Dear Ms. Dunham:

This letter is a follow-up to the June 24, 2008 meeting in which EPA provided an update on its methodology for determining lifecycle greenhouse gas emissions pursuant to the Renewable Fuel Standard (RFS), as amended by the Energy Independence and Security Act of 2007. RFA sincerely appreciated the opportunity to meet with you and other EPA officials, and we are grateful for the opportunity to comment on EPA’s “Key Assumptions” for this analysis for ethanol. EPA’s openness and candor during the rulemaking process is very much appreciated, and we look forward to continuing our interaction with the agency.

We understand these key assumptions will form the basis for EPA’s Regulatory Impact Analysis, but that they will also shape EPA’s determinations for the RFS regulations. We have attached comments based on the information EPA has provided, but it is difficult to provide a full analysis without a complete understanding of EPA’s entire methodology. As such, RFA is likely to submit additional comments, and may revise these comments, once it is able to review EPA’s entire model and methodology.

In general, we believe that several of the assumptions may be relying on outdated information that may not necessarily reflect the current state of the agriculture and ethanol industries. For example, there have been substantial advances in farming technologies and techniques over the last several decades that may not be accurately reflected in EPA’s estimates on corn yield increases, tillage practices, and fertilizer application. It is also unclear if EPA’s assumptions on energy use by ethanol plants reflect the most recent information, such as the information gathered through the RFA survey, which is currently being considered for incorporation into the GREET model.

These assumptions may result in underestimates of the reductions in greenhouse gas emissions that are actually being experienced today and that will likely be experienced in the future. This is particularly true as both the agriculture and ethanol

Sarah Dunham, Director
Office of Transportation and Air Quality U.S. Environmental Protection Agency July 31, 2008
Page 2

industry will be subject to more stringent requirements under the RFS and will continue to strive to become more efficient and to utilize more sustainable methods, as was the intent of the Act.

EPA’s assumptions and the attached comments reflect only some of the variations of ethanol plant processes. The ethanol industry continues to innovate, and we believe that EPA should consider inclusion of a provision to consider site-specific factors to give plants the opportunity to get credit for innovation above what EPA has taken into account.

RFA continues to have several questions with respect to EPA’s approach to the lifecycle analysis, particularly with regard to the assumptions and methodology EPA is using to determine the baseline lifecycle greenhouse gas emissions for gasoline and to determine land use changes. We would greatly appreciate an opportunity to review EPA’s assumptions and methodology for baseline gasoline, as well as ethanol, and would welcome additional explanation on EPA’s methodology for measuring land use changes. For your reference, we have also attached comments RFA has submitted to the California Air Resources Board (CARB) related to the work it is undergoing with respect to lifecycle greenhouse gas emissions.

RFA understands that this is a highly complex undertaking and, again, appreciates EPA’s efforts in this regard and its inclusion of stakeholders in that process.

Please do not hesitate to contact me if you have any questions. We look forward to continued dialog with EPA on this important matter.

Sincerely,

Bob Dinneen President and CEO

cc: Vincent Camobreco, EPA

Renewable Fuels Association Comments on EPA Key Assumptions Lifecycle Greenhouse Gas Emissions ‐ Ethanol July 31, 2008

I. FEEDSTOCK PRODUCTION

CORN YIELD ASSUMPTIONS

EPA ASSUMPTIONS: EPA assumes the U.S. average corn yield will be approximately 180 bushels (bu)/acre in 2022 (a 1.6% annual increase over the baseline year). EPA noted this estimate is consistent with USDA projections. EPA also assumes international corn yields also increase over time. Two examples are given (Brazil and Argentina), but outside of these examples, EPA does not provide adequate detail on country‐ or region‐specific yield assumptions. EPA suggests other factors affecting yield (such as the impacts of high crop prices, cultivation on marginal lands, and best management practices) will also be considered to the extent possible, presumably in sensitivity cases.

RFA COMMENTS: EPA’s corn yield assumptions are far more conservative than the projections of commercial seed producers and some agricultural economists. Though the baseline year is not stated, it appears the USDA yield projections are based on a 30‐year or longer trend analysis. Because significant advances in biotechnology, conventional corn breeding, and on‐ farm management practices have occurred in the last 20 years, the use of a 30‐year yield trend seems overly conservative. If the period in which biotech hybrids have been commercially available to U.S. farmers (1995‐present) is used as the basis for trend analysis, the 2022 average corn yield would be 193 bushels per acre,1 13 bushels per acre higher than USDA’s estimate. EPA is assuming annual yield growth of 1.6%, while average yields have actually grown 2.5% per year since 1995. What is the baseline year EPA is using for yield growth? Information from seed companies such as Monsanto and DuPont suggest the impact of marker‐assisted breeding and molecular breeding practices are likely to accelerate corn yield improvements even more quickly. Monsanto suggests U.S. average corn yields could reach approximately 265 bushels per acre by 2022 and 300 bushels per acre by 2030.2 While we are encouraged that EPA apparently will be considering accelerated yield growth scenarios in sensitivity cases, EPA should revise its trend yield analysis based on a more recent timeframe due to technological advances. Changes in agricultural technologies and practices since 1995 are more indicative of current and long‐ term practices. This is particularly true given the limitations in relying on “existing cropland” and increasing demands on land use, which will result in further innovations in farming technologies and techniques.

1 1995‐2022 trend yield analysis assumes 2008 average yield of 152 bu/acre (Informa Economics, July 11, 2008),

which is based on more current farmer surveys than the July USDA yield estimate.

http://uk.reuters.com/article/oilRpt/idUKN1139323820080711.

2 http://www.monsanto.com/pdf/investors/2008/06‐05‐08.pdf.

Monsanto presentation at Merrill Lynch Agricultural Chemicals Conference, June 5, 2008. 1

CORN RESIDUE REMOVAL RATES

EPA ASSUMPTIONS: EPA assumes corn residue removal rates of 50% for no‐till systems, 35% for reduced tillage systems, and 0% for conventional tillage systems.

RFA COMMENTS: Actual corn residue removal rates will depend on a variety of agronomic factors and are likely to vary widely from region‐to‐region. Though EPA’s working assumptions seem reasonable on their face, some estimates suggest more corn residue may be removed sustainably.3 EPA did not offer its assumptions on tillage practices in 2022 (i.e., share of total corn acres under no‐till, reduced till, conventional till systems, etc.). The adoption of no‐till and reduced tillage practices will be a key factor in determining the amount of corn stover available.

II. FERTILIZER USE

NITROGEN APPLICATION RATES

EPA ASSUMPTIONS: EPA assumes U.S. nitrogen application rates for corn are approximately 136 lbs./acre in the corn belt in 2022.

RFA COMMENTS: USDA has not surveyed nitrogen fertilizer use since the 2005 growing season, at which time national average application rates were 131.8 lbs./acre. In the last several years, U.S. farmers have increasingly adopted new technologies that make fertilizer application more efficient, such as GPS‐based precision fertilizer application systems and slow‐release fertilizers. The adoption rate of certain tillage practices and biotech‐derived corn hybrids also will dictate average nitrogen fertilizer application rates moving forward. In addition, seed companies are also developing corn hybrids that will more efficiently utilize nitrogen, effectively reducing the amount of nitrogen necessary to produce the same yield. Further, there is a trend toward using more animal manure for crop fertilization as a lower cost alternative to using synthetic fertilizers. We encourage EPA to consider the impact of these new and emerging technologies, as well as the effect of unprecedented nitrogen fertilizer costs on potential application rates.

We also noted that the “Key Assumptions” document provided by EPA does not include the agency’s current working assumptions on nitrogen fertilizer emissions rates. In our June 24 meeting, EPA stated that it is currently using the Intergovernmental Panel on Climate Change (IPCC) rates for emissions from nitrogen volatilization. RFA supports the use of the IPCC figure in EPA’s modeling work.

INCREMENTAL NITROGEN APPLICATION FOR HIGHER YIELDS

EPA ASSUMPTIONS: EPA assumes each 1% increase in corn yields over the baseline requires an additional 0.16% increase in nitrogen application.

3 A report released by the Biotechnology Industry Organization suggests 66% of residue may be removed in no‐till systems. BIO also states that up to 33% of corn residue can be removed in conventional tillage systems. http://www.bio.org/ind/biofuel/SustainableBiomassReport.pdf.

2

RFA COMMENTS: While this seems to be a reasonable assumption on the surface, it implies that nitrogen application will increase infinitum as yield increases. This ignores the fact that the corn plant can only utilize a certain maximum amount of nitrogen and that yield increases are not linearly correlated to nitrogen application beyond this maximum performance uptake level. We agree that the percentage increase in nitrogen application is much lower than the corresponding percentage increase in yield, but doubt that the relationship is linear.

III. PROCESSING

CORN ETHANOL DRY MILL ENERGY USE

EPA ASSUMPTIONS: For dry mills that are drying distillers grains, EPA assumes the following energy use values (table 1).

Table 1. EPA Assumptions for Corn Ethanol Dry Mill (Drying DG)

For dry mills that are not drying distillers grains, EPA assumes the following energy use values per gallon (table 2).

Table 2. EPA Assumptions for Corn Ethanol Dry Mill (Wet DG)

In its general notes, EPA suggests it will consider different process technologies (such as combined heat and power) and will perform sensitivity analysis around energy use values from other sources.

RFA COMMENTS: EPA is likely overestimating the energy use per gallon of dry mill ethanol plants. A 2007 RFA survey shows much lower thermal energy use values both for dry mill plants that are drying the majority of their distillers grains and for dry mill plants that are selling the majority of their distillers grains wet. According to data from the RFA survey:

• The average thermal energy use for responding dry mill plants that are drying 20% or less of their distillers grains output was 21,113 BTU/gallon. Average electricity use per

COMBUSTION TYPE

PROCESS ENERGY (BTU/gal.)

ELECTRICITY (BTU/gal.)

TOTAL ENERGY USE (BTU/gal.)

Natural Gas

38,717

3,242

41,959

Coal

43,447

3,125

46,572

Biomass

43,447

3,125

46,572

COMBUSTION TYPE

PROCESS ENERGY (BTU/gal.)

ELECTRICITY (BTU/gal.)

TOTAL ENERGY USE (BTU/gal.)

Natural Gas

24,764

1,675

26,439

Coal

26,445

1,720

28,165

Biomass

26,445

1,720

28,165

3

gallon for these plants was 1,752 BTU/gallon. Natural gas is the combustion fuel for all of these plants. The total energy value is approximately 14% lower than the assumption being used by EPA for natural gas‐fired dry mill plants marketing wet distillers grains.

  • For dry mill plants drying 80% or more of their distillers grains output, average thermal energy use was 32,427 BTU/gallon. Average electricity use per gallon for these plants was 3,204 BTU/gallon. Natural gas is the combustion fuel for 75% of the responding plants in this class, with coal being used as the combustion fuel for the remaining 25% of these plants. Based on this combined result, the total energy value is approximately 15% lower than EPA’s assumed value for natural gas‐fired dry mill plants drying distillers grains and approximately 23% lower than the EPA assumed value for coal‐fired dry mills drying distillers grains.
  • Dry mill plants that dried more than 20% but less than 80% of their distillers grains output, averaged 28,066 BTU/gallon for thermal energy use. Average electricity use per gallon for these plants was 1,968 BTU/gallon. Natural gas is the combustion fuel for all of these responding plants.Because EPA does not appear to offer an assumed energy use value for dry mills that are producing a mix of wet and dried distillers grains, we cannot make a direct comparison to EPA’s assumptions. Nonetheless, this data illustrates that EPA’s assumptions may be significantly overestimating current energy use for dry mill ethanol plants.The RFA data also illustrates that the variation in marketing of wet and dry distiller grains by ethanol plants may not be adequately captured by EPA’s assumptions. It should be noted that only one of the plants responding to the RFA survey dried 100% of its distillers grains output. Similarly, only one plant responding to the survey sold 100% of its distillers grains in wet form. All of the other responding plants dried some fraction of their distillers grains output, and marketed the remaining fraction as wet distillers grains. This indicates that there is great variation from plant to plant in the share of distillers grains that are dried versus the share that is sold wet. We encourage EPA to investigate mechanisms to account for this fact.

