Sugarcane Expansion: Land Use Changes and Social Impacts in the São Paulo State, Brazil

Transcription

1 Sugarcane Expansion: Land Use Changes and Social Impacts in the São Paulo State, Brazil Flávio Luiz Mazzaro de Freitas Master of Science Thesis KTH School of Industrial Engineering and Management Energy Technology EGI MSC Division of Energy and Climate Studies SE STOCKHOLM

2 Master of Science Thesis EGI 2010: 080MSC Sugarcane Expansion: Land Use Changes And Social Impacts In the São Paulo State, Brazil Approved Examiner Flávio Luiz Mazzaro de Freitas Supervisors Semida Silveira Commissioner Gerd Sparovek Göran Berndes Andrea Egeskog Semida Silveira Contact person Flávio Luiz Mazzaro de Freitas Andrea Egeskog Abstract There is a strong concern about the environmental and social impacts of land use changes caused by sugarcane expansion. This research aims to assess the land use changes caused by sugarcane expansion in the State of São Paulo in the last five years, as well as predicting land use changes in the coming years. In addition, this research evaluates the social impacts of sugarcane expansion. The assessment of land use changes was made through GIS analysis. First, the changes during the last five years were evaluated. Based on this information, the land use changes for the coming years were estimated. The social impacts of sugarcane expansion were evaluated by studying the correlation between Human Development Index (HDI) and the sugarcane expansion. The results confirm that sugarcane crop expanded about 1.85 million hectares between 2003 and About 62% of this expansion replaced areas used for agricultural crop in 2003, and about 34% replaced areas used for pasture in Three scenarios were created in order to estimate sugarcane expansion in the coming years. In the first scenario, sugarcane would expand about 0.9 million hectares in three years; in the second scenario, 1.1 million hectares in four years; and in the third scenario, 1.4 million hectares in six 2

3 years. In each scenario, about 70% of the expansion would take place in areas used for agricultural crops in 2003, and 40% in areas used for pasture in The sugarcane expansion caused a significant and positive impact on the income dimension of HDI for regions with a very low level of development. For regions of medium and high level of development, the HDI impact was not significant. In addition, a slightly negative impact on the longevity dimension of HDI was observed. Summary The ambition to reduce GHG emissions and oil dependency has increased the world demand for sustainable energy. In this context, many countries have given massive incentives to promote the production and consumption of ethanol. Brazil, which is the main producer and consumer of ethanol from sugarcane crops, has set ambitious targets to increase ethanol production. Thus, a wide sugarcane crop expansion for ethanol production is expected in the coming years. However, the benefits of replacing oil with ethanol are questioned by some authors, who suggest that a comprehensive analysis of the life cycle of ethanol shows that ethanol generates more GHG emissions than fuels based on petroleum. Other authors defend ethanol as a suitable option to reduce GHG emissions, claiming that studies which suggest negative impacts from replacing fossil fuel with ethanol refer to the worst case scenario, which is unlikely to occur with the Brazilian production technology. A key factor that challenges the sustainability of biofuels, and that demands clarification, is the direct and indirect land use changes caused by sugarcane expansion. In addition, sugarcane production in Brazil is heavily criticized due to the socio-economic impacts of the sugarcane sector. Some authors question the economic advantages of sugarcane industry, claiming that most of the profit reaches only a small share of the population. Others authors point to the health risks posed by sugarcane expansion. This research aims to analyze the sugarcane expansion in order to identify the trends of land use change caused by sugarcane expansion in the state of São Paulo, the largest sugarcane producing region in Brazil, as well as to make an assessment of the impact of sugarcane expansion on the human development index (HDI) of the region. The methodology chosen for this research is divided in three steps, starting with mapping down the sugarcane expansion that took place in the period This evaluation was based on the sugarcane mapping of INPE (Brazilian Institute for Spatial Research) and on the land use map of 2000 from PROBIO. Using this database, the land use changes for each year during the period were identified. In the second step, a model was developed to predict the geographical spreading of sugarcane crop in the coming years. The input data were distance to the sugarcane mill, density of sugarcane mills, suitability for sugarcane crop, and sugarcane mill size. In the last step of this study, an assessment was made in order to identify the impacts of 3

4 sugarcane expansion on human development, measured in HDI. This evaluation was carried out by correlating the sugarcane expansion during the period with the changes observed in HDI for the same period. The results show that in 2003 sugarcane crop represented an area of around 3 million hectares and by 2008 this area had increased to 4.8 million hectares. This sugarcane expansion took place mainly in areas previously used for agriculture (other crops) and pasture in the year In 2003, the class other crops represented 7 million hectares, 28.5% of the total area of the São Paulo State, which was the same size as the area occupied by the class pasture (Figure 4). By 2008, about 1.1 million hectares of the area belonging to the class other crops and 0.6 million hectares of the area belonging to the class pasture in 2003 had become the class sugarcane. The model developed to predict sugarcane expansion was able to explain more than 50% of the variation observed in the sugarcane expansion, which can be deemed satisfying considering the uncertainty involved in predicting land used change. Based on this model, three different scenarios were created to predict the future sugarcane expansion. Each scenario indicated that about 70% of the sugarcane expansion would replace areas used for agriculture in 2003, and only about 30% of the expansion would occur at the expense of areas belonging to the class pasture in that same year. The model was based on the assumption that the expansion would occur in line with the tendency in the last five years, under the same governmental policy, and without great variation in the price of the crops cultivated in the São Paulo State. However, these factors have significant influence on the land use changes caused by sugarcane expansion in the next years, and so changes in the governmental policy and variation in the crop prices can potentially affect the prediction of land use changes made in this study. The outcomes of the third step of the analysis, where the impacts of sugarcane on HDI were evaluated, suggest that sugarcane has a significant and positive effect on the income dimension of HDI. However, the improvement observed on the income dimension was not reflected in the dimensions education and longevity. Regions where sugarcane expansion took place showed a faster economic development than other regions. However, no positive correlation was found for the dimensions education and longevity. Rather, the results suggested that sugarcane expansion has a negative effect on longevity. The average longevity was always lower in the regions where sugarcane expansion was more intensive. However, this correlation was not statistically significant. 4

5 Acknowledgements I believe that, to reach success and happiness, we always need the support of others. We cannot do it all on our own. With this thought in mind, I would like, first of all, to say a great thank you to my friend and cosupervisor on this project, Professor Gerd Sparovek, to whom I owe a lot of what I have achieved in life. I am also very much thankful to my supervisor Göran Berndes and to my examiner Semida Silveira for the opportunity of working with them, and also for their advices and great ideas during my research project. My thanks also go to my co-supervisor Andrea Egeskog, who gave a great support during this thesis and for my colleagues André Assunção, Alberto Barreto, Leonardo Anchieta, Jane Lino, Israel Klug and Karen Leyton, who were always very helpful when I most needed them. Finally, I would like to thank the most valuable thing I have: my family and friends. Thank you very much for everything you have done for me. 5

6 Table of content Abstract... 2 Summary... 3 Acknowledgements Introduction Climate change and Biofuels The Boom of ethanol production in Brazil Threats and challenges Aim and scope of the study Land Use Changes Caused by Sugarcane Expansion Methodology Sugarcane expansion in the last five years Sugarcane expansion and suitability for mechanized sugarcane harvest Modeling of sugarcane spatial distribution in relation to the mill Prediction of land use changes Results and discussion Sugarcane expansion in the last five years Modeling of sugarcane spatial distribution in relation to the mill Estimation of sugarcane mills density in the future Estimation of sugarcane production per sugarcane mill in the future Prediction of sugarcane expansion per range of distance from the new mills Sugarcane replacement tendency Prediction of land use changes caused by sugarcane expansion SUGARCANE EXPANSION AND HUMAN DEVELOPMENT Methodology Results

