Reconciling Life Cycle Assessment and Indirect Land Use Change: Some Implications for RFS2

1. Reconciling Life Cycle Assessment and Indirect Land Use Change: Some Implications for RFS2 Bruce E. Dale Great Lakes Bioenergy Research Center Michigan State University EBI Biofuels Law and Regulation Conference Champaign-Urbana, Illinois April 25, 2012

2. Renewable Power is Critical for Human Well Being Rate of energy use (power consumed) strongly affects (determines?) national wealth and human development All rich societies use a lot of energy (~33% oil) “Energy efficiency” helps but is not an answer in itself Fossil energy use makes us rich today—what energy sources will make our grandkids rich? How will the billions of poor people in the world ever access enough energy to develop their potential? Of all forms of energy, liquid fuels are the most valuable and most problematic in terms of supply, price and price volatility Peak oil has already arrived- 2005 by my rear view mirror Only large scale, low cost, low carbon energy sources can reduce GHGs, provide energy security and long term wealth Thus cellulosic (and other sustainable) biofuels are not optional—they are essential How can we develop sustainable biofuel pathways?

3. Power Consumption and GDP (World Regions) Relationship between 2009 per capita primary energy consumption and GDP per capita for main regions of the world and income levels (as specified by the World Bank). GDP values are adjusted for Purchasing Power Parity and reported in current international $. GDP per capita is from the World Bank, and per capita primary power consumption is derived from other indicators provided by the World Bank: http://data.worldbank.org/indicator. Accessed on Jan. 4 2012. Information on whichcountriesareincluded in the classifications is available at: http://data.worldbank.org/about/country-classifications/country-and-lending-groups The regression line is derived with the constraint that 0 kilowatts per person = $0 GDP per capita.

4. Energy Consumption & Human Well Being are Linked Relationship between 2008 per capita primary energy consumption and human development indices (HDI) for 170 countries. Qatar is not shown with a per capita primary energy consumption of greater than 30 kilowatts per person. Based on a figure by Martinez and Ebenhack, 2008; and the inset is based on a figure from the Human Development Report Office (HDRO) of the United Nations (UN): http://hdr.undp.org/en/statistics/hdi/. Human development indices are also from the HDRO: http://hdr.undp.org/en/statistics/hdi/ . Per capitaprimaryenergyconsumption data arefrom the U.S. Energy InformationAdministration (EIA): http://www.eia.gov/cfapps/ipdbproject/IEDIndex3.cfm . All data were accessed on Nov. 17, 2011.

5. Energy Consumption & Human Well Being are Linked: NO Countries have Both High HDI and Very Low Energy Use Relationship between 2008 per capita primary energy consumption and human development indices (HDI) for 171 countries. Human development indices are also from the HDRO: http://hdr.undp.org/en/statistics/hdi/ . Per capitaprimaryenergyconsumption data arefrom the U.S. Energy InformationAdministration (EIA): http://www.eia.gov/cfapps/ipdbproject/IEDIndex3.cfm . All data were accessed on Nov. 17, 2011.

6. Some Basic Energy Facts • Services we need from energy (current primary sources of these services; fossil (black) and renewable (green)) • Heat (natural gas, coal, solar, wind, hydro, biomass) • Light/electricity (coal, natural gas, hydro/nuclear, solar, wind, hydro, biomass) • Mobility (oil—96%, ethanol, CNG, biomass)- most commerce • All energy services (all BTU, ergs, GJ) are not created equal—we value mobility (=oil) above all other energy carriers • Electricity/batteries can never provide more than about half of mobility needs—and they cannot support commerce at all • Only plant biomass can support all mobility demands • Commerce moves by trucks, ocean shipping, rail & jet aircraft • Economic chaos results when liquid fuel demand exceeds supply • Liquid fuels: not “energy” is the key economic security issue—and right now liquid fuels means refined oil products