    It also appears that EPA’s current approach to this issue does not account for other potential variations in plant design and operations found at individual plants, such as use of different forms of combustion fuel during the year. These variations, along with the marketing of a combination of dry and wet distiller grains, may reduce the plant’s energy use and greenhouse gas emissions. EPA’s modeling process should be flexible enough to account for these variations, as well as to take into account ongoing technological advances being utilized that reduce energy use and emissions, such as germ fractionation and carbon dioxide sequestration.

4

We are encouraged that EPA is considering other data sources for sensitivity analysis on this topic, but we think it is appropriate to consider use of the more updated RFA survey data for the base case. We also ask that EPA develop a rational analytical pathway that considers the increases in ethanol plant energy efficiency that are certain to occur between now and 2022.

In addition, we question whether coal‐ and biomass‐fired plants have the same energy use/gallon profile, as indicated by EPA assumptions. It is also unclear why the coal and biomass configurations use less electricity than a natural gas plant for dried distillers grains, but coal and biomass configurations use more electricity than natural gas plants for wet distillers grains.

AVERAGE CORN DRY MILL ETHANOL YIELDS

EPA ASSUMPTIONS: EPA assumes the average corn dry mill yields 2.71 gallons of ethanol per bushel of corn.

RFA COMMENTS: EPA is underestimating the average corn dry mill ethanol yield. The 2007 RFA survey found the average ethanol yield per bushel of corn is 2.81 gallons. We encourage EPA to consider other data sources for the base case and sensitivity cases.

DISTILLERS GRAINS PRODUCTION AND DISPLACEMENT

EPA ASSUMPTIONS: EPA assumes one bushel of corn produces 17 pounds of distillers dried grains (dry matter basis). EPA assumes one pound of distillers grains substitutes for 0.9 pounds of corn and 0.1 pound of soybean meal.

RFA COMMENTS: EPA’s assumptions on substitution are not consistent with other current estimates. It is unclear what the basis is for EPA’s assumption that 1 pound of distillers grains substitutes for 0.9 pounds of corn and 0.1 pounds of soybean meal. ADM animal research scientists, as well as literature on the issue, have found that the practical substitution ratios are affected by specie application and displacement materials involve more than just whole corn and soybean meal. There also is evidence that the replacement rate may be higher than a 1:1 ratio ‐‐ i.e., that 1 pound of distillers grains replaces more than 1 pound of conventional feed.4

EPA’s “Key Assumptions” document also did not offer current working assumptions regarding the co‐products of wet mill ethanol production, such as corn gluten feed, corn gluten meal and corn oil. How are these co‐products being treated?

CORN ETHANOL WET MILL ENERGY USE

EPA ASSUMPTIONS: EPA assumes the average corn wet mill uses 45,950 BTU/gallon of ethanol produced and that no electricity is purchased.

RFA COMMENTS: Corn ethanol wet mills responding to the RFA survey referenced above showed wide variability in energy use. The low end of the range was 28,795 BTU/gallon and the

4 “Use of Distillers By‐Products in the Beef Cattle Feeding Industry,” Klopfenstien, Erickson and Bremer, J. Animal Sci., 2008:86:1223‐1231.

5

upper end of the range was 63,422 BTU/gallon. Therefore, particularly for wet mills, we believe the EPA analytical system must be flexible enough to account for the wide variability in energy use among ethanol producers. It appears EPA is assuming all corn ethanol wet mills have the same configuration. Not all wet mills produce 100 percent of their electricity, particularly those that are fueled primarily with natural gas. Additionally, we would like EPA to clarify why it is using two different sources to estimate energy use for ethanol wet mills and dry mills. ASPEN/USDA is used for dry mills, while GREET is cited for wet mill energy use.

The comments regarding plant variation in the dry mills generally apply to wet mills as well.

CORN ETHANOL WET MILL ETHANOL YIELD PER BUSHEL

EPA ASSUMPTIONS: EPA assumes the average corn wet mill yields 2.5 gallon of ethanol per bushel of corn processed.

RFA COMMENTS: EPA cites the GREET model as the source of this assumption, but the actual

GREET model value is 2.62 gallons/bu. Additionally, the RFA survey showed the average wet

mill respondent averaged 2.74 gallons/bu. We encourage EPA to revisit this assumption. A/72609000.1

6

Bob Dinneen
President & CEO
Renewable Fuels Association One Massachusetts Avenue, NW Washington, D.C. 20001

June 27, 2008

Mr. John Courtis
Manager, Alternative Fuels Section California Air Resources Board 1001 “I” Street
Sacramento, CA 95812

Dear Mr. Courtis,

The Renewable Fuels Association (RFA) respectfully submits the attached comments in response to the California Air Resources Board Lifecycle Analysis Working Group’s
“Detailed California-Modified GREET Pathway for Denatured Corn Ethanol.” As the national trade association for the U.S. ethanol industry, RFA appreciates the opportunity to comment on CARB’s current approach to lifecycle analysis for corn-based ethanol. As you will see in the attached comments, we have questions and comments about several of the key assumptions CARB is using for its current lifecycle analysis approach to corn ethanol.

It appears that CARB’s value for corn farming input energy is based on a figure from the U.S. Department of Agriculture’s 1996 Agricultural Resource Management Survey (ARMS). A more recent ARMS survey with corn farming energy use data was conducted in 2001, and even its usefulness is limited because of the rapid adoption of new technologies and tillage practices in the past seven years. We believe CARB should update its farm input energy use values based on current practices and technologies.

CARB’s assumption for nitrous oxide emissions resulting from nitrogen fertilizer application is inconsistent with other research and appears arbitrary. Intergovernmental Panel on Climate Change findings suggest that 1% of nitrogen fertilizer is released as nitrous oxide, while CARB is using a factor of 2%. RFA supports the use of the 1% IPCC factor in the CARB model.

Further, we believe there is good reason for CARB to reevaluate its assumptions on carbon dioxide emissions related to lime application. The actual CO2 emission rates from lime vary

widely and depend on a number of factors. It is also notable that farmers who use limestone do not apply it annually and many do not ever apply lime.

In terms of energy input assumptions for ethanol production facilities, we encourage CARB to consider the results of a new industry survey on ethanol plant efficiency. The survey data were analyzed and published by Argonne National Laboratory in March 2008. The report clearly shows that energy use for ethanol processing has declined in recent years and that the values used by CARB, which were obtained from GREET, are likely too high.

RFA understands that there are a variety of methodologies and differences in opinion on the issue of co-product energy credits. However, we believe that because distillers grain typically contains higher protein and energy content than the feed products it is replacing, a pound-for- pound displacement assumption is incorrect. We encourage CARB to 1.) Engage animal scientists on this issue; and 2.) Review recent scientific literature on current distillers grains feeding practices.

Finally, RFA continues to be highly interested in the CARB’s current thinking on the subject of indirect land use change and its impact on the overall lifecycle. We understand that CARB may be reconsidering the land use change factor presented in its April 21, 2008, report. RFA encourages the agency to ensure the best science is brought to bear on this issue. We also believe it is important that land use metrics are applied equally to all fuel pathways and that the positive effects of possible land use changes are also considered.

We sincerely appreciate CARB’s consideration of these comments and look forward to further interaction with the agency as the fuel pathway methodologies are refined. We will continue to review information provided by CARB and respond with comments as appropriate.

Sincerely,

Bob Dinneen
President & CEO
Renewable Fuels Association

Comments of the Renewable Fuels Association on
“Detailed California-Modified GREET Pathway for Denatured Corn Ethanol”

Overview

This report was issued by the CARB Stationary Source Division on April 21, 2008 and is labeled Version 1.0. It is stated that the report remains under internal review and hence the results are subject to change.

This report is a well-to-wheels (WTW) analysis of greenhouse gas emissions associated with ethanol production from corn and its use in vehicles. It follows typical protocols for this type of analysis, namely separating the analysis into well-to-tank (WTT) and tank-to- wheels (TTW) sections. The WTT portion includes energy and greenhouse gas emissions (GHGs) from farming, agricultural chemicals, corn transport, ethanol production, ethanol transport and co-product credits. There is a small land use GHG category as well. The TTW portion encompasses the vehicle use phase. CO2 from the combustion of ethanol is not counted since it is renewable. Only the CO2 from the 2.5% gasoline denaturant is included. Nitrous oxide and methane from vehicle use of ethanol are not included since “ethanol is not typically used as a fuel by itself in California.”

As noted in the title of the report, the model used is a version of GREET (Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation) called the CA-modified GREET model. Modifications to the model have been made to reflect California specific conditions, rather that the U.S. national average values used in GREET. For this exercise, however, most energy use occurs outside California and hence U.S. national average values are used.

ARB has indicated that this is the first release of the report, and that they are performing additional work on many of the numbers used in the report, and that there will be a later release of the report with updated numbers. Our review evaluates both the methods and the numbers in the current report.

Comments

In the following comments, we first state the ARB assumption as indicated in the Corn Ethanol report. Then we state our comments on that assumption or estimate. 1

1. Corn farming input energy is assumed to be 22,500 Btu per bushel, which is stated to be 90% of the 1996 value. The 1996 value was based on the Agriculture and Resources management Survey (ARMS), which is conducted every several years by USDA.

1 In this document, footnotes are indicated with a superscript, and references are indicated with square brackets: [x]

Comment: What calendar year is ARB estimating the corn farming energy input for? Is it for calendar year 2006, which is the ARB-proposed base year from which the 10% LCFS reduction requirement is being estimated? We think the most recent corn farming survey in the ARMS is for 2001. If so, then the corn farming energy input value is out-of-date, and should be updated to a 2006 level.

Work conducted by Dale and Kim indicate a corn farming energy input of about 16,217 Btu per bushel, or 28% less than the GREET value.2[1] We believe this value is more current than the GREET value, and should be used in the California GREET model.

2. There is a land use change factor of 195 grams of CO2 per bushel. There is no indirect land use effect in the model.

Comments: We understand that this is an area that the ARB may change in the near future, and is conducting much more research on this issue. However, we offer the following comments at this time, and will offer more comments on land use issues if and when ARB releases its new analysis:

First, we think the science and data for developing the size of the indirect land use change and the effect of that change due to biofuels is currently too uncertain and inadequate to support the kind of estimates that ARB is attempting to make. We agree with the letter that was sent from Blake Simmons, PhD, et al to Mary Nichols on June 24th, stating that, “significant research is still required to develop reliable data training sets and validated LCA tools that can accurately guide policies such as the LCFS.” [2]

Second, in its effort to estimate the effects of the indirect land use change, ARB appears to not be considering the positive effects of possible land use change. For example, research by Oak Ridge National Laboratory and others indicates the biofuels can: (1) reduce recurring use of fire to clear land, thereby reducing GHG emissions, (2) reduce the pressure to clear more land, and (3) improve soil carbon. [3] These factors are not included in the GTAP modeling framework that ARB is using to project land use changes. In addition, research by Kauppi, et al, indicates that if annual per capita GDP is greater than $4,600, forest biomass stocks were increasing. [4] If communities around the world participate in growing additional crops, then it is possible that their improved standard of living would allow for increased yields, reducing the pressure to convert additional land, and thereby increasing forestation, rather than reducing it.

