Journal of Agricultural Science and Technology B 6 (2016) 400-410 doi: 10.17265/2161-6264/2016.06.005 D DAVID PUBLISHING

Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance

Vitor Augusto Cordeiro Milagres1 and Evandro Luiz Mendonça Machado2 1. Department of Soil Science, Luiz de Queiroz College of Agriculture, University of São Paulo, Piracicaba, 13418-900, São Paulo, Brazil 2. Forest Engineering Department, Federal University of Jequitinhonha and Mucuri Valleys, Diamantina, MG, 39100-000, Brazil

Abstract: Brazil is one of the countries with the greatest biodiversity, being covered by diverse ecosystems. Native trees commercially planted generate numerous benefits for communities, providing cultural, recreational, tourism riches, as well as ecological benefits, such as nutrient regulation and carbon sequestration. Thus, this work aimed to generate potential distribution modeling for the Brazilian forest species, to provide information that will serve as a strategy for conservation, restoration and commercial plantation of them, that is, encouraging the use of legal native species in the forest sector. Eleven tree species and 19 bioclimatic variables were selected. The software Maxent 3.3.3 was applied in the generation of the distribution models and the area under the curve of receiver operating characteristic (AUC) was used to analyze the model. The Jackknife test contributed to identify which bioclimatic variables are most important or influential in the model. The models showed AUC values ranged from 0.857 to 0.983. The species with higher AUC values were Araucaria angustifolia, scabrella and Euterpe edulis, respectively. The maximum temperature of warmest month showed the highest influence for the most species, followed by the mean diurnal range and annual precipitation. It was observed that for some species, there were restricted areas of environmental suitability, such as Araucaria angustifolia, Ilex paraguariensis and Mimosa scabrella. The models used could trace the potential distribution areas using the environmental variables, and these models contribute significantly to sustainable forest management.

Key words: Brazilian flora, Maxent, bioclimatic variables, distribution models, potential occurrence.

1. Introduction importance are being studied. One of the technologies used is the potential distribution models [3]. Some The Brazilian territory is covered by the most authors give it different names, such as ecological varied forest ecosystems, such as Amazon Rainforest, niche models [4] and species distribution models Atlantic Forest, Cerrado, Caatinga and others, which (SDMs) [5]. ranks it among the countries with the greatest These models aim to complement or infer diversity on the planet. In addition, Brazil has information on the geographic distribution of the plantations of native and exotic species distributed species. This approach is a numerical tool that throughout practically the entire national territory [1]. combines observations of species occurrence or Trees planted for industrial purposes are the source abundance with environmental parameters, predicting of hundreds of products and byproducts, and they distribution through landscapes, sometimes requiring produce diverse cultural, recreational, tourist and other extrapolation in space [6]. services related to research and regulation of water One of the desired characteristics of the SDMs is flow and nutrients, as well as generate climatic the transferability, that is, the possibility to indicate benefits with carbon sequestration [2]. potential areas of occurrence of the species beyond the Native species with ecological and economic known sites [7]. Potential distribution models can be useful in Corresponding author: Vitor Augusto Cordeiro Milagres, M.Sc. student, research fields: soil and nutrition. identifying suitable sites for the species due to climate

Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance 401 change and areas susceptible to invasive species [8], information on tree species for neotropical regions. By estimating the broad limits of the distribution of an 2010, there were over 1,216 checklists with 6,188 endangered species and identifying new suitable species [12]. Therefore, species were chosen referring habitat for their reintroduction [9]. only to the Brazilian territory. The productive capacity of a site can be defined as 2.2 Environmental Variables the potential to produce wood (or other product), in a given area, for a particular species or clone [10]. In this study, 19 bioclimatic environmental According to Clutter et al. [11], the models for variables (Table 2) were used with 1 km resolution at evaluating the local productive capacity are classified the equator extracted from WorldClim database. With as direct and indirect. the software DIVA-GIS version 7.5, the bioclimatic This study therefore aimed to generate potential variables were extracted to Brazil extension [13]. distribution models for Brazilian forest species, encouraging the use of legal native species for 2.3 Development of Models management, and contribute to the evaluation of The distribution models were generated for each productive capacity in an indirect way. In addition, species using the maximum entropy species this study will provide tools for forest producers to distribution modeling (Maxent). It is a find better combinations of trees for their farms, general-purpose machine learning method with a increasing their productivity and variability of species. simple and precise mathematical formulation, and it 2. Materials and Methods has a number of aspects that make it well-suited for species distribution modeling [14] (Fig. 1). 2.1 Obtaining Occurrence Data When comparing Maxent with other SDMs, such as Eleven tree species of Brazilian flora were defined GARP, BIOCLIM, for example, Maxent demonstrated (Table 1) with economic potential, whether wood or better predictability [6, 15-17]. Also, Maxent can non-wood product. The selected species are contained effectively and accurately model species with multiple in the species occurrence database, called TreeAtlan 2.1. forms of rarity, that is, with few points of occurrence, The TreeAtlan 2.1 program gathers information to the point of leading to new population discovery from herbarium around the world that contains [18].

