Ecological Archives E095-198-A2

Duncan N. L. Menge, Jeremy W. Lichstein, Gregorio Ángeles-Pérez. 2014. Nitrogen fixation strategies can explain the latitudinal shift in nitrogen-fixing tree abundance. Ecology 95:2236–2245. http://dx.doi.org/10.1890/13-2124.1

Appendix B. Additional evidence for differential regulation, stand-age distributions, inventory patterns with elevation, and sensitivity analyses.

FigB1

Fig. B1. Range and mean of % nitrogen (N) derived from fixation activity (%Ndfa) as a function of latitude. Data come from publications reviewed in Andrews et al. (2011), with latitudes taken from the original references (some latitudes are midpoints of latitude ranges in a given study). In each case N fixation was measured using the 15N natural abundance method, which estimates the percent of N that a potentially N-fixing plant acquires from fixing atmospheric N2 as opposed to soil N uptake. Data and fits shown are from the 19 data points that listed a range for a single genus, although fits through all 27 data points from Andrews et al. (2011) were similar. Solid circles represent actinorhizal N fixers, whereas open triangles represent rhizobial N fixers (not all of which are trees). (a) Range (maximum minus minimum) of %Ndfa in a given study reported in Andrews et al. (2011) or the original works where clarification was needed. Sigmoid and linear fits were compared using nls and lm functions in R; the sigmoid fit (%Ndfa range = (50.6 ± 5.0)*(1 – 1 / (1+exp(–(0.6 ± 0.4)*(Latitude – (42.9 ± 1.1)))))) was 8.2 AIC units better than the linear fit. (b) Mean %Ndfa in a given study reported in Andrews et al. (2011). The linear fit (%Ndfa mean = (8.6 ± 9.7) + (1.7 ± 0.3)*Latitude) was 3.14 AIC units better than the sigmoid fit.


FigB2

Fig. B2. Stand-age distribution of coterminous USA forests, estimated from U.S. Forest Inventory and Analysis (FIA) plots (N = 79,508). FIA defines stand age as "the average age of the live trees not overtopped in the predominant stand size-class," which is estimated by coring several trees within each plot.


FigB3

Fig. B3. Mexican nitrogen-fixing tree abundance and symbiotic type by elevation. Data are from the Mexican Inventario Nacional Forestal y de Suelos. Gray dots denote plots (N = 15,358) and black circles denote means of all plots within each 100 m elevation band (N = 40). Note that forest plots >2000 m elevation in Mexico span 19°–30° N, with peak elevation at 19°–20° N, so the trends seen here are not driven by the latitudinal trend in Fig. 2. (a) Nitrogen-fixing tree abundance. (b) Dominance of actinorhizal (as opposed to rhizobial) nitrogen-fixing trees. Gray dots are vertically jittered for visual clarity, as in Fig. 2. (c) Nitrogen-fixing tree abundance as a function of actinorhizal dominance.


FigB4

Fig. B4. Modeled latitudinal distribution of nitrogen (N)-fixing tree abundance when the steepness parameter for the N fixation strategy is not fixed. As a robustness check on our results in Fig. 4 (where the steepness parameter c3 is fixed at one per degree in Fig. 4a), we present here the model results for the case where c3 was estimated as a free parameter. This allows the steepness and the sign of the derivative (rise vs. fall in the fit) to change. Axes and symbols are defined as in Fig. 4. For the model with all N-limited habitat being severely N limited, the maximum likelihood parameter estimates (with 95% CIs) were h1 = 14% (11%–17%), h2 = 51% (0%–99%), h4 = 35° (24°–46°), s3 = 1.9°-1 (0.13°-1–27°-1),and s4 = 34° (30°–38°). For the model with all N-limited habitat being moderately N limited, the maximum likelihood parameter estimates (with 95% CIs) were h1 = 18% (14%–22%), h2 = 69% (0%–100%), h4 = 35° (24°–46°), s3 = 1.9°-1 (0.13°-1–27°-1),and s4 = 34° (30°–38°). The fit is similar to Fig. 4, but with wider confidence limits on the strategy distribution (panel a) toward the extremes of low and high latitude.


