NEMO_2011_SWB_V5

Using linear mixed models to evaluate forest sites capacity for the production of Pinus occidentalis, Sw. in La Sierra, Dominican Republic: 

Using linear mixed models to evaluate forest sites capacity for the production of Pinus occidentalis, Sw. in La Sierra, Dominican Republic Santiago W. Bueno, Ph.D . Pontificia Universidad Católica Madre y Maestra, Santiago NEMO 2011 October 3-4, 2011

Objectives: 

Objectives Develop curves and equations of site index to measure the productive capacity of natural stands located in three ecological zones within La Sierra (very humid subtropical forest, humid subtropical forest and subtropical dry forest) for the species Pinus occidentalis , using linear mixed models and data from 25 permanent plots. 09/11/2011 Indice de Sitio

Feature areas: 

Feature areas 09/11/2011 Indice de Sitio

Study area: La Sierra: 

Study area: La Sierra 09/11/2011 Site Index

Location of permanent plots: 

Location of permanent plots 09/11/2011 Indice de Sitio

Data: 

Data 25 Plots, 1 plot in each stand (9 in humid zone) (6 in intermediate) (10 in dry zone) Plots laid down randomly from 1984 until 1991. In the humid and intermediate zones all were established in 1988. In dry area: two in 1984, one in 1987, five in 1989 and two in 1991. Area of plots in has.(#): 0.125 (2), 0.1 (11). 0.0625 (12) Repeated measurements: from 2 to 7 09/11/2011 Indice de Sitio Mesurements 2 3 4 5 6 7 Plots 2 3 9 9 1 1

Variables: 

Variables DBH Age Average height of 15% tallest trees within each permanent plot. 09/11/2011 Indice de Sitio

Methods: Site Index estimation: 

Methods: Site Index estimation Repeated measurements on age and average height in each permanent plot were used to estimate the curves and the site index equations. Age (21 to 52 years) determined by ring count ( Burgt , 1993). In each permanent plot, heights of 15% tallest trees measured (162 trees with height ranging from11.50 to 31.25 m). A Schumacher type model in a mixed linear regression analysis framework and age base 40, was employed to generate the anamorphic site index equation and guide curves. 09/11/2011 Site Index

Slide 9: 

09/11/2011 Indice de Sitio In characterizing and modeling the spatial arrangement of the covariance structure we tested three candidates; Autoregressive, Compound Symmetric, Un-structured. Residual variation was assumed to be; random within plots (G Matrix). Fixed within trees (R Matrix). To select the best variance-covariance structure in the estimation of the fixed and random parameters we used the following criteria; -2 Log likelihood, Akaike’s Information Criterion (AIC), Bayesian’s Information Criterion (BIC). Methods: Site Index estimation

Methods: Analysis of indicator Variables: 

Methods: Analysis of indicator Variables To verify if different regression equations were needed in each of the three ecological zones the following dummy variables model was used. 09/11/2011 Indice de Sitio Where : Y: Natural log of tree height X: Inverse age Z: Indicator variable for the zones. ε : Error term β is: Coefficients to be estimated.

Methods: Analysis of indicator Variables: 

Inferential statistics for the analysis were base on the t partial statistic and the following hypothesis tests parallelism/ slope (1); equal intercepts (2); coincidence test (3) (the two regression curves coincide). 09/11/2011 Indice de Sitio Methods: Analysis of indicator Variables

Results: 

Results

Coefficients and Goodness-of-fit indicators: 

Coefficients and Goodness-of-fit indicators 09/11/2011 Indice de Sitio Parameter estimation for Site I ndex by zone and in general: M a ximum Likelihood Method following lineal mixed procedures Stand. Zone Covar. Par a meter Estimate Error P-Value -2LL AIC BICC Dry Auto regressive Intercept 3.368 0.084 <.0001 -138.9 -130.9 -129.6 Age -1 -13.087 1.890 <.0001 Intermid. Auto regressive Intercept 3.536 0.099 <.0001 -94.6 -86.6 -87.5 Age -1 -12.956 2.636 0.0001 H u mid Auto regressive Intercept 3.631 0.045 <.0001 -167.1 -159.1 -158.3 Age -1 -10.109 1.150 <.0001 Global Auto regressive Intercept 3.377 0.080 <.0001 -140.8 -139.7 -139.5 Age -1 -13.320 1.698 <.0001

Analysis of indicator variables: 

Analysis of indicator variables Pair wise Test comparison parallelism Intercept Coincidence Humid vs. Dry Non-significant P- Value: 0.5830 Non-significant P- Value: 0.1504 Significant P- Value: <.0001 Humid vs. Inter. Non-significant P- Value: 0.7172 Non-significant P- Value: 0.4855 Significant P- Value: <.0001 Inter. vs. Dry Non-significant P- Value: 0.7527 Non-significant P- Value: 0.9912 Significant P- Value: 0.0002 09/11/2011 Indice de Sitio Results from pair wise comparison on the three zones suggest that each region requires its own site index equation. Results from hypothesis test were similar for the three pair wise; regression lines on the three zones are parallel but no coincident .

Site Index in three zones: 

Site Index in three zones 09/11/2011 Indice de Sitio

Soil properties and Site Index: 

Soil properties and Site Index Soil Properties Plot Site Index Soil Depth Sand % Silt % Clay % 1 30 112 67 21 12 10 31 102 51 33 16 16 31 102 67 19 14 102 14 83 45 41 14 111 15 85 51 29 20 09/11/2011 Indice de Sitio

Results from stepwise OLS regression of soil characteristics on the dependent variable Site Index: 

Results from stepwise OLS regression of soil characteristics on the dependent variable Site Index Variable Parameter Estimate Standard Error Pr > t Intercept 12.294 4.295 0.008 Soil Depth 0.131 0.047 0.011 09/11/2011 Indice de Sitio

Slide 18: 

09/11/2011 Indice de Sitio

Slide 19: 

09/11/2011 Indice de Sitio

Anamorphic Site Index curves for P. occidentalis in the Humid zone: 

Anamorphic Site Index curves for P. occidentalis in the Humid zone 09/11/2011 Indice de Sitio

Conclusions: 

Conclusions The system of Site Index curves developed, is a simple mathematical expression of the biological concept of productivity . It allowed us to distinguish 13 grades of productivity in three ecological regions, using regression analysis characterized by linear mixed models. The predominant structure found to best fit the modeling of the variance-covariance was the unestructure . Stands with best productivity grade were those located in the Humid Zone, and were soils depth was higher. P. occidentalis silviculture could be more intensive in this life zone. The developed Site Index equations can be used to: Estimate future production, determine optimum rotation schedule, conduct timber financial analysis, estimate growth in height and volume, Objectively planning the use of forest land at La Sierra. 09/11/2011 Indice de Sitio