Third, ARB appears to have chosen to include indirect land use effects for biofuels grown from various crops or other feedstocks such as switchgrass, poplar, etc. However, ARB currently appears to be ignoring direct and indirect effects (land use and other effects) for other fuel pathways such as petroleum and electricity. The reasons for this are not clear. One example of direct land use effect for electricity is the use of coal to generate

2 The values for 8 counties in 7 different states range from 8,146 Btu/bu to 31,483 Btu/bu. See the data in Appendix 1, which is consistent with the information presented in Table 4 of the Reference 1.

electricity, where the coal comes from open mining operations in Wyoming and Montana. The ARB report “Detailed California Modified GREET Pathway for California Average Electricity” indicates that 15.4% of the electricity use in California comes from coal- fired facilities located out of the state (this 15.4% accounts for 48.4% of the GHGs from electricity). These coal-fired facilities are located in Nevada and Utah, and very likely use coal that comes from open mines in Wyoming and Montana, where the surface land has been stripped away to reveal the coal for mining. The ARB Electricity report does not discuss these land-use impacts.

Also, according to research conducted by Oak Ridge National Laboratory and others, the building of roads in tropical areas to install petroleum extraction facilities can lead to significant deforestation along the roads as the population expands along the roads and further. [3] Finally, ORNL references work that estimates that the Alberta tar sands operations have resulted in the clearing of 140,000 km2 of land in Canada. [3] To be consistent, these land use effects should be included for other fuel pathways if they are going to be included for biofuels. These issues are not discussed in the ARB report “Detailed California-Modified GREET Pathway for California Reformulated Gasoline Blendstock for Oxygenate Blending (CARBOB) from Average Crude Refined in California.” In order to be consistent with the primary purpose of AB 32 to reduce the potential global warming impact of greenhouse gases, the Low Carbon Fuel Standard regulatory activities should adequately consider all fuel related sources of greenhouse gases.

Finally, we have at least one major concern with the GTAP model that ARB appears to be using to evaluate land use changes. This concern is the fact that the model does not include co-product effects. [5] The primary co-product from dry milling ethanol plants are distillers grains, which are used in various forms as feed for ruminants, and replaces some of the grain used to feed cattle. Inasmuch as the corn used to make ethanol produces feed for cattle, this reduces the land needed to grow corn for cattle. This is discussed further in our comments on co-products.

3. It is assumed that 2% of the fertilizer nitrogen is released to the atmosphere as N2O.

Comments: Michigan State University’s work with the DAYCENT model using location- specific modeling information indicates that the range of nitrous oxide emissions is very large and depends on local soil type, temperature, rainfall and especially management practices. [1,6] It can be essentially eliminated, for example, using cover crops. The 2% value is on the high side of averages that MSU has calculated. The IPCC recommends a rate of 1%. We recommend that ARB use the IPCC rate of 1% instead of an arbitrary 2% rate.

4. It is also assumed that all carbon contained in lime is emitted to the atmosphere as CO2.

The CO2 emission rates from lime depend on the lime application rates. We think the lime application rate is far too high, but in general, the data on lime application rates are not very good. The lime application rate in GREET is 1,202 g/bushel. Work by Kim and Dale have estimated the rate as 32.39 kg/HA, which translates to about 87.4 g/bushel.3 [1]. A common error is to assume that the application rates given in sparse data are yearly values. Actually farmers never apply limestone on a yearly basis. If they apply limestone at all, it is every few years, not yearly. We think ARB’s estimates are far too large.

The lime application rates have a significant effect on WTT energy. With a 1,202 g/bushel lime rate, WTT energy for chemical input for a dry mill from GREET is estimated at 159,380 Btu/mmBtu (Table 2.01 of ARB report). If the lime rates were reduced to 87.4 g/bu as indicated in the Kim and Dale work, the WTT energy would be 119,492, or 25% lower .

5. The primary energy input for a dry mill plant for anhydrous ethanol is 34,889 Btu/gallon, or 457,046 Btu/mmBtu (Table 4.02). The primary energy input for a wet mill plant for anhydrous ethanol is 45,950 Btu/gallon, or 601,945 Btu/mmBtu (Table 4.03).

These energy use values for ethanol plants are obtained from GREET, and may be values based on older plants and surveys. The ARB report does not indicate what this estimate is based on, other than the GREET model.

RFA recently conducted a survey of 22 dry mill and wet mill plants. The survey data were analyzed by Argonne National Laboratory. [7] Average total primary energy use for dry mill plants was 31,070 Btu/gallon, or 410,124 Btu/mmBtu. This is 11% less than the GREET value. Average total primary energy use for dry mill plants was estimated at 47,409 Btu/gallon, or 625,798 Btu/gallon. This is 4% higher than the GREET value. We believe the RFA survey data for dry mills is appropriate, and should be used in the California GREET model. We have reason to believe that the wet mill plants responding to the survey may have been on the high side in terms of energy consumption. Further work is being done on wet mill energy consumption, which will be shared with ARB as soon as it is available.

6. Co-product energy credits for the dry mill are approximately half those of the wet mill, 96,137 vs. 200,986 Btu/mmBtu. It is noted in Appendix A (page 64) that “the weightings for displacing feed corn and soybean meal are different here compared to the original GREET which uses a much higher default co-product credit for dry mills.” The energy credit for wet mills reduces the total WTT energy by 20%.

3 See Table 1 of the reference, where the lime application rate is 32.39 kg/HA, which translates to 87.42 g/bushel at an average yield of 150 bushels per acre and 2.47 acres per HA.

We have at least two concerns with the co-product credits used by ARB. One concern is ARB’s assumption that distillers grains (DGs) replace conventional animal feed on a pound for pound basis (i.e., one pound of DGs replaces 1 lb of combined corn and soymeal). There is evidence that the replacement rate is higher than this; i.e. that 1 lb of DGs replaces more than 1 lb of conventional feed. If this is true, then the energy credit associated with DGs (estimated with the substitution method) is higher than ARB estimates. [8,9]

The second concern is that the GREET model does not include a land use credit for DGs. DGs replace both corn and soy meal utilized in animal finishing yards (feedlots). This replacement should not only have an energy credit, but should also significantly reduce the land area impact of ethanol. As indicated in point 2, this is a shortcoming of the GTAP model, and other land-use impact estimates as well. Therefore, in estimating direct or indirect land use changes due to ethanol, ARB should first estimate the land use credit due to DGs, using information on current and anticipated practices of use.

References

  1. “Life Cycle Assessment of Fuel Ethanol Derived from Corn Grain via Dry Milling,” Kim, S. and Dale, B., Department of Chemical Engineering and Materials Science, Michigan State University, Bioresource Technology 99 (2008), 5250-5260.
  2. Letter from Blake Simmons, et al, to Mary Nichols, June 24, 2008.
  3. “Global Land Use Issues”, Presentation by Keith Kline, et al, Oak Ridge National Laboratory, at the 5th Annual Forum of the California Biomass Collaborative, May 29, 2008.
  4. Kauppi, P.E., et al, “Returning Forests Analyzed with the Forest Identity”, Proc. Nat. Acad. Sci., USA 103, 17574 (2006).
  5. “Biofuels for all? Understanding the Global Impacts of Mulitnational Mandates”, Hertel, Tyner, and Birur, Revised May 1, 2008, Department of Agricultural Economics, Purdue University, GTAP Working Paper No. 51, 2008.
  6. Kim and Dale, “Life Cycle Assessment Study of Bipolymers (Polyhydroxyalkanoates) Derived from No-Till Corn”, International Journal of Life Cycle Analysis, 10 (3) 200-210 (2005).
  7. “Analysis of the Efficiency of the U.S. Ethanol Industry 2007”, Argonne, March 27, 2008.
  8. “Use of Distillers By-Products in the Beef Cattle Feeding Industry”, Klopfenstien, Erickson and Bremer, J Animal Sci, 2008:86:1223-1231.
  9. “Effect of Dietary Inclusion of Wet Distillers Grains on Feedlot Performance of Finishing Cattle and Energy Value Relative to Corn”, Vander Pol, Erickson, et al, Animal Science Department, University of Nebraska/Lincoln, 2006.

Appendix 1

Corn Farming Energy Inputs Source: Reference 1 and Authors

County

Farming, Btu/mmBtu

Farming, Btu/bu*

Hardin (IA)

51695

10608

Fulton (IL)

39696

8146

Tuscola (MI)

98282

20167

Morrison (MN)

95690

19636

Freeborn (MN)

68379

14031

Macon (Mo)

74963

15382

Hamilton (NE)

153426

31483

Codington (SD)

50117

10284

Average

79031

16217

Assumes 76,000 Btu per gallon of ethanol and 2.7 gallons ethanol per bushel of corn

Bob Dinneen
President & CEO
Renewable Fuels Association One Massachusetts Avenue, NW Washington, D.C. 20001

July 15, 2008

Mr. John Courtis
Manager, Alternative Fuels Section California Air Resources Board 1001 “I” Street
Sacramento, CA 95812

Dear Mr. Courtis,

The Renewable Fuels Association (RFA) respectfully submits the attached comments in response to the California Air Resources Board’s land use change workshop held on June 30, 2008.

As the national trade association for the U.S. ethanol industry, RFA appreciates the opportunity to comment on the information presented at the workshop and CARB’s current approach to lifecycle analysis. As you will see in the attached comments, we have questions and comments about the land use models, key assumptions, and fundamental approach CARB is using for its current lifecycle analysis of ethanol and other biofuels.

First, we continue to believe the current scientific, social, and economic understanding of land use change is woefully insufficient. Presentations at the workshop and the ongoing discourse surrounding land use change clearly suggest we are not currently able to estimate land use changes with any degree of certainty. The soundness and effectiveness of a policy framework based on concepts that are not fully understood would most certainly be called into question by stakeholders and consumers alike.

Additionally, we believe the Purdue University Global Trade Analysis Project (GTAP) model requires significant refinement and validation before it can be reasonably used in the development of a policy framework such as the Low Carbon Fuels Standard. The GTAP modeling results presented and discussed at the June 30 workshop clearly demonstrate that the model has substantial limitations and flaws. For example, the GTAP model’s land inventory does not include Conservation Reserve Program (CRP) lands or idle cropland.

Logically, CRP and idle cropland would be the first lands to be brought back into production if future increased crop demands cannot be met solely through increased yield per acre. Why does the GTAP model not include these lands?

Further, we encourage a thorough peer‐reviewed validation of GTAP that uses back‐casting to compare modeling results against real‐world empirical data. It is of great concern to us that such validation has not been performed as part of this process. Many of the assumptions underlying the collective understanding and modeling of global land use change—such as the idea that U.S. corn exports will be drastically reduced, or the idea that U.S. soybean production will be dramatically reduced—have certainly not proven true.