7 4. CONCLUSIONS BIBLIOGRAFY APPENDIX

8 1. Introduction 1.1. Climate change and Biofuels Today fossil fuels are responsible for 79% of the primary energy consumed in the world; the remaining 3% came from nuclear energy, while renewable sources are responsible for 18% (REN21, 2008). This high dependency on finite resources and the concerns regarding global warming have motivated a strong movement to replace energy based on fossil fuels with renewable energy (Escobar, 2009). Renewable energy is the energy that comes from natural sources which can be used infinitely. There are many different potential sources of renewable power generation, among them are the traditional biomass (wood, agricultural residues, waste, dung and others unprocessed biomass) and the hydropower generation, which together represent 89% of the renewable energy produced in the world (REN21, 2008). Biofuels represent 1.7% of this energy, which is a quite small share of the world s energy consumption; however this share has been growing quickly in the last years. In some countries biofuels are already quite important, for example, in the case of Brazil, biofuels replace about 50% of the gasoline consumption. Some authors (Sparovek, 2007; Goldemberg, 2008; Escobar, 2009) view the production of biofuels as an important means for reducing oil independency. Despite the presently insignificant contribution of biofuels to the world energy supply, they have a great potential to grow and many countries can easily produce them (Goldemberg, 2007). Biofuels in general are criticized because of the high amount of subsidies granted by the government to make their production feasible. In this context the Brazilian ethanol has received a lot of attention, since it is produced without subsidies and reaches the final consumers to a competitive price, usually cheaper than gasoline (Goldemberg, 2007) The Boom of ethanol production in Brazil In 2003, the Flexible Fuel Vehicles (FFV), which run on gasoline and ethanol in any proportion, were launched in Brazil. Due to the lower price of ethanol in relation to gasoline, FFV gave rise to a substantial increase in the ethanol production. Analyzing Figure 1, it is possible to observe that after 2003, when the ethanol production was 6 billion liters, there was an intense growth in ethanol production. In 2008 the production had tripled to 18 billion liters (Figure 1). A great part of this increase took place in São Paulo State, which is the area studied in this research. São Paulo, because of its geographical location, offers the perfect climatic conditions for growing sugarcane crop (raw material used to produce ethanol). Nowadays, about 60% of the Brazilian production of ethanol takes place in the São Paulo State, which is also the region that shows the highest level of technology and productivity in Brazil. 8

9 Ethanol Production (million liters) /04 04/05 05/06 06/07 07/08 08/09* Source: UNICA Crop year São Paulo State Others States Figure 1: Ethanol production in Brazil in the last five years Brazil is the fifth country in the world in terms of investments (national and foreign) in renewable energy (REN21, 2009). Part of this investment goes to the modernization of old sugarcane mills and another part to the construction of new sugarcane industries. Many countries in the world are seeking to reduce their oil dependency and their emissions of greenhouse gases, and have introduced incentives for biofuels consumption as an energy policy. As a result of this there will be a large international demand for ethanol and Brazil, having the second highest capacity for ethanol production in the world (REN21, 2009), will supply part of this demand. However, most of the increased ethanol production in Brazil is expected to be absorbed by the growing national consumption (Figure 2). Ethanol production (Billion liter) year Consumption Exportation Figure 2: Projection of the future ethanol production in Brazil. 9

10 This increase in ethanol production will require a large sugarcane expansion, which will for sure bring changes; great technological development is expected as well as economic development. However, negative impacts may also take place, depending on how this expansion will happen Threats and challenges Today there are many studies suggesting that the negative impacts of biofuels on the environment can be even bigger than their positive impacts (Scharlemann, 2008). Ethanol obtained through sugarcane expansion can also bring some environmental and social threats, and the challenge is to find ways to mitigate the negative impacts of ethanol production. The process of deforestation is a topic that has received great attention, especially in the case of the Amazon Rainforest. Some authors say that even if the expansion of the sugarcane crop does not take place in forested areas, it occupies the areas of other cultures, thus creating a deficit of land and an increased demand for converting forested areas into arable land (Scharlemann, 2008). However some authors disagree (Goldemberg, 2008; Escobar, 2009), suggesting that today most countries that have the potential to produce ethanol also have a poor agricultural sector with low productivity. Therefore, a small increase in the productivity makes it possible to maintain or even increase the production without occupying additional area. In the case of Brazil the agro-ecosystem most suitable for sugarcane expansion is the areas used for livestock production, which are covered with pasture (Goldemberg, 2008). Another topic that has been receiving great attention is the impacts of ethanol production on food prices. According to Scharlemann (2008), the sugarcane expansion competes for land with the food production, pushing down the production of those crops and consequently causing increases in the food prices. The overall most negative impact of ethanol production is the land use change caused by sugarcane expansion, thus studies that evaluate land use changes are required as databases for studies that evaluate the social and environmental impacts of ethanol production Aim and scope of the study The aim of this research is to evaluate the expansion of sugarcane in the São Paulo State, which is the main region for ethanol production in Brazil. This evaluation was divided into steps. The first step was to assess the sugarcane expansion for the last five years, generating a database that would be used in the second step, where the objective was to simulate the land use changes caused by sugarcane expansion in the next years. Finally, the social impacts of sugarcane expansion during the period were evaluated by studying the correlation between sugarcane expansion and changes in the Human Development Index (HDI). 10

11 2. Land Use Changes Caused by Sugarcane Expansion 2.1. Methodology The assessment of the land use changes caused by sugarcane expansion was performed in two steps. In the first one, the sugarcane expansion in the last five years, where the expansion has been more intense, was analyzed. This analysis generated the data needed in the second step, which aimed to estimate the land use changes for the coming years Sugarcane expansion in the last five years Land use map for 2003 In order to achieve the objectives of this study, the land use map for the year 2003, the first year of the assessed period, was required. Such a land use map (or related databases) is not readily available and it was therefore necessary to generate the land use map using the available database. Available databases Land use mapping: Database of Land cover map for the São Paulo State; the land use map from the project PROBIO (of the Brazilian Ministry of Environment), which represents the land cover of 2002; and which is divided in the following classes: o Pasture: This class represents areas covered with different species of grass which are used for livestock grazing; o Agriculture: This class represent the areas covered with any type of crops like orange, corn, soy, etc; o Native vegetation: This class represents the areas that are still covered with the original vegetation. For the São Paulo state, this means primarily savanna vegetation and Atlantic forest; o Reforestation: This class represents the areas taken up by forest planted for economic purposes. In the São Paulo state these areas are mainly covered with species from the genera Eucalyptus and Pinus. o Water bodies: represents areas covered with water; however, only big areas are taken into account. Small rivers and small lakes are not mapped in this class because of the resolution of this land use map. 11

12 o Urban area: represents urbanized areas. Sugarcane mapping: The sugarcane mapping was obtained from the project CANASAT (Sugarcane crop mapping in Brazil), coordinated by the National Institute for Space Research (INPE). This mapping has been made for the São Paulo State yearly between 2003 and 2008 using earth observation satellite images; Forest mapping: The Atlantic forest mapping made by the nongovernmental organization SOS Mata Atlântica. This mapping was made using remote sensing analysis for the year 20001:50000 scale. Assumptions Since the first year of the assessed period is 2003 (the first year of the CANASAT project) and the only land cover mapping available is for 2002 it was necessary to make the assumption that there were no significant land use changes from 2002 to 2003; For the same reason it was assumed that there was no significant deforestation of the Atlantic forest due to sugarcane expansion between 2000 and 2003; Procedures To add small fragments of forest that had not been mapped in the PROBIO land cover map, the Atlantic forest map was used, since it has a quite good accuracy for large and small fragments of forest. Finally, the class agriculture was subdivided into sugarcane and others crops. To do so, the sugarcane map of 2003 on the land use map from PROBIO was used. The tree maps used in this analysis came from tree different sources. Consequently they don t match each other completely and since they are of different scale, one was more accurate that the other. To combine these maps it was necessary to define a number of criteria: The class urban areas will never be replaced, since this class is well mapped in the PROBIO land cover map, and since it s unlikely that the sugarcane would expand to urban areas; The class forest from the Atlantic forest map will replace every class in the PROBIO land cover map except the class urban areas. This criterion was established because of the higher accuracy of the Atlantic forest map. The class sugarcane from the sugarcane map will replace every class in the PROBIO land use map except the class urban areas, as well as the class forest in the Atlantic forest map. This criterion is based on the rationale that the objective is to create a land use map representing the situation in Since the sugarcane map was made in 2003, later than the other two, it can be assumed to be more accurate. 12

