The U.S. Carbon Budget: Contributions from Land-Use Change

See allHide authors and affiliations

Science  23 Jul 1999:
Vol. 285, Issue 5427, pp. 574-578
DOI: 10.1126/science.285.5427.574


The rates at which lands in the United States were cleared for agriculture, abandoned, harvested for wood, and burned were reconstructed from historical data for the period 1700–1990 and used in a terrestrial carbon model to calculate annual changes in the amount of carbon stored in terrestrial ecosystems, including wood products. Changes in land use released 27 ± 6 petagrams of carbon to the atmosphere before 1945 and accumulated 2 ± 2 petagrams of carbon after 1945, largely as a result of fire suppression and forest growth on abandoned farmlands. During the 1980s, the net flux of carbon attributable to land management offset 10 to 30 percent of U.S. fossil fuel emissions.

The rate at which carbon is accumulating in terrestrial ecosystems in the United States is uncertain, as are the mechanisms responsible for the current sink. Estimates based on measured changes in wood volumes (forest inventories) (1–3) range between 0.079 and 0.280 petagrams of carbon per year (Pg C year−1). An estimate, based on atmospheric and oceanic data and models, and including southern Canada, calculated a sink of 1.7 ± 0.5 Pg C year−1 (4). Neither of these approaches identifies the mechanisms responsible for the sink (5). In contrast, although the changes in carbon associated with land-use change do not define the total net flux of carbon between land and atmosphere, they represent the portion of the flux that can be attributed to direct human activity, and it is this portion that is addressed by the United Nations Framework Convention on Climate Change and by the Kyoto Protocol. In this report, we estimate the annual flux of carbon in the United States attributable to changes in land use.

Our approach is based on two types of information: rates of land-use change and changes per hectare in carbon that follow a change in land use. We considered the conversion of natural ecosystems to croplands (cultivated) and pastures (not cultivated), the abandonment of croplands and pastures, harvest of industrial wood and fuel wood, and fire management, that is, the area annually burned by wildfires. Rates of agricultural clearing and abandonment and rates of wood harvest were obtained directly from the U.S. Department of Agriculture (USDA) for the period since 1945 (6) and largely from the U.S. Bureau of Census (7) for the period between 1700 and 1945. Areas burned each year were obtained from wildfire statistics of the U.S. Forest Service (8) for the period after 1930 and from pre-European burning rates (9) modified by settlement and logging practices for the years 1700–1930 (10). A more detailed description of the data, sources, and assumptions can be found elsewhere (11, 12).

We divided the United States into seven geographic regions (each region including two to five natural ecosystems) for a total of 13 different ecosystems, not including croplands and pastures. The areas and carbon stocks of ecosystems in 1700 were determined from both natural areas (13) and current ecosystems (14). The fractions of vegetation left alive, killed, and burned as a result of human activity and fire were defined for each type of land use and ecosystem. The efficiency of industrial wood harvest increased through time, so that more wood was removed per hectare and less left as slash (dead vegetation) (15, 16). Rates of forest growth after harvest, fire, and agricultural abandonment and rates of decay of organic matter for each ecosystem were obtained from the ecological literature (17). Changes in soil carbon included only the losses that resulted from cultivation and the reaccumulations that followed abandonment of cultivated land. These per hectare rates of carbon loss and accumulation after changes in land use were used in a bookkeeping model (18) to calculate annual changes in carbon for each type of ecosystem, land use, and region. The areas of forest burned each year, together with the total area of forest in each region and the rate of forest growth, constrained the average biomass that could be burned per hectare in a region's forests (19).

According to our analysis, changes in land use released about 25 Pg C to the atmosphere over the period 1700–1990, largely from the conversion of forests to agricultural lands and from cultivation of prairie soils (Fig. 1). The net release, overall, included a net uptake of 2.4 Pg C after 1945 as both the area and biomass of forests increased in response to fire suppression, reduced harvest of fuel wood, and the abandonment of farms. It also included an uptake in eastern forests, largely offset by releases of carbon from western regions where rates of industrial wood harvest increased.

Figure 1

(Left). Annual net sources and sinks of carbon resulting from different types of land use in the United States.

The net flux resulted from accumulations of carbon in regrowing vegetation and, to a lesser extent, soils and from releases of carbon from burning and decay (Fig. 2). Before European settlement, fires accounted for the largest sources (immediate burning and delayed decay) and sinks of carbon, although the net flux from fire was nearly zero as long as rates of burning were about constant. As wood harvest and agricultural clearing increased in the 18th century, emissions from fire, soil, and wood products increased, but annual accumulations also increased as the areas of forests recovering from fire and harvesting increased. Replacement of wood fuels with fossil fuels midway through the 19th century began to reduce the emissions from fuel wood harvesting, and fire suppression after 1930 further reduced the areas burned. Despite the recovery of forests in the east, increasing rates of industrial wood harvests in the west and south increased the generation of slash and wood products and hence emissions of carbon. The largest annual accumulations of carbon occurred in the 1950s and 1960s. By the 1970s, as a result of aging (slower growing) forests and a resurgence of fuel wood use, the net flux from land-use change was close to zero.

