The carbon sink in intact tropical forests

To quantify the carbon (C) sink in intact tropical forests, Pan et al. (2011) multiplied ‘...region-specific estimates of C density or change in C density times the associated areas represented by the region-specific estimates.’ A retrospective interpolation based on their numbers provides insight into the possible causes of a C sink in intact tropical forests. First, however, there are two reasons to believe that their estimated sink of 1.2 Pg C yr 1 between 1990 and 2007 is too large. The first problem concerns a mismatch between the sources used for forest area and carbon stocks. Forest area came from country reports to the Forest Resource Assessment of the United Nations Food and Agricultural Organization (FAO-FRA). FAO-FRA defines forest as land with ‘...trees higher than 5 m and a canopy cover of more than 10 percent...’ In contrast, carbon stocks came from repeated censuses of plots located in tall, closed canopy forests. Carbon stocks and change in carbon stocks are potentially much larger in tall, closed canopy forests than those in stunted, open forests. Three contrasting estimates of forest area illustrate the magnitude of the problem. The FAO-FRA country reports identified 14 860 000 km of intact tropical forest in 2000 (Pan et al., 2011). In contrast, two pantropical analyses of satellite imagery for 1997 identified just 10 720 000 and 11 160 000 km of dense tropical forest (>80% canopy cover) and evergreen and seasonal tropical forest, respectively (Achard et al., 2002; Hansen & DeFries, 2004). The 4 000 000 km discrepancy contributes to an overestimate of the C sink in intact tropical forests. A second problem concerns an overlooked source for carbon stocks. Pan et al. (2011) used two sources to quantify carbon stocks for intact tropical forests. Those two sources report aboveground biomass (AGB) of trees for repeated censuses of forest plots in Africa and Amazonia (Lewis et al., 2009a; Phillips et al., 2009). To estimate pantropical changes in carbon stocks from these data, Pan et al. (2011) extract AGB increases of 0.31% yr 1 for Africa and 0.28% yr 1 for Amazonia, assign all Neotropical forests the Amazonian value, assign Asian forests the mean of the African and Amazonian values, assume all biomass components change in proportion to AGB, assume C comprises 50% of biomass, and exclude soil C. The overlooked source also reports AGB change of trees for repeated censuses of tropical forest plots in Africa, the Americas, and Asia (Chave et al., 2008). Although AGB tended to increase, the increases averaged 0.15% yr 1 or just 48% and 53% of the African and Amazonian values used by Pan et al., respectively. Census effort, quantified as the sum of plot area (ha) multiplied by census interval (years), was actually greater for the neglected source (3156 ha years) than that for the selected Amazonian (2607 ha years) and African (1206 ha years) sources (Muller-Landau et al., in press). The omission of data from Chave et al. (2008) contributes further to an overestimate of the C sink in intact tropical forests. The leading cause of a large C sink in intact tropical forests is hypothesized to be increasing atmospheric CO2 concentrations (ca) (Lewis et al., 2009b). A retrospective interpolation of long-term forest biomass change suggests rising ca cannot be the sole cause of a C sink averaging 1.2 Pg C yr 1 between 1990 and 2007. The ca began rising around 1800 (Etheridge et al., 1996) and has affected forest C balance for more than 200 years. The retrospective interpolation explores the long-term implications for change in forest biomass C stocks (DFBC). The interpolation uses annual values of ca and assumes a linear relationship between DFBC and Dca, where Dca quantifies the level of CO2 fertilization as the difference between current and preindustrial ca in ppm. Ice core and Mauna Loa records provide annual ca values (Etheridge et al., 1996; www.esrl.noaa. gov/gmd/ccgg/trends), with values for missing years provided by Etheridge et al. (1996). Two paired values of DFBC and Dca constrain the DFBC-Dca relationship. The first is from 1800 when ca was 280 ppm, Dca was zero and, at appropriately large spatial scales, forest biomass can be assumed to have been at equilibrium (DFBC = 0). This same assumption is implicit whenever DFBC is attributed to CO2 fertilization in the absence of experimental manipulations of ca. Pan et al. (2011) provide the second pair of values for 1990 through 2007, when ca averaged 367.5 ppm, Dca averaged 87.5 ppm, and DFBC averaged 1.2 Pg C yr . The simplest DFBC Dca relationship consistent with the 1800 and 1990–2007 values is linear with intercept zero and slope 0.014 Pg C yr 1 Dca . I used annual Dca values and Correspondence: S. Joseph Wright, tel. +1-507-212-8132, fax +1-507-212-8148, e-mail: wrightj@si.edu