Transcript mcroberts.msts.pps

A brief introduction to statistical aspects of
the Forest Inventory and Analysis Program
of the USDA Forest Service
Ronald E. McRoberts
Patrick D. Miles
Forest Inventory and Analysis
North Central Research Station
USDA Forest Service
Forest Inventory and Analysis (FIA)
Mission:
To conduct forest inventories of the United
States to estimate:
 the extent (area) of forest land
 the volume, growth, and removal of forest
resources
 the health and condition of the forest
Forest Inventory and Analysis Regions
Pacific Northwest
Research
Station
Rocky Mountain
Research
Station
North Central
Research
Station
Northeastern
Research
Station
Southern
Research
Station
Strategic features
•
A standard set of variables with consistent
meanings and measurements
•
Field inventories of all forested lands
•
A national sampling design and plot configuration
•
A systematic, annual sample of each state
•
A national database with user friendly access
FIA: A 3-phase program
Phase 1:
Entails use of remotely sensed data to obtain initial plot land
cover observations and to stratify land areas with the objective
of increasing precision
Phase 2:
Entails field crew visits to locations of plots with accessible
forest to measure traditional suite of mensurational variables
Phase 3:
Entails field crew measurements of an additional suite
of variables related to the health of the forest on a 1:16
proportion of Phase 2 plots
Genesis of the FIA sampling design
With thanks to:
Tony Olsen
US EPA
Phase 3 (Forest Health Monitoring) hexagons
FIA plot configuration
14.63m
36.58m
0.40 ha
o
0
12
FIA Phase 2 observed variables
• Plot/subplot identification and location
• Observed condition (within subplots)
-
land cover, ownership, forest type,
stand age, size class, productivity class
-
origin, slope, aspect, physiographic class,
disturbance
• Observed tree attributes
-
location
-
species, status, lean, diameter, height,
crown ratio, crown class, damage, decay
FIA Phase 2 calculated variables
• Tree attributes
- volume
• Subplot attributes per unit area
- number of trees, volume, biomass
• By category
- species/species groups
- status: live, mortality, etc
Classification
Stratification
Using a forest/non-forest map
as a means of stratification
Number
of strata
Relative efficiency
IN
IA
MN
MO
1
1.00
1.00 1.00
1.00
2
2.00
1.68 2.82
2.33
4
3.94 1.89 3.22 2.89
Forest Health Monitoring (FHM)
- Detection monitoring through aerial and ground surveys
- Evaluation monitoring for particular situations
- Research on monitoring techniques
- Intensive site ecosystem monitoring.
FHM Ground Detection Monitoring
•
Fully integrated as Phase 3 of the FIA program
•
Indicators monitored:
-
tree crown condition
-
tree damage
-
ozone injury to vegetation
-
lichen diversity
-
vegetation diversity
-
soil chemistry and erosion
-
coarse woody debris
DWM Sample Design
Fuel loadings (tons/acre)
Output Products:
Spatial output products

National attribute maps

Ownership maps

Map-based estimation
confidential plot location
proprietary information
National forest biomass map
Small area mapping
and map-based estimation
•
Users want estimates at spatial scales for which
FIA does not report estimates
•
Users want to use FIA data to train satellite
image classifiers or as accuracy assessment data
•
Requires access to plot data and locations
•
Plot locations are confidential
protect integrity of sample
deter owner access denials
protect proprietary information
Small area mapping and map-based estimation
30 km
Proportion forest
30 km
Volume
Radius
(km)
Volume
3
6
9
12
15
Plots
1
6
10
16
25
Design-based
Mean
SE
1032.2
1026.6
871.6
899.4
807.6
Proportion forest area
3
1
6
6
9
10
12
16
15
25
1.000
0.833
0.700
0.625
0.600
Model-based
Mean
SE
t*
------405.0
281.7
286.5
210.4
916.7
886.1
841.6
833.2
843.6
162.2
80.5
54.1
39.7
30.5
----0.35
0.01
0.23
-0.17
------0.164
0.153
0.125
0.100
0.775
0.730
0.649
0.632
0.648
0.055
0.025
0.015
0.011
0.009
----0.63
0.33
-0.04
-0.48
Forest
Inventory
Mapmaker
Fuel Treatment
Evaluator
Area of forestland (hectares)
Total
Stand-size Large
Medium Small
Nonstoc
class
diameter diameter diameter ked
27007 BELTRAMI
338,156 102,547 131,199 102,264
27057 HUBBARD
161,793
27061 ITASCA
Total County code
42,087
80,775
2,145
38,324
607
535,561 172,963 175,067 184,739
1,035,511 317,598 387,042 325,327
2,792
5,544
FIA
Growth models
SpaRRS –
Spatial Resource
Support System
Forest
Inventory
Mapmaker
Geographic options
County retrievals →
Circular retrievals →
Polygon retrievals →
Generate tables, maps and data
Area of forestland(hectares).
Beltrami
Hubbard
Itasca
Total
Total
352,678
154,785
528,240
1,035,703
Large
Medium Small
diameter diameter diameter Nonstocked
79,134 147,005 118,948
7,591
42,361
62,335
49,711
378
114,651 194,064 213,273
6,252
236,147 403,404 381,932
14,221
Figure 1. Private timberland as a proportion of all land.
Percent of timberland with ash
Fuel Treatment Evaluator
•
Applies thinning prescription to each plot
•
Estimates torching and crowning index
for each plot before and after treatment
•
Estimates revenues for each plot by tree
component
•
Estimates harvest costs for each plot
Map results of silvicultural prescription
Initial biomass
Removed biomass
Remaining biomass
Graph results of silvicultural prescription
SpaRSS
Spatial Resource Support System
Focus of support:
•
assembly of relevant digital data layers
•
analyses based on the integration
of spatial data
•
comparison of results from
different integration approaches
•
comparison of results from
different decision alternatives.
Objective:
Identify forested areas of the USA
that satisfy three criteria:
•
high wildfire risk
•
close to rural communities
•
in need of economic assistance
Removable biomass from FIA data
Removable biomass
- upper 50th percentile
Upper 50% removable biomass
Condition classes 2 and 3
< 25 miles to rural community
Lower two economic classes
2002 S&PF EAP-NFP funding allocation
Outstanding statistical issues:

Sampling frame and variance estimation

Modeling issues
- model-based estimation
- propagation of error
- regional variations

Combining data from multiple panels

Plots, subplots, and microplots

Non-sampled areas
- Western Texas and Oklahoma
- Piñon-Juniper area (what is a tree?)
- Interior Alaska
Technical statistical issues
 Stratification schemes
- plots that sample multiple strata
- squeezing more precision from stratifications
•
The effects of measurement error
 Map accuracy assessment
 Modeling issues
- design-based versus model-based estimation
- propagation of error
- regional variants
Summary
Forest Inventory and Analysis
U. S. Forest Service
• Nationally consistent inventory
• Emphasis on spatial products
• Resource support tools
- Mapmaker: data access
- Fuel Treatment Evaluator
- Spatial Resource Support System
• Statistical issues
8 FIA sessions at MSTS*
Wed
Thu
am
am
Forest inventory and monitoring policy
Remote sensing applications
pm
pm
Statistical applications
Assessing forest sustainability
am
am
Diversity, change, and stability
Approaches to forest health monitoring
pm
pm
Forest health criteria and indicators
Carbon accounting applications
* FIA scientists will also be speaking in non-FIA sessions
Remember …………..
………….. only YOU can prevent forest fires!