RECENT REGIONAL TRENDS Of LAND USE AND LAND COVER TRANSfORMATIONS IN BRAZIL

Regional trends of land use/land cover transformation in Brazil during 2001—2012 were analyzed in the following order: 1) identification of the types of transitions for different land use and land cover categories and aggregated groups of transformation processes based on the Global Land Cover Facility datasets, 2) analysis of national agricultural and forestry statistics to find out the principal socioeconomic drivers, 3)  land cover and land use data merging to elaborate comprehensive typology of land use/land cover changes on a regional level. The study evealed 96 types of transitions between land cover categories, aggregated into 10 groups corresponding to driving processes. It was found that the main processes of land cover transformations is related to both natural and anthropogenic origins. Cropping and deforestation are anthropogenic processes, flooding and draining are the principal natural ones. Transformation of cultivated lands and reforestation are combined natural and anthropogenic. The contribution of natural factors is higher in the states of the North (Amazonia) and the Northeast macroregions; in the Center-West and the South anthropogenic factors make larger contribution. We have also detected considerable land use/land cover changes caused by agricultural development in densely populated states of the Southeast and the South. In both macroregions planted area expands due to increase of soybeans and sugar cane production, while area of pastures is shrinking. The trends of transformations of agricultural land use revealed as a result of statistical data analysis, match with transitions of land cover categories belonging to the aggregated group of cropping processes. Transformations of land cover types with predominance of shrub vegetation were the most problematic to interpret because of lack of comparable statistical data on pastures. KEY wORDS: land use/land cover regional dynamics, Brazil, MODIS, agriculture, forestry CITATION: Oksana Klimanova, Alexey Naumov, Yulia Greenfieldt, Rachel Bardy Prado, Daria Tretyachenko (2017) Recent regional trends of land use and land cover transformations in brazil. Geography, Environment, Sustainability, Vol.10, No 4, p. 98-116 DOI-10.24057/2071-9388-2017-10-4-98-116 SU ST A IN A BI LI TY


INTRODUCTION
Conversion of virgin lands for agriculture followed by soil degradation, indiscriminate use of fertilizers and water pollution, loss of biodiversity, along with climate change has exerted strong pressure on terrestrial and aquatic ecosystems, compromising their sustainability.Over the past 50 years, humans were changing ecosystems more rapidly and extensively than in any comparable period of time before.This has resulted in a substantial and largely irreversible loss in the diversity of nature on Earth (MEA 2005).
The above mentioned transformations, mostly related with development of agriculture and forestry, require elaboration of appropriate action strategies, e.g.agrienvironmental sustainability (Stallman 2011).Aiming the sustainability of rural landscapes as reconciliation of ecosystems preservation and satisfaction of demand in food, fiber, energy, water, raw materials and other goods and services, is a fundamental approach for studies of dynamics and functions of ecosystems and landscapes (Barkmann et al. 2004).
Modern studies of land use and land cover (LULC) transformation processes and specific features of land resources utilization are based on the analysis of remote sensing data of various spatial resolutions.Land cover means a present-day mosaic of vegetation cover, both natural and anthropogenic, the last one including agricultural crops, of settlements, industries and infrastructure, of lands without vegetation (fallow, glaciers, etc.) and water surfaces (Channan et al. 2014).Satellite imagery is a tool for highly precise identification of areas where land cover transformations are caused by particular anthropogenic and/or natural factors.However the interpretation of obtained results and the discovery of reasons of LULC changes, especially at the regional level, are possible only through the complex analysis of natural and socio-economic features of the territories.
In this context, characteristics of agriculture are considered as research priority for LULC studies.Geographical analysis reveals spatial heterogeneity of agricultural development, and provides bases for assessment of its role in land cover changes within different regions.This could be done with a great level of precision as statistical data on land use by administrative-territorial units become available for the increasing number of countries.As estimated in 2006, agricultural censuses were held in 134 countries of the world, and most of them contained data by states, provinces and similar units.For 2020, the UN Food and Agricultural Organization (FAO) plans to launch global data collection through the World Agricultural Census (World Programme FAO for the Census of Agriculture).
In recent years, several attempts were made to integrate the statistical and remote sensing data in order to evaluate geographical variety of agricultural production systems.An example is the project on global mapping of the differences in productivity of the main crops fulfilled by US and Canada researchers (Monfreda et al. 2008;Ramankutty et al. 2008).Factors influencing the yield differences, such as water and nutrients supply, were also taken into account.However, the yield was considered as a function of biological productivity, and the socio-economic features of agricultural production were ignored.A number of similar global projects are carried out under the auspices of FAO: the GLC2000 (studying the global land cover for the assessment of the state of ecosystems), the Global Livestock Information System, and the so-called Agro-maps project, involving compilation of universal cartographical database on agriculture at the "subnational" level (Agro-maps).The International Food Policy Research Institute (IFPRI) also has launched the program of global localization of harvest data (Global Mapping... 2006).
Joint analysis of remote sensing and agricultural statistics is also applied at the national level.E.g., in Brazil, the Companhia Nacional do Abastecimento uses remote sensing and analysis of the municipal statistical data for estimate crops productivity, being the GeoSafras (yield forecasting) and SIGABrasil (mapping of agricultural indicators) projects two basic components of this research (CONAB).
LULC studies are quite common at the level of one country, but usually they emphasize only main drivers of land use and land cover changes.E.g. publications on land use dynamics in Brazil focus on deforestation, expansion of soybeans and other commercial crops (Naumov 2005;Lapola et al. 2014).More often, rather short periods of time, usually 3-5 years are covered (Richards et al., 2014).At the same time, large-scale studies concerned with the detailed analysis of land cover changes basing on the remote sensing data for certain territories are common (Ferreira et al. 2009).
The objective of our paper is to evaluate regional trends of land use/land cover transformations in Brazil during 2001-2012, focusing on agricultural and forestry development as their driving forces.The methodology was based on the GIS assisted extrapolation of the big array of statistical data for small administrative-territorial units to the natural (soil and landscape) areas identified through the interpretation of space imagery.