7. Comparative Value of U.S. Energy Sources over Time Prices are adjusted to 2005 US$ using the GDP price deflator. All values are for domestic production. Coal is free-on-board price and includes all types of coal. Natural gas is the wellhead price. Crude oil is the first purchase price average for the entire country. Prices were reported as $/short ton for coal, $/thousand cubic feet for natural gas, and $/barrel for crude oil. The conversions used for each were: coal (19.858 million BTU per short ton), natural gas (1.025 million BTU per thousand cubic feet), and crude oil (5.800 million BTU per barrel). All values and conversion factors are from the U.S. Energy Information Administration. http://www.eia.gov/kids/energy.cfm?page=about_energy_conversion_calculator-basicsAccessed Jan. 5, 2012.

8. Crude Oil Prices from 1861 to 2010 Values for 1861-1944 are the US Average values; 1945-1983 are Arabian Light posted at RasTanura; 1984-2010 and 2012 are Brent dated. Based on a table from the BP Statistical Review of World Energy 2011 and data is from the associated database: http://www.bp.com/statisticalreview. Accessed Jan 13, 2012. Current price of crude oil is in current US$ and was obtained from the US Energy Information Administration: http://www.eia.gov/dnav/pet/pet_pri_spt_s1_d.htm. Accessed Jan 24, 2012.

9. Worldwide Crude Oil Production – Subdivided into World Regions and Top 10 Producers in 2010 Peak Oil Data is from the BP Statistical Review of World Energy 2011 and the associated database: http://www.bp.com/statisticalreview. Accessed Jan 13, 2012. Numbered countries are the top ten worldwide oil producers.

10. Commodity Prices Prices are from the World Bank GEM Commodity Index Database: http://data.worldbank.org/data-catalog/commodity-price-data. Accessed: Jan 16, 2012. Prices were reported as 2005 US$/bbl for crude oil (spot price average of West Texas Intermediate, Brent, and Dubai), 2005 US$/mt for urea (E. Europe, bulk) and steel rebar, and 2005 US$ (2005 = 100) for food and metals & minerals.

11. Slide from Prof. Michael O’Hare This “indirect” GHG emission incurs a “carbon debt” assessed against biofuels of up to 900 years

12. Why Are We Making Biofuels? • Use surplus agricultural commodities, increase crop prices & benefit rural communities (policy since 1970s) • Reduce dependence on imported oil (policy since 1970s) • Reduce greenhouse gases and provide other environmental services (policy last decade) • Biofuels do accomplish all these objectives • Reduce direct GHG compared to gasoline & diesel • Reduce dependence on imported oil • Increase agricultural commodity prices & help rural communities • It is a little late to discover that biofuels are at least partially designed to increase crop prices, Congress has had that policy objective for decades

13. What does the Law Say? Section 201 of EISA (42 U.S.C. 211(o)(1)(H) defines lifecycle greenhouse gas emissions as follows: (H) LIFECYCLE GREENHOUSE GAS EMISSIONS. The term “lifecycle greenhouse gas emissions” means the aggregate quantity of gas emissions (including direct emissions and significant indirectemissions such as significant emissions from land use change), as determined by the Administrator, related to the full fuel lifecycle, including all stages of fuel and feedstock production and distribution, from feedstock generation or extraction through the distribution and delivery and use of the finished fuel to the ultimate consumer, where the mass values for all greenhouse gases are adjusted to account for their relative global warming potential. The law correctly describes the fuel lifecycle (large red oval above). However, ILUC occurs outside the fuel life cycle…

14. It isn’t life cycle assessment just because someone says it is.

15. Some Life Cycle Assessment Standards: In Plain English • Use the most recent/most accurate data possible • Select the reference system: what exactly are we comparing? • Make it easy for others to check your data and methods= transparency • Set clear system boundaries (physical & temporal)—must be equal or comparable for reference system and/or reference product of interest • Multi-product systems must allocate environmental costs among all products • Perform sensitivity analysis: how much do results vary if assumptions or data change?

