CO2 Fertilization and Land Use

Since the inception of the grant, Steve Pacala’s group has worked to reduce uncertainty about the future of the land sink. There are two major sources of uncertainty. The first is the future impact of CO2 fertilization of the biosphere. If CO2 fertilization is as large as is assumed in most models, then the sink will approximately double in size over the next half century. However, if fertilization is absent, then the sink is expected to decrease. The difference between these two alternatives is a factor of two in terms of the size of the carbon and climate problem: seven wedges of mitigation required with weak or absent fertilization and 3-4 wedges with strong fertilization. During the last several years, Pacala’s group has repeatedly analyzed forest inventory data looking for evidence that CO2 fertilization is as strong as models currently say it should be. This is a difficult task because expected increase in plant productivity is only 0.1% per year. Nonetheless, the large sample sizes available in forest inventories provide estimates with the necessary precision and two independent methods now show weak or absent fertilization effects in the United States.

Pacala’s team has recently completed a factorial modeling experiment to study the causes of the terrestrial carbon sink over the last 300 years. The experiment was designed to examine the relative effects of CO2 fertilization, climate change, agricultural clearing and abandonment, and timber harvesting. The results show that secondary forestry (the cycle of harvest and re-growth of managed forest) is responsible for a previously unexplained feature of the sink – the sharp peak in its magnitude at north temperate latitudes. The group has also worked steadily to improve the models – they now have a prototype version with a global nitrogen cycle and will add a phosphorus cycle this year.


Impact of Climate Change on the Terrestrial Carbon Cycle

The other major source of uncertainty about the future of the sink concerns feedback between climate change and carbon storage. As the climate warms, plant productivity and the rate of decomposition both increase. Because these two increases have opposite effects on the sink (the first increases it while the second decreases it) the net impact of warming depends on the precise balance of the two, which is difficult to predict. A large amount of atmospheric carbon is at stake here – some models show that the biosphere could become a large net source in the next century and increase atmospheric CO2 by 250 ppm.

Uncertainty about the effects of warming on the biosphere is exacerbated by the large variation among species in climate sensitivity. Models and inverse methods typically assume that plant species may be divided into a few functional types (i.e. evergreen conifers, temperate broadleaf trees, cool season grasses). All species within a functional type are assumed to have the same climate sensitivity, and so all differences between species are assumed to be between functional types. In fact, the reverse is true; climate sensitivities are more variable within than between functional types. The false assumption of homogeneous functional types has been necessitated by the inability to measure climate sensitivities for each and every species. However, two new developments suggest that it should now be possible to build a species-level model for forests in the temperate zone (the biomes thought to be responsible for most of the sink), and to greatly improve models of tropical forests.

Many countries of the world have joined the US in creating national forest inventory programs that document the births, growths and deaths of millions of trees. Collectively, these programs provide the largest ecological data sets ever collected. They provide real examples of the very quantities that the equations in forest models predict – the births, growths and deaths of individual trees. It should thus be possible to develop an inverse method to estimate the equations from the inventory data, by adjusting the constants and functional forms in equations so as to best predict the observed births, growths and deaths. This method would enable the production of predictive models of all of the different kinds of forest in every location in countries with national inventories and would revolutionize forest ecology.

All of this was impossible for technical reasons as of last year, when Pacala and colleagues discovered a new mathematical technique to decrease the computer run time for forest models by many orders of magnitude. With the new technique, forest models are fast enough to allow the development of an inverse method.

The group spent the year designing and building most of the machinery for an inverse method and deriving mathematical results. The machinery is almost ready, and the researchers have many promising preliminary results. They can now routinely estimate parameters for all of the tree species in the U.S. in all locations. The extension of the method to Canada and Eurasia should be relatively straightforward because the forests there are similar to those in the U.S. Pacala’s team also has a strategy for extending the method to tropical forests, and has just begun to work with the extensive inventory data from Panama.