Download presentation
Presentation is loading. Please wait.
Published byKory Murphy Modified over 9 years ago
1
University of Washington Experiences conducting large scale growth and yield simulations using FIA inventory plot data and the Forest Vegetation Simulator 2013 Western Mensurationists Meeting June 23-25, 2013 Dr. James B. McCarter University of Washington and North Carolina State University Photo Credit: Logging of Douglas fir trees in Washington state. (iStockphoto/Phil Augustavo)
2
University of Washington Outline Introduction Projects – Western States Fire Risk/Carbon – WA Biomass Assessment – US Biomass/Life Cycle Assessment Challenges/Lessons/Conclusions
3
University of Washington Introduction USDA Forest Service FIA inventory data has been the source of information for a series of large scale biomass and carbon assessment projects. An overview and challenges associated with three projects will be presented. Western States (Fire Risk/Carbon) WA BiomassUS Biomass/Lifecycle States11149 Plots16,6076000125,194 Variants15520 Simulations166,070 *3~12,000,000~3,129,850+ ?? DataFIAGNN, WA DNR GIS, derived GIS FIA, US GIS Year2009-201320112012-2013
4
University of Washington Western States Fire Risk/Carbon The Fire and Carbon Project examined carbon sequestration and fire risk for 11 western states. Each candidate inventory plot (latest inventory and target forest type) for the 11 western states was simulated under 10 management alternatives. 11 Western States – AZ, CA, CO, ID, MT, NM, NV, OR, UT, WA, WY Target Forest Types: DF, PP, S/F, H/C, OP, MC, HW, P/J 16,607 FIA Plots [latest complete cycle or last periodic (2)] 15 FVS Variants ~166,070 pathways for each generation of runs
5
University of Washington Western States - Plot Locations
6
University of Washington Western States - FVS Variant Coverage State# FIA Plots FVS Variants AZ1626CR (1626) CA1754CA (441), NC (20), SO (200), WS (875) CO1808CR (1807), UT (1) ID1143CI (509), EM (4), KT (121), NI (402), TT (95), UT (12) MT1735EM (965), KT (454), NI (315), TT (1) NM2223CR (2223) NV331CI (1), CR (291), WS (39) OR2497BM (638), CA (138), EC (106), NC (62), PN (419), SO (557), WC (577) UT1984CI (34), CR (13), UT (1937) WA1483BM (30), EC (607), NI (185), PN (326), WC (335) WY1441CR (793), TT (582), UT (66)
7
University of Washington Western States – Scenarios Simulated No Management (Base) Current Typical Management (different management for FS, States, and Private) Fire Reduction Scenarios – Fire Risk: Crowning Index classified into High (CI CI 50) – Fire1 - High (BA 45, from below); Mod (½ BA from below). Assumes follow up fuel treatments. – Fire2 - High (BA 45, from below); Mod (½ BA, from below). Applied when stand at risk. – Fire3 - Same as Fire1, except thinning not done if < 40 TPA – Fire4 - High (BA 40, proportional thin); Mod (½ BA, proportional thin) – Fire5 - High or Mod (Thin all trees <9”) Active Management High Revenue Carbon Sequestration Scenario Wildfire (overlay – each scenario above run with stochastically scheduled wildfire) – one run of historical fire, one increased fire
8
Western States - Results Treatment Effects on CI Mean Carbon w/ SD Mean Carbon w/Max & SD Mean Stand, Products, and Landfill Carbon
9
University of Washington Western States – Results Gifford Pinchot NF LCA No ManagementActive Management
10
University of Washington Western States - Spatial Results No ManagementActive Management 2007 2107
11
University of Washington Western States - Conclusions Combination of # plots and growth model variants allows for each scenario to be run as a batch for whole state The simulation results represent a modeling database (>75 GB, zipped) that can be mined for management alternatives that provide a synergy between management objectives. – Determine forest types and regions where sequestration provides more benefit and at lower fire risk – Determine forest types and regions where fire risk reduction treatment are most effective
12
University of Washington WA Biomass Assessment The WA Biomass project used Landscape Ecology, Modeling, Mapping & Analysis (LEMMA) Project outcomes as starting inventory condition for a state wide biomass assessment for WA Department of Natural Resources. 1 State 5,999 unique FCIDs 194,500,000 - 30m Pixels (105,421,583 forested) 5 FVS Variants ~12,000,000 pathways
13
University of Washington WA Biomass – Input Spatial Information: – LEMMA (Landscape Ecology, Modeling, Mapping & Analysis) Project outcomes for starting vegetation map. Results from two LEMMA projects provided 30m pixel resolution forest cover (2000 eastside, 2006 westside) – WA State Parcel based ownership, forest type, management zone, transportation, hydrology, elevation, slope, etc. Inventory Information: LEMMA tree list data. – Imputed (GNN) tree lists from FIA, FS CVS, and other – Updated with harvest information 2000-2009 (WA Dept. Revenue) and apply growth and regeneration to create 2010 inventory – Apply management scenarios based on forest type, ownership, and management zone based on information gleaned from Forest Operations Surveys
14
University of Washington WA Biomass – Spatial Analysis Input Layers & Geospatial database Source spatial data Derived spatial data
15
University of Washington WA Biomass - GNN Spatial Map Stand Basal Area in 2006 Darker green is higher stand basal area.
