linear coverage: approx. 600 km Total area covered: 19,387 ha (2.3 % of western Kenai lowlands) Total cost: $60K ALASKA Mobilization: 10% Reflights: 10% Total size of raw binary data: 50 GB Biomass estimation within lidar strips used area-based regression approach* * Li, Andersen & McGaughey.2008. A comparison of statistical methods for estimating forest biomass from light detection and ranging data. West. J. of Applied Forestry. 23(4): 223-231. USDA Forest Service PNW Research Station Forest Inventory and Analysis Example: Median Lidar-based lidar intensity individual by species, tree measurement, Kenai KenaiPeninsula, Peninsula,AK AK Field Spc: Birch Field Ht: 18.8 m Lidar Spc: Hardwood Lidar Ht: 18.1 m Lidar Lidarcrown surface trees Field–measured segmentation model (> 12.5 cm) 50 50 40 All Data 40 50 Late May Lidar Acquisition 40 Early May Lidar Acquisition 30 20 White spruce Lidar individual tree species classification 20 Lidar Ht: 11.37 m Birch 30 Median Lidar Intensity 30 Lidar Spc: Conifer 20 Median Lidar Intensity Field Ht: 11.4 m Median Lidar Intensity Field 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 Example: LiDAR-assisted sampling design for biomass estimation, Kenai Peninsula, Alaska (cont.) Apply regression model to LiDAR metrics at each 13-m grid cell to obtain predictions of biomass over entire LiDAR coverage 250 200 150 100 50 Mean LiDAR height Coefficient of variation of LiDAR canopy heights LiDAR-derived canopy cover Coefficients vary by forest/species type adj. R2 = 0.68 0 Lidar-predicted Biomass (MG/ha) Plot-level LiDAR structural metrics* 300 LiDAR structural metrics are well-correlated with aboveground biomass at plots 0 50 100 150 200 250 300 FIA Subplot Biomass (MG/ha) *Li, Andersen & McGaughey. 2008. A comparison of statistical methods for estimating forest biomass from light detection and ranging data. West. J. of Applied Forestry 23(4): 223-231. USDA Forest Service PNW Research Station Forest Inventory and Analysis USDA Forest Service PNW Research Station Forest Inventory and Analysis Tanana Inventory Pilot: Statistical estimators and products Design-based estimation (Bechtold & Patterson, 2005 - aka “green book”) can be used for plot-based inventory estimates in TVSF & Tetlin NWR Active research underway developing statistical properties of 2-level designs integrating field plots & LiDAR sampling (Gregoire et al., CJFR 2011; Stähl et al., CJFR 2011; Mandallaz et al. CJFR 2013) Model-assisted (i.e. approximate design-unbiased) vs. model-based approaches 2-stage (i.e., cluster sampling) vs. 2-phase sampling designs Simulation analyses are helping to further develop and refine variance estimators for complex LiDAR-assisted inventory designs (Ene et al., RSE 2012) Inventory products: Report (PNW-GTR) describing inventory design and forest resources in TVSF & Tetlin NWR Maps of inventory attributes (w/ uncertainty) for entire Tanana unit (via Bayesian hierarchical modeling (Finley et al.) Peer-reviewed methods paper(s) USDA Forest Service PNW Research Station Forest Inventory and Analysis Thank you! USDA Forest Service PNW Research Station Forest Inventory and Analysis