13 The final result was a land use map for the year 2003 showing the classes pasture, agriculture (other uses), agriculture (sugarcane), water, urban areas, native vegetation, and reforestation areas. The Boolean function used to generate the land cover map for 2003 is shown below. Variables Sugarcane map of No presence of sugarcane crop 1- Sugarcane crop present Atlantic forest map of No presence of Atlantic forest 1- Atlantic forest present Land cover map of Not mapped 1- Pasture 2- Agriculture 3- Native vegetation 4- Reforestation 5- Water body 6- Anthropic 7- Urban area Sugarcane map 2003 Atlantic forest map 2000 Land use map 2002 Land use map 2003 if then Sugarcane if else then Sugarcane if else then Sugarcane if else then Sugarcane if else then Sugarcane if else then Sugarcane if else then Sugarcane if else then Urban area if else then Sugarcane if else then Sugarcane if else then Sugarcane if else then Sugarcane if else then Sugarcane if else then Sugarcane if else then Sugarcane if else then Urban area 13

14 if else then Native vegetation if else then Native vegetation if else then Native vegetation if else then Native vegetation if else then Native vegetation if else then Native vegetation if else then Native vegetation if else then Urban area if else then Missing data if else then Pasture if else then Agriculture if else then Native vegetation if else then Reforestation if else then Surface of water if else then Anthropic if else then Urban area else Error Land use changes Once the land cover map for 2003 had been generated and the sugarcane crop mapped for each year of the period , it was possible to identify the changes caused by sugarcane expansion from 2003 to 2008 in relation to the land cover map of In order to do so each one of the sugarcane maps for the period was overlaid with the land use map of 2003, making it possible to identify what crops had been replaced by the sugarcane expansion and where sugarcane had retreated and been replaced by other land uses. The Boolean function describing this process is presented below. Variables Land Use map Not mapped 1- Pasture 2- Agriculture 3- Native vegetation 4- Reforestation 5- Surface of water 6- Anthropic 7- Urban area Sugarcane map i 14

15 0- No present of sugarcane crop 1- Sugarcane crop present Land Use map 2003 Sugarcane map i Land use replaced if 0 1 then No replacement if else 1 1 then Pasture if else 2 1 then Agriculture if else 3 1 then Native vegetation if else 4 1 then Reforestation if else 5 1 then Surface of water if else 6 1 then Anthropic if else 7 1 then No replacement if else 0 0 then No replacement if else 1 0 then No replacement if else 2 0 then No replacement if else 3 0 then No replacement if else 4 0 then No replacement if else 5 0 then No replacement if else 6 0 then No replacement if else 7 0 then No replacement else error Sugarcane expansion and suitability for mechanized sugarcane harvest This section aims to establish the relationship between sugarcane expansion and the suitability for mechanized sugarcane harvest (SM), in order to assess if expansion tends to take place in suitable areas. Slopes steeper than 12% are not suitable for mechanized harvest, due to limitations of the sugarcane harvesting machine. Sugarcane expansion can therefore be expected to take place in areas where the declivity is less than 12%. This assessment was realized in two steps. The first step was to identify the suitability for SM and the second step was to overlay the SM areas with the areas where expansion or retreat of sugarcane crop had taken place. The SM was identified through the Digital Elevation Model SRTM (DEM) from the Brazilian Agricultural Research Corporation (EMBRAPA). Based on the DEM a declivity map was generated and combined with the land use maps. The Boolean function used in this process is showed below. 15

16 Variable Slope: Continuous variable from 0% to Slope (%) Suitability for mechanized sugarcane harvest (SM) if <= 12 then Suitable if else > 12 then Not suitable To assess the relation between the sugarcane expansion and the suitability for mechanized sugarcane harvest, the map of suitability was overlaid with the land use changes caused by sugarcane expansion. In this way it was possible to identify the proportion of each land use replaced by sugarcane that was suitable or not suitable for mechanization. This process was made using the following Boolean function. Variables Land use replaced by sugarcane expansion 0- Not replaced 1- Pasture 2- Agriculture 3- Native vegetation 4- Reforestation 5- Surface of water 6- Anthropic 7- Urban area Suitability for mechanized sugarcane harvest 0- Not suitable 1- Suitable Suitability for mechanized sugarcane harvest Land use replaced Land use replacement in relation to Suitability for mechanization (SM) if 1 0 then No replacement If else 1 1 then Pasture/Suitable If else 1 2 then Agriculture/Suitable If else 1 3 then Native vegetation/suitable If else 1 4 then Reforestation/Suitable If else 1 5 then Surface of water/suitable If else 1 6 then Anthropic/Suitable If else 0 0 then No replacement If else 0 1 then Pasture/Not suitable 16

17 If else 0 2 then Agriculture/Not suitable If else 0 3 then Native vegetation/not suitable If else 0 4 then Reforestation/Not suitable If else 0 5 then Surface of water/not suitable If else 0 6 then Anthropic/Not suitable else Error Modeling of sugarcane spatial distribution in relation to the mill The purpose of this section is to identify a variable that can explain how sugarcane expands. The sugarcane land share is defined as the % of a specific region that is covered by sugarcane. Four potential parameters were identified and tested for their suitability to explain sugarcane land share: i) The distance from the sugarcane mill; ii) the density of sugarcane mills; iii) the mill s size or capacity for sugarcane production; iv) physical suitability for sugarcane production. The parameters will be further explained in this section. To facilitate comprehension, this section was divided into six steps. Step 1: Calculation of the sugarcane land share in relation to distance to the sugarcane mills - The first step in determining the spatial distribution of sugarcane in relation to the mills was the calculation of a distance raster representing all installed mills in São Paulo in 2006 for which the location was known. A distance raster is a raster file which calculates the distance in relation to one or more references, which in this case were the installed sugarcane mills. The resolution of the raster was 100 meters, which means that the distance to the sugarcane mills was calculated for each 100 meters. The distance raster was then divided into the following distance ranges: 0 5 km; 5 10 km; km; km; km; km; km; km; and more than 40 km. Step 2: Sugarcane land share per distance range - In the next step, the map showing the distance in relation to installed sugarcane mills created in step 1 was combined with the sugarcane map of Combining the maps made it possible to assess the sugarcane land share for each range of distance. The Boolean function used for this procedure is shown below. Variables Map showing the distance from sugarcane mills: Continuous variable. Sugarcane map No sugarcane crop present 17