We also question whether current modeling efforts account for the statutory restrictions on certain land use changes included in the 2007 Energy Independence and Security Act. The Act’s “renewable biomass” provision clearly precludes biofuels derived from feedstocks coming from previously forested or non‐agricultural lands. How do GTAP and other models account for this and other regulatory restrictions on land use change?

As stated in our June 27, 2008, comments to CARB, we continue to believe it is important that land use (and other indirect effect) metrics are applied equally to all fuel pathways and that the positive effects of possible indirect effects are also considered.

We sincerely appreciate CARB’s consideration of these comments and look forward to further interaction with the agency on land use change issues. We welcome a further dialog on this subject and look forward to responses to any of the questions offered in the attached comments. We will continue to review information provided by CARB and respond with comments as appropriate.

Sincerely,

Bob Dinneen
President & CEO
Renewable Fuels Association

Comments of the Renewable Fuels Association
On
California Air Resources Board
June 30, 2008, Workshop Presentations and Materials Submitted July 14, 2008

I. Introduction

The California Air Resources Board (ARB) on June 30 held a Low Carbon Fuels Standard (LCFS) workshop on the land use effects of corn‐based ethanol. There were two presentations made on the GTAP model—one by ARB and one by Purdue University, the developers of the model. After the workshop, a spreadsheet was provided on the LCFS website that utilized GTAP modeling results to estimate the carbon release associated with different types of land conversion.

Four scenarios were analyzed by ARB:

Scenario A. Scenario B. Scenario C. Scenario D.

Increase in ethanol from 1.75 bgy to 15 bgy, normal forest/pasture conversion values
Increase in ethanol from 13 bgy to 15 bgy, normal forest/pasture conversion values

Increase in ethanol from 1.75 bgy to 15 bgy, lowest forest conversion values in U.S. and elsewhere
Increase in ethanol from 1.75 bgy to 15 bgy, pasture conversion values for all land in U.S., normal forest/pasture conversion values outside of U.S. (so‐called “CRP case”)

Results indicated a range of land use impacts from 39 grams of CO2‐ equivalent/Megajoule (Scenario D) to 117 g CO2 eq/MJ (Scenario A). ARB indicated that there was much more work to do in a number of areas, and welcomed comments and questions on the models, inputs, methods, and results. Comments and questions were requested by July 15.

We have reviewed the two presentations and the spreadsheet provided. We also have interacted with representatives from both Purdue University and the University of California‐‐Berkeley (UCB). We have a number of comments and questions on the analyses presented at the June 30 workshop. Our major comments are summarized below. Following the major comments, we have included detailed comments and questions on GTAP, and detailed comments/questions on the spreadsheet which utilizes the GTAP outputs.

We should note that we are continuing to evaluate both GTAP and the spreadsheet, and may have additional comments as we learn more.

II. General Concerns with Indirect Land Use Change Analysis

Prior to presenting our major comments and concerns on the materials presented, we want to reiterate an overarching concern we have with the current status of the science of quantifying “indirect” land use changes.

Life‐cycle analysis (LCA) is a standards‐driven procedure for determining the environmental impacts of products and processes. Credible LCA is data dependent. However, there are no data in ARB’s current analysis conducted by Purdue and UC Berkeley on the land use changes that actually occur as more corn is processed to ethanol. These analyses are, in fact, highly speculative and uncertain scenarios for what might happen as a result of increased corn demand.

Even if there were data connecting increased corn demand for ethanol with global land use changes, ethanol produced in the U.S. would not be “responsible” (in a strict LCA sense) for anything but its own environmental profile. “New” corn produced in Brazil by clearing savannah to satisfy animal feed demand is responsible for its own environmental profile as an animal feed, not as an ethanol feedstock. It is arbitrary and unreasonable to make individuals who are producing biofuels responsible for the tenuous, uncertain land use decisions of other individuals many thousands of miles away who are producing animal feed.

This is clearly different from the situation in which tropical wetlands and forests are actually converted to oil palm production to provide oil for biodiesel production. Direct land use change as a result of biofuel production is a legitimate subject for LCA and carries a reasonable level of certainty. In contrast, indirect land use change supposedly caused by biofuel production is tenuous, uncertain and highly speculative.

III. Specific Major Comments and Concerns

Our major comments are listed below and summarized in the paragraphs following the list.

  1. The analyses do not appear to consider the restrictions on land use for biofuels imposed by the 2007 Energy Independence and Security Act (EISA).
  2. The GTAP model requires much further development before it can be reasonably used in this analysis.
  3. The sensitivity case presented on GTAP outputs did not test the sensitivity of the results to the major input assumptions. Consequently, the range of results presented by ARB is not at all indicative of the range of possible results.
  4. Some of the outputs of GTAP are misleading. For example, estimates of forest converted to crops may not be current forests, but forests that have not even

been planted yet. This can lead to serious errors in estimating land conversion emissions

  1. The emission rates for land conversion come from only one source, and it appears those values have not been critically reviewed by ARB for use in this analysis.
  2. There does not seem to be any verification of GTAP predictions. We assumed the results would be verified using other models, back‐tests of GTAP, or empirical data. However, it appears this verification has not occurred.
  3. The ARB analysis to date focuses solely on putative agricultural expansion as a driving force for land use change worldwide. However, there is significant academic literature on land use change, and the actual factors driving land use change are by no means as simple as the scenario being studied by ARB using GTAP.

Based on these and other comments and concerns as presented in the remainder of this document, we conclude that the range of estimates provided by ARB for the land use impacts of corn ethanol – from 39 to 117 g CO2 eq/MJ – cannot be trusted as an initial range of results that will be refined later. Our recommendation is that ARB (1) attempt to validate the GTAP projections with other independent models (FASOMGHG perhaps) and with real data (2) exercise GTAP over a wider range of input assumptions, (3) critically evaluate what the outputs of GTAP mean relative to the base case, (4) and critically evaluate the data and procedures being used to estimate emissions from land converted.

IV. Expansion on Major Comments

1. The analysis does not consider the restrictions on land use for biofuels imposed by the Energy Independence and Security Act (EISA).

Sec. 201(I)(i) of the Energy Independence and Security Act defines “renewable biomass” as: “Planted crops and crop residue harvested from agricultural land cleared or cultivated at any time prior to the enactment of this sentence that is either actively managed or fallow, and non‐forested.”

Based on this language, it is our presumption that feedstocks coming directly from previously forested or other non‐agricultural lands will not qualify for the federal Renewable Fuels Standard program. This provision discourages farmers from clearing forested and non‐agricultural lands and it seems highly unlikely that there will be a market for biofuel feedstocks that don’t meet the “renewable biomass” criteria.

The GTAP modeling results show a substantial amount of U.S. forested land being converted to crop production. Do the GTAP modeling results account for the EISA

“renewable biomass” provision and its likely influence on the farmer’s land use decision‐making process? If not, how does ARB propose to integrate this consideration into its modeling?

2. The GTAP model requires much further development before it can be reasonably used in this analysis

A significant limitation of the GTAP model is that it does not currently include (in its inventory of land) the Conservation Reserve Program (CRP) land in the U.S., nor does it include idle cropland that is not part of CRP. This means that when the model tries to find land for crops, it can only take it from forest and pasture. Scenario D, which ARB and others indicate is a “CRP scenario” because all the land is assumed to be pasture land (no forest), is not really a CRP scenario at all. The reason it is not a CRP scenario is because the amount of land converted in this scenario is estimated from the relative rents and productivities of pasture, forest, and crops. It is highly likely that if Purdue were to input CRP land or idle land into the model appropriately, the amount of land converted would be significantly reduced from Scenario D, thereby reducing the carbon impact. Idle lands fall in this same category. There is good available documentation on the amount of CRP land available in the U.S. We are searching for documentation on the amount and status of idle or unused land.

3. The sensitivity case presented on GTAP outputs did not test the sensitivity of the results to the major input assumptions. Consequently, the range of results presented by ARB is not at all indicative of the range of possible results.

The sensitivity cases presented for the GTAP model focused only on the reasons for the expansion of biofuels in the U.S. – oil prices or public policy measures. The conclusions drawn from these were that the reason for the expansion had some effect, but that it was small. Another conclusion presented was that the “model response was fairly robust – 0.70 hectares per 1000 gallons of ethanol.” However, the analysis ignored four other input factors that would have had a much more significant effect on land use outputs per 1000 gallons of ethanol, as follows:

  1. CRP and idle lands, as discussed above
  2. The impact of different assumptions regarding projected improvements incrop yields
  3. The impact of different assumptions regarding the treatment of distillersgrains (DGs) displacement
  4. The impact of the model’s rent and productivity assumptions for differenttypes of land

No sensitivity analyses were performed with these critical inputs. Consequently, the conclusion that the model is “robust” is false. Much more time should be spent in performing detailed sensitivity analyses of GTAP model outputs to these inputs. We

can provide additional information on these inputs, but first we need to better understand what the assumptions in the model are. For example, we know that GTAP estimates a U.S. coarse grain yield of 151 bushels/acre in the base case, and only about 154 bushels/acre in the expanded ethanol case. We think this increase in yield is far too low for the time period from 2001 to 2015, but we want to understand the various assumptions used to make these estimates before we provide more details.1

4. Some of the outputs of GTAP are misleading. For example, estimates of forest converted to crops may not be current forests, but forests that have not even been planted yet. This can lead to serious errors in estimating land conversion emissions. In fact, since there are no actual carbon emissions from land conversion in such instances, but only carbon capture that is forgone, the way this must be treated from a life cycle basis is completely different. No one‐time emission will be incurred, but rather a distributed reduction in carbon capture over many years.

The GTAP model estimates that in Scenario A, an additional 3.6 million acres of forest will be converted to crops more than would be converted in the base case. However, forest cover has been increasing in the U.S. in spite of federal and state biofuels use requirements. Thus, this estimate of forest converted is not existing forest converted, but “future forests” that have not yet been planted. In estimating the emissions from these converted future forests, UCB has been assuming they are existing forests, resulting in one‐time conversion emissions as well as future uptake emissions. If the estimate is for future forest, then only the future uptake should be counted under the current procedures utilized by UCB. 2 This output is misleading; ARB should clarify how much of this forest is existing forest, and how much is future forest that has not yet been planted.

5. The emission rates for land conversion come from only one source, and do not have appear to have been critically reviewed by ARB for use in this analysis.

The emission rates for lands converted come from only one source (Houghton) and were also used in the Searchinger, et al analysis. These come from the types of lands converted in various countries in the 1990s. It is not clear that this data is appropriate for use for biofuels expansion in the 2001‐2015 time period. Also, some of the methods used to apply these data are very questionable. For example, UCB assumes that for forests, all mass above the ground is converted to CO2. However, nearly all forests are first harvested for lumber, and the lumber is used to build

1 We know that GTAP assumes that the lands coming into corn production may be more marginal lands that may have lower yields than current lands, and this is mitigating some of the improvements. But how the productivities of the various lands are being estimated in GTAP is not clear.
2 We are not agreeing that future uptake should be counted. The counting of future uptake is dependent on the type of forest assumed and the number of years over which it is counted. The number of years is mostly an arbitrary assumption.

houses, furniture, etc. Not all this lumber would be converted to CO2 in 30 years. The researchers appear to have dismissed this as de minimus effect without performing any calculations. Another concern is that all the carbon is attributed in the analysis to biofuels such as ethanol, even though the land is harvested for lumber first, and may serve as pasture second, and finally may be utilized to grow crops. With such multiple uses over time, it does not seem appropriate to allocate the entire one‐time carbon loss (or carbon uptake for 30 years) solely to biofuels.