STUDY AREA
Extensive territory of Brazil (8.5 million sq.km) consists of both densely populated areas and natural ecosystems of global value in different biomes (Fig. 1).It is rapidly colonized and developed, generally because of agricultural expansion.Understanding the current LULC dynamics will enable assessment of anthropogenic impact on the natural-territorial complexes and could serve as a base for the forecast of the future LULC changes.(Ferreira et al. 2014).

MATERIAL AND METHODS
The methodological scheme of the study includes three main stages: 1) identification of the types of transition between the land use and land cover categories and of aggregated groups of transformation processes on the basis of the Global Land Cover Facility (GLCF) data, 2) analysis of national agricultural and forestry statistics aiming to find out the principal socioeconomic LULC drivers, 3) integration of both kinds of above mentioned data and elaboration of a complex typology of LULC changes.Understanding LULC transformation at different territorial levels requires to use remote sensing data (space imagery along with databases derived from their interpretation).The use of global land cover databases projected at the regional level seems quite adequate for such big countries as Brazil, which territory ranks fifth in the world.The state of land cover in Brazil during 2001-2012 was evaluated using the open data of the Global Land Cover Facility (GLCF) obtained through the MODIS satellite survey with the resolution of 5'x5' .These sources were considered as the most reliable basing on the comparison of a series of heterochronous global land cover data (Alekseeva et al. 2017).The data were analyzed using the environment of the ArcGIS Desktop for Desktop Spatial Analyst.
The legend to the land cover map compiled on the basis of MODIS data includes 17 classes, corresponding to the classification of the International Geosphere-Biosphere Program (IGBP) (Loveland et al. 2000).The classes were identified according to the height of trees and shrubs, canopy coverage of a forest stand, etc. (Table 1).
To find out social and economic drivers of LULC transformation, the analysis of national agricultural statistics was carried out.Changes in the harvested area under main grains (soybeans, corn, sorghum, wheat, rice, beans and others) and under other commercial crops (soybean, sugar cane, cotton, coffee) were analyzed.The abovementioned crops, except rice, account for a considerable share of cultivated land in Brazil, and the essential changes of growing and harvested area were typical for all of them during the studied period.According to the Ministry of Rural Development of Brazil, in 2009 soybeans covered 37.3% of all acreage under seasonal cultivation, sugar cane -14.6 rice -4.9 and cotton -1.4% (Estatísticas... 2011).Sugar cane is cultivated on the same field during 4-5 years; it grows up after harvest and is poorly distinguishable from perennial crops on the space imagery.Rice is particularly interesting for our study, because the expansion of this irrigated crop causes a specific type of land cover transformation.Corn accounted for 23.4% of seasonally cultivated lands, however in Brazil it is most often grown up in one-year crop rotation with soybean.
Because of overlapping of cropping area under these two crops it was not possible to consider data on corn cultivation separate from soybean.Coffee ranked first among perennial crops accounting for 35.4% of the total area (Estatísticas... 2011).Statistical data on afforestation were also used.
Unfortunately, data on the area of pastures in Brazil are not published annually; they are available only in census years which do not coincide with the first and the last years of the period under our study.Therefore it was impossible to analyze the dynamics of areas under pasturing.
The processes described on the basis of statistical data analysis, i.e. expansion of agriculture and creation of forest plantations, do not cover all the types of changes identified through the use of MODIS data.However, we decided it was feasible to consider these two processes as the main drivers of LULC changes in Brazil.

