全球变化知识资源中心

Liska et al. reply —

The soil organic carbon (SOC) model that we used¹ was parameterized with data from arable land under normal farming conditions in North America, Europe, Africa and Asia², but the equation is insensitive to changes in tillage, soil texture and moisture. The model has reasonable accuracy, however, in predicting changes in SOC, residue remaining and CO₂ emissions from initial SOC, carbon inputs from residue, and daily temperature^{1, 2}; the shoot-to-root ratio used in the geospatial simulation was 0.29 (that is, root carbon is 29% of total aboveground carbon), which did not underestimate carbon input to soil (Supplementary Fig. 2 in ref. 1). There is more theoretical confidence in the conserved nature of SOC oxidation due to temperature^{1, 2, 3, 4, 5} relative to other factors such as tillage^{6, 7, 8}. In a recent comparison of three SOC models (CENTURY, DAYCENT and DNDC), predictions were close to or within the range of uncertainty of estimates derived from soil measurements, showing that these models tend to produce similar results from residue removal⁵. (A range of soil measurements have also shown net SOC loss from residue removal^{1, 5}.) The model also agreed well with CO₂ emissions measurements from an AmeriFlux field site¹, which since 2000 has been funded with $7,370,000 from the US Department of Energy, the US Department of Agriculture and NASA, leading to over 85 peer-reviewed publications.

The question for life cycle assessment (LCA)¹ is: what is the net change in SOC compared with a counterfactual situation where residue is not removed? It seems that the logic of this question has not been recognized by the US Department of Agriculture⁹ or US Department of Energy¹⁰. Simulations with 2, 4 and 6 Mg ha⁻¹ yr⁻¹ residue removal in the Corn Belt, corresponding to ~25, ~50 and ~75–100% of corn residue produced in a single year, respectively, each resulted in a net SOC loss compared with no removal, which is difficult to measure in soil in less than 5 years but can be estimated confidently using models^{1, 3, 5}. Importantly, when SOC losses are normalized for the energy in the biofuel derived from residue, roughly equivalent CO₂ intensities are estimated regardless of the amount of residue removed (Fig. 2c in ref. 1) — a central finding of our research.

The question for LCA is also not: how could these systems be in the future? The question is, however: how are these systems performing now, and how are they going to perform in the near term? The lignin coproduct is burned to provide energy for biofuel processing, and currently no electricity exports or other coproducts exist in the Poet's Liberty project (http://poet-dsm.com/liberty). Potential electricity output from burning lignin could also be 69% lower than the estimate previously provided (that is, −17 g CO₂ equivalent MJ⁻¹ versus −55 g CO₂ equivalent MJ⁻¹)^{1, 10}. The lignin oxidized in biofuel processing is the SOC that is lost, because that lignin would have oxidized more slowly in soil^{1, 2, 3, 4}.

Standards for LCA are under development and in a state of flux. Owing to the complexity of LCA, a wide range of values can be produced in these assessments due to arbitrary variability in spatial and temporal parameter values, modelling assumptions, timeframes and system boundaries^{11, 12}. Consequently, our analysis focused on quantifying uncertainty in one primary variable: net SOC loss to CO₂ from residue removal¹. The 30-year time interval precedent set by Searchinger et al. is arbitrary and biases results in favour of biofuel producers^{12, 13}. Precedents used by the US Environmental Protection Agency may not favour near-term emissions reductions, and existing precedents will probably be revised. To accurately represent current climatic conditions and SOC dynamics, temperature measurements from 2001 to 2010 were used¹, because older data do not represent increased temperatures and future projections are more uncertain. The model¹, however, was also used to estimate SOC changes from 2010 to 2060 with estimated increases in crop yields and temperatures from the IPCC's Fifth Assessment Report climate simulations (representative concentration pathway 8.5 emissions scenario)¹⁴. When compared with no residue removal, removal of 3 Mg ha⁻¹ yr⁻¹ of residue from continuous corn was estimated to lose ~0.22 Mg C ha⁻¹ yr⁻¹ on average in the first 10 years in three counties in Nebraska and Iowa; for the first 30 years, this value was reduced by ~52% on average to ~0.11 Mg C ha⁻¹ yr⁻¹ (ref. 14).

Yet, to dilute SOC emissions over 30 years or more does not represent actual CO₂ emissions over the first 10 years, and presenting longer-term lower values can be deceptive. Sanchez et al. noted, “Policymakers may find it appropriate to focus on more certain, near-term climate impacts, in which case a short horizon for fuel warming potential is sufficient.”¹² If residue is removed for biofuel, these systems could produce more CO₂ emissions than gasoline for more than 10 years (ref. 1) and then possibly reduce emissions in 20 to 30 years, after agricultural SOC stocks have significantly decreased and crop yields have probably declined. Alternatively, SOC loss from residue removal can be widely recognized, and appropriate management can be used to compensate for lost carbon and increased CO₂ emissions¹.