UNDERSTANDING THE DYNAMICS OF THE BRAZILIAN MARKET FOR NATIVE FOREST WOOD: A CASE STUDY FOR MINAS GERAIS STATE, BRAZIL

Understanding the Brazilian market for native wood helps forest managers and can guide conservation policies. This study aimed to evaluate the spatial and temporal distribution of supply and demand of the Brazilian native wood for four products namely, charcoal, roundwood, fi rewood, and sawn wood. We used the consumed wood volume from 2006 to 2016 in Minas Gerais (MG) tracking its supplier in Brazil accounting for 31 million cubic meters. These data were used as input for calculating mean centers and performing grouping analysis. The wood supply centers for the energy sector in Brazil have moved from Cerrado to the Caatinga biome. This change is consistent with historical patterns of agricultural expansion in Brazil. The largest mesoregions producing wood were: East of Mato Grosso do Sul State, South of Goiás State, and Center-South of Bahia State. The solid wood supply centers have moved from the North to Midwest regions but remained within the Amazon biome for almost all the years. This pattern is likely a consequence of timber regulation practices in the Amazon. The mesoregions: Madeira-Guaporé and East of Rondônia State, Southeast and Northeast of Pará State, and North of Mato Grosso State were the most important suppliers of solid wood to MG. Wood consumption for energy is centered in the Center and North regions of MG State, especially in Sete Lagoas, a hub of pig iron and cement industries. Solid wood products are dispersed in MG State, being Belo Horizonte, Conselheiro Lafaiete, Ubá, and Uberlândia the distribution centers of native wood for the furniture industry within MG. The spatial statistics helped to exploit the activity of the native wood market in Brazil and provides essential information for decision-makers that was not previously available.


INTRODUCTION
The Brazilian native wood market incorporates multiple ventures hold diff erent positions in the production chain, varying in structure and capacity, but dependent of a common natural resource: the wood (Kies et al., 2009;Canova and Hicley, 2012;Köhl et al., 2015;Bösch et al., 2015). In order to maintain a continuous fl ow of wood and other forestderived benefi ts, it is essential to understand market dynamics and nuances shaping relationships between the native wood supply and demand (Syrbe and Walz, 2012).
The wood supply depends on forest biophysical conditions and the potential changes imposed to its structure, which may have anthropic or natural causes. If wood supply signifi cantly decreases, companies and society`s demands may not be met (Burkhard et al., 2012). On the other hand, the status of a forest depends not only on its ability to supply wood but also on society's needs (Paetzold et al., 2010). In this context, variation in the human desired level of provision, increases or decreases the demand for wood and other forest products, impacting economic factors, policies, among others (Curran and de Sherbinin, 2004).
Transport wood over long distances, which may also increase its price, contributes to the existence of a geographical discrepancy between supply and demand regions (Syrbe and Walz, 2012). This trend is typical in Brazil, where consumer centers are mainly located in the southeast part of the country, considerably distant from the supplier regions, often located in the Amazon Region (IBAMA, 2014;IBGE, 2018). As a consequence, harvest environmental impacts are perceived in places far from the large population centers (Burkhard et al., 2012). Identifying regions that are being over explored for wood provision would help to guarantee a continuous fl ow of supply. These specifi c areas need to be conserved and managed according to their productive potential (Van Jaarsveld et al., 2005).
Due to this spatial characteristic of the native wood market, mapping supply and demand distributions and its changes over time potentially provide direct measures of the human activity over the ecosystem (Burkhard et al., 2012). This visualization can be used to support decision-making concerning conservation policies and to allocate production resources in order to take market opportunities as advantage . Unfortunately, this type of research is rare in Brazil and the scientifi c and political challenges that arise require innovative responses and approaches.
Wood-based industries tend to be clustered on specifi c regions, forming networks and regional production centers (Bösch et al., 2015) that may vary through time and space. Spatial statistics tools can include the spatial component of the native wood market and provide a better understanding about supply and demand, revealing not only changes but also its trends (Kies et al., 2009). In this perspective, this study investigates the spatial and temporal dynamics regarding supply and demand of native wood in Brazil, from 2006 to 2016, considering Minas Gerais State (MG), the second-largest consumer of native wood in the country, as the demand center.

