Appendix S1: Detailed information about the data used
Global forest database
The publically available (http://www.ua.ac.be/main.aspx?c=sebastiaan.luyssaert&n=35884) global forest database contains 558 forests of which 268 provided an estimate for tree biomass production, i.e., the sum of stem, branch, foliage, coarse root and fine root production (termed TNPP1 in Luyssaert et al. 2007). Of these 268 forests, 49 also provided independent estimates for gross primary productivity (GPP), and this is the dataset used for the current study (Table S1). Gross primary productivity of these 49 forests was estimated via eddy covariance (n=36), or modelled independently of biomass production estimates with site-specific parameterization (n=13).
Total plant biomass production – gap-filling
The original tree biomass production presented in the database does not represent total plant biomass production, as production of reproductive organs, biomass loss to herbivores and understory biomass production are not included. In addition, it has been shown that foliage production is underestimated in tropical forests when estimated via litterfall collection, because an important fraction of the litter does not reach the litter baskets or is decomposed in the baskets before being collected (see e.g., Clark et al. 2001a). Because GPP, typically derived from eddy covariance measurements, is estimated for the entire ecosystem, we need total ecosystem plant biomass production rather than the original (underestimated) tree biomass production when estimating the carbon fraction that is allocated to biomass. When not available for the site, we therefore supplemented the biomass production estimates with estimates of production of reproductive organs, biomass loss to herbivores and understory biomass production. For tropical forests we also applied a correction for within-canopy decomposition of leaf litter when foliage production was estimated via litter collection. Estimates of production of unmeasured biomass components were made as follows:
1. Production of reproductive organs
We used the data available in our database to estimate the production of reproductive organs. We plotted reproductive biomass production versus aboveground biomass production (data not shown) and used this relationship to estimate reproductive organ production whenever this biomass component was not measured. Estimates for reproductive biomass production were small (Table S2; Fig. S1) and did not affect patterns observed for the ratio of biomass production to GPP. Gap-filling of reproductive organ production is thus performed merely for completeness.
2. Understory biomass production
For the current study, understory biomass production was needed for 36 of the 49 study sites (i.e., whenever GPP referred to the entire ecosystem as measured via eddy covariance). Understory biomass production estimates were available for 25 of these 36 forests (TableS2). Forests for which no estimate of understory production was available were all located in the temperate or tropical zone. Unfortunately, we could not find a useful relation between understory biomass production and other variables in our database (e.g., leaf area index; data not shown). For that reason, we estimated understory biomass production for these 11 forests based on observational estimates of understory production from 66 temperate and 7 tropical forests in the database. Specifically, understory production was estimated as a*original tree biomass production, with a=0.043 and a=0.073 in temperate and tropical forests, respectively (a represents the average of the understory biomass production as a fraction of original tree biomass production for these forests in our database).
The uncertainty of these understory estimates following this calculation is obviously quite large, but the contribution of understory production to total biomass production remained small in temperate and tropical forests (Fig. S1). In the boreal forests, on the other hand, understory production (which was measured in all boreal forests of our dataset) constituted a substantial fraction of total biomass production (Fig. S1). This demonstrates the importance of taking understory biomass production into account when analyzing the ratio of total biomass production to GPP. Last, we also want to mention that the biomass production-to-GPP ratio did not differ between the 36 forests for which understory was included and the 13 forests for which understory was neglected in both GPP and biomass production estimates (p=0.47; we also found no interaction effect with nutrient availability classification; p=0.95).
3. Litterfall decomposition in tropical forests
Within-canopy decomposition of leaf litter in tropical forests was previously discussed by Clark et al. (2001a). Similar to Malhi et al. (2009), we used the estimate provided by Clark et al. (2001a) for missing litter (i.e., 12% of foliage production) to take into account within-canopy decomposition of leaf litter when foliage production was estimated via litter collection. This litterfall decomposition was a minor fraction of total biomass production (Table S2) and its effect on biomass production-to-GPP ratio of tropical forests was minor.
4. Biomass losses to herbivores
Herbivory estimates were available for only 10 forests in our database, and we thus needed to estimate this missing biomass production for most forests, using information available in other publications. Cebrian & Lartigue (2004) showed that, on a global scale, herbivory scales with aboveground biomass production. Via the publically available database compiled by Cebrian & Lartigue (2004), we could establish a relation between herbivory and aboveground biomass production in temperate and tropical forests (data not shown). For boreal forests, no such relation could be made because of a lack of data in the database of Cebrian & Lartigue (2004). We therefore estimated biomass loss to herbivores in boreal forests analogously to Keeling & Philips (2007):
Insect herbivory=2.5% of needle production or 1.59% of broad leaf production
Vertebrate herbivory=5.6 g m-2 y-1 for deciduous forests, and 2.98 g m-2 y-1 for coniferous forests.
