1 / 55

Model Comparison for short-term growth projections on a Western Oregon property

Model Comparison for short-term growth projections on a Western Oregon property. GMUG Meeting June 2, 2010. G&Y Models Survey. *. 15 Respondents (Industry + TIMOs + Consultants), multiple responses allowed. G&Y Model Used Percent SPS/FPS 8 28% FVS 7 24% ORGANON 5 17%

Download Presentation

Model Comparison for short-term growth projections on a Western Oregon property

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Model Comparisonfor short-term growth projections on a Western Oregon property GMUG Meeting June 2, 2010

  2. G&Y Models Survey * • 15 Respondents (Industry + TIMOs + Consultants), • multiple responses allowed G&Y Model Used Percent SPS/FPS 8 28% FVS 7 24% ORGANON 5 17% In-House 5 17% Other 2 7% Excel 1 3% DFSim 1 3% * PACIFIC NORTHWEST TIMBERLAND INVESTMENT SURVEY RESULTS, MARCH 2010, SIZEMORE & SIZEMORE, INC.

  3. Western Oregon South of Portland, North of Eugene Stands in Valley & up west slope of Cascades Data - Location

  4. Data - Remeasurement Count of Plots:

  5. Data – Age & SI

  6. Site Classes * Prior to 1997, SI class assigned based on growth rings

  7. Stand Statistics (Init)

  8. Merch Specs

  9. 4,382 trees across 15 species DF (3,469), WH (604) Tree Statistics

  10. SIS (SPS) FPS (v6.73) FVS (12/16/2008) PN, WC ORGANON (v8.2) NWO, SMC Growth Models

  11. SIS (SPS) Client coefficient file FPS Library 11 FVS Max SDI: DF 600, WH 720, RA 300 ORGANON (v8.2) Max DF SDI = 600 Growth Models - Options

  12. Measured vs. Predicted Average Growth Basic Error Statistics Error CDF’s Error Significance Tests Linear Regressions of Errors 50-Year Growth Projections Comparison Methods

  13. Period length is not uniform across plots The same set of data was used across models Relative size of errors important Comparison Caveat

  14. High level view of growth PAI Focus on four youngest age classes due to sample size Measured vs. Predicted Average Growth

  15. Measured vs. Predicted Average Growth

  16. Measured vs. Predicted Average Growth

  17. Measured vs. Predicted Average Growth

  18. Measured vs. Predicted Average Growth

  19. Measured vs. Predicted Average Growth

  20. Performance? SIS, PN, FPS, NWO all about same so far SMC not as good WC least accurate so far Measured vs. Predicted Average Growth

  21. Basic Error Stats - Stands • Errors: Predicted - Observed • Negative = under-prediction • Positive = over-prediction • Errors not annualized** • Stand-level and Tree-level • FPS – no grown treelist output

  22. Basic Error Stats - Stands

  23. Basic Error Stats - Stands

  24. Basic Error Stats - Trees

  25. Basic Error Stats - Trees

  26. Basic Error Stats - Trees

  27. Basic Error Stats • Performance? • SIS smallest mean errors for all but BA (2nd smallest) and average variability • PN probably second best choice • NWO was least variable • SMC and WC falling out of the race

  28. Error CDF’s - Stand • Cumulative Distribution Functions • Picture of error distribution by error size • More information than a mean bias

  29. Error CDF’s - Stand

  30. Error CDF’s - Stand

  31. Error CDF’s - Stand

  32. Error CDF’s - Stand

  33. Error CDF’s - Stand

  34. Error CDF’s - Tree

  35. Error CDF’s - Tree

  36. Error CDF’s - Tree

  37. Error CDF’s - Tree • Performance? • NWO and SIS had largest % of small errors • FPS and SMC next best, with PN close behind • WC doesn’t seem to be a good fit

  38. Significance Tests • Paired t-test vs. Wilcoxon signed rank test • All data • 5-7 yr vs. 8-10 yr Projections • Initial Vol <= 25 MBF vs. Initial Vol > 25 MBF • Unthinned vs. Thinned

  39. Significance Tests

  40. Significance Tests

  41. Significance Tests

  42. Significance Tests

  43. Significance Tests

  44. Significance Tests • Performance so far? • Shorter projections generally better (higher p-values) • Less significance in higher volume stands than lower • Thinned vs. unthinned unclear • SIS and PN least significant differences, followed by NWO

  45. Simple Linear Regression • Oi = β0 + β1*Pi + εi • Nature of the bias if it is present, telling us whether it is constant or changing and to what degree.

  46. SLR Results Significance at α = 0.05

  47. SLR Results • Performance so far? • NWO did best, especially when recall low variability • FPS and SMC the next best • SIS and PN several significant slopes and intercepts

  48. Long-Term Projections • Long-term remeasurement data uncommon • Grow young stands and plot against older inventory data • Less for identifying a ‘best’ so much as pointing out models to avoid

  49. Long-Term Projections

  50. Long-Term Projections

More Related