16
University of Washington WA Biomass – Forestland Parcel Database All owner information and parcel geometry for the Biomass Database is derived from the 2009 Washington State Parcel Database (Parcel Database). The Parcel Database contains parcel data from 42 different source data providers, including Washington’s 39 Counties, The Washington State Department of Natural Resources (DNR), The Washington Department of Fish and Wildlife (WDFW), and the United State Bureau of Land Management (BLM).
17
University of Washington WA Biomass – Spatial Database Applying the management alternatives, staggered by cycle results in ~ 12 million simulations (~ 4 days on 4 core computer) Simulation results summarized, biomass estimates calculated, and results loaded into spatial database Spatial database can be used to calculate potential biomass based on management alternative by forest type, ownership, management zone and provide economics based on transportation constraints, haul distance and potential biomass price. Biomass Calculator created to provide interface to examine results.
18
University of Washington WA Biomass - Putting It All Together
19
University of Washington WA Biomass – Feedstock Areas
20
University of Washington WA Biomass – Biomass Calculator
21
University of Washington WA Biomass - Conclusions Combination of ownership, GNN source pixels, FVS variant, management constraints results in 550,288 unique FCIDs Simulations run as county level batches, entire state too large for FVS – 2 GB file limit (database and text files) – Balance between image loads and file sizes Biomass Calculator provides public access to assessment results Feedstock areas, biomass prices sensitivity, transportation sensitivity, and competition between facilities can be examined
22
University of Washington US Biomass/Life Cycle Assessment The Biomass and Carbon Assessment project examines biomass availability and carbon sequestration for the continental US. This project expands the carbon sequestration assessment of the Fire and Carbon Project to the continental US. In addition it provides biomass estimates and area by forest type based on FIA population estimation factors. 49 States 125,194 FIA Plots (119,010 with trees, 376 species) 148 Forest Types, 37 Provinces, 1160 Ecological Subsection Codes 20 FVS Variants ~3,129,850+ pathways
23
University of Washington US Biomass/LCA – Plot Locations
24
University of Washington US Biomass/LCA – Biomass FIA Database Biomass: Total, Merch, NonMerch
25
University of Washington US Biomass/LCA County level simulations work for this dataset – Run into FVS file size limits when examining model behavior for variants with many habitat types (segment by location/forest to get around this) Ongoing project, still solving some issues: – Appropriate regeneration – Site quality (Habitat type, EcoClass, and Site Index) – Consistent biomass estimates across regions and species
26
University of Washington Comparing Projects Western States (Fire Risk/Carbon) WA BiomassUS Biomass/Lifecycle States11149 Plots16,607~6000125,194 Variants15520 Simulations166,070 * 3~12,000,000~3,129,850+ ?? DataFIAGNN, WA DNR GIS, derived GIS FIA, US GIS SpatialNoYesPartial – area represented Year2009-201320112012-2013
27
University of Washington Challenges/Lessons/Conclusions Large scale simulations are difficult, but possible – Which model/variant, what site quality/habitat code, variant differences, data differences, time differences, etc… These projects require a lot of data: – Need fast computers, 64 bit (>3 GB RAM), large/fast hard drives (10,000 RPM, SDD), and fast networks, but computer resources are CHEAP compared to analytical capability and keeping up with technology Knowledge required to use data sources, models, and perform additional analysis (numbers ≠ information) Data sources and models not always well matched: – Bad codes in database, typos – Codes in database not supported by model
28
University of Washington
29
Challenges/Lessons/Conclusions Large scale simulations are difficult, but possible – Which model/variant, what site quality/habitat code, variant differences, data differences, time differences, etc… These projects require a lot of data: – Need fast computer, 64 bit (>3 GB RAM), large/fast hard drives (10,000 RPM, SDD), and fast networks, but computer resources are CHEAP compared to analytical capability and keeping up with technology Knowledge required to use data sources, models, and perform additional analysis (numbers ≠ information) Data sources and models not always well matched: – Bad codes in database, typos – Codes in database not supported by model
