Presentation of Laura Barcellos Antoniazzi for the “Workshop on the Impact of New Technologies on the Sustainability of the Sugarcane/Bioethanol Production Cycle” Apresentação de Marcos Laura Barcellos Antoniazzi realizada no “Workshop on the Impact of New Technologies on the Sustainability of the Sugarcane/Bioethanol Production Cycle” Date / Data : May 14 - 15th 2009/ 14 e 15 de maio de 2009 Place / Local: ABTLuS, Campinas, Brazil Event Website / Website do evento: http://www.bioetanol.org.br/workshop3

1. ICONE Brazilian Land Use Model
and its application for ethanol impacts
Laura Barcellos Antoniazzi
Researcher, ICONE
(lantoniazzi@iconebrasil.org.br)
Sao Paulo, May 15th 2009

2. Outline
1. Introduction.
2. General structure of the land use model
 Diagrams explaining the information flows, connections among variables and
effect-cause relations.
 Assumptions.
3. Summarized results.
 Historic data.
 Supply and demand projections for Brazil.
 Area allocation at regional level.
 Improvements under development
3. Applications
2

3. Introduction
 There are many concerns worldwide about the social and environmental impacts
of biofuels productions and expansion.
 Land Use Change (LUC), as well as Indirect Land Use Change (ILUC), are now
being taking into account in many policies aiming to promote biofuels.
 ICONE started to work on an Agriculture Projection Model in the beginning of 2008
in a joint effort with the Food and Agricultural Policy Research Institute (FAPRI),
from Iowa State University, which is part of the Center for Agriculture and Rural
Development (CARD).
 Although the Brazilian model follows the same structure of the FAPRI Models, it
was adapted to the specific conditions and situations of the Brazilian
agricultural sector.
 The Model aims to capture the Brazilian land use dynamics observed in the past
and to forecast new dynamics
3

4. Agricultural Land use in Brazil
(Agricultural census)
P = Preliminary
Source: IBGE

5. General structure of the land use model
 The model comprises two general sections: supply and demand and land allocation.
 Supply and demand, for a given year, is calculated at a national level. Supply includes
production and initial stocks (in the case of crops), and demand includes domestic
consumption, net trade and final stocks.
 Supply (regional) = Demands (national)
 Land allocation calculations are integrated to the supply side. Area is calculated at a regional
level, as a function of the expected market profitability of the product and of the competing
products.
 The amount of land allocated for a given crop, in a given region, depends on the response to
expected market profitability, which means that regions with higher market returns will have
higher planted area.
 Allocation of land across regions Brazilian production national supply
 The calculation of the expected planted area is also calibrated with the lagged area (area of
the previous year) in order to avoid strong oscillations in the planted area.
Confidential. Do not quote or cite unless authorized. 5

6. General structure of the land use model
 Planted forests are also included in the area allocation section of the model. For the version
we are using now, projections of land allocation for planted forest are exogenous.
 The model comprises 6 macro-regions in Brazil:
 Region 1: South (States of Rio Grande do Sul, Santa Catarina e Parana);
 Region 2: Southeast (States of Sao Paulo, Minas Gerais, Espirito Santo e Rio de Janeiro);
 Region 3: Center-West Cerrado (States of Goias and Mato Grosso do Sul and the Cerrado area in
Mato Grosso);
 Region 4: Amazon North (States of Amazonas, Para, Amapa, Acre, Rondonia, Roraima and the
Amazon area in Mato Grosso);
 Region 5: Coastal Northeast (States of Ceara, Alagoas, Sergipe, Pernambuco, Rio Grande do Norte
and Paraiba);
 Region 6: Mapito e Bahia (States of Maranhao, Piaui, Tocantins e Bahia).
 The regions are independent in the model in the sense that land allocation equations for each
crop are different among regions. However, given that the total production must be equal to
the demand, if a given crop looses area in a certain region, for a given regional yield, other
region will offset it with an increase in area.
Confidential. Do not quote or cite unless authorized. 6