Estimated biomass losses to herbivores were small (=~2% and never higher than 7% of total biomass production; Table S2; Fig. S1), and hardly influenced our estimates of biomass production (Table S2; Fig. S1) and the ratio of biomass production to GPP.
Gap-filled versus original biomass production-to-GPP ratio
Most biomass components that were not taken into account in the original dataset (i.e., TNPP1 in Luyssaert et al. 2007) constituted a minor fraction of biomass production, except for understory production in boreal forests (Fig. S1; Table S2). Consequently, the effect on the biomass production-to-GPP ratio was small for most sites (Fig. S2) and similar patterns emerged for the gap-filled biomass production-to-GPP ratio (further termed biomass production efficiency; BPE) as for the original tree biomass production-to-GPP ratio. Stepwise regression analysis (see below for more info) revealed very similar results for original compared to gap-filled BPE (Fig. S3 in analogy with Figure 1 of the manuscript). Effects of both nutrient availability and management were highly significant (p<0.01 and p<0.01, respectively), while pvalues for climate zone, forest type and stand age were all higher than 0.1. We thus conclude that the previously unaccounted biomass components did not affect the observed differences in BPE, and our observation of higher BPE in nutrient-rich compared to nutrient-poor forests, and for managed versus unmanaged forests is robust.
Uncertainty estimates
Measuring biomass production in forest ecosystems is not an easy task, and is accomplished in different ways (allometric relations, harvests, etc.), each with their own difficulties. Luyssaert et al. (2007) therefore included uncertainties in biomass production in the forest database, based on site-location and the method used for estimating biomass production (see Luyssaert et al. (2007) for detailed information). In a similar way, they also attributed uncertainties to GPP and Ra. For the current study, we included an uncertainty of 100% for the biomass production components that were estimated as described above. Uncertainties for biomass production and BPE (original and gap-filled) are presented in Fig. S4 and S5. Weighted means, the corresponding ANOVA analysis, and weighted regressions revealed similar results as the unweighted analysis (both showing a highly significant effect of nutrient availability). This similarity emphasizes the robustness of our results.
Table S1: Detailed information for the forests used in our analyses, including gross primary production (GPP), biomass production efficiency (BPE) and the standard error (SE) on BPE and GPP. Whenever multiple years of data were available, GPP and BPE represent the average over all years. Inter-annual variation is included in the calculation of SE. Site names refer to the names as given in the global forest database (http://www.ua.ac.be/main.aspx?c=sebastiaan.luyssaert&n=35884). Abbreviations for climate are Bo (boreal), Te (temperate), and Tr (tropical); for forest type: B (broadleaved), C (coniferous) and M (mixed); for management practice UM (unmanaged), M (managed), RD (recently disturbed) and F (fertilized). For our analyses, recently disturbed and fertilized forests are included in the managed group because too few forests were classified as RD or F to treat them as separate groups. Age represents the average stand age (in years) as reported in the different publications of the specific site. Because drought may be an important factor explaining variation in the biomass production-to-photosynthesis ratio, we computed the Gaussen index to test for potential drought effects. The Gaussen index is computed as MAP/2MAT, with MAP=mean annual precipitation (mm per year) and MAT=mean annual temperature (Kelvin); the lower the number, the dryer the site. We found no significant relation between biomass production efficiency and Gaussen index (p=0.58; ANCOVA with Gaussen index as a covariable). Therefore, we assume that drought did not affect the patterns in BPE observed for the forests in this dataset. The last column indicates the key publications in which biomass production, gross primary production and/or site information is given.