6. There does not seem to be any verification of GTAP predictions using other models, back‐tests of GTAP, or empirical data.

There is a clear need for verification and validation of the GTAP model results, either using historical data, other models, or both. For example, as noted in point 4, the model predicts that significant forests are converted in expanding from 1.75 bgy to 15 bgy. This should be validated with independent data. Also, the FASOMGHG model results presented by Murray at the EDF workshop on July 2 appear to significantly contradict the GTAP modeling results. Further work should be done to understand why the two models predict vastly different results for the United States.

7. The ARB analysis to date focuses solely on putative agricultural expansion as a driving force for land use change worldwide. However, there is significant academic literature available on land use change. The actual factors driving land use change are by no means as simple as the scenario being studied by ARB using GTAP.

The literature cited at the end of this document indicates that there are a multitude of reasons for land use changes, and it is seldom just one factor as indicated with GTAP. Taken as a whole, since there are a multitude of reasons, it is speculation to attribute significant indirect land use changes solely to increased corn production for biofuels production. At a minimum, these other reasons should be taken into account in the method of allocating the carbon released during a land use change.

These are the seven primary concerns we have at this time. We appreciate the opportunity to comment on this analysis. We will provide additional comments when we learn the answers to our questions.

V. Questions on Dr. Tom Hertel’s Presentation at June 30 Workshop

All references in this section are to the presentation entitled “Implications of U.S. Biofuels Production for Global Land Use” presented June 30, 2008

Page 5: Do these values come from the Energy Information Administration? What is the source of these numbers?

Page 6: The ethanol percent of coarse grain sales rises from 5.6% in 2001 to 35% in 2015, while the animal feed fraction drops from 48.2 to 27.4%. We assume these are gross sales (or similar), and do not take into account co‐products like distillers grains from dry mill corn ethanol production that make up some of the loss in animal feed.

Page 7: We note that coarse grain production rises by 9.8%, and livestock production decreases by 1.5%. Are there GHG credits associated with a reduction in livestock production? We realize that Purdue is not estimating GHGs for ARB, but wonder whether UCB will be including livestock reduction credits.

Page 9: Do we understand Dr. Hertel’s analysis to mean that there is a 17.6% increase in ethanol from sugarcane in Brazil with $115 oil, and a 5% increase in ethanol from sugarcane in Brazil with $60 oil? Why does the price of oil have such a disproportionate effect on Brazilian ethanol production and whether ethanol is imported into the United States?

Page 11: In the U.S., when cropland is increased, how does the model choose between forest and pasture to make up this land? In other words, why doesn’t the model just choose all forest, or all pasture?

Page 13: For the “world” why is the percent increase in cropland lower for the $107 oil than for the $60 oil? Is this because at $107 oil, there is less demand for energy at this price?

Page 15: The top of the slide indicates a 2 bgy increment‐‐ shouldn’t this be a 13.25 bgy increment, as indicated in slide 14?

Page 18: We are not clear how the values at the bottom of this slide are estimated (4.7%, ‐2.6%). Please elaborate.

Slides 23 and 26: What is included in “coarse grains”? For example, in slide 26, the yield is 151 bu./acre. Is this for corn, or is it some average of all “coarse grains”? Since the yields of coarse grains are so different from one another, why average them?

Slide 27: This slide indicates a 2.3% increase in yield of coarse grains. Again, is this all coarse grains, or just corn? We want to understand how you take the time dimension into account in this estimate. We understand that you shock the model in going from 1.75 to 15 bgy. This implies no time change. And yet, in the real world, it is not “shocked” like this with a big demand all at once, but this demand increase occurs between 2001 and 2015, or over 14 years. Over the last 20 years, corn yields have increased at a far faster pace than 2.3% over 14 years. We do understand that you are also taking into account the use of perhaps less productive lands, as well as yield increases on current lands. But on the current more productive lands, what is the assumed yield increase between 2001 and 2015? How are technology

developments that apply to yields accounted for when the model is “shocked”? Finally, when the model is shocked for ethanol, is it also shocked for baseline conditions in 2015 that may be completely different from today (increased food demand, higher oil prices)?

Slide 32: Please explain how the numbers are estimated for “effective land”, “productivity adjust”, and “physical land” in this slide. Also, where do the “rents” come from (data sources, references)? How much variation is there in these rents? How sensitive are the model results to these rents. What correlation is there between potential crop yields from forest and grazing land and their current rents?

Slide 35: We assume that the conclusion that 23 million acres are needed to boost average crop land the equivalent of 5.4 million acres flows from the data on slide 32. This says that if crops are from converted pasture and forest, that we will need over 4 times as much land (23/5.4). Another way of stating this is that if one assumes 151 bushels/acre from current cropland, then the average converted pasture/commercial forest will only yield 38 bu/acre, based on the current rent analysis presented in slide 32. Is there any evidence of corn yields being this low in the U.S. currently? Why would anyone convert pasture or forest to cropland if they expected their yield to be only 25% of the average yield in the nation? We think this is an area where the omission of idle land and CRP land becomes very significant.

Slide 38: This slide indicates that the (land) impacts depend in part on the source of expansion: mandates or oil prices. The slide also indicates that once the focus is on corn‐only ethanol, that the land cover results are “fairly robust across the scenarios”, e.g., 0.7 ha/1000 gallons.” We have two comments on this:

VI.

• •

First, the source of the expansion seems to have very little effect on the results (our conclusion on this is from slides 11, 13, and 15).

Second, the conclusions ignore four other very important determinants of the land impacts: (1) omission of idle land and CRP land in the U.S., (2) effect of estimated projected yield improvements, and (3) GTAPs rent and productivity analysis and its impacts on land use, and (4) the treatment of distillers grains. These factors are much more important than the source of the expansion, and are not tested for “robustness” at all.

Additional GTAP Modeling Questions

Questions in this section are in reference to the presentation entitled “Implications of U.S. Biofuels Production for Global Land Use” presented June 30, 2008, as well to the GTAP model in general.

1. The model does not currently contain CRP land or land that is idle. What effect does this have on the analysis of the type of land converted in the U.S.? What is

the current inventory of CRP land in the U.S.? What is the current inventory of idle land? Does ARB intend to add these types of land to the GTAP model or analysis? Why or why not?

2. Scenario A indicates that in the U.S., 3.6 million hectares of “forest” would be converted to agriculture between 2001 and 2015, when ethanol volumes increase from 1.75 bgy to 15 bgy. By this year (2008), nearly 9 bgy of ethanol will be used, implying conversion of about 2 million hectares of “forest”. Is there independent verification that 2 million Ha of forest has been converted between 2001 and 2008? If so, where has this occurred?

3. Forested areas are actually increasing in the U.S. and have been for a number of years. Is the forest that is estimated by GTAP to be converted existing forest, or is it areas that are planned to be forest under the baseline at some time in the future?

4. What is GTAP’s definition of “forest” from an ecosystem standpoint? Are these primary forests or commercial forests? What is the definition of “pasture”? Does GTAP differentiate between grassland and pasture? Why or why not?

5. In estimating area converted to agriculture from forest and pasture, GTAP adjusts for productivity based on average rents. Where do the rent data come from for the U.S., and what time period is it from? We have the same question for the rest of the nations in GTAP.

6. Also related to the productivity adjustment, what evidence is there that economic rents are the correct method to adjust the acres converted? Does this take into account intensive farming techniques that could be applied to converted forest (advanced seed, irrigation, fertilizer, etc.)? Why or why not? We have the same question with respect to adjustments made to pasture lands. This question extends to areas outside the U.S. as well.

7. We understand that the effect of distillers grains (DGs) from ethanol plants on land use was only added to the model in the last month or so. What are GTAP’s assumptions on the types and mass of feed that 1 lb of DGs replaces? What other assumptions were made to incorporate the DG effect on land use into GTAP? What is the underlying evidence or sources for these assumptions?

8. For Scenarios A‐D, where (geographically) are the predominant areas that the forest is located in the U.S. that GTAP assumes is converted to agriculture? Where in the U.S. are the predominant areas of pasture that GTAP assumes is converted to agriculture?

9. For each of the scenarios, how does GTAP balance or choose between forest and pasture to be converted?

VII. UC­Berkeley Spreadsheet Questions

Questions in this section are in reference to the U.C. Berkeley spreadsheet posted on the CARB Low Carbon Fuels Standard web site following the June 30 workshop.

1. In the plot below, we compare the high forest, low forest, and pasture values used by UCB. What is the reason for the high degree of variability between the high and low forest values (some of the high values are 13 times as high as the low values)? What is the reason why some of the pasture values are higher than some of the low forest values (Europe, Soviet Union, China/India/Pakistan)?

UCB Land Conversion Values

1200 1000 800 600 400 200 0

2. We understand that “grassland” values are being used to estimate emissions from land that GTAP designates as “pasture”. What is the reason for this? We assume that areas designated by GTAP as pasture would be used to graze animals, and would have relatively short grasses, while areas designated as “grasslands” would have much taller grasses. Also, there is evidence that some pastureland was previously planted to crops, leading us to believe that the carbon stored in these lands would be less than in native grassland. Doesn’t this assumption result in overestimating the emissions from converted “pastures”?

3. Why is there so much variation in carbon values of “pasture” between different countries? We note in the plot above that the “pasture” values range from 44 to 199.

Forest-Higher Forest-Lower Pasture

Mg CO2eq/Ha

Canada
Africa

Europe Soviet Union

Latin America
North Africa and Middle East

Developed Pacific China/India/Pakistan

Southeast Asia United States

Rest of World

4. When estimating carbon conversion values, UCB is first estimating the one time conversion of 1⁄4 of the underground carbon and all of the above ground carbon. Then, UCB estimates lost uptake of carbon for the forest (or pasture) for 30 years, and adds the two numbers together to obtain the total carbon loss. Why did the analysis use 30 years? Why not 50 or 100? Why not 5 years? The choice of number of years for uptake losses is arbitrary.

5. Why does UCB include all of the forest above ground as carbon loss? When forests are harvested, much of the aboveground mass is used in building (homes or furniture or other wood products), so it would not be lost for a long time. Why wasn’t this factor taken into account?

6. When forest is cleared, it is harvested for lumber first, generally used for pasture second, and is sometimes converted to crops. Why is all of the carbon associated with clearing a forest being attributed only to biofuels? Shouldn’t some of the carbon also be associated with lumbering and cattle operations?

7. What types of forests are assumed to be converted—are they commercial, primary, or a combination of both? If commercial forest is being converted, isn’t this just forest that would have been converted more slowly anyway? Don’t the conversion rates depend on the Net Ecosystem Exchange (NEE) of the type of forest?

References

1. Proximate Causes and Underlying Driving Forces of Tropical Deforestation, Geist and Lambin, February 2002 Issue of Bioscience

2. The emergence of land change science for global environmental change and sustainability, Turner, Lambin, and Reenburg, PNAS, vol 104, no 52, December 26, 2007

3. Dynamics of Land‐use and Land‐Cover Change in Tropical Regions, Lambin, Geist, and Lepers, Annual review of Environmental Resources, 2003

4. Bioenergy Expansion and Indirect Land Use Change: An Application of the FASOMGHG Model, Murray, presentation at EDF Forum, July 2, 2008

November 21, 2008

Mr. John Courtis
Manager, Alternative Fuels Section California Air Resources Board 1001 “I” Street
Sacramento, CA 95812

Dear Mr. Courtis,

The Renewable Fuels Association (RFA) respectfully submits the attached comments in response to the California Air Resources Board’s workshop held October 16, 2008.