Table 1 List of species chosen for the potential species distribution model. Family Scientific name Common name Uses Anacardiaceae Anacardium occidentale Cashew tree Culinary, medicinal, CC Aquifoliaceae Ilex paraguariensis Yerba mate RDA, beverage, chemical industry Araucariaceae Araucaria angustifolia Araucaria CC, culinary, medicinal, afforestation Arecaceae Euterpe edulis Juçara palm Culinary, CC, handicraft Arecaceae Euterpe oleracea Acai palm Culinary, CC, handicraft Calophyllum Calophyllum brasiliense Guanandi CC, medicinal Euphorbiaceae Hevea brasiliensis Rubber tree Natural rubber, culinary Hymenaea courbaril Jatobá CC, RDA, culinary Fabaceae Mimosa scabrella Bracatinga RDA, CC, charcoal, beekeeping, animal feed Fabaceae Schizolobium amazonicum Parica CC, RDA, medicinal Meliaceae Swietenia macrophylla Brazilian mahogany CC, landscaping RDA: recovery of degraded areas; CC: civil construction.

402 Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance

Table 2 Nineteen environmental variables used in the potential distribution of species and their codes. Variables Descriptions BIO1 Annual mean temperature BIO2 Mean diurnal range (mean of monthly (max. – min.) temperature) BIO3 Isothermality (BIO2/BIO7) × 100) BIO4 Temperature seasonality BIO5 Max. temperature of the warmest month BIO6 Min. temperature of the coldest month BIO7 Temperature annual range (BIO5 – BIO6) BIO8 Mean temperature of the wettest quarter* BIO9 Mean temperature of the driest quarter BIO10 Mean temperature of the warmest quarter BIO11 Mean temperature of the coldest quarter BIO12 Annual precipitation BIO13 Precipitation of the wettest month BIO14 Precipitation of the driest month BIO15 Precipitation seasonality BIO16 Precipitation of the wettest quarter BIO17 Precipitation of the driest quarter BIO18 Precipitation of the warmest quarter BIO19 Precipitation of the coldest quarter * Quarter: period of 13 weeks (three months) consecutive.

The idea of Maxent is to estimate a target in relation to predicted absences predictions [20]. The probability distribution by finding the probability area under the ROC curve (AUC) was used to distribution of the maximum entropy subject to a set evaluate the distribution models. It ranges from 0 to 1, of constraints that represent the incomplete and predicted values above 0.5 are acceptable [21]. information about the target distribution [14]. 2.4 Mapping Distribution Maxent also evaluates through the Jackknife test. In this analysis, each variable is deleted and a new model DIVA GIS 7.5 was used to create the maps [13]. is created with only the other variables. Furthermore, The points of presence were superimposed on the other models are created using each variable expected areas of occurrence to evaluate the separately [19]. Thus, it is known which bioclimatic performance of the models [22]. For each cell’s variables are the most important or influential in the probability, the Maxent indicates a percentage ranging model. A low gain variable, close to 0, presents from 0 to 100, indicating environmental suitability (not probability of occurrence) [23]. prediction as poor as a random model. However, the variables with gains close to 1 present information 3. Results highly correlated to the occurrences and they perform 3.1 Prediction Maps of Species good predictions [20]. To analyze the performance of the model, that is, its The resulting maps are referents of areas of wide accuracy, a technique known as receiver operating occurrence for the species studied (Figs. 2 and 3). characteristic (ROC) curve was used. The curve is It was observed that for some species, there were made plotting on one axis—the proportion of true restricted areas of environmental suitability, such as presences of total presences predicted and specificity, Araucaria angustifolia (Fig. 2b), Ilex paraguariensis and on the other axis—the proportion of true absences (Fig. 3b) and Mimosa scabrella (Fig. 3c).

Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance 403

Fig. 1 Process of potential distribution modeling of the species [22].

(a) (b) (c)

(d) (e) (f)

Fig. 2 Potential distribution model of Anacardium occidentale (a), Araucaria angustifolia (b), Calophyllum brasiliense (c), Euterpe edulis (d), Euterpe oleracea (e) and Hevea brasiliensis (f), containing the environmental suitability in Brazil with their occurrence points. : 0.0000-0.1806; : 0.1806-0.3613; : 0.3613-0.5419; : 0.5419-0.7225; : 0.7225-1.0000. Modelling was performed using Maxent with 19 environmental variables.

404 Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance

(a) (b) (c)

(d) (e)

Fig. 3 Potential distribution model of Hymenaea courbaril (a), Ilex paraguariensis (b), Mimosa scabrella (c), Schizolobium amazonicum (d) and Swietenia macrophylla (e), containing the environmental suitability in Brazil with their occurrence points. : 0.0000-0.1806; : 0.1806-0.3613; : 0.3613-0.5419; : 0.5419-0.7225; : 0.7225-1.0000. Modelling was performed using Maxent with 19 environmental variables.

A relevant factor for the restriction of areas is the The species with higher AUC were Araucaria reduced number of records (below 100). However, angustifolia, Euterpe edulis and Mimosa scabrella, species, such as Euterpe oleracea (Fig. 2e), and with lower AUC values were Hymenaea courbaril,

Schizolobium amazonicum (Fig. 3d) and Swietenia Schizolobium amazonicum and Anacardium macrophylla (Fig. 3e), despite the low occurrence, occidentale. presented a notable area of occurrence widely 3.3 Standardization of Models distributed. It was highlighted that the species Anacardium In this study, these distinct patterns were observed: occidentale, Calophyllum brasiliense and Hymenaea (1) Pattern I: species with few occurrence records courbaril showed greater abundance. These species and wide distribution and adaptation, such as are found from the north to the south of the country, Schizolobium amazonicum and Swietenia covering several biomes with high environmental macrophylla; resilience. (2) Pattern II: species with many occurrence records and wide distribution and adaptation, such as 3.2 Evaluation of the Models Hymenaea courbaril, Anacardium occidentale, The AUC values found by the Maxent program Calophyllum brasiliense and Hevea brasiliensis; ranged from 0.857 to 0.983 (nine species with values (3) Pattern III: species with many occurrence records in 0.9-1.0 and two species with values in 0.8-0.9) and restricted areas of environmental suitability, (Table 3). such as Araucaria angustifolia, Ilex paraguariensis,

Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance 405

Table 3 AUC values to the potential distribution models of the 11 species. No. Species AUC 1 Anacardium occidentale 0.900 2 Araucaria angustifolia 0.983 3 Calophyllum brasiliensis 0.912 4 Euterpe edulis 0.975 5 Euterpe oleracea 0.928 6 Hevea brasiliensis 0.960 7 Hymenaea courbaril 0.894 8 Ilex paraguariensis 0.970 9 Mimosa scabrella 0.982 10 Schizolobium amazonicum 0.857 11 Swietenia macrophylla 0.947

Table 4 Order of the most important environmental variables for each species, according to results of the Jackknife statistics. Environmental variables No. Species 1st 2nd 3rd 1 Ilex paraguariensis BIO18 BIO3 2 Mimosa scabrella BIO18 BIO3 3 Euterpe edulis BIO2 BIO3 BIO5 4 Araucaria angustifolia BIO12 BIO2 5 Calophyllum brasiliense BIO11 BIO2 6 Swietenia macrophylla BIO3 BIO4 7 Anacardium occidentale BIO12 BIO4 BIO2 8 Euterpe oleracea BIO7 BIO12 9 Hevea brasiliensis BIO2 BIO4 BIO12 10 Hymenaea courbaril BIO4 BIO17 11 Schizolobium amazonicum BIO4 BIO11 BIO16

