Designated Forestland Presentation - UW GIS Certificate Program

Estimating Market Value of Designated Forestland:

Estimating Market Value of Designated Forestland Sponsor:  Luke Rogers, UW College of the Environment Students: Robbie Andrus, Trevor Wong University of Washington GIS Certificate Program Spring 2011

Project Description:

Project Description Designated Forestland (DFL) – Created to maintain forested lands for timber and ecological benefit – DFL = 20+ acres of timber + no improvements – Average = Operability + Land Grade Problem: "Designated forestland” (DFL) is currently valued using a table of statewide averages from the Department of Revenue. – No individual assessment – Only forest lands not required to be appraised by county assessor Goal: Develop model to estimate market sales price of DFL lands. – Model should be based on land’s “highest and best use” 2 | 15

Risks & Deliverables:

Risks & Deliverables Risks Experiment - An educated trial and error process Is there any relationship at all between the value of one assessed forest land parcel vs. another? Assessor’s true market value Compare to assessor’s true market value for another source Deliverables Model to assess market value of parcels in DFL program (R 2 > .75) Cartographic examples of model and model process A final presentation Documentation of model and project workflow 2 | 15

Current DFL Assessment:

Current DFL Assessment 3 | 15 Parcel comparison DFL non-DFL ID P42244 P42236 Zone IF IF Site Class 4 3 UGA Dist 3 2 Rds Dist 2 2 Slope 12 13 Acres 160 158.49 TMV $24,000 $319,300 TMV/Acre $150 $2,021

Slide 5:

4 | 15

Skagit Zoning + DFL:

Skagit Zoning + DFL Parcel Stats: Total Parcels in Skagit Co.: 78,925 DFL Parcels: 1372 - only 137 w/o package deals Skagit Zoning Federal Lands UGA’s DFL Timberlands 5 | 15

Slide 7:

Model 1: Attribute Model Skagit Zoning Federal Lands UGA’s DFL Timberlands 6 | 15

Slide 8:

7 | 15

Slide 9:

Model 1.5 Forest Cover Attribute Model Forest Cover Model 8 | 15 Source: National Land Cover Dataset Restrictions: Covered by forest No public lands Near DFL Parcels

Slide 10:

Model 2 An investigation of neighboring parcels 9 | 15 Restrictions: New study area No public lands No DFL Parcels No UGA

Slide 11:

Model 2: Total Market Value 10 | 15

Slide 12:

Euclidean Allocation 11 | 15

Slide 13:

Focal Mean (1 mi) 12 | 15

Slide 14:

Model 2: Total Market Value 10 | 15

Slide 15:

Interpolation: IDW and Krige 13 | 15

Slide 16:

14 | 15

Best Effort in R:

Best Effort in R Summary of the Linear Regression model (built using lm): Call: lm(formula = RLG_Mod_Wgt_Ac ~ ., data = crs$dataset[crs$train, c(crs$input, crs$target)]) Residuals: Min       1Q   Median       3Q      Max -2.16862 -0.48981 -0.03274  0.56682  2.40408 Coefficients: Estimate  Std. Error t value Pr(>|t|) (Intercept)     9.53120805  0.44381397  21.476  < 2e-16 *** IDW3            0.00018021  0.00004139   4.354 3.03e-05 *** RLG_TOTAL_AREA -0.60002975  0.08794881  -6.822 5.23e-10 *** --- Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Residual standard error: 0.9183 on 109 degrees of freedom Multiple R-squared: 0.3757, Adjusted R-squared: 0.3642 F-statistic: 32.79 on 2 and 109 DF,  p-value: 7.086e-12 2 | 15

Slide 18:

18

Slide 19:

Model 2.1 Forest Cover Neighbors Model 8 | 15 Source: National Land Cover Dataset Restrictions: Revised study Area 50% Forest Cover No public lands Near DFL Parcels 131 Forested Parcels

Slide 20:

20

Slide 21:

Model 3: Revised Study Area TMV vs. TMV - Focal 10 | 15 Restrictions: Smaller Study Area ¼ mi UGA 1 mi DFL TMV $1,000-$18,000

Slide 22:

Model 3: Revised Study Area TMV vs. TMV - Interpolation 10 | 15 Restrictions: Smaller Study Area ¼ mi UGA 1 mi DFL TMV $2,000-$15,000 Land use restriction

Slide 23:

23

Slide 24:

The Data Dilemma Large Sample Size: Include many parcels Model 2.0 / 3.0 Parcels included may not be applicable e.g. Trailer Park Small Sample Size: Exclude many parcels Model 2.1 / 3.1 Significantly reduced sample size Lots of statistical noise High margin of error 15 | 15 County DFL Parcels Forest Cover Parcels (Model 1.5, 2.1) Privately Held Parcels (Model 2.0, 3.0, 3.1 ) Skagit 137 157 1800 Kitsap 231 N/A N/A Okanogan 108 N/A N/A

Slide 25:

25 Project Documentation

Slide 26:

Deliverable Comparison Original vs. Reality Model to assess market value of parcels in DFL program (R 2 > .75) 1.0 - Attribute Model (R 2 = .02) 1.5 - Forest Cover Model (R 2 = .03) 2.0 - Nearest Neighbor Model (R 2 = .04) 2.1 – Forest Cover Nearest Neighbor Model (R 2 = .05) 3.0 – Revised Nearest Neighbor Model (R 2 = .41) Cartographic examples of model and model process Successfully completed A final presentation Here it is!! Documentation of model and project workflow Check!! 15 | 15

Slide 27:

Summary Conclusions: Currently accept Null Hypothesis No relationship Why? Deficiencies in data Few comparable sales Forestland-to-Forestland Transactions do not generate usable sales data Perhaps we could know more about Statistics Goals Achieved: Data: Acquisition, Import, & Manipulation ModelBuilder Build Troubleshoot Output Cartographic Representation 15 | 15

Slide 28:

Methodology Difference Robbie : Using DFL Sales Prices: Historical Sales Data + DFL Attributes (size, location) = Predicted Market Value Trevor : Using TMV Assessed Value: Annually Assessed TMV Market Values + TMV Attributes (size, location) = Predicted Market Value Best R 2 = 0.108 Best R 2 = 0.112

Slide 29:

Thank You Sponsor: Luke Rogers, Our Instructors Suggestions ? Questions?