Figure 2

(Right).Annual losses and accumulations of carbon in different terrestrial pools as a result of land management in the United States.

The errors resulting from uncertainties in rates of land-use change, from aggregated estimates of biomass, growth, and decay, and from simplifying assumptions in the structure of the model may be estimated by comparing the results of successive analyses (Fig. 3). For the three analyses in which changes in land use are reasonably well documented, estimates of the long-term (1700–1990) carbon flux range between 24 and 30 Pg C. For shorter intervals, the errors are larger. The difference between our estimate that includes fire and an earlier estimate for North America that did not include fire (18) is as large as 0.2 Pg C year−1 before 1900. A comparison of our estimates with and without fire shows that including fire and fire suppression changed estimates of flux by as much as 0.05 Pg C year−1. Surprisingly, fire reduced the estimated emissions during the 1800s. The reduction occurred because the biomass of forests cleared for agriculture was less if fire was included in the analysis. The errors are estimated to be less than ±0.1 Pg C year−1in recent decades (20). The analysis that shows a large sink for carbon beginning in 1920 is based on generous assumptions about woody encroachment (21) and early fire exclusion (22). Although detailed site-specific data document these processes, the aerial extent and the historical pattern of the changes are unknown. The curve represents an upper limit for the sink strength of U.S. ecosystems.

Figure 3

Annual net flux of carbon in the United States as a result of changes in land use.

Most other estimates of flux available for comparison pertain to the 1980s and 1990s. Our analysis for the 1980s shows an average annual accumulation of 0.037 Pg C (Fig. 4). This net uptake results from the uptake of carbon by growing vegetation (0.280 Pg C year−1) and emissions from the decay of wood products (0.104 Pg C year−1) and slash (0.091 Pg C year−1), some of which had accumulated before 1980. The uptake of 0.280 Pg C year−1 occurred on 200 × 106 ha of recovering forest and woodlands (about 65% of the U.S. forest areas). The average rate of growth (1.4 Mg C ha−1 year−1) was low, indicating that many of the forests and woodlands were aging and their rates of growth were declining. For comparison, net annual uptake in a 55- to 75-year-old New England forest varied between 1.4 and 2.8 Mg C ha−1year−1 (23).

Figure 4

The amount of carbon in terrestrial carbon pools of the United States in 1990 (petagrams of carbon), changes in these pools (petagrams of carbon per year), and exchanges between these pools and the atmosphere during the 1980s (petagrams of carbon per year).

The rate of carbon accumulation attributed to land-use change in our study is somewhat less than that determined from direct measurement of wood volumes in forests (forest inventories) (24) (Table 1). Despite the relatively high variability of estimated changes for soil carbon and wood products, the absolute differences between analyses based on forest inventories are <0.2 Pg C year−1, about the error estimated for our analysis.

Table 1

Average annual changes in the amount of carbon held in forests and wood products during the 1980s (petagrams of carbon per year). Positive values indicate an accumulation on land.

View this table:

The approaches based on land-use change and on forest inventories are largely independent, and differences in their results may suggest the importance of management in determining the accumulation of carbon in forests (25). Carbon accumulations in our analysis are calculated by applying growth rates to forest areas previously harvested. Carbon accumulations obtained from forest inventories, on the other hand, are based on measured rates of growth, which include both recovery from earlier harvests and other factors. If CO2 fertilization, climatic change, or N deposition were responsible for an enhanced rate of growth, for example, one would expect analyses based on forest inventories to show higher rates of growth (and a greater rate of carbon accumulation) than the analysis based on land-use change (in which rates of growth were held constant over time) (26). This difference in accumulation rates is observed. The growth attributable to recovery from previous harvests accounts for only 20 to 30% of the growth measured (Table 1). The other 70 to 80% may be attributable to other factors. Remarkably, the uptake of carbon calculated recently with a process model including variations in temperature, precipitation, and increased CO2accounts for the rest of the difference (0.078 Pg C year−1) (27). The agreement is fortuitous given the errors associated with the estimates, but it suggests that both changes in land use and the responses of ecosystems to environmental changes are important in explaining the uptake of carbon observed in U.S. forests during the 1980s.