Site name / GPP / SE GPP / BPE / SE BPE / Climate zone / Forest type / Manage-ment / Age / Gaussen index / ReferencesBartlett / 1053 / 145 / 0.52 / 0.14 / Te / B / M / 80 / 2.27 / (Ollinger & Smith 2005)
Bornhoved Alder / 2420 / 433 / 0.37 / 0.08 / Te / B / UM / 60 / 1.37 / (Dilly et al. 2000; Kutsch et al. 2001)
Bornhoved Beech / 1324 / 433 / 0.53 / 0.19 / Te / B / M / 110 / 1.02 / (Dilly et al. 2000; Kutsch et al. 2001)
Cascade Head (1) / 1400 / 677 / 0.51 / 0.30 / Te / C / UM / 150 / 4.43 / (Runyon et al. 1994; Williams et al. 1997)
Cascade Head (1A) / 1558 / 677 / 0.55 / 0.29 / Te / B / F / 50 / 4.43 / (Runyon et al. 1994; Williams et al. 1997)
Caxiuana / 3630 / 295 / 0.34 / 0.05 / Tr / B / UM / 100 / 4.01 / (Carswell et al. 2002; Aragao et al. 2009)
Collelongo / 1127 / 58 / 0.60 / 0.07 / Te / B / M / 102 / 2.08 / (Schulze 2000; Valentini et al. 2000)
Dooary / 2001 / 132 / 0.69 / 0.07 / Te / C / RD / 15 / 1.51 / (Black et al. 2004; Black et al. 2007)
Flakaliden C / 1000 / 115 / 0.54 / 0.09 / Bo / C / M / 36 / 1.37 / (Bergh et al. 1999; Valentini et al. 2000; Law et al. 2002)
Frazer old / 915 / 714 / 0.52 / 0.48 / Bo / C / UM / 245 / 1.34 / (Ryan 1991; Murty et al. 1996)
Frazer young / 977 / 714 / 0.26 / 0.32 / Bo / C / M / 40 / 1.34 / (Ryan 1991; Murty et al. 1996)
Hainich / 1594 / 67 / 0.42 / 0.05 / Te / B / UM / 270 / 1.41 / (Knohl et al. 2003)
Harvard / 1279 / 43 / 0.43 / 0.10 / Te / B / UM / 60 / 1.87 / (Curtis et al. 2002; Law et al. 2002; Urbanski et al. 2007)
Hesse / 1434 / 43 / 0.70 / 0.06 / Te / B / M / 32 / 1.69 / (Valentini et al. 2000; Granier et al. 2008)
Jacaranda/K34 / 3040 / 208 / 0.35 / 0.06 / Tr / B / UM / NA / 4.56 / (Chambers et al. 2004; Malhi et al. 2009)
Juniper / 302 / 683 / 0.49 / 1.37 / Te / C / RD / NA / 0.39 / (Runyon et al. 1994; Williams et al. 1997)
Kannenbruch Alder/Ash / 1594 / 185 / 0.43 / 0.16 / Te / B / M / 68 / 1.27 / (Kutsch et al. 2005)
Kannenbruch Beech / 1470 / 185 / 0.49 / 0.18 / Te / B / M / 114 / 1.27 / (Kutsch et al. 2005)
Kannenbruch Oak / 1794 / 185 / 0.63 / 0.16 / Te / B / M / 210 / 1.27 / (Kutsch et al. 2005)
Metolius / 1141 / 88 / 0.38 / 0.07 / Te / C / M / 148 / 0.86 / (Law et al. 2000; Law et al. 2001)
Metolius young / 727 / 77 / 0.53 / 0.14 / Te / C / M / 15 / 0.98 / (Law et al. 2000; Law et al. 2001)
Morgan Monroe / 1452 / 81 / 0.70 / 0.06 / Te / B / M / 75 / 1.93 / (Curtis et al. 2002; Ehman et al. 2002)
Pasoh / 3230 / 120 / 0.47 / 0.10 / Tr / B / UM / 100 / 3.04 / (Clark et al. 2001b; Hirata et al. 2008)
Popface alba / 2230 / 492 / 0.64 / 0.15 / Te / B / F / 2 / 1.27 / (Gielen et al. 2005)
Popface euamericana / 1966 / 492 / 0.73 / 0.20 / Te / B / F / 2 / 1.27 / (Gielen et al. 2005)
Popface nigra / 2424 / 492 / 0.76 / 0.17 / Te / B / F / 2 / 1.27 / (Gielen et al. 2005)
Prince Albert SSA (SOAS) / 1226 / 45 / 0.36 / 0.06 / Bo / B / UM / 69 / 0.76 / (Gower et al. 1997; Gower et al. 2001; Barr et al. 2002)
Prince Albert SSA (SOBS) / 857 / 90 / 0.34 / 0.09 / Bo / C / UM / 107 / 0.66 / (Kimball et al. 1997; Malhi et al. 1999; Gower et al. 2001)
Prince Albert SSA (SOJP) / 665 / 91 / 0.37 / 0.12 / Bo / C / UM / 66 / 0.77 / (Kimball et al. 1997; Malhi et al. 1999; Gower et al. 2001)
Qianyanzhou Ecological Station / 1870 / 330 / 0.54 / 0.21 / Tr / C / M / NA / 2.74 / (Ma et al. 2008)