18 1- Presence of sugarcane crop Distance map from sugarcane mills (km) Sugarcane map 2006 Sugarcane occupation per Distance from mills if 0 to 5 1 then Sugarcane within 0 to 5 km if else 5 to 10 1 then Sugarcane within 5 to 10 km if else 10 to 15 1 then Sugarcane within 10 to 15 km if else 15 to 20 1 then Sugarcane within 15 to 20 km if else 20 to 25 1 then Sugarcane within 20 to 25 km if else 25 to 30 1 then Sugarcane within 25 to 30 km if else 30 to 35 1 then Sugarcane within 30 to 35 km if else 35 to 40 1 then Sugarcane within 35 to 40 km if else More than 40 1 then Sugarcane within further than 40 km if else 0 to 5 0 then No Sugarcane within 0 to 5 km if else 5 to 10 0 then No Sugarcane within 5 to 10 km if else 10 to 15 0 then No Sugarcane within 10 to 15 km if else 15 to 20 0 then No Sugarcane within 15 to 20 km if else 20 to 25 0 then No Sugarcane within 20 to 25 km if else 25 to 30 0 then No Sugarcane within 25 to 30 km if else 30 to 35 0 then No Sugarcane within 30 to 35 km if else 35 to 40 0 then No Sugarcane within 35 to 40 km if else More than 40 0 then No Sugarcane within further than 40 km else Error Step 3: Suitability for agriculture in relation to distance from installed sugarcane mills - In order to create an indicator for the suitability for agriculture, the physical restrictions for agriculture expansion were identified. Three restricting factors were identified: the areas covered with water, urbanized areas, and the areas with a declivity greater than 25% (Montardo, 2009). The urbanized areas and the water bodies were extracted from the land use map of 2002 and in order to identify the areas with a slope suitable for agriculture, the slope map was classified into two classes, the first class being 0-25% declivity, representing the areas suitable for agriculture with conservation techniques, and the second class being more than 25% declivity, representing the areas not suitable for agriculture. These data were combined with the map showing distance from installed mills, this way it was possible to know how much of the space is unsuitable for agriculture. The following Boolean function was used in this procedure. Variables Land use map Water 1- Urban area 18

19 2- Other land use Slope: Continuous variable. Distance map from sugarcane mills: Continuous variable. Land use map 2002 Slop e (%) Distance from sugarcane mills map Suitability for sugarcane cultivation if 0 or 1 <=25 0 to 5 then Not suitable wit in 0 to 5 km if els e 0 or 1 <=25 5 to 10 then Not suitable within 5 to 10 km if else 0 or 1 <=25 10 to 15 then Not suitable within 10 to 15 km if else 0 or 1 <=25 15 to 20 then Not suitable within 15 to 20 km if else 0 or 1 <=25 20 to 25 then Not suitable within 20 to 25 km if else 0 or 1 <=25 25 to 30 then Not suitable within 25 to 30 km if else 0 or 1 <=25 30 to 35 then Not suitable within 30 to 35 km if else 0 or 1 <=25 35 to 40 then Not suitable within 35 to 40 km if else 0 or 1 <=25 More than 40 then Not suitable further than 40 km if else 0 or 1 >25 0 to 5 then Not suitable within 0 to 5 km if else 0 or 1 >25 5 to 10 then Not suitable within 5 to 10 km if else 0 or 1 >25 10 to 15 then Not suitable within 10 to 15 km if else 0 or 1 >25 15 to 20 then Not suitable within 15 to 20 km if else 0 or 1 >25 20 to 25 then Not suitable within 20 to 25 km if else 0 or 1 >25 25 to 30 then Not suitable within 25 to 30 km if else 0 or 1 >25 30 to 35 then Not suitable within 30 to 35 km if else 0 or 1 >25 35 to 40 then Not suitable within 35 to 40 km if else 0 or 1 >25 More than 40 then Not suitable further than 40 km if else 2 <=25 0 to 5 then Suitable within 0 to 5 km if else 2 <=25 5 to 10 then Suitable within 5 to 10 km if else 2 <=25 10 to 15 then Suitable within 10 to 15 km if else 2 <=25 15 to 20 then Suitable within 15 to 20 km if else 2 <=25 20 to 25 then Suitable within 20 to 25 km if else 2 <=25 25 to 30 then Suitable within 25 to 30 km if else 2 <=25 30 to 35 then Suitable within 30 to 35 km if else 2 <=25 35 to 40 then Suitable within 35 to 40 km if else 2 <=25 More than 40 then Suitable further than 40 km if else 2 >25 0 to 5 then Not suitable within 0 to 5 km if else 2 >25 5 to 10 then Not suitable within 5 to 10 km if else 2 >25 10 to 15 then Not suitable within 10 to 15 km if else 2 >25 15 to 20 then Not suitable within 15 to 20 km if else 2 >25 20 to 25 then Not suitable within 20 to 25 km if else 2 >25 25 to 30 then Not suitable within 25 to 30 km if else 2 >25 30 to 35 then Not suitable within 30 to 35 km if else 2 >25 35 to 40 then Not suitable within 35 to 40 km if else 2 >25 More than 40 then Not suitable further than 40 km else Error 19

20 Step 4: Voronoi diagram for the location point of installed sugarcane mills - The Voronoi diagram is a tool used to divide the space determined by distances between points (Aurenhammer, 1996), as can be seen in the illustration below. Figure 3: Illustration of a Voronoi diagram The diagram was included in this analysis primarily to serve as an indicator of the density of sugarcane mills. The Voronoi diagram can be used as such an indicator based on the rationale that the larger the area of the Voronoi cell, the lower the density of mills in that region. So the area of a Voronoi cell is inversely proportional to the density of sugarcane mills. Since the Voronoi diagram will be used to explain the spatial distribution of sugarcane in relation to the mill, it was necessary to overlay it with the distance from the installed sugarcane mills in order to eliminate the areas of each Voronoi cell that overpasses 40 km (number identified by the author in previous research as the distance limit for sugarcane expansion). Finally the maps produced in steps 2 and 3 were divided according to the Voronoi diagram, making it possible to determine the distance for each Voronoi cell or for each sugarcane mill. Step 5: Production capacity of each sugarcane mill - Another parameter used for modeling the sugarcane spatial distribution was the production capacity of each mill. This is an important parameter, since the larger the capacity of the mill, the greater the amount of sugarcane crop necessary to supply the mill. To determine the distribution of production capacity, a database of sugarcane production for the year 2006 from UNICA was used. Step 6: Statistical analysis - The statistical analysis was employed to assess the influence that density of mills, suitability for agriculture and the mill s production capacity have on the spatial distribution of sugarcane. This was made in the form of a regression analysis using the following equation: 20

21 is the percentage of land occupied by sugarcane; is the intersect; is the coefficient for ; is the Voronoi cell area (ha); is the coefficient for ; is the area unavailable for agriculture (%); is the coefficient for ; is the production of sugarcane of the mill (tons); and is the error or residual that represents what is not handled by the proposed model Prediction of land use changes Based on the modeling of the spatial distribution of sugarcane and data on the sugarcane expansion in the last five years, it was possible to estimate the future land use changes caused by sugarcane expansion once all the sugarcane mills, for which the geographical location is already known, have been installed. The procedure used in this analysis is described below. Step 1: Combining the locations of old and new sugarcane mills - In this step the map showing the old sugarcane mills was combined with the map showing the new ones in order to estimate the future scenario when all the sugarcane mills will have been installed. The map showing the locations of new sugarcane mills was from 2009 and was obtained from the website of UDOP (Union of Biofuel Producers); Step 2: Distance raster - in this step a distance raster was created based on the sugarcane mills locations map created in step 1. The distance raster was then divided into the following distance ranges: 0 5 km; 5 10 km; km; km; km; km; km; km; and more than 40 km. Step 3: (Land use + sugarcane 2008) x Distance from old and new sugarcane mills - in order to determine the current situation of sugarcane spatial distribution and to establish a starting point for the prediction of land use changes, the sugarcane map of 2008 was combined with the land use map of 2002 (only pasture and agriculture). It was also overlaid with the distance map created in step 2. It is important to clarify here that, as this analysis is using the land use map from 2002, the prediction of land use changes described in this analysis will be the changes in land use in relation to the year The Boolean function is shown below. Variables Sugarcane map of No presence of sugarcane crop 2- Sugarcane crop present Land cover map of Pasture 1- Agriculture 2- Other land use Distance from sugarcane mills map: Continuous variable. 21