FigB5

Fig. B5. Modeled latitudinal distribution of nitrogen (N)-fixing tree abundance using a younger stand age distribution for data below 35° latitude. To explore the uncertainty associated with forest stand age in Mexico, we re-ran the latitudinal analysis using succession-weighted N fixer abundances that would arise from each stand below 35° latitude (slightly outside the border of Mexico to be conservative) being half as old as in the USA (this figure) or twice as old as in the USA (Appendix B, Fig. B6; also see Appendix A, Table A1). The best fit models in each case were the same as in Fig. 4. For all N-limited habitat being severely N limited, maximum likelihood parameter estimates (with 95% CIs) were h1 = 14% (11%–18%), h2 = 38% (0%–99%), h4 = 33° (28°-38°), and s4 = 33° (27°–39°). For all N-limited habitat being moderately N limited, maximum likelihood parameter estimates (with 95% CIs) were h1 = 15% (12%–19%), h2 = 50% (0%–100%), h4 = 33° (29°–38°), and s4 = 33° (28°–38°). The fit is very similar to Fig. 4. Axes and symbols are defined as in Fig. 4.


FigB6

Fig. B6. Modeled latitudinal distribution of nitrogen (N)-fixing tree abundance using an older stand age distribution for data below 35° latitude. To explore the uncertainty associated with forest stand age in Mexico, we re-ran the latitudinal analysis using succession-weighted N fixer abundances that would arise from each stand below 35° latitude (slightly outside the border of Mexico to be conservative) being half as old as in the USA (Appendix B, Fig. B5) or twice as old as in the USA (this figure; also see Appendix A, Table A1). The best fit models in each case were the same as in Fig. 4. For all N-limited habitat being severely N limited, maximum likelihood parameter estimates (with 95% CIs) were h1 = 16% (13%–19%), h2 = 48% (0%–100%), h4 = 39° (14°–65°), and s4 = 34° (32°–37°). For each stand being twice as old and all N-limited habitat being moderately N limited, maximum likelihood parameter estimates (with 95% CIs) were h1 = 27% (22%–33%), h2 = 64% (0%–100%), h4 = 40° (6°–73°), and s4 = 34° (32°–37°). The fit is similar to Fig. 4, but with wider confidence limits at low latitudes for the habitat distribution (panel b) and N fixer abundance (panel c). Axes and symbols are defined as in Fig. 4.


FigB7

Fig. B7. Modeled latitudinal distribution of nitrogen (N)-fixing tree abundance when some obligate N fixers exist at lower latitudes. As a robustness check on our results in Fig. 4, we present here the model results when the minimum percent of N fixers that are obligate is assumed to be 10% rather than 0%. For the model with all N-limited habitat being severely N limited, the maximum likelihood parameter estimates (with 95% CIs) were h1 = 15% (12%–19%), h2 = 41% (0%–99%), h4 = 33° (28°–38°),and s4 = 33° (29°–38°). For the model with all N-limited habitat being moderately N limited, the maximum likelihood parameter estimates (with 95% CIs) were h1 = 20% (15%–24%), h2 = 55% (0%–100%), h4 = 33° (28°–38°), and s4 = 33° (29°–38°). The fit is extremely similar to Fig. 4, suggesting that the latitudinal pattern is equally well explained when some lower-latitude N fixers are obligate. Axes and symbols are defined as in Fig. 4.


FigB8

Fig. B8. Modeled abundance as a function of the cost of being facultative. Each point is the USA forest stand age-weighted mean of 400 years of modeled succession for a given cost of being facultative. In each simulation, facultative N fixers, obligate N fixers, and non-fixers were all in competition, unlike the simulations for Fig. 3, in which one of the two N fixers was competing with the non-fixer. Data are displayed as the (a, b) % of biomass occupied by N fixers (facultative and obligate combined) as opposed to non-fixers and the (c, d) % of N fixer biomass that is obligate as opposed to facultative. Simulations were run for (a, c) severely and (b, d) moderately N-limited habitat.


Literature cited

Andrews, M., E. K. James, J. I. Sprent, R. M. Boddey, E. Gross, and F. B. dos Reis Jr. 2011. Nitrogen fixation in legumes and actinorhizal plants in natural ecosystems: values obtained using 15N natural abundance. Plant Ecol. Divers. 4:131–140.


[Back to E095-198]