7. Fig. 1 Brazilian Biomes and States
RR
AP
AP
AM
MA CE
PA RN
PB
PE
AC PI AL
RO MT TO SE
BA
GO
MG
MS ES
Amazon Forest
Atlantic Forest SP RJ
Savanna PR
Steppe
Pantanal wetland SC
South Grassland
RS

8. Fig. 2 Macro-regions used in the Land Use Model
North Amazonia
Center West Cerrado
MAPITO and Bahia
Northeast coast
Southeast
South
Source: UFMG, ICONE. Confidential. Do not quote or cite unless authorized. 8

9. Figure 3. Land Use Model: Interactions Among
Sectors
Rice
Corn
Ethanol
Sugarcane
Sugar
Industry and
Cotton biodiesel
Soybean oil
Drybean
Pork
Soybean
Soybean Poultry (eggs
meal
and chicken)
Pasture Beef
Source: ICONE
Confidential. Do not quote or cite unless authorized. 9

10. Figure 4. General Structure of the Land Use Model
Exogenous macroeconomic data
- Population;
- World and national GDP;
- World oil price and domestic gasoline
price;
Domestic - Exchange rate;
consumption - Inflation rate;
- Fertilizer price index;
- Vehicle fleet.
Costs
Net exports
Demand
Final stocks
(t-1) Expected
Price Area
return
Production
Supply Yields
Initial stocks
Exogenous Endogenous
variable variable
Source: ICONE. Confidential. Do not quote or cite unless authorized. 10

11. General structure of the land use model
 One key output of the model is the total land allocated for agriculture and
pastures. If this total, plus the exogenous planted forest area, is increasing over
time, more natural land is brought for production purposes. This excess allocation of
land can be explained by the combination of two factors:
 Increase of cattle herd in regions with agricultural frontiers (regions 4 and 6 of
the model), with a simultaneous reduction or stabilization of cattle herd in the
traditional areas. This can be interpreted as an indirect effect due to crops
expansion;
 Expansion of crops in the frontier, which is a direct effect.
In order to measure the indirect effect it is necessary to isolate these two causes
of crops and pasture expansion in the frontier regions.
11

12. Data Gathering and Preparation
 In order to run the regressions and to calculate the parameters (elasticities and coefficients)
used for the projections, a 13 year database was organized, from 1996 to 2008. Due to the
lack of information with respect to prices and costs of production, data from 1997 and 2007
were used for the estimations.
 With respect to the variables that are solved endogenously in the model, the following
variables are included in the database:
 Supply and demand balance sheets for Brazil;
 Regional information on production, planted area, yields, prices received by farmers, costs of
production, cattle herd structure, animal production (beef, chicken, eggs, pork and milk).
 The model requires a set of exogenous macroeconomic variables (presented in the figure 2).
We are using exogenous macroeconomic variables supplied by FAPRI.
 The model is prepared to project ethanol exports and domestic demand endogenously, but
can also work with exogenous scenarios, such as different countries´ mandates. ICONE is
ready also to create exogenous scenarios of ethanol world trade, calculating supply and
demand for the main producers and consumers (Unites States, European union, Japan, etc.).
12

13. Assumptions
 Equilibrium price is obtained when supply is equal to demand, in a given year, for a
given activity (crops or animal products).
 Area allocated to a given crop, in a given year, is a result of the market equilibrium.
Producers respond, in terms of planted area, according to the expected market
return (costs of production of the current year and prices received in the previous
year).
 Prices received by farmers and prices paid by consumers follow the same trend
over time.
 The model assumes full availability of capital for investments and credit for working
capital, which means that it does not capture negative effects of low availability of
credit in the market.
 Crops yields and total recoverable sugar factors, although regional, are projected as
a time series trends.
 Prices are solved at a national level and are transmitted to the regions using
transmissions coefficients estimated by regressions. Improvements in infrastructure,
which would may lead to increase prices and to reduce costs of production can be
simulated as scenarios.
Confidential. Do not quote or cite unless authorized. 13

14. The two model´s versions
 First Version
 Results are ready and have already been used for some projects. Results validation
among analyzed sectors will be carry out soon.
 The first version doesn't considered land availability, so that total agricultural land
use can expand indefinitely.
 Equations for Allocation Area are independent in each Region. Historical planted
areas are considered as an explanatory variable.
 Improved Version
 Area allocated for each crop and pasture will be defined by two equations
1. Expansion Equation: it will define total agricultural area and it is subject to land
availability.
2. Shares Equation: for each crop and pasture, its share on the total agricultural area
is the dependent variable. Returns of all crops are used as explanatory variables.
 Equations for Allocation Area are simultaneously in each Region.