RESULTS AND DISCUSSION
Structure and main process related to the LULC dynamics The comparison of data for 2001-2012 made it possible to identify the types of land cover changes; each of them was assigned a two-digit index according to the IGBP classification (Table 1).The first digit corresponds to the land cover type in 2001, and the second -to the same in 2012.
For example, type "2/9" means the change "evergreen broad-leaved forest -savanna"; 12/5 -"croplands -mixed forest".In total, 96 types of changes were identified and subsequently combined into 10 groups according to the processes causing the changes (Fig.   Each type of changes and group of processes were characterized by: 1) area of changes; 2) share of each type of changes and group of processes in the area of each state; 3) share of each type of changes and group of processes in the total area of all types of changes in particular state.Similar approach was already used on the global scale, data generalized for the main biomes (Klimanova et al. 2017).Final results of the complex analysis of LULC changes are presented in Table 3.
The area of changes revealed by comparison of land cover data on MODIS space imagery for all categories of lands in the states of Brazil differs significantly: in 3 states of 27 (Mato Grosso, Bahia, Pará) it exceeds 100.000 sq.km, in 8 states it is less than 10.000 sq.km.Minimum area of changes registered was 1.600 sq.km in the Federal District Brasilia.However, the total area of this units just 5.800 sq.km.Minimum share of the area of change in the total area of the state is 0.9% (Amazonas) and the maximum -36.2% (Paraná).In the North and Northeast there are states with low, average and high degree of land cover change, and in all states of the South the share of the area of changes is the highest in the country.
Regional differences are also legibly traced in the basic processes causing land cover changes.In the states of the North deforestation makes the greatest contribution to these changes.The contribution of agriculture is the most valuable in the state of Tocantins, which is located on the transition zone between the Amazon forest (port.Amazônia) and Savanna (port.Cerrado) biomes.In the states of Northeast processes of shrub vegetation change (considered in the our analysis as "others") have the greatest value; these states differ markedly from the rest of the country.Interpretation of driving forces of these changes is complicated; perhaps, they are caused by the natural transformation of shrubby vegetation.We can also assume that a considerable proportion of areas here were converted into pastures, which could hardly be distinguished from shrubby savannas or grasslands, especially during first years.In some states of the Northeast (Maranhão and Piauí) deforestation plays the most important role, in some other states of this macroregion (Pernambuco and Alagoas) the main process is conversion of deforested lands to agriculture.
The states of the Center-West which lay in the area of predominance of the Savanna biome, since 1980s are characterized by the most intense agricultural development, causing drastic changes in the structure of land use.In the state of Mato Grosso the deforestation is very important accounting for a half of area of all land cover changes.
In the South and Southeast the role of agricultural development is also high, and the considerable proportion of already cultivated land is characterized by the transformation of land cover as a consequence of replacement by traditional crops by the new ones, also of conversion to cultural pastures.
The contribution of flooding or, on the contrary, draining as driving forces of transformation processes is more complicated to determine and requires verification.The pattern of these processes could be related to the annual changes of river flow and respective expansion or shrinking of wetlands.It could be implicitly proven by the greatest contribution of such processes to the structure of land cover changes in the states of the North, where they are obviously caused by natural factors.Reforestation could be considered as a result of both expansion of forest plantations establishing and natural vegetation dynamics.