16. Direct vs. Indirect Land Use Change • Direct land use change is: • Supply chain oriented—specific, defined system • Can identify/hold responsible individual actors • Data driven—real situations described • Testable in real time—grounded in reality • Indirect land use change is: • Market oriented—whole world is system • No individual actors/no one is “responsible” • Model and assumption driven- hypothetical situations described: “scenarios” • Not truly testable– predicts future • ILUC shifts responsibility for GHG emissions from the actual polluter to the biofuel producer

17. What Are Life Cycle (LCA) Models? • Set of “accounting” procedures for determining and comparing the environmental impacts of different products • Goal is environmental improvement • For example, disposable diapers vs. reusable diapers • It turns out that disposable diapers have lower environmental impact than reusable diapers • LCA exists to make proper comparisons • Like all accounting systems, there are rules that must be followed • Unfortunately, much current “LCA” analysis for biofuels is deeply flawed: it doesn’t meet basic standards for LCA

18. iLUC Predictions from Previous Studies

19. Our Work on iLUC • Effect of management decisions on iLUC: Result: iLUC has no unique value and can even be negative (net carbon sequestration) if best land management practices are used • Test iLUC predictions with global historical data through a “bottom-up”, data-driven, statistical approach Result: no evidence for iLUCin the US from historical data (confirmed by ORNL and Ford Motor Company) • Allocate iLUC between ethanol and human dietary preferences Result: iLUC decreased to ~ 4 gCO2/MJ • Ensure mathematical rigor in iLUC calculation by complying with the allocation principle Result: iLUC estimates based on global economic models probably violate LCA allocation principle and are thus mathematically invalid

20. Allocate iLUC between Ethanol and Human Dietary Preferences Global use of coarse grain • About 62% of coarse grain used as animal feed • 34% of coarse grain used as human food, seed & non-food purposes • The remainder (4%) of coarse grains produced ends up in various waste streams

21. Current Practices in iLUC • GHG emissions associated with iLUC have been assigned only to biofuel production irrespective of the actual use of the corn (or coarse grain) • This is a value choice… it is not based on science • What if other value choices were considered? • For example: iLUC does not differentiate between real nutritional needs and dietary preferences • What are the basic functions of animal-derived foods? • Nutrition & Taste(dietary preference)

22. Two Questions for iLUC Current Practices • Should biofuel production be held responsible for the environmental consequences of the new converted croplands for coarse grain production that ends up in waste streams? • Coarse grain produced in the converted croplands due to biofuel production follows the same pattern of the current usage of coarse grain • 4% of grain goes to the waste streams • It seems more reasonable and more useful to assign these consequences to crop supply chain management rather than to biofuel production.

23. 2) Should biofuel production be held responsible for lifestyle food choices or instead for its effect on the supply of basic human nutrients? • About 62% of coarse grain produced in the converted croplands will be used as animal feed to produce animal-based foods • LCA is function oriented – Nutrient and Taste (lifestyle food choice) • From a nutritional perspective, vegetable-based alternatives for current animal-based foods are available • Biofuel production is not responsible for food choices, but only for the nutrients lost when crops are diverted to biofuel production. Thus, “Nutrition versus Biofuel”, not “Food versus Biofuel”.

24. Background for Allocation in iLUC • The choice between animal-based foods and equally-nutritious alternatives is a personal decision which depends on personal and socio-cultural characteristics (dietary preferences) • Biofuel production should take full responsibility for the environmental consequences of the newly converted croplands involved in producing grain • for food for direct human consumption, • for seed production, • for other non-food purposes, and • for the basic nutrients lost to biofuel production • but not for animal feed production, which depends on personal food preferences, rather than actual nutritional requirements, • not for the waste streams