22 Sugarcane map 2008 Land use map 2002 Distance from sugarcane mills map Current sugarcane land share (2008) and land use 2002 if to 5 then Pasture within 0 to 5 km if else to 10 then Pasture within 5 to 10 km if else to 15 then Pasture within 10 to 15 km if else to 20 then Pasture within 15 to 20 km if else to 25 then Pasture within 20 to 25 km if else to 30 then Pasture within 25 to 30 km if else to 35 then Pasture within 30 to 35 km if else to 40 then Pasture within 35 to 40 km if else 0 0 More than 40 then Pasture further than 40 km if else to 5 then Agriculture within 0 to 5 km if else to 10 then Agriculture within 5 to 10 km if else to 15 then Agriculture within 10 to 15 km if else to 20 then Agriculture within 15 to 20 km if else to 25 then Agriculture within 20 to 25 km if else to 30 then Agriculture within 25 to 30 km if else to 35 then Agriculture within 30 to 35 km if else to 40 then Agriculture within 35 to 40 km if else 0 1 More than 40 then Agriculture further than 40 km if else to 5 then Other land uses within 0 to 5 km if else to 10 then Other land uses within 5 to 10 km if else to 15 then Other land uses within 10 to 15 km if else to 20 then Other land uses within 15 to 20 km if else to 25 then Other land uses within 20 to 25 km if else to 30 then Other land uses within 25 to 30 km if else to 35 then Other land uses within 30 to 35 km if else to 40 then Other land uses within 35 to 40 km if else 0 2 More than 40 then Other land uses further than 40 km if else to 5 then Sugarcane within 0 to 5 km if else to 10 then Sugarcane within 5 to 10 km if else to 15 then Sugarcane within 10 to 15 km if else to 20 then Sugarcane within 15 to 20 km if else to 25 then Sugarcane within 20 to 25 km if else to 30 then Sugarcane within 25 to 30 km if else to 35 then Sugarcane within 30 to 35 km if else to 40 then Sugarcane within 35 to 40 km if else 1 0 More than 40 then Sugarcane further than 40 km if else to 5 then Sugarcane within 0 to 5 km if else to 10 then Sugarcane within 5 to 10 km if else to 15 then Sugarcane within 10 to 15 km if else to 20 then Sugarcane within 15 to 20 km 22

23 if else to 25 then Sugarcane within 20 to 25 km if else to 30 then Sugarcane within 25 to 30 km if else to 35 then Sugarcane within 30 to 35 km if else to 40 then Sugarcane within 35 to 40 km if else 1 1 More than 40 then Sugarcane further than 40 km if else to 5 then Sugarcane within 0 to 5 km if else to 10 then Sugarcane within 5 to 10 km if else to 15 then Sugarcane within 10 to 15 km if else to 20 then Sugarcane within 15 to 20 km if else to 25 then Sugarcane within 20 to 25 km if else to 30 then Sugarcane within 25 to 30 km if else to 35 then Sugarcane within 30 to 35 km if else to 40 then Sugarcane within 35 to 40 km if else 1 2 More than 40 then Sugarcane further than 40 km else Error Step 4: Suitability for agriculture in relation to distance from old and new sugarcane mills - in this procedure the same rationale as for step 3 of this section was used, with the difference that here the suitability for agriculture was considered in relation to the distance from old and new sugarcane mills, based on the raster described in step 2 in this section. Step 5: Voronoi diagram for the location points of old and new sugarcane mills - This step was based on the same rationale as used in step 4 of section However, now the Voronoi diagram was generated based on the location point of old and new sugarcane mills. Step 6: Identification of sugarcane expansion - since the aim of this part of the study is to estimate the land use changes caused by sugarcane expansion in the coming years, it is necessary to know how much the sugarcane crop will expand. Torquato (2006) estimates that sugarcane will expand in São Paulo State from 390 million tons in 2008 to about 580 million tons in Step7: Estimation of sugarcane production for each mill - To estimate the future production of each sugarcane mill, the database provided by UNICA was used, where it is possible to find the current sugarcane production for most sugarcane mills in the São Paulo State. However, for some sugarcane mills, especially the recently installed ones, it was not possible to find data on the current production, and it was necessary to make an estimation of their production. To do this estimation, the production for the whole São Paulo State () was obtained from the IBGE database. Then, the sum of the production was calculated for the sugarcane mills (p 1) for which the current 23

24 production was known. The difference between P and p was distributed over the sugarcane mills for which the sugarcane production is not known. This is illustrated in the mathematical expression below. p2 is the estimated production of each sugarcane mill for which the production is not known; P is the production of sugarcane in the state of São Paulo; p1 is the total sugarcane production of the sugarcane mills for which the production is known; and n is the number of sugarcane mills for which the sugarcane production is not known. Step 8: Projection of sugarcane production for each sugarcane mill - In order to estimate the future situation when all the new sugarcane mills are installed and working in full capacity, it was necessary to estimate the sugarcane production of existing sugarcane industries. In order to do that, the growth rate for sugarcane production had to be estimated. Since it s hard to determine the future growth rate of sugarcane production, three different scenarios were created representing different increase rates for the sugarcane production of existing sugarcane mills. The following scenarios were considered: Scenario 1: the existing sugarcane mills will keep increasing the sugarcane production at an annual growth rate of 2%; Scenario 2: the existing sugarcane mills will keep increasing the sugarcane production at an annual growth rate of 3%; Scenario 3: the existing sugarcane mills will keep increasing the sugarcane production at an annual growth rate of 4%; Scenarios of annual rates of increase in sugarcane production lower than 2% were not included, because analyzing the past shows that production has always been between 2% and 5% except in 2008 where it was around 10%. Considering the current prosperity of the sugarcane sector, an annual increase rate in sugarcane production lower than 2% would be unlikely. Conversely, considering the estimations of Torquato (2006), it s unlikely that the annual growth rate of sugarcane production from existing sugarcane mills would increase more than 4%, since new sugarcane mills are already under construction. Once they start working, they will take over part of the increasing production. These scenarios were then applied to the production of each existing sugarcane mill calculated in step 7 in this section. Subsequently, the difference between the production of sugarcane from existing sugarcane mills and the production suggested by Torquato (2006) was estimated. This difference gave the production which might be supplied by new sugarcane mills. 24

25 Step 9: Prediction of the land share occupied by sugarcane in the future - In this step of the analysis the following information was used: the density of sugarcane mills in the future (step 5 of this section); the projection of sugarcane production for each mill (step 8 of this section); and the suitability for sugarcane production (step 3 of section 2.2.2). This information was applied in the model equation for determining sugarcane land share, created in step 6 of section The calculation was made for the different scenarios. Step 10: Prediction of sugarcane expansion - Once the future distribution of sugarcane is known (step 9 of this section), correlating it with the current sugarcane distribution (step 3 of this section), makes it possible to identify the spatial distribution of sugarcane expansion. This information was calculated for each distance range of each Voronoi cell. For some regions the current area occupied by sugarcane was larger than the predicted one. The following criteria were defined to handle such situations: If the predicted area is larger than the current one, it will be assumed that no expansion will take place in that region; The difference between the current and the predicted area will be deducted from other regions which are in the same distance range. Step 11: Prediction of land use changes from 2008 based on the land use map of Once the sugarcane expansion is known for each range of distance from the sugarcane mills, it is necessary to estimate what kind of land use will be replaced by sugarcane. This analysis was made based on two kinds of information, namely: The sugarcane replacement tendency, identified by analyzing the sugarcane expansion in the last five years. The current land uses available for sugarcane expansion in each region. This information was produced in step 3 of this section Results and discussion Sugarcane expansion in the last five years The results show that in 2003 the sugarcane crop represented an area of around 3 million hectares and by 2008 the area had increased to 4.8 million hectare (Figure 4). The analysis shows that this sugarcane expansion took place mainly in areas used for agriculture (other crops) and pasture in By 2003 the class other crops represented 7 million hectares, 28.5% of the total area of the São Paulo State, which was the same as the area occupied by the class pasture (Figure 4). In 2008 about 1.1 million hectares of the area belonging to the class other crops and 0.6 million hectares of the area belonging to the class pasture in 2003 instead belonged to the class sugarcane. Also, in the year 2008 about 70,000 hectares of the area belonging to the class native vegetation in 2003 instead belonged to the class sugarcane. 25