Mimosa scabrella and Euterpe edulis; possible to represent the fundamental niche of the 11 (4) Pattern IV: species with few records and species studied. restricted areas of environmental suitability, such as This result can be evaluated using two components: Euterpe oleracea. the ROC curve and the Jackknife test. The first The results of the Jackknife tests are in relation to provides a standard quantitative measure of model the importance of environmental variables for each performance. The higher the AUC value, the better the species. BIO5 showed a higher gain for a larger model [24]. On the other hand, in the Jackknife number of species, followed by BIO2 and BIO12 analysis, each variable is deleted and a new model is (Table 4). created with only the other variables. In this way, it is In this study, 10 test variables did not present possible to visualize which bioclimatic variable had significant influence in the models, among them, such the greatest influence on the model [19]. as annual mean temperature (BIO1) and the The AUC indexes found were satisfactory. precipitation of the coldest quarter (BIO19). According to Metz [25], an evaluation test may adopt AUC values as: excellent (0.9-1.0), good (0.8-0.9), 4. Discussion fair (0.7-0.8), poor (0.6-0.7) and fail (0.5-0.6). In this Through the chosen environmental variables, it was study, the AUC values ranged from 0.983 to 0.857,

406 Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance with the Araucaria angustifolia, Mimosa scabrella of them [30]. This was observed for the species and Euterpe edulis species having the highest values, Anacardium occidentale, Calophyllum brasiliense and respectively. Hymenaea courbaril. These were the ones with the Considering the data used in this work, the results highest occurrence and consequently the greatest suggest that the Maxent program presented high ecological niche. predictive power. Comparing different models The model indicates that the environmental generated (GARP, BIOCLIM, Maxent and SVM), it conditions are similar to the places where the known was concluded that the model generated by Maxent species occurs [23]. In the same way, Bracatinga was the most efficient, with the highest AUC value (Mimosa scabrella) is native to the colder climates of among all the algorithms used [26]. Brazil, that is, rainy temperate, constantly humid with Maxent is one of the few algorithms that make a average temperatures of the month warmer and cooler prediction consistent with low sample numbers (n < inferior to 22 °C and 18 °C, respectively [31]. From 30), and can be used with data below 10 points of their distribution models, it is observed that the occurrence [27]. Other models, such as BIOCLIM species are sensitive to high temperatures, and with stabilizes after 50 points [28]. For some species, a this, it is verified that the areas with greater small number of occurrence points may be sufficient environmental suitability are regions with lower to characterize the environmental niche, whereas in temperatures, as it happens in the south of the country. other cases, the same number of occurrence may be In addition, the species obtained similar results in inadequate to represent the range of conditions under other variables of greater importance, such as which the species exists [23]. One can therefore precipitation of warmest quarter (BIO18) and understand why some species have restricted areas isothermality (BIO3). Such information confirms the and other large areas. The importance of each record sensitivity of both to temperature. will depend on whether this location represents an Species, such as Hymenaea courbaril and exclusive environment, not represented by the other Calophyllum brasiliense are considered as plastic sampling points [23]. species, since they occur in different regions of the The variable with greater training gain is the one of country. Calophyllum brasiliense grows well in greater importance and the one that most influenced in alluvial, clayey, sandy-loamy or sandy soils with acids the modeling of the potential distribution of the (pH 4.5-6.0), and presents excellent adaptation to both species [22]. In the generation of the models, the wet and dry environments [32]. Due to its high variable that influenced most species was the adaptability, the species can be an alternative maximum temperature of the warmest month (BIO5), plantation for farmers. as observed for Ilex paraguariensis and Mimosa The Guanandi has light to moderately dense wood, scabrella. According to Oliveira and Rotta [29] in the good durability and resistance, which allows its use in climatic mapping for Ilex paraguariensis, using the civil and naval construction, barrels for wine storage classification of Köppen, it is evidenced that the and general carpentry [33]. Hymenaea courbaril predominant distribution of Yerba mate is covered by occurs from the north to the southeast, both in high climatic types Cfb, that is, temperate climate with and medium fertility soils [34]. Its wood is used in mild summer. hydraulic works, bodywork, poles, barrels, various The probability of relative occurrence from a constructions, furniture, laminates, among other forms combination of environmental variables depends on [35]. Both species have a high environmental the density of species records in the fundamental niche amplitude, therefore, being a common characteristic of

Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance 407 the species more found in the Brazilian territory. systems have grown widely in Brazil. Paricá (Schizolobium amazonicum) was more Consortia of tree species with palm trees show influenced by temperature seasonality (BIO4). This themselves to be increasingly promising. Examples of variable represents the coefficient of variation of the this are rubber tree plantations (Hevea brasiliensis) mean temperature, expressed as a percentage of the with the Acai (Euterpe oleracea) or the Juçara palm average annual temperature. It was observed a (Euterpe edulis). feasibility of a large part of the Brazilian North to Analyzing Figs. 2d and 2f, it is possible to observe Paricá. This region has a spatial and seasonal the viability of both cultivations for the states of homogeneity of temperature, which does not happen Espírito Santo (Southeast) and Acre (North), where in relation to precipitation [36]. their effectiveness is already proven. The consortium Although it shelters the largest commercial growing between rubber tree and palm tree was investigated by area in the country, the 3D figure suggests that the Bovi et al. [40] and it was concluded that production species still has potential for exploration. The wood could be feasible with adequate spacing. has potential use in the manufacture of phosphorus The Parica (Schizolobium amazonicum) consortium sticks, shoe heels, toys, concrete forms, laminates, with the Brazilian mahogany (Swietenia macrophylla) plywood, cellulose and paper, due to its easy has also been promising. Similar regions of processing [37]. environmental suitability are observed for the species, The mean diurnal range (BIO2) was the most especially in the north of the country. Santos et al. [41] important variable for the species Anacardium analyzed the Paricá, mahogany and other crops occidentale and Euterpe edulis. The main producing consortium and the economic viability, and presented regions of cashew in the world present climate Aw, a positive profitability for this system. In addition, it characterized in tropical, hot, humid, with rains from was observed that the consortium exceeded the November to April and dry, or too little rainfall monoculture results with adequate management. intensity, in other months, according to Köppen In addition to what has been discussed so far, the classification [38]. The same was observed in the evaluation of productive capacity may be done by potential distribution of the species. Anacardium means of direct methods (relation between dominant occidentale shows greater suitability in Northeast and height and age) or indirect methods, based on Southeastern Brazil. In the present study, it was indicative vegetation and edaphoclimatic factors [42]. observed that the species is susceptible to prolonged This work is expected to contribute to the indirect periods under temperatures below 22 °C, since young methods, since it used climate data for model are harmed by cold [38]. generation. The Euterpe edulis species presented a greater 5. Conclusions adaptation in the Atlantic Forest region, from the Northeast to the South of the country. Juçara palm It was possible to trace potential occurrence areas, develops well in tropical and subtropical regions, with using environmental variables appropriate to the high precipitation, i.e., regions with high precipitation species. Species distribution models provide important and no pronounced drought [39]. Due to its contribution to the design and implementation of environmental requirements, Brazil is one of the few strategies for sustainable forest management. countries that present adequate climatic conditions for The relevance of the use of predictive modeling of the cultivation and commercial exploitation of this distribution for the forest agribusiness is also plant [39]. highlighted. Species, such as Anacardium occidentale,