MAIN DRIVERS Of LULC CHANGES
The calculations reflecting the dynamics of agricultural and planted forest areas by the states of Brazil made it possible to reveal the main types of changes in agricultural and forestry LULC.The results of this analysis can be found in Table 4.   where soybean and cotton are expanding, also due to increase of the area of sugar cane plantations in a coastal strip along the Atlantic Ocean.At the same time, the area under seasonal crops decreased significantly in the states of Alagoas, Pernambuco, Rio Grande do Norte and Ceará.In the state of Bahia, particularly in its western part, the area of coffee plantations increased.In this state, and also in the state of Maranhão, large areas of forest plantations (307.000 and 168.000 ha respectively) were created during the first decade of the 21 st century.

AGRICULTURE
In the states of the Southeast different directions of the dynamics of the LULC changes was detected.In the state of Minas Gerais, area under grains (mainly soybeans), sugar cane, and both coffee and forest plantations increased considerably.The total increase of the area under seasonal crops and sugar cane amounts to about 1% of this state area; coffee and forest plantations area registered 2.5% increase.After 1970-s, when the large-scale colonization of Amazonia started, the problem of deforestation in Amazonia draw great attention of the world community.
Nowadays, Brazilian government implemented public policies and environmental control which enabled reducing the deforestation of the Amazon (Laurance et al. 2002;Chazdon 2008;INPE 2015).In 2016, the Environmental Monitoring Program of the Brazilian Biomes was created by the Ministry of the Environment (MMA), whose purpose is to expand the mapping and monitoring of deforestation and land cover for all Brazilian biomes.
It is also important to stress, that regional differences in deforestation rates in Brazil depend on the nature protection legislation and public policies.Forest Code (BRASIL 2012) established a legal protection norm of 80% of total area of each rural land property for the Amazon forest biome, of 35% for the areas under Savanna, located in the socalled Legal Amazon region, and of 20% for the other areas under Savanna for the rest of biomes.Moreover, the protection of native vegetation in public reserves is consistently low outside the Amazon and below 5% for most biomes.As the largest part of area under native vegetation in each Brazilian biome is still found in private ownership, it is hard to control the accomplishing of the nature protection legislation (Ferreira et al. 2012), and to get the data for LULC analysis.In total, from 21 to 30 million ha is the estimated size of areas in Brazil, where natural vegetation should be restored and/ or protected (Sparovek et al. 2010;Soares-Filho 2014) The analysis of statistical data on protected areas of Brazil allows to evaluate their possible contribution, in particular, on reforestation processes (Protected Planet 2015).During 2002-2012, 17 natural reserves, 80 national parks, 205 managed natural areas, 83 protected areas with sustainable use of natural resources were established, mostly in the North and Southeast macroregions.However, the status of approximately 700 protected areas is not defined yet.We can assume, that the reforestation processes were driven mainly by expanding of protected areas in Amazonian states and in three states of the Northeast (Maranhão, Pernambuco, Piauí).
Forest plantations, created for production of commercial timber, cellulose and wood charcoal, which rapidly expands in Brazil, is another driving force of the LULC changes.In 2003 total national area of forest plantations in Brazil reached 1.6 million ha, 73.7% under eucalyptus and 25.3% under pine, with minor share of araucaria plantations (mostly in the state of Paraná) and acacia (in the state of Amapá) (BRASELPA 2003).By 2012, total area under plantations of eucalyptus and pines in Brazil reached 6.7 million ha (ABRAF 2013).