26 8 7 Area (Million hectares) Year Agriculture (Other crops) Agriculture (Sugarcane) Pasture Native vegetation Anthropic Reforestation Urban Water Figure 4: Land use changes from 2003 to 2008, assuming all the land use change that took place in the São Paulo State was connected to sugarcane expansion. Thus, sugarcane expansion during the period took place mainly on land belonging to the classes other crops (e.g., used for crops such as soy and corn) and pasture in A smaller share of the sugarcane expansion took place on land belonging to the land use class native vegetation in However, considering the scale of this study and the possible errors that can arise from combining different data sources, it is not possible to state with very high confidence that sugarcane expansion during the period to some extent replaced native vegetation. Similarly, results suggesting that there was some sugarcane expansion on areas belonging to the class water bodies in 2003, can be a consequence of errors arising from the combination of different data sources. However there was a significant sugarcane expansion on one area belonging to the class water bodies located in the west of the São Paulo State (Figure 5), and most probably this areas was wetland which was drained after The expansion on urban areas, anthropic use and reforestation areas was very small and will not be considered due to the scale of this analysis. 26

27 Figure 5: Sugarcane expansion on an area that was covered by water in The result was quite unexpected, since the expansion occurred mainly in areas occupied by agricultural crops while it was expected that most expansion would occur in areas with pasture (Goldemberg, 2008; Escobar, 2009). The rationale behind this assumption was that pasture land is usually much cheaper and the profitability gains of changing from pasture to sugarcane is much higher compared to any other agricultural crop. One fact that can potentially explain these results is that sugarcane can only be collected within a limited distance from the mill. Hence, the mills are restricted by the distance and cannot be selective about the areas where they expand their cultivation of sugarcane. Another explanation resides in the difference between crop farming and cattle farming. Possibly there s a lower barrier for crop producing farmers to convert to sugarcane production than for cattle producers, since the crop farmers can use part of their existing infrastructure for sugarcane production, while the cattle producer cannot. A third possible explanation is that the land use classification from PROBIO of 2002 could have overestimated the class agriculture In the data processing, where this class was subdivided into sugarcane and other crops, this might have led to the class other land being overestimated (since the area belonging to the class sugarcane was known). The methodology used to produce this land use map was based on satellite images classification, which is recognized as a feasible way to estimate land use. However, for images that cover wide areas there is the risk of changing weather conditions, such as a cloud cover, affecting the quality of the image and resulting in variations in the images that can confuse the software used to make the classification. An important consideration is that this assessment has not indentified the net land use change caused by sugarcane expansion, since it wasn t possible, through the methodology adopted in this study, to identify the 27

28 indirect land use changes caused by sugarcane expansion. Indirect land use changes takes place when sugarcane expands on areas close to the sugarcane mill, replacing other crops that in their turn replace other land uses far from the mill. So, it is important to emphasize that the results generated in this study give an evaluation of the direct land use changes caused by sugarcane expansion since 2003, which is different from the net land use changes caused by sugarcane expansion. Sparovek et al. (2008), assessing the net land use changes caused by sugarcane expansion, suggest that sugarcane expansion did not significantly replace land used for other crops in the period This observation in combination with the outcomes of this study supports the theory that even though sugarcane expanded mainly to areas occupied with the land use crops, as observed in this study, these crops were moved to areas far from the sugarcane mill, replacing areas occupied with pasture. The net land use changes would therefore be sugarcane replacing pasture. Sugarcane expansion and suitability for mechanization Figure 6 shows the annual sugarcane expansion and whether the expansion took place on land that was classified as suitable for mechanical harvest. As can be seen, the sugarcane expansion from 2003 to 2008 took place mainly in areas suitable for mechanization. Only about 2 % of the sugarcane expansion took place on lands classified as not suitable for mechanization. 2,5 Area (Million hectates) 2,0 1,5 1,0 0,5 0, to to to to to to 2008 Period Suitable Not Suitable Figure 6: Expansion of sugarcane in relation to the suitability of mechanization. Figure 7 shows the distribution of areas covered with sugarcane plantations that were converted to another land use, in relation to land suitability for mechanization. Note that the scales of Figure 6 and Figure 7 are different: the conversion of sugarcane plantations to other land uses is small compared to the area claimed by sugarcane expansion. 28

29 Area (Million hectares) 0,50 0,45 0,40 0,35 0,30 0,25 0,20 0,15 0,10 0,05 0, to to to to to to 2008 Period Suitable Not Suitable Figure 7: Retrocession of sugarcane in relation to the suitability of mechanization As can be seen, most of the land where sugarcane plantations were converted to other uses did support mechanization. Thus, the data could not confirm that most of the discontinuation of sugarcane plantations was due to mechanization requirements, forcing the sugarcane producers using lands not suitable for mechanization to stop their practice Modeling of sugarcane spatial distribution in relation to the mill This modeling was based on data from 2006, and at that time the sugarcane crop represented an area of 3.35 million hectares. An important share of this area 58% - was located 5-15 km from the mill, making this distance range very important for analyzing the sugarcane expansion of new sugarcane mills. 99% of the land occupied by sugarcane was located within 30 km from the mill (Figure 8). 29

30 Area of Sugarcane (million hectares) 1,4 1,2 1,0 0,8 0,6 0,4 0,2 0,0 0 to 5 5 to to to to to to to to 50 Distance (km) Figure 8: Sugarcane distribution in relation to the distance from sugarcane mills The distribution of sugarcane is highly correlated with the distance from the sugarcane mill. In the distance range 0-5 km, the average share of land occupied by sugarcane is around 45% (Figure 9). This number falls almost linearly to 4% in the distance range km (Figure 9). After 30 km the sugarcane land share is insignificant (Figure 9). Sugarcane land occupation (%) 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% 0 to 5 5 to to to to to to to to 50 Distance (km) Figure 9: Sugarcane land share in relation to the distance from the sugarcane mill Correlation between Sugarcane mills density (SMD) and Sugarcane Land Share (SLS) In 2006, there were 179 sugarcane mills installed in the São Paulo state, according to the Biofuel Producers Union (UDOP). These sugarcane mills were distributed all over the state, except for the Southeast, where the 30

31 climate is unfavorable for sugarcane production. The region of Ribeirão Preto has the highest density of sugarcane mills, followed closely by the region of Piracicaba. The region of Pontal do Paranapanema is the region with lowest density (Figure 10). Figure 10: Location point of installed sugarcane mills in Analyzing the correlation between the sugarcane mills density (SMD) in the analysis, the Voronoi cell area was used as an indicator of SMD and the share of land occupied by sugarcane crop, it is possible to conclude that SMD has a strong influence on the distribution of sugarcane crop. This influence is stronger in areas close to the sugarcane mills and becomes less significant the farther one gets from the sugarcane mills. In the range 0-5 km, the SMD is responsible for around 42% of the variations in the share of land occupied by sugarcane. The corresponding number is 45%, 47%, 37%, 39%, 34% and 22% for the distance ranges 5-10 km, km, km, km, km, and km, respectively. The statistical analysis for these distance ranges, with a significance of 5%, confirms that the sugarcane land expansion is highly dependent on SMD. For the distance range km, SMD cannot significantly explain the variations in sugarcane land distribution. The correlation between the sugarcane land share and the SMD for the different distance ranges is shown in Appendix 1Appendix 1: Correlation between the sugarcane land share and sugarcane mills density for the different ranges of distance.. 31