25. How to Allocate iLUC between Biofuels and Dietary Preferences: Function-Oriented • Thus biofuels are responsible for GHG emissions due to iLUC associated with “NUTRIENTS” lost due to U.S. ethanol production regardless of the origin of these nutrients– whether vegetable or animal-based. • The major nutrient of animal based foods is protein (neglecting the micro nutrients also available from animal proteins). • From a nutritional perspective, vegetable-based alternatives are equivalent to animal-based foods. • Even though supplements for some micro nutrients (e.g., calcium, vitamin B12) would be needed for vegetarian diets, croplands used for those supplement productions are ignored because 1) those supplements are produced via a fermentation process and 2) no process information of fermentation is available at this time. • Price competition between vegetable- and animal-based proteins is ignored

26. Assigned to crop supply chain management Waste stream 0.08 million hectare (4.6%) Assigned to dietary preferences Nutrition for animal Converted croplands Coarse grain 1.72 million hectare Animalfeed 1.26 million hectare (73.4%) Nutrition for human Vegetable-based nutrition for human 0.32 million hectare Food, seed, other purposes Assigned to corn grain based ethanol 0.38 million hectare (22.0%) 0.70 million hectare Diagram for the Assignment of Converted Croplands based on Tyner et al., 2010

27. Results from Nutrient-based Allocation • Allocating between actual corn uses reduces the converted croplands and GHG of iLUC by 60% and 72%, respectively. • Converted croplands: from 1.72 to 0.7 million ha • GHG of iLUC: from 14.5 to 3.9 g CO2/MJ

28. Influence of Personal Preferences on Allocation of Impacts www.glbrc.org

29. Influence of Personal Preferences on Allocation of Impacts www.glbrc.org

30. Ensuring Mathematical Rigor in iLUC Calculations used with Life Cycle Assessment Land use Land use • Global agriculture economic models consist of a multi-input/output system, involving numerous economic activities. • Models isolate the activity of interest from the whole economic system by “shocking” an exogenous (independent) variable • Shock (∆): marginal or incremental change Incremental change B0+∆ Biofuel production Incremental or Marginal change ∆:Shock B0 B0+∆ B0 Marginal change year B0+∆ (|∆|<<1) Biofuel production

31. Similarities between Isolation and Allocation • In iLUC calculations, the endogenous variables (i.e., dependent variables) are assigned to various exogenous variables (i.e., independent variables). • In the LCA context, the endogenous variables in the economic model are the ‘burdens’, and the exogenous variables are the ‘functions’ delivered by a system. • If we are to use these economic models and LCA models together, we must do so in a way that preserves the mathematical integrity of both

32. Absolute Requirement for Allocation in LCA The sum of the allocated burdens must be equal to the total burdens before the allocation…the allocation process cannot alter the actual environmental burdens Therefore the sum of the allocated endogenous variables must be equal to the total unallocated endogenous variables. Where L is the endogenous variable (or burden), e.g., cropland use. Xi is the exogenous variable (or function), e.g., biofuel demand, food, animal feed, etc. Ai is the allocation factor for the ith exogenous variable.

33. Relation to Shocking Process • The shocking process in the economic models attempts to estimate the allocation factors, Ai. Where ∆Xbiofuel is the marginal (or incremental) change in biofuel demand. • When ∆Xi, are very small (i.e., infinitesimal or approaching zero), • Then the endogenous variable becomes • The shocking process in the economic models is mathematically identical to the marginal allocation approach in LCA (Azapagicand Clift, 1999). • In LCA, the marginal allocation approach cannot be used for nonlinear systems because a new equilibrium state is identified

34. Mathematical Rigor in Economic Models when Combined with LCA Endogenous variables: Biofuels, food, animal feed, etc. Exogenous variable: Cropland use, etc. • The isolation process is mathematically correct only if either of the following equations is satisfied Or • Otherwise, the isolation process cannot be used with LCA. Maximize Utility Function Maximize Utility Function … Maximize Utility Function Maximize Utility Function

35. To Recap • Testing indirect land use change from the historical data might reduce uncertainties in iLUC estimates or perhaps form the basis for better policies or standards for biofuels. • Bioenergy production is one of many factors driving land use change and thus a proper allocation among these drivers is required. Much more needs to be done with allocation. • Regulatory agencies should apply appropriate mathematical rigor in developing and applying their mathematically-based regulations. Thus far, the necessary mathematical rigor does not seem to have been applied to iLUC analyses done with LCA. • Instead of imagining how bad biofuels might be, why not design biofuel systems to be as sustainable as possible?