CONCLUSIONS AND fINAL REMARKS
The complex analysis of LULC changes in Brazil during 2001-2012 comprised land cover dynamics study based on satellite imagery data, and land use dynamics study, based on agricultural statistics data.For both approaches, the results were generalized at the regional level, being the states of Brazil basic units of research.Most of these states, surpass by area medium and even large European countries, which allowed to draw general conclusions on LULC changes and their driving forces.Nevertheless, we got reliable results, which can be discussed and developed in the further studies.
According to the data on conditions of land cover, in 2001-2012 the main processes of its transformation in the states of Brazil were driven by both natural and anthropogenic factors.Agriculture and, in large extent, deforestation are among the processes of anthropogenic origin, flooding and unwatering are the principal natural processes.Changes of vegetation cover on agricultural lands and reforestation are the processes of mixed natural and anthropogenic nature as they depend from both natural dynamics of vegetation, along with creation of forest plantations and conversion of pastures.
In general, the contribution of natural factors on LULC changes in Brazil is higher in the North (except Tocantins) and the Northeast (except Piauí and Pernambuco) macroregions.In Center-West and South macroregions anthropogenic factors make larger contribution.The latter are: development of agriculture, generally colonization as the country still has the considerable resources of undeveloped lands, and only 1/3 of its territory is used for cropping; conversion of pastures into agricultural lands; conversion of natural pastures into improved ones.
Comparison of the data on particular types of land cover changes and on the LULC transformation processes related with agricultural development shows for the most densely populated and developed states of Brazil the general correlation in trends, rather than the complete coincidence.According to the MODIS data, the areas with an increase of forest cover are larger by size, than statistics gives for forest plantations expansion.We assume that this is an evidence of the dualistic nature of this process, which depends from both natural and anthropogenic factors.Generally, the size of area of changes in the structure of cultivated lands is statistically less, than the area of transitions included in the "cultivation" category estimated assessed through the remote sensing data.
The suggested technique of comparative analysis of land use changes basing of the remote sensing data of medium spatial resolution and the data of agricultural statistics has shown its potential for identification of the main drivers of LULC transformation at the macro-regional level.Based of comprehensive geographical approach, it could be applied for different purposes, such as environmental assessment and territorial planning.

Table 2 . The main processes of LULC transformation and corresponding types of changes
* Including minimum and zero plowing.** To avoid the distortion of results because of relatively small size of areas of settlements, the impact of urbanization process is not specially considered in this study.

Table 3 . Structure and the main processes of land cover changes in the states of Brazil during 2001-2012
* Total area of changes within each state equals 100%.**Otherprocessesincludenumberedas6,7,9,10 in the Table2.***Hereaftertypes of transitions match with the processes listed in Table2.***Comparison of the sum of cropland and planted forest area changes estimated by analysis of statistics (CAS and FAS) and remote sensing data (FAS and FAM), %.SUSTAINABILITYcover changes in Brazil, being this sector a trigger of the recent economic growth of the country and providing 28% of the total national value of exports(FAO 2012).Its contribution to the LULC changes can be considered as of highest value for the states of the Center-West and the South, but it was also significant for different states in all other macroregions.In the North (Amazonia) the LULC changes were not so significant in relation to the size the states, which make part of this macroregion.The only exception is the state of Tocantins, where the cultivated area under seasonal crops during 2001-2012 nearly doubled and has reached more than 0.5 million ha, or nearly 2% of the total

Table 4 . Main types of changes in agricultural and forestry land use in the states and regions of Brazil, 2001/2002 -2011/2012 1
The most developed parts of this state have experienced conversion of pastures into agricultural lands; at the same time, unused slopes of the Plateau of Brazil started to be used for new plantations of coffee, which made the Minas Gerais state ranking first in Brazil by the volume of coffee harvest.Expansion of the area of sugar cane plantations by 1.3 million ha, or by 40% compared with 2005, was the main process of land use changes in the state of São Paulo, harvesting this state ¾ of this crop in Brazil.At the same time, the area of coffee plantations decreased both in São Paulo and in the Espírito Santo states.A considerable increase of the area under forest plantations in Minas Gerais, São Paulo and Espírito Santo states (855.000,240.000 and 78.000 ha respectively) is worth attention.It can be explained by both the growing demand for wood charcoal, and the development of and the cellulose and paper production.by 5% in Mato Grosso do Sul and Goiás states.The area under sugar cane has also considerably expanded here.In all states of this macroregion large forest plantations were established, reaching those nearly 1 million ha in the state of Mato Grosso do Sul.
1 Acronym, meaning the first letters from the names of four states: Maranhão, Tocantins, Piauí and Bahia.109SUSTAINABILITYalmost