imputating snag data to forest inventory for wildlife habitat modeling kevin ceder college of forest...

Imputating snag data to forest inventory for wildlife habitat

modelingKevin Ceder

College of Forest ResourcesUniversity of Washington

GMUG – 11 February 2008

Why impute snag data?

• Snags are an important habitat element and needed for habitat assessments.

• These data are often not collected in forest inventory

• The Large-Landscape Wildlife Assessment models will need these data

Why use Nearest-Neighbor?

• Non-parametric requiring no assumptions of underlying functional form

• Retains the variance/covariance structure of the input data in the output data

The Questions

1) Can snag data be imputed using kNN techniques with stand and site data?

2) How well do the results fit observed data?

3) Which distance measure performs best? 4) What is the effect of increasing

neighborhood size?5) How do the results compare with random

sampling?

The Process

• The database– FIA integrated database version 2.1– Data for private forests in western Washington

(1510 plots)– Both tree and snag data collected between

1989 - 1991– Representative of the forest targeted for the

LLWA project

The Process

• The tool - – The yaImpute package for kNN imputation

• Raw, Euclidean, Mahalanobis, MSN, MSN2, ICA, and randomForest distance measures

• k = 1, 2, 3, 4, 5, 10• For k>1 imputed data are distance weighted means of

neighbors

– 9999 permutations of the data for comparisons with random sampling

• k = 1, 2, 3, 4, 5, 10• For k>1 imputed data are distance weighted means of

neighbors using Euclidean distance

The Statistics

• Goodness of fit • Comparison with random

yyRMSD oi

yybias oi

yyMAD oi

The Input Data – Tree and site data (xData)

N = 1510 Min Max Mean

Trees per Acre (TOT_TPA) 6.7 2920.5 475.8

Basal Area per Acre (TOT_BA, sqft/ac) 0.0 397.8 119.8

Quadratic Mean Diameter (QMD, in) 0.1 28.5 7.6

Mean Height (MEAN_HT, ft) 1.0 147.9 43.7

Stand Age (AGE, yr) 5 215 37

Site Index (SITE_INDEX_FIA, feet @ 50 yr) 44 180 112

Slope (SLOPE, %) 0 99 24

Aspect (ASPECT_DEG, deg) 0 130 155

Elevation (ELEV_FT, ft) 3 4724 869

The Input Data – Snag data (yData)

N = 1510 Min Max Mean

Snags per Acre (SNAG_TPA_TOTAL) 0.0 96.8 4.8

Basal Area (SNAG_BA, sqft/ac) 0.0 9.7 0.5

Quadratic Mean Diameter (SNAG_QMD, in) 0.0 10.5 2.7

Mean Height (SNAG_ MEAN_HT, ft) 0.0 161.0 14.9

• 695 of 1510 plots did not have snags present

Results

1) Can snag data be imputed using kNN techniques with stand and site data?

Results

RMSD SPA BA QMD Mean Ht

Min 7.0 0.6 2.5 18.9

Max 11.2 1.0 3.7 28.4

Mean 9.0 0.8 3.0 23.4

Results

BIAS SPA BA QMD Mean Ht

Min -1.7 -0.2 -0.6 -4.0

Max -0.1 0.0 0.1 0.3

Mean -0.5 0.0 -0.1 -0.8

Results

MAD SPA BA QMD Mean Ht

Min 3.5 .3 1.8 11.3

Max 6.2 0.6 2.7 18.1

Mean 5.1 0.5 2.3 15.1

Results

Marginally…

• High RMSD and MAD relative to mean snag measures in the data

• Observed vs imputed plots show poor patterning

Results

2) Which distance measure performs best?

3) What is the effect of increasing neighborhood size?

Results

2) Which distance measure performs best?• All are generally similar• randomForest imputations provide lower

RMSD and MAD but under-predict more than others

3) What is the effect of increasing neighborhood size?

• Increasing k reduces RMSD and MAD• Little effect on bias• Slightly decreased range in imputed values

with k = 10

Results4) How do the results compare with random

sampling?

RMSD k 1 2 3 4 5 10

SNAG_TPA_TOTALNN 11.24 9.96 9.15 8.70 8.62 8.35