As the national trade association for the U.S. ethanol industry, RFA appreciates the opportunity to comment on the information presented at the workshop and CARB’s current approach to lifecycle analysis and land use change effects. As you will see in the attached comments, we have prepared detailed comments about the land use models, key assumptions, and fundamental approach CARB is using for its current lifecycle analysis of ethanol.

In general, we continue to believe the current understanding of the causes and effects of indirect land use change is woefully insufficient. The ongoing discourse and research surrounding land use change issues clearly suggest we are not currently able to estimate indirect land use changes (particularly international land conversions) with any degree of certainty. The soundness and effectiveness of a policy framework based on concepts that are not fully understood would most certainly be called into question by stakeholders and consumers alike.

Additionally, we continue to believe the Global Trade Analysis Project (GTAP) model employed by CARB for this analysis requires significant refinement and validation before it can be reasonably used in the development of a policy framework such as the Low Carbon Fuels Standard. Our attached comments are quite detailed in this regard, as we continue to gain a better understanding of the model.

Among the major concerns we have with the GTAP modeling used to produce the results presented October 16 are: underestimation of the productivity of converted land in the U.S.; underestimation of average grain yields due to the absence of a factor accounting for technology improvements; underestimation of the significant land use “credit” provided by

distillers grains (the feed co-product of grain ethanol); and omission of Conservation Reserve Program land and idle cropland from the land inventory.

One other particular concern with the GTAP model is that it does not include a time element. To simulate ethanol expansion, the model is “shocked” for a 13.25 billion gallon ethanol increase (simulating the increase in ethanol between 2001 and 2015). The model must “handle” this extreme adjustment instantaneously. In the real world, market conditions change, new technologies are introduced and dynamic adjustments are made every year. In other words, the “shock” is much slower and sufficiently more complex in the real world, with potentially much different effects than simulated by the model.

Further, we have concerns about the data (from Woods Hole Research Center) being used by CARB to estimate carbon emissions from converted lands. The Woods Hole data are derived from research examining Latin American native grassland with relatively high carbon storage rates, but these data are being applied to non-native grassland and pasture with much lower carbon storage rates in the United States. We believe more accurate data on emissions rates from U.S. grassland and pasture is available through Colorado State University and the U.S. Environmental Protection Agency.

As stated in our July 15, 2008, comments to CARB, we continue to believe it is important that indirect land use (and other indirect effect) metrics are applied equally to all fuel pathways and that the positive effects of possible indirect effects (e.g. reductions in enteric methane emissions from livestock due to increased feeding of distillers grains) are also considered.

We sincerely appreciate CARB’s consideration of these comments and look forward to further interaction with the agency as it continues development of the Low Carbon Fuels Standard regulation. We welcome a further dialog on this subject and look forward to responses to any of the comments offered in the attached document. We will continue analyze the GTAP model, review the information provided by CARB, and respond with comments as appropriate.

Sincerely,

Bob Dinneen
President & CEO
Renewable Fuels Association

Comments from the Renewable Fuels Association
to California Air Resources Board
Regarding October 16 Workshop Materials and GTAP Model

November 21, 2008

On October 16, 2008, the California Air Resources Board (CARB) released a draft regulation for the California Low Carbon Fuels Standard (LCFS) and a document entitled “Supporting Documentation for the Draft Regulation for the California Low Carbon Fuels Standard.” Our comments are primarily focused on information presented in the supporting documentation report.

Our main comments focus on CARB’s current estimates of greenhouse emissions resulting from land use changes (LUC) due to corn ethanol expansion. CARB’s analysis of LUCs for corn ethanol is contained in Appendix A of the supporting documentation report. Basically, CARB ran the Global Trade Analysis Project (GTAP) model through a number of different sensitivity cases using various elasticities to estimate a range of land use change impacts. GTAP was used for estimating land use changes and the locations of those changes, and the Woods Hole data was used to estimate emission rates for converting different types of land (e.g., forest vs. grassland). The land use change estimates ranged from 20 to 88 g CO2 eq./MJ, with a median estimate of about 35 g CO2 eq./MJ. We note that this represents a factor of more than 4X between the low and high estimate.

We still have a number of concerns with how the GTAP modeling is being conducted, and also with certain applications of the Woods Hole emissions data. These concerns are summarized below, and subsequently expanded upon.

  1. CARB likely underestimates the productivity of land being converted to crops in the United States (i.e. “marginal” land).
  2. Due in part to item 1, and considering the fact that there is no factor to account for observed and future technology improvements in yield independent of price, the projected crop yields are too low in the most recent GTAP analysis. Because the model is “shocked” with 13.25 billion gallons of new ethanol production instantaneously, and yield values do not take into account the improvement in yields between 2000 and 2015, the model is converting too much land to crops as a result.
  3. The GTAP model may not be accounting for natural declines in wheat and cotton in the U.S. expected between 2001 and 2015. Empirical data indicates lost production of wheat and cotton in the United States over the past several years has not entirely been made up for in other locations.

1

  1. The above three factors cause exports of corn and soybeans to decline significantly in the modeling. Empirical data shows exports have not declined in the period from 2001 to 2007.
  2. The distillers grain (DG) land use “credit” being used in the GTAP modeling is likely too low and needs to be modified, taking into account the recent analysis of DG feed displacement performed by Argonne National Laboratory.
  3. The land conversions in GTAP do not adequately take into account the economic cost of converting forest and native grasses to cropland.
  4. There does not appear to be Conservation Reserve Program land or idle cropland in the GTAP database used for the analysis described in the October 16 documentation.
  5. Woods Hole data for native grassland with high carbon storage rates are being used to estimate emissions from non‐native grassland and pasture in the U.S. with lower carbon storage rates.
  6. Emissions for forest area assume all mass above ground is converted to CO2 immediately, when some is likely to be used in building products that would not be converted for a long time.

These concerns are expanded upon below.
Comment 1: CARB likely underestimates the productivity of land being converted to

crops in the United States (i.e. “marginal” land).

CARB refers to this factor as the “elasticity of crop yields with respect to area expansion.” CARB indicates that “although this is a critical input parameter, little empirical evidence exists to guide the modelers in selecting the appropriate value. Based on the judgment of those with experience in this area, the modelers selected a value of 0.66. For purposes of the sensitivity analysis this parameter was varied from 0.25 to 0.75. This input variable produced by far the greatest variation in the output GHG variable: 77%.”

When CARB varied this parameter from 0.25 to 0.75, the GTAP model produced the two extremes in LUC emissions, 88 and 20 g CO2 eq./MJ (the price‐yield elasticity was held at 0.4 for this sensitivity analysis).

RFA believes there is empirical data to guide the selection of this important parameter, especially for the U.S. Through our analysis of land use patterns, it has become evident that land devoted to wheat and cotton in the U.S. is declining somewhat, and corn is replacing these crops in some of these areas. In addition, corn‐on‐corn cropping systems are increasingly replacing traditional corn‐soybean

2

rotations. Literature suggests the corn‐corn pattern does involve a modest decline in corn yields from a corn‐soybean system, but the expected decline for this rotation is not in the range of 25‐75%. Finally, farmers may convert some idle land or cropland pasture to corn. Many farmers will crop land for a given period, and then convert it to pasture or fallow the land to regain nutrients and carbon. When the land is re‐cropped after fallowing, yields tend to rise.

To evaluate the potential yield of corn replacing cotton and wheat, we examined USDA corn yield data for states with the highest cotton and wheat output. The corn yields in these states were a volume‐weighted average of 20% below the corn yields in the top 10 corn producing states. The details of this analysis will be described in a forthcoming land use change report by Air Improvement Resource (AIR). As a result, we believe that there is data available in the U.S. that indicates the elasticity of crop yields with respect to area expansion should be 0.8 or higher.

We have not found data for areas outside of the U.S., but that is a different matter. One of the major flaws with the current GTAP model is that it applies the same expansion elasticity to all regions, all agricultural ecological zones (AEZs) within a region, and all crops. This is a parameter that should be input by region, AEZ, and crop (e.g., coarse grains should have a different elasticity value than oilseeds).

Comment 2: Due in part to the issues described in Comment 1, and considering the fact that there is no factor to account for observed and future technology improvements in yield independent of price, the projected crop yields are too low in the most recent GTAP analysis.

The GTAP model used for the October 16 report is based on a 2000/2001 database. To simulate ethanol expansion, the model is “shocked” for a 13.25 billion gallon ethanol increase (simulating the increase in ethanol between 2001 and 2015, for example). The model must “handle” this extreme adjustment instantaneously, while in the real world, conditions change every year and dynamic adjustments are made every year. In other words, the “shock” is much slower in the real world, with potentially much different effects than simulated by the model.

Nevertheless, the model outputs the change in yield for different crops in response to the shock. This yield is a function of two factors: the elasticity of crop yields discussed in comment 1, and the price‐yield elasticity. CARB ran a sensitivity analysis of the price‐yield elasticity, with values ranging from 0.6 to 0.1, while the elasticity of crop yields was fixed at 0.5. In this analysis, LUC impacts varied from 29 to 57 g CO2 eq./MJ, not as sensitive as the elasticity of crop yields, but still quite sensitive. The higher value (0.6) would indicate a higher response of crop yields to crop prices. For its pending report, AIR examined the yield increases before and after the shock, and compared these yields to historical and projected yields obtained from USDA for the time period from 2000‐2001 to 2015‐2016, which the model is trying to represent. The results are shown in the figure below.

3

170
165
160
155
150
145
140
135
130
125

U.S. Coarse Grain Yield, USDA Corn vs GTAP (USDA: Actual through 2007, Projected beyond 2007) (GTAP: Projected beyond 2001)

USDA (

ctual/Proj

GTAP, 0.6 Price Yield

ected)

A

2000 2002 2004 2006

2008 2010

Year

2012 2014 2016

Air Improvement Resource, Inc.

Note: 2001‐2007 USDA yield plots are actual recorded values. 2008‐2015 yield plots are USDA projections from “Agricultural Long Term Projections to 2017”

Analysis of GTAP output shows that for this scenario, yield values increase by 3.27% in the production region defined as “U.S.” The base yield is 138 bu./acre, so a 3.27% increase is 4.5 bu./acre, and, thus, the expected 2015‐16 yield in the U.S. is 142.5 bu./acre. This is far too low, as USDA historical yields for the 2004‐2007 time period are much higher (in the 150+ bushel/acre range). USDA’s projections to 2015‐16 show a yield of approximately 170 bu./acre, or 20% higher than the GTAP 2015‐16 yield value generated by the 13.25 billion gallon ethanol shock. This underestimation of yield in GTAP results in much more land being converted than is likely to be the case.

Part of the reason the GTAP yields stay low in the U.S. under this scenario is because the elasticity of crop yields with area expansion is set to 0.5. To evaluate only the price‐yield effect, we reset the elasticity of crop yields to area expansion to a value of 1.0, left the price‐yield elasticity at 0.6, and ran the 13.25 billion gallon shock through GTAP to examine the coarse grain yield increase in the U.S. The results show a coarse grain yield increase of just 3.9%, from 138 bu./acre to 143.4 bu./acre. This is still far below the USDA projection, and a source for significant concern.