32 Correlation between sugarcane production per mill and SLS Data on the production of each sugarcane mill was extracted from the database of UNICA. Analyzing this database it was possible to observe a wide variation among the mills as regards sugarcane production, where the mill with the highest sugarcane production produced million tons of sugarcane while the one with the lowest sugarcane production produced million tons of sugarcane. Studying the effect of sugarcane production on the share of land occupied by sugarcane it s possible to say that in the ranges 0-5 km, 5-10 km, km, km, km, km, km, and km, sugarcane production answers for 31%, 33%, 23%, 23%, 23%, 14%, 0% and 3%, respectively, of the variations in the sugarcane land share, as can be seen in Appendix 2. In general, the sugarcane production has a weaker influence on the sugarcane land distribution than the SMD, however, it is still highly significant. According to the statistical analysis (5% significance), the share of land occupied by sugarcane is dependent on sugarcane production for the distances 0-25 km. For distances over 25 km, the sugarcane land share was independent of sugarcane production. Another parameter analyzed was the suitability for sugarcane production (SSP). In general, the state of São Paulo has a quite high SSP. Most observations show a SSP close to 100%, and only a few observations showed less than 90%. Due to this low variation it was not possible to prove that the sugarcane land share is dependent on SSP, which means that in this model, the sugarcane spatial distribution will be considered as independent of SSP. Model equation for sugarcane land distribution The last step in creating a model for sugarcane spatial distribution was to combine all explanatory variables which have shown statistical significance into one single equation. The model thus included the variables density of sugarcane mills and the production per mill. The equations that describe the share of land occupied by sugarcane for each range of distance are shown in Table 1. Table 1: Model equation for sugarcane land distribution Dist. (km) Model equation R 2 0 to , " $% " $% " " 5 to " $% " $% " " 10 to " $& " $% " " 32

33 15 to to to to " $% " " $% " " $& " " $% " = Sugarcane land share (%); = Voronoi cell area (km 2 ) indicator of sugarcane mills density; = Sugarcane production per mill (1000 tons); As can be seen in the table above, the modeling has higher accuracy for short distances and is less precise the farther one gets from the sugarcane mill. It can be explained by the fact that the sugarcane land share is more significant at short distances than at longer distances. Once a model has been elaborated that describes the sugarcane spatial distribution with the help of the variables density of sugarcane mills and production of sugarcane per mill, it is possible to estimate the sugarcane land share at any point in time in the future, if the density of sugarcane mills and the production of each sugarcane mill are known for that period. So, the challenge here was to estimate those two variables Estimation of sugarcane mills density in the future By 2008, there were about 193 sugarcane mills and 15 of them were installed recently. According to the UDOP, there are 20 new sugarcane mills, which will be installed in the state of São Paulo in the near future. Based in that information, it is possible to estimate that soon there will be 213 sugarcane mills in the São Paulo State (Table 2). Table 2: Number of sugarcane mills in 2008 and in the future. Situation 2008 Near future Installed Mill New Sugarcane Mill 20 Recently installed mill Total

34 Most of the recently installed sugarcane mills are located in the western part of the São Paulo State. In addition, most of the new sugarcane mills are going to be built in this region as well. So, in a future scenario, there will be a quite significant increase in the density of sugarcane mills in the western part of the São Paulo State. Figure 11 below shows the current location of old sugarcane industries as well as the location of the recently installed mills and the location of new sugarcane mills. Figure 11: location of installed and new sugarcane mills Estimation of sugarcane production per sugarcane mill in the future Analyzing the estimation made by Torquato (2006), it is possible to see that there will be a quite large sugarcane expansion in the coming years. This estimation was made in 2006, and so far, it shows a good accuracy in relation to the real production in the state of São Paulo. To illustrate this fact, the actual production in the state of São Paulo during the first two years (IBGE) is shown beside Torquato s (2006) estimations in Figure

35 700 Sugarcane production (million tons) ,6 398,4 423,3 445, ,5 576,2 0 Year Real sugarcane production Prediction (Torquato) Figure 12: Prediction of sugarcane expansion for the coming years (Torquato, 2006) Obviously, for longer periods, estimations tend to be less accurate. This estimation was the one with the clearest methodology, which is why it was used as basis for estimating the future sugarcane production per sugarcane mill. Taking into account the first scenario created in this study, where it s assumed that existing sugarcane mills would keep increasing their production in a rate of 2%, all the new sugarcane mills will have been built and be operating at maximum capacity in three years, by Most sugarcane expansion, about 71%, would be led by new sugarcane industries and 29% by old sugarcane mills (Figure 13). In the second scenario (where old sugarcane mills keep increasing their production at an annual rate of 3%), the new sugarcane mills will have been built in four years and the sugarcane expansion by old mills and new mills will be pretty much the same. In the third scenario (annual increase in the sugarcane production of 4%), new sugarcane mills will be operating in full capacity by 2015, in six years. Most of the sugarcane expansion will be led by old sugarcane industries, as can be seen in the Figure

36 Scenario 3 (4%, 6 year) Scenario 2 (3%, 4 year) Scenario 1 (2%, 3 year) Sugarcane expansion (million tons) Installed Mills Recently Installed Mills New Sugarcane Mills Figure 13: Sugarcane expansion lead by installed and new sugarcane mills. Figure 14 shows the current sugarcane production as well as sugarcane production for the studied future scenarios. In each scenario, old sugarcane mills represent about 78% of the whole sugarcane production in the São Paulo State and new sugarcane mills (computing here also the sugarcane mills recently installed) represent about 22% of the sugarcane production. Scenario 3 (4%, 6 year) Scenario 2 (3%, 4 year) Scenario 1 (2%, 3 year) 2008/ Sugarcane production (million of tons) Installed Mills Recently Installed Mills New Sugarcane Mills Figure 14: Sugarcane production in the future different scenarios Prediction of sugarcane expansion per range of distance from the new mills Once the future density of sugarcane mills and the future production of sugarcane per mill have been estimated, it is possible, through the model presented in section 2.3.2, to estimate the future sugarcane land share; and correlating this with the current sugarcane land share it was possible to determine the expected sugarcane expansion for the coming years. The expansion is expected to happen mostly within the distance 36

37 range 0-15 km from the sugarcane mill, where about 80% of the expected sugarcane expansion will take place (Figure 15). 0,45 Sugarcane expansion (million hectares) 0,40 0,35 0,30 0,25 0,20 0,15 0,10 0,05 0,00 0 to 5 5 to to to to to 30 0 to 5 5 to to to to to 30 0 to 5 5 to to to to to 30 Scenario 1 (2%, 3 years) Scenario 2 (3%, 4 years) Scenario 3 (4%, 6 years) Distance from the sugarcane mills (km) Figure 15: Prediction of sugarcane expansion in relation to the distance from the sugarcane mill. In each scenario created in this study, it was observed that the sugarcane expansion will happen mainly in the western part of the São Paulo State, especially in the northwest of the state (Figure 16)Erro! Fonte de referência não encontrada.. The region of Ribeirão Preto is the region which will present the lowest level of sugarcane expansion, due the current large presence of sugarcane plantations in that region. The expected sugarcane expansion was estimated for each Voronoi cell, this information is presented in Figure 16Erro! Fonte de referência não encontrada.. Now that the spatial distribution of sugarcane expansion is known, it is necessary to identify what type of crop will be replaced by sugarcane. This question will be addressed in the following sections. 37

38 a) Scenario 1 b) Scenario 2 c) Scenario 3 Figure 16: Localization of the predicted sugarcane expansion for scenarios Sugarcane replacement tendency In order to estimate the future land use changes it was necessary to analyze the land use changes in the last five years in order to identify the sugarcane replacement tendency or what type of land use is most commonly replaced through sugarcane expansion. Based on the estimated map of land use in 2003 (which is the land use map of the sugarcane map of 2003), the observation can be made that the class other crops and the class pasture show the same percentage of occupied land in 2003 (Figure 17a). Then, analyzing the land use replaced by sugarcane expansion from 2003 to 2008, it can be observed that sugarcane mainly replaced areas belonging to the classes pasture and other crops (Figure 17b). The map also suggests expansion replacing native vegetation, areas of reforestation, water bodies and anthropic uses; however, considering the level of accuracy of the methodology adopted in this analysis, it is not possible to affirm that the sugarcane expansion onto these types of land actually occurred. Only the expansion in areas belonging to the classes pasture and other crops was conclusive. 38