The estimates of uptake of carbon from direct measurements of forest growth and from changes in land use are both considerably lower than the annual sink of 1.7 ± 0.5 Pg C year−1 inferred from atmospheric data and models (4). That estimate includes Canadian lands south of 51°N, so the sink for the United States alone is presumably less. Some of the difference may also be explained by the time periods included in the analyses. Our estimate and those based on forest inventories were for the 1980s. The analysis based on atmospheric data and models was based on a shorter interval, 1988–1992, that included the eruption of Mt. Pinatubo and anomalously low growth rates of atmospheric CO2. Thus, the large sink may have been transient. Other estimates of a terrestrial carbon sink in northern mid-latitudes, based on analyses of atmospheric gradients in CO2, 13CO2, and O2, vary between 2 and 3.5 Pg C year−1 (28) and suggest a U.S. sink of 0.3 to 0.5 Pg C year−1 if the sink is assumed to be distributed evenly over northern lands (29).

The accumulations of carbon estimated from forest inventories and land-use change may be low. The analyses do not consider all lands. Forest inventories, in addressing only forests, ignore 70% of the land surface, and natural ecosystems not directly harvested, cleared, or burned were ignored in our analysis. Could such ecosystems be accumulating substantial carbon? It seems unlikely. The total area of these unmanaged lands was only 220 × 106 ha (24% of the land area of the United States) in our analysis, too small to account for much of the difference between 0.2 Pg C year−1(from land-use change or forest inventories) and ∼1 Pg C year−1 [from (4)]. The sink would have to average 3.6 Mg C ha−1 year−1 or 85% of the average net primary production of terrestrial ecosystems in the contiguous United States (27). Furthermore, forest inventories include the accumulation of carbon in unmanaged as well as managed forests, thus reducing further the area not considered in either analysis.

Perhaps forest soils, which are not measured in forest inventories, are accumulating carbon. We have already noted the large uncertainty in estimates of change in soil carbon (Table 1). In our analysis, we included the accumulation of carbon in soils recovering from abandoned agriculture but not the possible loss (and subsequent recovery) of carbon in forest soils affected by logging. Although some studies have reported such a loss, the data are not consistent (30). Furthermore, a recent analysis of soils in a 55- to 75-year-old New England forest found that soils accounted for only 7 to 15% of the net sink measured for the stand (31). If the estimate is representative, the accumulation of carbon in forest soils is intermediate between those calculated from forest inventories (2, 3) (Table 1).

The evidence is more compelling that carbon is accumulating in agricultural soils as a result of changes in management, but our bookkeeping model did not consider these changes. We calculated that the net loss of soil organic carbon from cultivation and abandonment accounted for about 23% of the long-term net flux attributable to land-use change. Each hectare cultivated lost about 25% of its initial carbon to a depth of 1 m, and during the 1980s, soils were a net sink of 0.002 Pg C year−1 as a result of agricultural abandonment (Fig. 4). However, increased crop productivity since 1960 and, more recently, conservation tillage and the Conservation Reserve Program are likely to have increased further the storage of carbon in agricultural soils. Using the estimates of Donigian et al. (32), we calculated an additional rate of accumulation of 0.107 Pg C year−1 for the 1980s. Adding conservation tillage (33) and the Conservation Reserve Program (34) increased the estimated sink in managed soils to 0.138 Pg C year−1 (Table 2).

Table 2

Average annual emissions (+) and accumulations (−) of carbon (petagrams of carbon per year) during the 1980s from forestry and changes in land use. Changes in soil carbon are included in cultivation and abandonment of croplands. Wood products are included in the emissions from harvests.

View this table:

We also estimated the amount of carbon that might have accumulated as a result of woody encroachment in nonforest ecosystems (21) and the “thickening” of western pine forests that began with fire suppression initiated before systematic collection of data on wildfires (22). All of these processes together (changes in agricultural soils, woody encroachment, and early fire suppression) give an upper estimate for carbon sequestration of 0.35 Pg C year−1 for the entire United States (Table 2). Two of the largest sinks (woody encroachment and early fire suppression) are also the most uncertain, whereas the sinks best documented (from changes in agricultural areas and wood harvests) are small. It is quite possible that the total terrestrial sink during the 1980s averaged only 0.15 Pg C year−1, and thus it is difficult for us to imagine that an additional sink three to ten times larger (4) has been overlooked. A possible increase in the carbon buried in sediments as a result of human-induced erosion (35) might increase the terrestrial carbon sink further, but the magnitude of carbon burial through erosion is unknown. In contrast to these terrestrial carbon sinks attributable to management of the landscape, the annual release from fossil fuels in the United States averaged 1.230 Pg C year−1 during the 1980s and was 1.447 Pg C in 1996 (36). Management of terrestrial ecosystems appears to have offset only 10 to 30% of the carbon released from fossil fuels.

  • * To whom correspondence should be addressed. E-mail: rhoughton{at}


View Abstract

Stay Connected to Science

Navigate This Article