15. Net Growth of Agricultural Land Uses Area and Cattle Herd,
2002-06 (1,000 ha and heads)
Sugarcane Other crops Total used Cattle
State Pasture (ha)
(ha) (ha) area (ha) Herd (hd)
São Paulo 622 -224 -882 -484 -909
Minas Gerais 153 389 -625 -82 1,644
Paraná 74 850 -1 287 -284
Mato Grosso do Sul 41 1 -985 -210 558
Goiás 34 576 -2,041 -1,431 545
Bahia 26 492 143 661 912
Mato Grosso 25 1,634 -1,437 0 3,881
Maranhão 16 298 -463 -148 1.835
Pará 3 115 2,502 2,620 5,311
South-Centre 949 3,226 -5,971 -1,920 5,435
Total 1,000 5,446 -5,385 1,061 18,383
Source: PAM/IBGE, Agricultural Census/IBGE and PPM/IBGE.

16. Brazil: Crops Area (excluding 2nd crop)
and Production and Meat Production
Performance
Crops Area and Production
(million tons and million ha) Meat (million tons)
900 160 12 18
800 140 16
10
700 120 14
600 8 12
100
500 10
80 6
400 8
60
300
4 6
200 40
4
100 20 2
2
0 0
0 0
5
9
1
3
7
7
90
92
94
96
98
00
02
04
06
08(e)
/0
/0
/9
/0
/0
/9
00
04
06
98
02
96
Production Planted Area
Chicken Beef Pork
Source: CONAB; USDA; ABIEC; ABIPECS; ABEF. Elaboration: ICONE

17. Brazil: Livestock dynamics
Catle Herd in the Model´s Regions
Pasture Area in the Model´s Regions
60000
70000 South
Southeast
Center West Cerrado
60000 North Amazonia 50000
MAPITO and Bahia
Northeast coast
( 1 ,0 0 0 h d s )
50000
(1,000 ha)
40000
40000
30000
30000
20000
20000
South
Southeast
10000 Center West Cerrado
10000 North Amazonia
MAPITO and Bahia
Northeast coast
0 0
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

18. Brazil: Sugarcane, Sugar and Ethanol
Production
Sugarcane Production and Area Sugar and Ethanol Production
Million ton Million ha Million ton Billion ltr
600 8 35 60
7 30
500 50
6 25
400 40
5
20
300 4 30
15
3 20
200 10
2
5 10
100
1
0 0
0 0
90/91
92/93
94/95
96/97
98/99
00/01
02/03
04/05
06/07
90/91
92/93
94/95
96/97
98/99
00/01
02/03
04/05
06/07
Production Planted Area Sugar Ethanol
Source: CONAB; IBGE; UNICA. Elaboration: ICONE

19. Brazil: Ethanol Supply & Demand
(billion liters)
60 60
Production Exports Domestic Consumption
50 50
40 40
30 30
20 20
10 10
0 0
96/97
97/98
98/99
99/00
00/01
01/02
02/03
03/04
04/05
05/06
06/07
07/08(e)
Source: CONAB; IBGE; UNICA. Elaboration: ICONE

20. Overview of Sugarcane in Brazil

21. Applications
 The model is very flexible and can be used for different types of simulations and
scenarios.
 The model is also prepared to simulate impacts of variables that are not explicit in
the model, such as improvements in transportation infrastructure (construction of an
ethanol pipeline, for example), costs reduction due to he adoption of new
technologies (second generation ethanol and GMOs adoption), among others
applications.
 Useful for specific sector analysis
 Regional land use results can be disaggregate for smaller scales (World Bank´s
Brazil Low Carbon Study)
 Land use changes can be converted into carbon emissions.
 It is possible to connect the Brazilian Model with world models.
 It is used to compare land use results from other models (GTAP- CARB)