30
University of Washington Knowledge required to use data sources properly FIA data used from PLOT, COND, TREE, and SEEDLING tables. Other tables required to select proper records! – Need to preserve plot layout information (subplots) for proper weighting of results Not all FIA plots are growth model (FVS) “friendly” – What happens when you “loose” a plot from the analysis? Can you get the model or data “fixed” in analysis time frame? Can you drop the plot from the analysis? Inventory data passed through GNN process did not have all FIA data fields available – Had to fill in habitat types, which dictate growth potential for pixels
31
University of Washington Knowledge required when using models Site Quality Selection Regeneration Growth Results Evaluation Growth Model Constraints
32
University of Washington Site Quality Selection – Different habitat types (in Western US) have different behaviors – which selected matters! – Range of possible results can be considerable – Habitat type in FIA does not always map cleanly to FVS growth model
33
University of Washington Regeneration Relatively well known for industrial land and plantations, not as well known for less intensive and more diversified management objectives Appropriate regeneration of hardwood stands and currently bare or understocked plots can be challenging
34
University of Washington Regeneration Different site/habitat and regeneration have varying results at same location
35
University of Washington Growth Results Evaluation Expected stand development for range of site classes from DF yield tables (McArdle et al. 1949)
36
University of Washington Growth Results Evaluation Most stands exhibit expected behavior with rapid increase for young stands and leveling off for older stands Stands above SDI=600 are suspect. Mixed stands with significant tolerant species component can be higher
37
University of Washington Growth Results Evaluation Most stands exhibit expected behavior with increase in younger stands and leveling off of older stands. Several suggest suspect behavior: 1.rapid increase and then crash, 2.sustained increase over entire period.
38
University of Washington Growth Model Constraints MaxSDI values matter (not always set low enough in FVS habitat types (PN-PSME-2 DF MaxSDI=950, EC-PSME-34 MaxSDI=767) Maximum SDI values by habitat type and species extracted from FVS variants, then examined and adjusted based on habitat manuals and published max values by species
39
University of Washington Challenges/Lessons/Conclusions Biomass equations Biomass calculation method selected can be critical to results
40
University of Washington Challenges/Lessons/Conclusions FVS Variant specific issues: – EM variant does not like “large amounts” of regeneration – CR variant is “somewhat fragile”, several math overflow errors exposed by running FIA plots. – Emergence of “new” underflow errors across variants Some variation is the result of suspect growth projections from the growth model: – exploring “dark corners” of FVS modeling database
41
University of Washington Challenges/Lessons/Conclusions We need to be sure to minimize the impact of assumptions/choices required for the analysis – Simulations after harvest (regeneration) no longer analysis of inventory data, overall results strongly influenced by regeneration – If you harvest the majority of plots at the beginning of an analysis, then you are doing an analysis of your regeneration choices and how they respond in the growth model, not an analysis based on initial plot data (FIA) New challenge is sifting through the outputs for information Model users should be cooperators with model developers toward model improvements
42
University of Washington Thank You Data and Tools Used: – FIA (http://fia.fs.fed.us/) – FVS (http://www.fs.fed.us/fmsc/fvs/) – Python (www.python.org) – R (http://cran.r-project.org/) – ArcGIS (www.esri.com) – MS Office (Access and Excel)
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.