Data acquisition
Data of native wood consumption in MG State, from 2006 to 2016, were compiled from the DOF System (Document of Forest Origin), based on "Portaria MMA nº 253/2006" published in 08/18/2006. This document carries information about the commercialization and legal transport of native forests resources in Brazil since 2006. It includes wood origin in the country and destination all over MG State for energy-related (charcoal and fi rewood) and solid wood products (roundwood and sawn wood). Origins and destinations were imported into GIS environment, ArcGIS® 10.6.1 software (ESRI -Redlands, CA, USA). Records were manually checked for quality control in order to ensure information fi delity when creating the geospatial database with 770,413 records ( Figure 1).

2.2.Spatial analysis
The Brazilian geographic space was divided according to administrative limits of its mesoregions, in order to analyze the wood origin in the entire country and its destination in municipalities of MG State ( Figure. 1). Data spatial pattern was analyzed using two spatial statistics: (1) mean center and (2) grouping analysis. The fi rst one gives a measure of spatial central tendency, while the second one a measure of spatial autocorrelation.

2.2.1.Mean center
Mean center, or centroid, (Hart, 1954;Yue et al., 2003;Wang et al., 2010;Ayhan e Cubukcu, 2010;Arima et al., 2011;Lausch et al., 2013;Martins et al., 2018) is a statistic that presents only the data tendency and not the exact point of the greatest value. The main applications refer to analyzing spatiotemporal change patterns measured by the central tendency value. For this reason, the weighted mean center was used with wood volume as the weighting factor. The model is defi ned as (Equation 1): Eq.1 where (X w ; Y w ) are the weighted mean center coordinates; x i and y i are the coordinates of the feature i; n is the number of features; w i is the weight of the feature i. Compilado de: http://ibama.gov.br/fl ora-e-madeira/dof/relatorios-dof#info-transporte.
Weighted mean center of wood origin and destination were calculated for each year of investigation (2006 to 2016), based on the wood volume (m 3 ) for each mesoregion and municipality, respectively.

2.2.2.Grouping analysis
Grouping analysis identifi es similarity or divergence between sampling units from diff erent spatially defi ned locations and separates statistically diff erent clusters (Wu et al., 2015). This analysis was done based on wood volume for each category (charcoal, fi rewood, roundwood and sawnwood) and by using the Euclidean distances between features, without a maximum distance threshold. Four groups were created: very high, high, moderate and low. R 2 was calculated to show data variation and grouping accuracy (Equation 2 to 4).

Eq.2
Where SST represents the diff erence between groups and SSE is the similarity within a group (Equation 3 and 4): where n is the number of features; n c is the number of groups; n i is the number of features in group i; n v is the number of variables used to group the features; V Ij k is the value of variable k of feature j in group i; (V k ) is the average of k; e (V I k ) is the average of k in group i (ESRI, 2019).

Map production
Spatial statistics results were presented as maps.
Features were projected into Albers conical projection. For wood origin locations in Brazil, central meridian was -55° and standard parallels were -1° e -27°. For the wood destination locations in MG State, central meridian was in -45.45° and standard parallels -15.68° and -21.47°.

3.1.Dynamics of the Brazilian native wood destined to Minas Gerais State
The volume of native wood allocated to MG State market was about 31 million cubic meters from 2006 (when DOF started being used) to 2016. The main products derived from this wood were charcoal (87%) and sawn wood (13%). Trades in charcoal, roundwood and sawn wood signifi cantly decreased between 2008 and 2009, with falls of 45, 94% and 99%, respectively. Specifi cally, the sawn wood market followed the economic cycles of expansion and recession, showing recovery in 2010 and a further decline after 2014 ( Figure 2).

