Laughlin, D.C. Appendix S3. SEM results using individual plant functional traits
Appendix S3. SEM results using individual plant functional traits
Westoby et al. (2002) and Reich et al. (2003) proposed that plant strategies be considered
as spectrums of coordinated/correlated traits, rather than single traits. Dimensionality reduction
techniques (such as principal components analysis, PCA) are useful because they extract
information about plant strategy spectrums by taking into account all the available trait
information, rather than relying on a single measure, such as SLA. In the words of Lee and
Verleyson (2007), “…instead of arbitrarily removing one variable…another way to reduce the
number of variables would be to find a new set of transformed variables. This is motivated by the
facts that dependencies between variables may be very complex and that keeping one of them
might not suffice to catch all the information they both convey.”
The structural equation model (SEM) presented in the main text used a communityweighted mean leaf trait axis to represent the location of the community along the leaf
economics spectrum. Therefore, the use of the leaf trait axis in this context is novel and
represents an analytical advance in studies that relate multidimensional plant communities to
ecosystem processes. The first principal component extracted from the species-trait matrix
represented species-level covariation in SLA, LDMC, leaf [N], and fine root [N] (Laughlin et al.
2010). The average value of this axis was then calculated for each community using species
relative abundances as weights. The community-weighted mean PCA leaf axis was strongly
correlated with community-weighted mean SLA, LDMC, leaf [N], and fine root [N] (r = 0.82, 0.58, 0.88, 0.80, respectively).
Here I show that the SEM results are qualitatively similar if individual communityweighted mean traits are used instead of the community-weighted mean PCA axis to represent
1
Laughlin, D.C. Appendix S3. SEM results using individual plant functional traits
the plant strategy spectrum. For each model presented here, I replaced the community-weighted
leaf trait axis with the community-weighted mean traits (i.e. SLA, leaf [N], LDMC, and fine root
[N]).
The model with SLA (Fig. S3.1) fit the data well (χ2 = 39.7, df = 32, P = 0.16) and all
pathways were qualitatively similar to the model presented in Fig. 3 in the main text. The sign of
the pathway between SLA and nitrification potential was positive, yet marginally non-significant
(P = 0.057).
The model with leaf [N] (Fig. S3.2) fit the data well (χ2 = 39.1, df = 32, P = 0.18) and all
pathways were qualitatively similar to the model presented in Fig. 3 in the main text. The sign of
the pathway between leaf [N] and nitrification potential was positive and highly significant (P =
0.0003).
The model with LDMC (Fig. S3.3) fit the data well (χ2 = 38.1, df = 32, P = 0.22) and all
the pathways were qualitatively similar to the model presented in Fig. 3 in the main text. The
sign of the pathway between LDMC and nitrification potential was negative because LDMC was
negatively correlated with the leaf trait axis (i.e. LDMC is negatively correlated with SLA and
leaf [N]). However, the pathway from LDMC to nitrification potential was not significant (P =
0.54).
The model with fine root [N] did not fit the data well initially (χ2 = 58.2, df = 32, P =
0.0031). I obtained a good fitting model by adding pathways between soil total N and fine root
[N], and between soil pH and fine root [N]. This model (Fig. S3.4) fit the data marginally well
(χ2 = 43.0, df = 32, P = 0.0584). The sign of the pathway between fine root [N] and nitrification
potential was positive, but non-significant (P = 0.15).
2
Laughlin, D.C. Appendix S3. SEM results using individual plant functional traits
This analysis demonstrates that leaf [N] and SLA were the two most important traits that
influenced the results obtained in this paper, suggesting that by themselves, leaf [N] and SLA are
potentially useful predictors of nitrification potential in these forest soils. The model with leaf
[N] strongly agreed with the model presented in the main text, and this is likely because
community-weighted mean leaf [N] exhibited the strongest correlation with the leaf trait axis
compared to the other traits. The model with SLA also generally agreed with the model
presented in the main text, and this is likely because community-weighted mean SLA exhibited
the second strongest correlation with the leaf trait axis compared to the other traits. The models
with LDMC and fine root [N] were still qualitatively similar (the signs on the pathways were
what would be expected), but the significance level of the pathways was changed.
References
Laughlin, D.C., Leppert, J.J., Moore, M.M. & Sieg, C.H. (2010a) A multi-trait test of the leafheight-seed plant strategy scheme with 133 species from a pine forest flora. Functional
Ecology, 24, 493-501.
Lee, J.A. & Verleyson, M. (2007) Nonlinear dimensionality reduction. Springer.
Reich, P.B., Wright, I.J., Cavender-Bares, J., Craine, J.M., Oleskyn, J., Westoby, M. & Walters,
M.B. (2003) The evolution of plant functional variation: traits, spectra, and strategies.
International Journal of Plant Science, 164 (3 Supplement), S143-S164.
Westoby, M., Falster, D.S., Moles, A.T., Vesk, P.A. & Wright, I.J. (2002) Plant ecological
strategies: some leading dimensions of variation among species. Annual Review of
Ecology and Systematics, 33, 125-159.
3
Laughlin, D.C. Appendix S3. SEM results using individual plant functional traits
Figure S3.1. Structural equation model with community-weighted mean specific leaf area (SLA)
instead of the leaf trait axis (χ2 = 39.7, df = 32, P = 0.16). Standardized path coefficients can be
interpreted as partial correlation coefficients that range from -1 to 1. Coefficients of
determinisms (R2) are shown for every response variable in this set of eight linear equations.
4
Laughlin, D.C. Appendix S3. SEM results using individual plant functional traits
Figure S3.2. Structural equation model with community-weighted mean leaf nitrogen content
instead of the leaf trait axis (χ2 = 39.1, df = 32, P = 0.18). Standardized path coefficients can be
interpreted as partial correlation coefficients that range from -1 to 1. Coefficients of
determinisms (R2) are shown for every response variable in this set of eight linear equations.
5
Laughlin, D.C. Appendix S3. SEM results using individual plant functional traits
Figure S3.3. Structural equation model with community-weighted mean leaf dry matter content
(LDMC) instead of the leaf trait axis (χ2 = 38.1, df = 32, P = 0.22). Standardized path
coefficients can be interpreted as partial correlation coefficients that range from -1 to 1.
Coefficients of determinisms (R2) are shown for every response variable in this set of eight linear
equations.
6
Laughlin, D.C. Appendix S3. SEM results using individual plant functional traits
Figure S3.4. Structural equation model with community-weighted mean fine root nitrogen
content instead of the leaf trait axis (χ2 = 43.0, df = 32, P = 0.0584). Standardized path
coefficients can be interpreted as partial correlation coefficients that range from -1 to 1.
Coefficients of determinisms (R2) are shown for every response variable in this set of eight linear
equations.
7