4

Yield (Bushels/Acre)

One conclusion from this is that the price‐elasticity function does not explain all of the yield increases that are anticipated. The model is shocked, coarse grain prices increase somewhat, and the elasticity function predicts a slightly higher yield (but not enough). We believe there is a technology factor in yield that is not necessarily explained with price. This would mean that either the price‐yield elasticity value needs to be increased to explain this technology driver, or perhaps a separate factor should be added that would be a technology driver. Either way, the current yield increases in the U.S. being modeled by GTAP on the 13.25 billion gallon ethanol shock are far too low, as demonstrated by actual average yields from the past four years and the projected yield for 2008 of 153.8 bu./acre.

We did try to increase the yield in GTAP by setting the yield expansion elasticity to 1.0 and increasing the price yield elasticity well above 0.4 or 0.6. However, the model applies this price‐yield elasticity to every crop in every region. The GTAP model should allow the user to apply different improvements to different crops and different regions. We are attempting to program this characteristic into GTAP so that we can vary price yield elasticity by crop (e.g., oilseeds vs. coarse grains) in the U.S.

Comment 3: The GTAP model may not account for reductions in wheat and cotton in the United States.

This issue is based on analysis of trends, just like the previous issue. Information from USDA and other sources indicates that land devoted to cotton and wheat in the U.S. has been declining over the long term, due to a reduction in the demand for wheat (along with productivity improvements), reduction in the demand for cotton, and a shift from cotton growing in the U.S. to some being grown in China and India. Since the GTAP model starts with a 2000/2001 database, and the model is shocked for 13.25 billion gallons, the model may not be appropriately accounting for this change. The model appears to assume that the demand for cotton and wheat are essentially constant, and is therefore forced to make up the loss in these crops elsewhere.

Comment 4: The three factors described in Comments 1­3 cause exports to decline significantly in the modeling.

Since the factors discussed in comments 1 and 2 result in yields that are too low for the U.S., and the situation described in comment 3 may not be not properly accounted for, U.S. exports drop significantly on the shock, and the regions outside of the U.S. must make up for the drop in exports. These regions do so by converting land to coarse grains and other crops. However, since yields are lower outside the U.S., more land is converted to meet these shortfalls than would be converted inside the U.S. For this reason, it is very important that GTAP model the U.S. situation as accurately as possible with respect to land elasticity and price‐yield elasticity.

5

Comment 5: The distillers grain (DG) land use credit is too low and needs to be modified, taking into account the recent analysis of this issue performed by Argonne National Laboratory.

The GTAP report “Biofuels and their Byproducts: Global Economic and Environmental Implications” (June 2008) indicates that DGs are being modeled as a substitute for coarse grains (see flow diagram on page 12 of the GTAP report) in the livestock sectors of the model. GTAP is using an elasticity of substitution of .30 between coarse grains and DGs. This value was selected by examining the price changes of coarse grains and DGs over the time period of 2001‐2006 when ethanol production was rising sharply. Results of simulations with and without coproducts indicate that incorporating these effects reduces the increase in the demand for corn land from 9.8% to 6.3%, a reduction of 36%.

A recent report by Argonne National Laboratory on the use of ethanol co‐products in all livestock sectors indicates that 1 lb. of DGs replace around 1.28 lbs. of base animal feed, Of the feed replaced, 0.96 lbs. is corn and 0.29 lbs. is soy meal.1 There are two important implications for GTAP in the Argonne report. One is that the GTAP model should be modified so that DGs replace not only coarse grains, but also replace some amount of oilseed meal (in the livestock section of the model). Since soybean yields are lower per acre than corn yields, this will have significant land use implications. In other words, referring to page 12 of the GTAP report referenced above, the oilseed part of the feed model should be modified in a similar way as coarse grains were for byproducts. Then, the model will have to allocate a portion of the DGs to coarse grains and oilseeds, according to the allocations developed by Argonne.

The second implication of the Argonne work is that DGs replace base feed on a greater than 1‐to‐1 basis. It appears this fact is not being included in the GTAP model simply by evaluating historical data of the elasticity of substitution between coarse grains and DGs. Therefore, some factor will need to be incorporated into GTAP for this relationship as well.

We estimated the impacts of the Argonne work on land use changes using inputs from the California GREET report for corn ethanol.2 The report indicates that the DG yield per gallon of anhydrous ethanol is 6.4 lbs. Assuming 151 bu./acre (USDA value for 2007), and 2.6 gal/bu. (GREET input), this results in 2,513 lbs. DGs per acre. The Argonne co‐products report indicates that this amount of DG will replace 3,217 lbs. of feed, consisting of 2,445 lbs. of corn meal and 772 lbs. of soy meal. Again using USDA’s corn and soy yields for 2007 of 8,456 lbs./acre for corn (151 bu./acre * 56 lbs./bu.) and 2,502 lbs. per acre for soy (42 bu./acre and 44 lbs. of soy meal/bu.),

1 “Update of Distillers Grains Displacement Ratios for Corn Ethanol Life-Cycle Analysis,” Arora, Wu, and Wang. Argonne National Laboratory. September 2008.
2 “Detailed California-Modified GREET Pathway for Denatured Corn Ethanol,” Stationary Sources Division, ARB, April 21, 2008.

6

the corn acres replaced are 0.29 acres, and the soy acres replaced are 0.42 acres, for a total of 0.71 acres replaced by the DGs produced from making ethanol from one acre of corn.3 Thus, 71% of the acres devoted to ethanol are replaced by the resultant DGs. This is significantly higher than the current GTAP assumption of about 36%. Most of this difference is due to the fact that GTAP is not currently assuming that DGs replace any soy meal.

Comment 6: The land conversions in GTAP may not adequately take into account the cost of converting forest and grasses to cropland.

The land conversions between cropland, pasture and forest are governed at least in part by the elasticity of land transformation across cropland, pasture, and forestry. This value “was set to the relatively low value of 0.2, based on historical evidence for land cover change in the U.S. over the 1982‐1997 period,” according to the supporting documentation. We are not sure that this value properly evaluates the costs of converting land from forest to crops and from grass to crops. Research conducted by Colorado State University for the U.S. EPA in estimating conversion of land to cropland in the U.S. indicates that most of the land converted in the last decade to crops in the U.S. has been non‐native grassland such as pasture or fields that have been idled, and not forest or native grassland. 4 CARB’s “Scenario A” in Appendix A indicates that GTAP expects that 40% of the land converted in the U.S. to be forest, and 60% to be pasture. Other scenarios in this appendix indicate a range of 31% to 50% forest converted. We will be providing further information on forest conversion in the forthcoming AIR land use report.

Comment 7: There does not appear to be CRP land or Idle Land in the GTAP database.

In our comments on the previous workshop (June 30, 2008), we indicated that CRP land and idle land should be included in the GTAP model land use database. To our knowledge, this has not yet been done, but we understand CARB, U.C.‐Berkeley, and Purdue University may still be working on this.

This issue is important because it affects the mix of land converted to crops. Idle land and CRP land are both areas of land that previously grew crops. If this land is not available in the model, then the model will instead convert forest, pasture, and other crops to corn. The inappropriate conversion of forest will raise emissions. The inappropriate conversion of pasture will cause a false reduction in livestock output.

3 Note that in this estimate, we have estimated that 100% of the corn is converted to corn meal, but 73% of the soybean bushel of 60 lbs. is converted to soy meal because 26% of the mass has been extracted in the form of soy oil and other materials. (Source: Chicago Board of Trade “Soybean Crush Reference Guide”). Also, the ethanol yield of 2.6 gal./bu. may be too low – two recent studies of ethanol processing efficiencies indicate that the yield may be between 2.7 and 2.8 gal./bu. This would increase the DG land credit from 71% to 77%. (Sources: “Analysis of the Efficiency of the U.S. Ethanol Industry in 2007”, May Wu, Argonne, March 27, 2008; and “U.S. Ethanol Industry Efficiency Improvements, 2004 through 2007”, Christianson and Associates, August 5, 2008)

4 Personal Communication with Dr. Steve Ogle, Colorado State University, November 14, 2008. 7

The inappropriate conversion of other crops will mean that production needs to be made up elsewhere, when this is not likely the case.

A good source of data on idle cropland is the 2003 National Resources Inventory (NRI). 5 This data source is also used by the Colorado State University CENTURY model mentioned earlier. The table below shows trends in cultivated and non‐ cultivated cropland. CRP land, pasture land, range land, and forest land are separate from these categories in the NRI.

Cultivated and non­Cultivated Cropland by Year (millions of acres)

Year

Cultivated

Non‐cultivated

Total

1982

375.8

44.1

419.9

1992

334.3

47.0

381.3

1997

326.4

50.0

376.4

2001

314.0

55.5

369.5

2003

309.9

58.0

367.9

These data show that the agriculture industry had 58 million acres of non‐cultivated cropland in 2003. It is unclear whether this land is part of the GTAP land inventory for the U.S., but based on the modeling results it seems unlikely. Much of the non‐ cultivated cropland would be utilized for expansion of crops before forest or native grass is converted.

Comment 8: Woods Hole Research Center data for native grassland with high carbon storage rates are being used to estimate emissions from non­native grassland and pasture in the U.S. with lower carbon storage rates.

The emissions rate for grassland converted to cropland being used in GTAP is a value of 110 Mg CO2 eq./Ha. This comes from the Woods Hole data, and was developed in Latin America for natural or native grassland in that region. 6

ARB is currently applying this rate of 110 Mg CO2 eq./Ha to conversion of all grassland in the U.S. and elsewhere, whether it is native grassland, pasture, or idle farmland. However, it is inappropriate to apply this emission rate to U.S. pasture or idle farmland. Native grassland, since it has been undisturbed for perhaps hundreds of years, would store much more carbon than pasture and idle farmland.7 And, it is

5 2003 Annual NRI – Land Use, USDA.
6 “Changes in the Landscape of Latin America Between 1850 and 1985 II. Net Release of CO2 to the Atmosphere”, R.A. Houghton, et al, Forest Ecology and Management, 38 (1991). This study indicates that 10 Mg of C/ha is above ground for grassland, and 80 mg of C/Ha is below ground, and that by conversion of the land, 25% of the root carbon is released (10+25%*80 = 30 Mg/ha). This is then converted to CO2 by multiplying by the ratio of molecular weights of CO2 to C (3.67).
7 Personal Communication with Dr. Steve Ogle, CSU, November 14, 2008.

8

very unlikely that widespread conversions of native grassland are taking place in the U.S. Thus, a different emissions rate must be used for grassland conversion in the U.S., and for pasture conversions outside the U.S.

The Colorado State University (CSU) CENTURY model was used to estimate the emissions from converting land to cropland for the most recent EPA Greenhouse Gas and Sinks Report. 8 According to CSU, most of this land converted was grassland. Using information in various Annexes to this report which show total emissions and total land converted, the average emission rate is about 16 Mg CO2 eq/Ha. This is far less than the 110 Mg CO2 eq./Ha being used by CARB. Our review of the EPA report indicates that this is a much more detailed and better method of estimating carbon releases from land conversions in the U.S. than using estimates for native grassland in tropical areas. It should also be used for pasture conversions outside of the U.S., since these are also not “native grasslands.”

Comment 9: Emissions for forest area assume all mass above ground is converted to CO2.

The emission rates being used for forest converted in the model assume that all forest is converted to CO2. In reality, much of the forest mass is harvested before conversion. Some of this mass is used to produce furniture or to build houses and other products, where it would not be converted to CO2 for many years. ARB should subtract some mass from forest conversion for these products. AIR is evaluating data on these fractions and will supply what we have a later date.