3.2.Spatial analysis of the native wood origin
The spatiotemporal change in the mean centers of native wood origin presented two diff erent patterns: oriented for charcoal and sawn wood, and not oriented ("zigzag pattern") for fi rewood and roundwood ( Figure 3A to 3D). Along the 11 years, charcoal and fi rewood production centers have moved from the Midwest region, in Cerrado biome, towards the Northeast of Brazil, in Caatinga biome, with fi rewood without a clear distribution pattern. The migration of round timber production centers occurred from the North Region, in Amazon biome, towards the Midwest, in Cerrado, without a clear distribution pattern too. For sawn wood centers, the shift was from North to Midwest region of Brazil, remaining in the Amazon biome for almost all years.
Grouping analysis of wood origin in Brazil was highly accurate (R 2 > 0.96). The production of charcoal and fi rewood in Midwest, North and Northeast regions of Brazil were defi ned as high and very high, mainly in the East of Mato Grosso do Sul State, South of Goiás State and South-Central of Bahia State mesoregions ( Figure 4A to 4D). The high and very high production classes of roundwood and sawn wood were concentrated in the Midwest and North regions of Brazil. Specifi cally, the mesoregions Madeira-Guaporé, East of Rondônia State, Southeast and Northeast of Pará State, and North of Mato Grosso State were the most important solid wood suppliers to MG State. Together, these mesoregions provided approximately 81 and 41% of all roundwood and sawn wood for the state, respectively (Table 1).

3.3.Spatial analysis of the native wood consumption in Minas Gerais State
The spatiotemporal change for native wood destination mean centers did not present an oriented pattern for any product ( Figure 3E to 3H). Spatial shifts in charcoal and sawn wood mean centers were relatively low (63 and 44 km, respectively), remaining around the urban core of MG State: West and Belo Horizonte Metropolitan. For fi rewood and roundwood, mean centers have moved from West and Vale do Rio Doce to North and Triângulo Mineiro, respectively.
Regarding the grouping analysis of consumption sites (R 2 ≥ 0.89), it was found that most charcoal and fi rewood were concentrated in a few municipalities within MG State ( Figure 4E to 4H). Remarkably, Sete Lagoas municipality consumed, approximately, 36% of all native wood charcoal produced in this period. Demands for roundwood and sawn wood had a diff erent pattern, with the highest consumption values being evenly distributed over geography. Municipalities of Belo Horizonte, Conselheiro Lafaiete, Uberlândia and Ubá stood out for the high demand of solid wood (Table 1).

Dynamics of native wood supply in Brazil
The native wood market destined to MG State experienced a signifi cant decline, from 2006 to 2016. Several factors may have driven this reduction in wood consumption during this period, especially the economic downtrend (Köhl et al., 2015). The decrease in charcoal, roundwood, and sawn wood supply, especially from 2008 to 2009, is consistent with this statement (Figure 2). Certainly, the 2008 global economic recession aff ected the forest sector in many developing countries (FAO, 2012; Arima et al., 2014). In Brazil, an exploratory case study in the Amazon reported that this fi nancial crisis signifi cantly reduced native wood production, leading to a general unemployment in many forest-dependent regions (Canova and Hickey, 2012). The continuous decrease in native wood market after the crisis recovery, in 2010, may also be explained by the set of policies implemented by the Brazilian government, such as the internal wood market regulation, the creation of protected areas, the suppression of illegal deforestation and the defi nition of credit barriers to municipalities on deforestation edges (Lapola et al., 2014). As an example, two highly broadcasted operations carried out in 2008 resulted in the apprehension of nearly 20,000 m³ of illegally harvested logs (Arima et al., 2014). The advances in forest exploitation control and licensing strictness have contributed to an increasing change in the production system, expanding the opportunities to establish plantation areas with exotic species, such as those of Eucalyptus genus, which are especially replacing native wood species for charcoal production (IBGE, 2017).
Native wood consumption in MG State previous to the fi rst economic recession (2007)(2008) was considerably high, according to consumption dynamics from 2006 to 2016 (Figure 2). When applying the average central statistic, it is clear that during this period, the native wood origin mean centers were located in the Midwest and North regions of Brazil, corresponding to Legal Amazon and Cerrado areas (Figure 3). At that time, the country was at the end of an intense deforestation process driven by widespread agriculture and livestock expansion mainly in Cerrado and Amazon Regions, with peak years occurring between 1990 and 2000 (Lapola et al.,  (Santos and Santana, 2009;Ribeiro et al., 2011).
The high accuracy of grouping analysis concerning supply and demand of native wood (expressed by R 2 ≥0.89) indicates that the cluster analysis was eff ective in separating groups of wood producers and consumers (Wu et al., 2015). Regarding the origin sites, there was high charcoal production in the East of Mato Grosso do Sul State mostly due to deforestation by livestock purposes , which was stimulated by global demand and supportive government policies at the beginning of the decade (Fialho, 2014). The high production of fi rewood in Midwest and Northeast of Brazil, particularly in Bahia State, in turn, is a consequence of fi rewood dependence as an energy source for local domestic and economic purposes. Native biomass is relatively inexpensive in these regions, which favors its use as an alternative fuel (fi rewood) (Coelho Junior et al., 2018). The great production of solid wood in Madeira-Guaporé, East of Rondônia, Southeast and Northeast of Pará, and North of Mato-Grosso mesoregions is historically driven by the expansion of pastures and soybean plantations (Arima et al., 2011;Lima et al., 2019), associated with a traditional forestry sector leveraging the economy of many Amazonian municipalities (Ribeiro et al., 2011;IBGE, 2017).