39 Based in the information shown above, it can be assumed that sugarcane expansion will occur only on areas belonging to the classes pasture and agriculture(other crops). Analyzing the estimated land cover map of 2003 and land uses replaced by sugarcane expansion in the last five years, the estimation can be made that the sugarcane replacement tendency is 35% onto areas belonging to the class pasture in 2003 and 65% onto areas belonging to the class agriculture(other crops), as it is shown in Figure 17c. a) b) c) Figure 17: a) Share of area based on the estimated land cover map for the year 2003; b) Share of area based on sugarcane expansion in the last five years; c) Tendency of other land uses to be replaced by sugarcane expansion Prediction of land use changes caused by sugarcane expansion The prediction made in this study suggests that most of the sugarcane expansion in the next years will occur on areas used for agriculture (other crops) in In every scenario created in this study, about 65% of the sugarcane expansion will replace areas used for agriculture in 2003, and only about 35% of the expansion will take place on areas belonging to the class pasture in 2003 (Figure 18). In scenario 1, about 0.69 million hectares of areas belonging to the class agriculture (other crops) in 2003 will be converted to sugarcane crop over a period of three years. For the class pasture the number is 0.26 million hectares. Scenario 2 suggests that there will be a sugarcane expansion of 0.70 million hectares onto 39

40 areas belonging to the class agriculture (other crops) and 0.30 million hectares onto areas belonging to the class pasture. And in scenario 3 sugarcane will replace about 0.84 million hectares of areas belonging to the class agriculture (other crops) and 0.40 million hectares of areas belonging to the class pasture. 0,45 0,40 0,35 0,30 0,25 0,20 0,15 0,10 0,05 0,00 0 to 5 5 to to to to to 30 0 to 5 5 to 10 Land use changes (million hectares) 10 to to to to 30 0 to 5 5 to to to to to 30 Scenario 1 (2%, 3 years) Scenario 2 (3%, 4 years) Scenario 3 (4%, 6 yeas) Distance from the sugarcane mills (km) Pasture Agriculture Figure 18: Prediction of land use change caused by sugarcane expansion in the next years Using the methodology adopted in this study, it could be estimated that within the near future (3 to 6 years), the sugarcane expansion will keep replacing mainly areas used for agriculture in The study was based on the assumption that the expansion will follow the tendency in the last five years, with the same governmental policy, and without great variation in the price of the crops cultivated in the São Paulo State. However, both factors have strong influence on the land use changes caused by sugarcane expansion for the next years, so changes in the governmental policy and variation in the crops prices can potentially affect the prediction of land use changes made in this study. 3. SUGARCANE EXPANSION AND HUMAN DEVELOPMENT 3.1. Methodology The method used for assessing the impacts of sugarcane expansion on human development was to study the correlation between the expansion of sugarcane and the impacts on the local population measured in the Human Development Index (HDI). In order to make this assessment it was necessary to evaluate a long period, since such impacts only become visible over a longer period of time. Therefore, the studied period was , in the cases where there 40

41 were databases available that allow such evaluation. The databases were provided by the Institute of Applied economic research (IPEA) and the Brazilian Institute of Geography and Statistics (IBGE), and are as follows: Production of sugarcane for the years 1973, 1980, 1991 and 2000 from IPEADATA; Human Development Index (HDI) for the years 1970, 1980, 1991 and 2000, from IBGE; Population for the years 1970, 1980, 1991 and 2000, from IBGE; These data are offered for the municipality level, however, as the period of analysis is quite long and it is known that along this period there were changes in the perimeter of some municipalities, it was necessary to use the minimum comparable areas (MCA), which is the smallest area that had not suffered any change in the perimeter during the studied period. In total, the state of São Paulo had 645 municipalities and 568 MCA in the period The sugarcane expansion in the period had to be estimated, since the database used only offers data on the sugarcane production from 1973 to In order to make this estimation, the sugarcane expansion for the period of 1973 to 2000 was calculated. It was calculated as the difference in production between those two points in the time using the expression below. '( )$ ' ' ) '( $) is the sugarcane expansion from 1973 to 2000; ' ) is the production of sugarcane in 1973; ' is the production of sugarcane in The estimation of sugarcane expansion from 1970 to 2000 was made based on the assumption that the expansion from 1970 to 2000 took place linearly, using the follow expression: '()$ '( )$ )* '( )$ is the sugarcane expansion from 1970 to 2000; the number 27 refers to the number of years for which the sugarcane expansion is known, and the number 30 refers to the number of years for which the sugarcane expansion will be estimated. Since the impacts on MCAs with a low level of HDI can be expected to be greater than for those that already have a high level of HDI, it wouldn t be possible to The impact of sugarcane expansion is likely to differ greatly between the MCAs with a low level of HDI and those with a high level of HDI. It is therefore not possible to compare one to another, and so the MCAs were 41

42 divided into four equal parts based on their level of HDI in 1970, creating four groups. These four standard MCA groups were then measured separately. Subsequently, the changes in HDI from 1970 to 2000 were calculated for all the MCAs within each group, using the following expression: +,- )$,-,- ) +,- )$ denotes the changes in HDI from 1970 to 2000;,- ) the HDI in 1970; and,- the HDI in Then, to allow for comparisons among the MCAs within each group, the indicator SEI (Sugarcane expansion per inhabitant) was created. This was done by dividing sugarcane expansion by the average population for the studied period, using the following expression: '( )$ '(. )$ ( ) ) '(. )$ is the sugarcane expansion per inhabitants from 1970 to 2000; ) is population in 1970 and is the population in The effect of sugarcane expansion on the HDI was analyzed through linear regression using the following expression, where the response variable is +,- )$ and '(. )$ is the explanatory variable. +,- 1$2 '(. 1$ Results In 1970 the first year of the studied period the HDI in the São Paulo State was 0,598; however, there was a large variation between the municipalities. The HDI level was highest in the region around Campinas and São Carlos, and lowest in the regions of Pontal do Paranapanema and Apiaí. Figure 19 shows the normal distribution of HDI for the different MCAs of the São Paulo State and the limits between the respective MCA groups. The MCAs belonging to cluster 1 (C1) are those with an HDI lower than 0,410; cluster 2 (C2) gathers the MCAs with an HDI in the range 0,410-0,452, while cluster 3 (C3) gathers the MCAs in the range 0,452-0,507. The MCAs with the highest level of HDI are found in cluster 4 (C4), with an HDI of 0,507-0,723. Figure 20 shows the spatial location of C1, C2, C3 and C4 in the São Paulo State. 42

43 Figure 19: Normal distribution of HDI for the different regions of São Paulo State Figure 20: Spatial location of C1, C2, C3 and C4 in São Paulo State From 1970 to 2000 the sugarcane production increased with about 700%, from 23,4 million tons in 1979 to 189 million tons in Most of this growth took place in MCAs in the groups C3 and C4, where sugarcane production increased with about 45 million tons (Figure 21). In the MCAs in group C2, sugarcane production expanded with about 24 million tons, while in C1 this expansion was around 8 million tons (Figure 21). Analyzing the expansion of sugarcane in relation to the population it is possible to see that C2 and C3 are the regions that showed the highest production of sugarcane per inhabitant, about 15 tons (Figure 21) C1 C2 C3 C4 MCA cluster Expansion of sugarcane production (million tons) Population (million inhabitants) Expansion of sugarcane production per inhabitants (ton/inhabitant) Figure 21: Expansion of sugarcane production, population and the expansion of sugarcane production per inhabitant for each cluster. 43