36. Acknowledgements • The DOE Great Lakes Bioenergy Research Center (DOE Office of Science BER DE-FC02-07ER64494) • Michigan AgBio Research • Biomass Conversion Research Laboratory, Michigan State University

37. Problems with Existing Biofuel LCA Analyses • Use the most recent/most accurate data possible • Predicting agriculture & technology in 2022 using economic models • We can reasonably ask for tests of the models against past history • So far, they seem to fail such tests • Select the reference system: what exactly are we comparing? • Compares future biofuels with petroleum fuels in 1999-2005 • Make it easy for others to check your data and methods= transparency • It is NOT easy to check results, very complex, linked models • Set clear system boundaries (physical & temporal)—must be equal for reference product of interest • Indirect effects are assessed only against biofuels, not petrofuels • Multi-product systems must allocate environmental costs among all products • Entire environmental “cost” of indirect land use change is assessed against biofuels, in spite of the fact that we use land to provide food, feed, fiber, timber, etc… • Perform sensitivity analysis: how much do results vary if assumptions or data change? • Some really important ones have been missed

38. How Can We Move Forward? • Most of us want energy security & environmental progress • Find some shared objectives and common sense solutions • Commend Searchinger and others for raising a crucial issue: how we use land to meet needs • Disagree strongly with the basic premise of ILUC • Polluter should pay: foreign or domestic • We should think globally, and act locally • Suggestion: • Assess fuels on actual carbon content • Conduct annual GHG accounting of direct supply chain, incentivize technological improvement • Protect world’s forests by other mechanisms, financial incentives? • I fear we will continue to drift in our fuel choices

39. The Fundamental Premise of ILUC is Wrong • Price increases drive models and ILUC, therefore: • Any increase in agricultural prices is “bad” • Conservation programs (CRP) that take land out of production are “bad” because they raise prices • Domestic agricultural land should never be converted into forest/grassland • Agricultural communities around the world should stay poor forever • Some consequences of indirect effects analysis: • Domestic industries responsible for the environmental performance of competitors around the world • Protects existing uses of land (mostly for animal feed) vs. any new uses • Undermines value of real life cycle analysis • Applied consistently, indirect effects analysis will doom all petroleum alternatives

40. ILUC & the Allocation Principle of Life Cycle Analysis • ILUC does not allocate GHG impacts of land use among all land uses, it favors existing uses versus any new (eg, biofuel) uses • Most human use of land is for animal feed, not “food” (10 times other uses) • ILUC allocates the entire land use “cost” of biofuel production to the biofuel • Thus, animal feed production is assumed to be “sustainable” but biofuel production is not • Again, Dr. Jason Clay (WWF) “Why would we want sustainable biofuels and not sustainable feeds?” • Also biofuel production is not sustainable but the following land use practices are sustainable: • Tropical hardwood harvest • Urban sprawl • Tobacco, opium and marijuana production • Cotton and linen for clothing (polyester, anyone?) • Pulp, paper and building materials (cinder blocks)

41. Apply Indirect Effects to Electric Vehicles • Let’s power battery electric vehicles (BEVs) with electricity from wind or solar • Large greenhouse gas benefits, right? • No. By the magic of indirect effects, BEVs would have huge, permanent indirect GHG emissions • Consider the logic: • BEVs drive up demand for electricity (or grain) • Market responds by replacing that electricity (or grain) • With a coal burning power plant (clearing new land) • Indirect effects, consistently applied, make BEVs “responsible” for those coal emissions (land clearing) • Doesn’t matter if BEVs are actually powered by wind or solar—(the actual or direct supply chain doesn’t matter) • Thus, by indirect effects, BEVs would have large negative GHG consequences (carbon debt)

42. Indirect Land Use Change (iLUC) • Changes in United States corn supply caused by ethanol fuel production could increase • corn (coarse grain) production areas in other countries due to reduced US corn exports and/or • croplands for soybean in other countries due to reduced US soybean exports, resulting from increased corn production in the United States. • Associated land conversion events might release GHGs which should be assessed against corn ethanol’s GHG profile.