Dynamics of native wood demand in Minas Gerais State
The spatiotemporal dynamics of native wood consumption in MG State is directly associated with the development of wood-demanding industries within the state. The mean centers of charcoal and sawn-wood were strongly infl uenced by the high consumption of Sete Lagoas and Belo Horizonte (capital) municipalities, where most of the points were concentrated over the years. Such pattern/behaviour demonstrates the large industries resilience when facing supply reduction after the economic recession. Furthermore, it also shows the effi ciency in changing the production systems of forest plantations , and governmental programs to guarantee investments in civil construction in order to mitigate the crisis impacts (Young, 2012).
The grouping analysis regarding native wood demand shows distinct patterns when products related to energy (charcoal and fi rewood) are compared to solid wood (roundwood and sawn-wood). Energy consumption groups are located in the Center and North of MG State, while solid wood clusters are dispersed. In general, productive clusters are conditioned to a complex set of regional factors such as availability and costs of qualifi ed workforce, connection with large markets and local technological progress (Kies et al., 2009;Li et al., 2020). In this research, the infl uence of regional factors on wood consumption was not studied, however, this may support extra explanation for the observed spatial pattern.
The main use of charcoal is for pig iron production, ferroalloy, and cement for civil construction (Fontes et al., 2005;Gomes et al., 2007). Consumption of this product in MG State was concentrated around Sete Lagoas and Divinópolis, precisely where pig iron, steel centers, and cement industries are located (Rezende et al., 2005). Firewood highest demands occurred close to origin centers in the Brazilian Northeast region, due to its low value-added, which does not justify its transport over long distances (Coelho Junior et al., 2018). The greater dispersion of solid wood consumer groups suggests an existence of small to medium primary wood processing industries in the state. The higher demand in some micro-regions, such as Belo Horizonte, Conselheiro Lafaiete, Ubá and Uberlândia are explained for their importance as input suppliers (IBGE, 2008) and/or industrial potential, mainly concerning the furniture manufacturing sector (Bastos and Almeida, 2008).

CONCLUSION
This study provides crucial and previously inaccessible information to understand the forest sector in Brazil through spatial statistics application and historical data. This approach allowed us to show oscillations in native wood market, map spatial shifts and clusters for supply and demand, concerning roundwood, sawn wood, fi rewood and charcoal. Thus, historical dynamics may reveal the eff ect of forest-related factors, including economic recessions and control policies, whose eff ect suggests the gradual replacement of native wood production by planted forest in Brazil. The highly productive suppliers for native wood that have been identifi ed , indicate the most suitable sites for upcoming forest policies concerning regional development, traditional market maintenance and deforestation control. Writing, review and editing -Thales Guilherme Vaz Martins, Geraldo Gonçalves dos Reis, and Maria das Graças Ferreira Reis.

ACKNOWLEDGEMENTS
This study was fi nanced in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior -Brasil (CAPES) -Finance Code 001.