Ecological Archives C006-105-A4

Ekananda Paudel, Gbadamassi G. O. Dossa, Marleen D. de Blécourt, Philip Beckschäfer, Jianchu Xu, and Rhett D. Harrison. 2015. Quantifying the factors affecting leaf litter decomposition across a tropical forest disturbance gradient. Ecosphere 6:267. http://dx.doi.org/10.1890/es15-00112.1

Appendix D. Model summaries.

Table D1. Summary of linear mixed-effects model for percent mass loss (log transformed) from leaf litter bags installed across a tropical forest disturbance gradient. Independent variables included, installation season (SEASON), disturbance category (FOR_TYPE), litter type (LIT_TYPE), bag type (BAG_TYPE) and all two-way interactions. Installment season (SEASON) had 2 levels: Dry = dry season (baseline) and Wet = wet season. Forest type (FOR_TYPE) had 3 levels: CC = mature forest (baseline), OC = regenerating forest and OL = open land. Litter type had 2 levels: Dipterocarpus (baseline) and Parashorea. Bag type (BAG_TYPE) had 2 levels: C_Mesh = coarse mesh (baseline, fauna access) and F_Mesh = fine mesh (fauna exclusion). The number of days since installation (DAYS) was modeled as a second order polynomial for the fixed time effect, as this was a repeated measures design, and we included a DAYS:SEASON interaction. Random effects included a random intercept for PLOT/SUBPLOT, as the subject level term, and a random slope for installment season to account for the heteroscedasticity of variance.

Random effects:

Fixed effects:

Table D2. Summary of linear mixed-effects model for percent mass loss (log transformed) from leaf litter bags installed in a shade-house mesocosm experiment, with three levels of simulated canopy cover and three litter-bed treatments. Independent variables included, installation season (SEASON), simulated canopy cover (SIM_CC), litter-bed disturbance category (FOR_TYPE), litter type (LIT_TYPE) and all two-way interactions. Installment season (SEASON) had 2 levels: Dry = dry season (baseline) and Wet = wet season. Simulated canopy cover had three levels: full-shade (baseline, two layers of shade-cloth), half-shade (HS, one layer of shade-cloth) and open-canopy (OP, no shade-cloth). Liter-bed forest type (FOR_TYPE) had 3 levels: CC = mature forest (baseline), OC = regenerating forest and OL = open land. Litter type had 2 levels: Dipterocarpus (baseline) and Parashorea. The number of days since installation (DAYS) was modeled as a second order polynomial for the fixed time effect, as this was a repeated measures design, and we included a DAYS:SEASON interaction. Random effects included a random intercept for ROW:COL/PLOT, as the subject level term, and a random slope for installment season to account for the heteroscedasticity of variance.

Random effects:

Fixed effects:

Table D3. Summary of linear mixed-effects model for percent mass loss (log transformed) from leaf litter bags installed across a tropical forest disturbance gradient. Independent variables included, installation season (SEASON), plant species composition (NMDS1 and NMDS2), litter type (LIT_TYPE), bag type (BAG_TYPE) and all two-way interactions. Installment season (SEASON) had 2 levels: Dry = dry season (baseline) and Wet = wet season. Plant species composition (trees, herbs, lianas) was represented by the first two axes of an NMDS ordination (see Fig. 5). Litter type had 2 levels: Dipterocarpus (baseline) and Parashorea. Bag type (BAG_TYPE) had 2 levels: C_Mesh = coarse mesh (baseline, fauna access) and F_Mesh = fine mesh (fauna exclusion). The number of days since installation (DAYS) was modeled as a second order polynomial for the fixed time effect, as this was a repeated measures design, and we included a DAYS:SEASON interaction. Random effects included a random intercept for PLOT/SUBPLOT, as the subject level term, and a random slope for installment season to account for the heteroscedasticity of variance. Note response data were the same as in Table A3 but the analysis replaced FOR_TYPE with plant species composition data.

Random effects:

Fixed effects:

Table D4. Summary of linear mixed-effects model for percent mass loss (log transformed) from leaf litter bags installed across a tropical forest disturbance gradient. Independent variables included, installation season (SEASON), soil/topography (PCA1 and PCA2), litter type (LIT_TYPE), bag type (BAG_TYPE) and all two-way interactions. Installment season (SEASON) had 2 levels: Dry = dry season (baseline) and Wet = wet season. Soil (see Appendix C: Table A2) /topographic data was represented by the first two axes of a principal components ordination (see Fig. 7). Litter type had 2 levels: Dipterocarpus (baseline) and Parashorea. Bag type (BAG_TYPE) had 2 levels: C_Mesh = coarse mesh (baseline, fauna access) and F_Mesh = fine mesh (fauna exclusion). The number of days since installation (DAYS) was modeled as a second order polynomial for the fixed time effect, as this was a repeated measures design, and we included a DAYS:SEASON interaction. Random effects included a random intercept for PLOT/SUBPLOT, as the subject level term, and a random slope for installment season to account for the heteroscedasticity of variance. Note response data were the same as in Table A3 but the analysis replaced FOR_TYPE with soil/topographic data.

Random effects:

Fixed effects:

[Back to C006-105]