mattocci27 / LMAms

MIT License
0 stars 0 forks source link

variance partition #30

Closed mattocci27 closed 1 year ago

mattocci27 commented 2 years ago

Jeremy: After we see the new (simpler) Fig. 5 results, we should consider combining sections 3 and 4 into a single section on LMA variance. Currently, we show that LMAm and LMAs increase in concert across groups of leaves (e.g., both are low in GLOPNET deciduous, and both are high in GLOPNET evergreen). Does this positive correlation also occur within each group (e.g., within GLOPNET deciduous leaves), or are LMAm and LMAs independent within groups (as in Fig. 1)? Here some questions we may want to address: (1) GLOPNET variance components: (1a) How much of the total LMA variance is due to variance in LMAm vs. LMAs? [we already have this result] (1b) Within each component (LMAm and LMAs), how much variance is due to the difference between evergreen and deciduous leaves, and how much variance occurs within leaf habit? [I think we can use ANOVA to get the between-group variance fraction (R2), and the within-group percent would be 1 minus R2] In summary, for GLOPNET, we can partition total LMA variance in LMAm vs. LMAs components, and then we can use ANOVA to partition LMAm and LMAs variance into between group (evergreen vs. deciduous) and within group components.

(2) Panama variance components: (2a) How much total LMA variance is due to LMAm vs. LMAs? [I don't think we currently have this value for sun and shade combined.] (2b) Within LMAm and within LMAs, how much variance is due to difference among groups (sun/shade and wet/dry) vs. variance within groups? [This would be a 2-factor ANOVA. As in GLOPNET case, 1 minus R2 is the fraction of variance that occurs within groups. The R2 could be partitioned into site (dry/wet) and light (sun/shade) effects. So for Panama, we could estimate the fraction of LMAm and LMAs variance due to site, light, and within-group differences.]

Finally, we can ask: (3) What are implications of variance components for understanding mass- vs. area-dependence of Amax? [This would be simplified version of Fig. 5 we discussed via email.]

I think Fig. 6 can remain unchanged.