Conclusion

This concludes our comments at this time. We are continuing to evaluate GTAP and emissions rate data for land conversion from different sources. We will have more specific comments on GTAP in the near future. We also continue to review other sections of the draft LCFS regulation ad supporting documentation and may have comments on other aspects of the pending regulation in the near future.

8 “Inventory of U.S. Greenhouse Gases and Sinks: 1990-2006”, USEPA, April 15, 2008. 9

ATTACHMENT 2
RFA COMMENTS
EPA ANPR REGULATING GREENHOUSE GAS EMISSIONS UNDER THE CLEAN AIR ACT
NOVEMBER 28, 2008

For Immediate Release: June 24th 2008

Contact:
Blake A. Simmons, Ph.D. PH: 925 337 6154 Manager, Energy Systems Department
Sandia National Laboratories
Livermore, CA 94551

Scientists across the nation counsel California to regulate Low Carbon Fuel Standard on hard data only and not on speculative indirect impacts where no data or consensus exists.

Citing a severe lack of hard empirical data, an assembly of scientists and researchers across the nation urged California to adopt a Low Carbon Fuel Standard based on commonly understood direct impacts on carbon dioxide emissions and further study highly controversial and speculative “indirect land use changes” before incorporating any of these indirect impacts in that standard.

In a letter sent to the Chair of the of the Air Resources Board, the scientists and researchers state: “Given that our only options for sustainably powering transportation with a significant reduction in transportation related greenhouse gas emissions are biofuels, batteries, and hydrogen, a presumptive policy implementation based on the current understanding of indirect impacts will have a significant chance to hurt real progress on reducing carbon emissions and decreasing our reliance on fossil fuels.

We propose that a sound policy approach would be to base the initial LCFS on existing data sets that possess scientific consensus. These include the direct impacts of renewable biofuels production. The scientific and economic communities can then take advantage of the necessary time over the next five years to fully understand, gather, and validate the indirect impacts of biofuels production with empirical evidence that will enable the implementation of a sound policy that can address any indirect impacts.”

The scientists and researchers, signing on their own and not on behalf of their institutions, include scientists from Universities and National Laboratories across the nation. The full text of the letter appears below.

Mary D. Nichols, Chairman California Air Resources Board 1001 “I” Street
P.O. Box 2815

Sacramento, CA 95812 June 24, 2008

Dear Chairwoman Nichols,

We are writing regarding the California Air Resources Board’s (ARB) ongoing development of the Low Carbon Fuel Standard (LCFS). As you are well aware, the Governor issued Executive Order S-1-07 on January 18, 2007, which calls for a reduction of at least 10 percent in the carbon intensity of California’s transportation fuels by 2020.

As researchers and scientists in the field of biomass to biofuel conversion, we are convinced that there simply is not enough hard empirical data to base any sound policy regulation in regards to the indirect impacts of renewable biofuels production. The field is relatively new, especially when compared to the vast knowledgebase present in fossil fuel production, and the limited analyses are driven by assumptions that sometimes lack robust empirical validation.

As an example of the confusion that this lack of reliable data produces, there has been significant attention to a recent article by Searchinger and coworkers in Science Express (“Use of U.S. Croplands for Biofuels Increases Greenhouse Gases through Emissions from Land Use Change,” February 7, 2008). This article attempted to address the issues of fuel ethanol’s effects on greenhouse gas (GHG) emissions by including GHG emissions from potential land use changes arising from ethanol production. It has prompted a large response from the scientific community, pointing out apparent errors and/or gaps in the analysis presented. For example, Searchinger et al. estimated that U.S. corn ethanol production (between 15 billion and 30 billion gallons) would result in a requirement for an additional 10.8 million hectares of crop land worldwide; 2.8 million hectares in Brazil, 2.3 million hectares in China and India, and 2.2 million hectares in the United States, with the remaining hectares in other countries. Searchinger et al. maintain that the United States has already experienced a 62% reduction in corn exports. In reality, U.S. corn exports have remained relatively constant at around 2-billion-bushels-per-year since 1980. In 2007, when U.S. corn ethanol production increased dramatically to approximately 6 billion gallons, corn exports increased to 2.45 billion bushels — a 14% increase from the 2006 level (excerpt taken from Wang’s response to Searchinger, 2008). Searchinger also ignored the fact that the protein in corn still goes on for use as cattle feed as it cannot be converted to ethanol, with the result that there is no reduction in protein available for feeding animals, the major (about 60%) market for corn.

The traditional tools used by researchers, including Searchinger et al., to determine the direct and indirect impacts of renewable biofuel production are life cycle analysis (LCA) coupled with land-use change (LUC) projections. The results produced by the majority of the LCA models are

highly sensitive to LUC assumptions, as well as baseline projections and test cases that have very limited scope. These sensitivities highlight how common LCA models can be applied to the same problem but produce significantly different, and often contradictory, results. There remain great uncertainties and challenges in combining LUC and LCA models that make their use highly problematic, particularly if the outputs of these models are used as a basis for policy decisions, or for comparing indirect impacts between fuel types. Some of the problems include the lack of large-scale, reliable data sets from field and process trials of growing, harvesting, and converting dedicated energy crops into biofuels. These data are needed as “training sets” for the LCA models. Moreover, without validation of the results produced by the LCA models, they should not be considered as based in fact, but rather based on statistical correlations. Thus it is extremely difficult to make a comparison of the direct and indirect impacts between fossil fuels and renewable biofuels.

Significant research is still required to develop reliable data training sets and validated LCA tools that can accurately guide policies such as the LCFS. Renewable biofuels remain a relatively new field of study with significant gaps in our current understanding that will only be filled with research over an extended period of time. Given that our only options for sustainably powering transportation with a significant reduction in transportation related greenhouse gas emissions are biofuels, batteries, and hydrogen, a presumptive policy implementation based on the current understanding of indirect impacts will have a significant chance to hurt real progress on reducing carbon emissions and decreasing our reliance on fossil fuels. We propose that a sound policy approach would be to base the initial LCFS on existing data sets that possess scientific consensus. These include the direct impacts of renewable biofuels production. The scientific and economic communities can then take advantage of the necessary time over the next five years to fully understand, gather, and validate the indirect impacts of biofuels production with empirical evidence that will enable the implementation of a sound policy that can address any indirect impacts.

It is clear that building a LCFS is a significant undertaking. Many states and countries will look to this regulation as a template for reducing the impact of transportation fuels in other parts of this country and overseas. It is therefore critical that we keep the underlying need for innovation in mind, and base the LCFS upon data obtained from robust and mature tools and empirical validation.

We are writing this letter as researchers in the field of biomass to biofuel conversion, but do not represent the official views of the Department of Energy, the United States Department of Agriculture, or the National Laboratories.

Thank you in advance for your consideration of this important issue.

Sincerely,

Blake A. Simmons, Ph.D.

Manager, Energy Systems Department, Sandia National Laboratories, Livermore, CA and Vice-President for Deconstruction, Joint BioEnergy Institute, Emeryville, CA

Jay D. Keasling, Ph.D.

Professor, Departments of Chemical Engineering and Bioengineering, University of California, Berkeley, CA
and Director, Physical Biosciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA

and Chief Executive Officer, Joint BioEnergy Institute, Emeryville, CA

Harvey Blanch, Ph.D.

CSTO, Joint BioEnergy Institute, Emeryville, CA
and Merck Professor of Biochemical Engineering, Department of Chemical Engineering, University of California, Berkeley, CA

Paul D. Adams, Ph.D.

Deputy Division Director, Physical Biosciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA

Todd W. Lane, Ph.D.

Member of the Technical Staff, Sandia National Laboratories, Livermore, CA

Christopher Shaddix, Ph.D.

Sandia National Laboratories, Livermore, CA

William J. Orts, Ph.D.

Research Leader, Bioproduct Chemistry & Engineering, USDA-ARS-WRRC, Albany, CA

R. Michael Raab, Ph.D.

Massachusetts Institute of Technology, Department of Chemical Engineering, Cambridge, MA

Brad Holmes, Ph.D.

Sandia National Laboratories, Livermore, CA

Rick Gustafson, Ph.D.

Denman Professor of Bioresource, Science and Engineering, College of Forest Resources, University of Washington, Seattle, WA

Lonnie Ingram, Ph.D.

Distinguished Professor of Microbiology, University of Florida
and Director of the University of Florida Center for Renewable Chemicals and Fuels, Institute of Food and Agricultural Sciences, Gainesville, FL

Mohammed Moniruzzaman, Ph.D.

VP Research & Development, BioEnergy International, Quincy, MA

Masood Hadi, Ph.D.

Member of the Technical Staff, Sandia National Laboratories, Livermore, CA

David Reichmuth, Ph.D.

Member of the Technical Staff, Sandia National Laboratories, Livermore, CA

Swapnil Chhabra, Ph.D.

Research Scientist, Lawrence Berkeley National Laboratory, Berkeley, CA
and Adjunct Professor, Department of Bioengineering, University of California, Berkeley, CA and Vice President for Technology, Joint BioEnergy Institute, Emeryville, CA
and Head, Berkeley Center for Structural Biology, Berkeley, CA

Bruce E. Dale, Ph.D.

Distinguished University Professor, Dept. of Chemical Engineering & Materials Science, Michigan State University, East Lansing, MI

Charles E. Wyman, Ph.D.

Ford Motor Company Chair in Environmental Engineering, Center for Environmental Research & Technology (CE-CERT), Riverside, CA
and Department of Chemical and Environmental Engineering, Bourns College of Engineering University of California, Riverside, CA

and Adjunct Professor of Engineering at Dartmouth College, Hanover, NH

Stephen R. Kaffka, Ph.D.

Co-Director, California Biomass Collaborative
and Extension Agronomist, Department of Plant Sciences, University of California, Davis, CA

Mike Henson, Ph.D.

University of Massachusetts, Amherst, MA

Keith Kretz, Ph.D.

VP, R&D Operations and Services, Verenium Corporation, San Diego, CA

Jeffrey L. Blanchard, Ph.D.

University of Massachusetts, Amherst, MA

Randolph T. Hatch, Ph.D.

President, Cerex, Inc., Wellesley, MA

Susan Leschine, Ph.D.

University of Massachusetts, Amherst, MA

Ken Copenhaver

Program Director, Institute for Technology Development, Savoy, IL

Dean Dibble

Member of the Technical Staff, Sandia National Laboratories, Livermore, CA

Seema Singh, Ph.D.

Member of the Technical Staff, Sandia National Laboratories, Livermore, CA

Rajat Sapra, Ph.D.

Member of the Technical Staff, Sandia National Laboratories, Livermore, CA

ATTACHMENT 3
RFA COMMENTS
EPA ANPR REGULATING GREENHOUSE GAS EMISSIONS UNDER THE CLEAN AIR ACT
NOVEMBER 28, 2008

ATTACHMENT 4
RFA COMMENTS
EPA ANPR REGULATING GREENHOUSE GAS EMISSIONS UNDER THE CLEAN AIR ACT
NOVEMBER 28, 2008

NOVEMBER 28, 2008
EPA ANPR REGULATING GREENHOUSE GAS EMISSIONS UNDER THE CLEAN AIR ACT

Sarah Dunham, Director
Office of Transportation and Air Quality Transportation and Climate Division U.S. Environmental Protection Agency Ariel Rios Building
1200 Pennsylvania Avenue, NW
Mail Code: 6401A
Washington, DC 20460