Tree Inventory Overview - i-Tree

Download Report

Transcript Tree Inventory Overview - i-Tree

Sampling in i-Tree
Concepts, techniques
and applications
Introduction
Sampling is so pervasive in
i-Tree that we have factored it
out for a separate discussion
Overview
Concepts
Techniques
Applications
Concepts I
Random sample
Data collection in which every member of the
population has an equal chance of being selected
 Population = the set of people or entities to which
findings are to be generalized.
 The population must be defined explicitly before a
sample is taken
Can sometimes break population into subgroups
(stratification) for better numbers
Mind tricks easily, so need rigorous method
Source: http://www.negrdc.org/counties/madison/comprehensiveplans/newcomp/maps/8_01ExistLandUseMadisonCo.jpg
Concepts II
Variance
 = (SD)2
 Measure of how spread out the distribution is,
i.e., how much individual samples vary
 The less the individual measurements vary from
the mean (average), the more reliable the mean
 In an urban forest, different traits to investigate
(variables) may have different variances
 Species distribution (high?) vs. population size (low)
 Hurricane debris (high?) vs. ice storm debris (low)
Source: Dave Nowak
and Jeff Walton,
personal communication
(DRG data)
Concepts III
Sample size
Will need to be larger
 the weaker the relationships to be detected
 the higher the significance level being sought
 the smaller the population of the smallest subgroup
 the greater the variance of the variables
Can be smaller as these factors change,
especially as variance goes down
Source: Dave Nowak, personal communication
Concepts IV
Standard error (SEM)
The Standard Error (Standard Error of the Mean)
calculates how accurately a sample mean
estimates the population mean.
Formula: SEM = SD/N , where SD = “standard
deviation” of the sample, and N = sample size.
Note that as SD goes down or N goes up, SEM
gets smaller—i.e., estimate becomes better.
Commonly represented by “±” after a number.
Source: blogaloutre http://www.ontabec.com/fatigue.jpg
Techniques I
Get random
numbers
Tables
Telephone book (final
digits!)
Electronic
randomizers
 Online
 Desktop
 PDA
Techniques II
Select plots
Use map techniques
 Grid overlay for maps/photos
 Simple edge rulers also work
Pick randomly from list
 Street, with replacement
 Block number
Create random coordinates
 Spreadsheet
 GIS
Techniques II
Easy way to get
random list of street
segments
Bring TIGER/Line files
as shape file from ESRI
into a GIS
Details in Appendix B of
the Manual
Techniques III
Reserve
Create more plots than needed
Something like 10%
Take replacements from list in order
when plot must be thrown out
 Non-existent
 Unfindable
 Inaccessible
No bias!
Application I
Inventory types
Complete Inventory
 Costly, time-consuming
Partial Inventory
 Complete inventory of some forest segment
Sample Inventory
 Randomly-selected trees inventoried for large-
scale interpretation
Cost-efficient
 Good for planning
 Not suitable for day-to-day field management

Application I
Sample inventory benefits
Increase public safety
Facilitate short- and long-term planning
Improve public relations
Justify budgets
Estimate tree benefits
Large gain for small investment
i-Tree promotes the value of sampling
Applications II
Manual sampling techniques
valid, but tedious for larger areas
i-Tree v. 1.0 will include
applications to automate the
process for two types of plots:
Linear (street) plots/segments
 STRATUM/MCTI, SDAP
Spatial (park, any area) plots
 UFORE
Applications II
Linear plot selector
 STRATUM/MCTI
 SDAP
Final testing
Requirements
 ArcMap 8.3 or 9.0
 Polygon file delimiting study area boundary
 Road shape file (TIGER/Line data)
Applications II
Spatial plot selector
 UFORE
 Final testing
Requirements
 ArcMap 8.3 or 9.0
 Polygon file delimiting study area boundary
 Raster-based file of strata (e.g., land uses)
within study area
 Digital aerial photos (optional)
Final sampling thoughts
Sampling is our friend
Both tool and product
in i-Tree
Understanding of
validity of what i-Tree
offers will depend
critically on
understanding the
process and capability
of sampling