43. Historical Test: Five Conditions • iLUC due to US biofuel production happens in regions, which meet each of the following five conditions. • For a specific region, average areas of croplands used for corn (coarse grain) plus soybean production in the 2000s must increase compared to the 1990s. • For a specific region, average areas of arable lands in the 2000s must also increase compared to the 1990s. • For a particular geographical region, average areas of natural ecosystem lands in the 2000s decline compared to the 1990s. • For a particular region, average corn (coarse grain) plus soybean imports from United States in the 2000s decline significantly compared to imports during the 1990s. • Annual percentage changes in croplands for corn (coarse grain)plus soybean in a specific region are positively correlated with annual percentage changes in harvested areas for corn and soybean for biofuel production in the United States.

44. Testing United States • Cotton, corn plus soybean and oats are significantly correlated with the annual percentage change in croplands for biofuel production in the United States at p < 0.05. • However, negative correlations are observed only for cotton, which is not a major food source. • No arable land increases from the 1990s are observed in the United States. • Furthermore, no declines in natural ecosystem lands in the United States have been observed since 1998. • Therefore, the US data do not support the idea that iLUC occurred within the 48 contiguous states as a result of US biofuel production. • Correlation is not causation—but lack of correlation greatly weakens the causal argument

45. Details: Results from Testing Conditions 1 - 4 • Three regions meet conditions 1 – 4 • Brazil, Oceania countries, and Sub Saharan Africa • Other regions do not meet conditions 1 – 4 • For examples, Malaysia and Indonesia (MAI) region shows no increases in croplands for corn (coarse grain) plus soybean despite satisfying conditions 2 – 4. • Palm oil, but not corn (coarse grain) or soybeans, is the major crop planted in the converted natural ecosystem followed by coffee, coca bean and natural rubber from 1992 to 2008. • Rest of South East Asia (SEA), and South and Other Americas (SOA) regions meet conditions 1 – 3, but do not meet condition 4. • Rice and beans are the major crops grown on increased arable lands in the SEA region • Soybean is the major crop grown on increased arable lands in the SOA region, particularly in Argentina. Large increases in soybean areas in Argentina occurred from 2001 to 2005 because of its floating currency policy, biotechnology and double cropping

46. Results from Testing Condition Five • No regions that have a significant correlation to biofuel production in the United States (at p < 0.05) • Increased soybean production plays a major role in increased croplands in Brazil: Policies, biotechnology and growing soybean demand from China expanded soybean area in Brazil. • Wheat, barley and rapeseed are major crops for the expansion of croplands in Oceania, particularly Australia: primary reasons for the expansion of croplands for rapeseed: better varieties, improved agronomy, crop monitoring program (Canola Check) and good prices • Population pressure is a major driver for agricultural expansion in Sub Saharan Africa, where extensification is a dominant trend

47. Findings from Testing Historical Data • Biofuel production in the United States up through the end of 2007 has probably not induced indirect land use change • 1) crop intensification may have absorbed the effects of expanding US biofuel production or • 2) the effects of US biofuel production expansion may be simply negligible, and not resolvable within the accuracy of the data. • Oladosu et al (2011) also show little support for large land use changes or diversion of corn exports due to the US ethanol production during the past decade through a systematic decomposition analysis of the empirical data from 2001 to 2009. Or • A contrary interpretation is that this empirical test simply fails to detect ongoing indirect land use change from the historical data