408 Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance

Calophyllum brasiliense and Hymenaea courbaril [8] Araújo, M. B., Alagador, D., Cabeza, M., Nogués-Bravo, D., and Thuiller, W. 2011. “Climate Change Threatens were the ones with the highest occurrence. Others, European Conservation Areas.” Ecol. Lett. 14 (5): such as Araucaria angustifolia, Ilex paraguariensis 484-92. and Mimosa scabrella were specific species of [9] Pearce, J., and Lindenmayer, D. 1998. “Bioclimatic Southern Brazil, and their planting in other regions is Analysis to Enhance Reintroduction Biology of the Endangered Helmeted Honeyeater (Lichenostomus not recommended. In addition, it is observed that the melanops cassidix) in Southeastern Australia.” Restor. species with higher AUC values were Araucaria Ecol. 6 (3): 238-43. angustifolia, Mimosa scabrella and Euterpe edulis, [10] Oliveira, M. L. R., Leite, H. G., Nogueira, G. S., Garcia, respectively. S. L. R., and Souza, A. L. 2008. “Classification of the Productive Capacity of Unthinned Stands of Eucalypt This work can be a tool to evaluate the productive Clones.” Pesqui. Agropecuária Bras. 43 (11): 1559-67. capacity of the place, using climatic data as an indirect (in Portuguese) method. It is necessary for the forest management the [11] Clutter, J. L., Fortson, J. C., Pienaar, L. V., Brister, G. H., incorporation of the knowledge between the and Bailey, R. L. 1983. Timber Management: A Quantitative Approach, 3rd ed.. New York: John Wiley interactions of the climate with the forest. & Sons, Inc.. References [12] Oliveira Filho, A. T. 2014. “NeoTropTree, Neotropical Tree Flora: A Database Involving Biogeography, [1] Ribeiro, E. S., Sousa, R. A. T., Dias, M., Dorval, A., and Diversity and Conservation.” Federal University of Minas Costa, R. B. 2011. “Marketing of Wood from Natural Gerais. Accessed August 20, 2016. Forests in the State of Mato Grosso from 2004 to 2009.” http://www.icb.ufmg.br/treeatlan. Floresta e Ambient 18 (4): 365-75. (in Portuguese) [13] Hijmans, R., Cruz, M., Rojas, E., and Guarino, L. 2001. [2] IBÁ. 2016. Brazilian Tree Industry 2016 Report. DIVA-GIS Version 1.4: A Geographic Information [3] Souza-Muñoz, M. E., Giovanni, R., Siqueira, M. F., System for the Analysis of Biodiversity Data. Lima: Sutton, T., Brewer, P., Pereira, R. S., Canhos, D. A. L., International Potato Center (CIP). and Canhos, V. P. 2011. “OpenModeller: A Generic [14] Phillips, S. J., Anderson, R. P., and Schapire, R. E. 2006. Approach to Species’ Potential Distribution Modelling.” “Maximum Entropy Modeling of Species Geographic Geoinformatica 15 (1): 111-35. Distributions.” Ecol. Modell. 190 (3-4): 231-59. [4] Peterson, A. T., Soberón, J., Pearson, R. G., Anderson, R. [15] Elith, J., Graham, C. H., Anderson, R. P., Dudík, M., P., Martínez-Meyer, E., Nakamura, M., and Araújo, M. B. Ferrier, S., Guisan, A., Hijmans, R. J., Huettmann, F., 2011. Ecological Niches and Geographic Distributions. Leathwick, J. R., Lehmann, A., Li, J., Lohmann, L. G., Princenton, NJ: Princenton University Press. Loiselle, B. A., Manion, G., Moritz, C., Nakamura, M., [5] Guisan, A., Tingley, R., Baumgartner, J. B., Nakazawa, Y., Overton, J. M. M., Peterson, A. T., Naujokaitis-Lewis, I., Sutcliffe, P. R., Tulloch, A. I. T., Phillips, S. J., Richardson, K., Scachetti-Pereira, R., Regan, T. J., Brotons, L., McDonald-Madden, E., Schapire, R. E., Soberón, J., Williams, S., Wisz, M. S., Mantyka-Pringle, C., Martin, T. G., Rhodes, J. R., and Zimmermann, N. E. 2006. “Novel Methods Improve Maggini, R., Setterfield, S. A., Elith, J., Schwartz, M. W., Prediction of Species’ Distributions from Occurrence Wintle, B. A., Broennimann, O., Austin, M., Ferrier, S., Data.” Ecography 29 (2): 129-51. Kearney, M. R., Possingham, H. P., and Buckley, Y. M. [16] Giovanelli, J. G. R., Siqueira, M. F., Haddad, C. F. B., 2013. “Predicting Species Distributions for Conservation and Alexandrino, J. 2010. “Modeling a Spatially Decisions.” Ecol. Lett. 16 (12): 1424-35. Restricted Distribution in the Neotropics: How the Size [6] Elith, J., and Graham, C. H. 2009. “Do They, How Do of Calibration Area Affects the Performance of Five They and Why Do They Differ on Finding Reasons for Presence-Only Methods.” Ecol. Modell. 221 (2): Differing Performances of Species Distribution Models?” 215-24. Ecography 32 (1): 66-77. [17] Tarkesh, M., and Jetschke, G. 2012. “Comparison of Six [7] Paglia, A. P., Rezende, D. T., Koch, I., Kortz, A. R., and Correlative Models in Predictive Vegetation Mapping on Donatti, C. 2012. “Species Distribution Models (SDM) in a Local Scale.” Environ. Ecol. Stat. 19 (3): 437-57. Biodiversity Conservation Strategies and Climate Change [18] Williams, J. N., Seo, C., Thorne, J., Nelson, J. K., Erwin, Ecosystem Based Adaptation.” Natureza & Conservação S., O’Brien, J. M., and Schwartz, M. W. 2009. “Using 10 (2): 231-4. (in Portuguese) Species Distribution Models to Predict New Occurrences

Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance 409

for Rare Plants.” Divers. Distrib. 15 (4): 565-76. Salamanders.” Org. Divers. Evol. 11 (5): 409-23. [19] Lima, L. M. C. 2014. “Geographic Distribution Modeling [31] Carpanezzi, A. A., Laurent, J. M. E., Carvalho, P. E. R., for Hydromedusa maximiliani (Mikan, 1820) (Testudines: Pegoraro, A., Baggio, A. J., Zanon, A., Oliveira, E. B., Chelidae), Brazil.” Masters thesis, Federal University of Iede, E. T., Rotta, E., Sturion, J. A., Pereira, J. C. D., Juiz de Fora. (in Portuguese) Graça, L. R., Rauen, M. J., Carpanezzi, O. T. B., and [20] Coelho, G. L. N., Carvalho, L. M. T., and Gomide, L. R. Oliveira, Y. M. M. 1988. Technical Manual of 2016. “Predictive Modeling Distribution of Pioneer Bracatinga (Mimosa scabrella Benth.). Curitiba: Species in the State of Minas Gerais, Brazil.” Pesq. EMBRAPA-CNPF, 6. (in Portuguese) Agropec. Bras. 51 (3): 207-14. (in Portuguese) [32] Silva, K. E. 2005. “Jacareúba, Calophyllum brasiliense [21] Souza, R. A., and Marco Júnior, P. 2011. “The Efficiency Cambess.” Informativo Técnico Rede de Sementes da of Potential Species Distribution Models for the Amazônia No. 11. Acessed June 19, 2017. Determination of Potential Areas of Occurrence of https://www.inpa.gov.br/sementes/iT/11_Jacareuba.pdf. Deforestation in the Brazilian Amazon.” In Proceedings (in Portuguese) of the VIII Congress of Research, Education and [33] Filho, A. N. K., Marzollo, L. G., Lopes, A. J., and Extension, 1-5. (in Portuguese) Wendling, I. 2007. Production of Guanandi Seedlings. [22] Branquinho, F. G. G. 2014. “Potential Distribution Technical Announcement, EMBRAPA Florestas, Modeling of Useful Brazilian Plants.” Masters thesis, Colombo. (in Portuguese) Federal University of Jequitinhonha and Mucuri Valleys. [34] Costa, W. S., Souza, A. L., and Souza, P. B. 2011. (in Portuguese) “Jatobá, Hymenaea courbaril L.: Ecology, Management, [23] Pearson, R. G., Raxworthy, C. J., Nakamura, M., and Silviculture Technology of Native Species of the Atlantic Peterson, A. T. 2007. “Predicting Species Distributions Forest.” Native Species of the Atlantic Forest No. 2, from Small Numbers of Occurrence Records: A Test Federal University of Viçosa. Acessed June 19, 2017. Case Using Cryptic Geckos in Madagascar.” J. Biogeogr. http://www.sifloresta.ufv.br/bitstream/handle/123456789/ 34 (1): 102-17. 11139/Jatoba_Ecologia,%20manejo,%20silvicultura%20 [24] Pinto, M. A. S., Gonçalves, A. P. S., Santos, S. A. P., e%20tecnologia%20de%20especies%20nativas%20da%2 Almeida, M. R. L., and Azevedo, J. C. M. 2014. 0Mata%20atlantica%20.pdf?sequence=1&isAllowed=y. “Biological Invasion of Corythucha ciliata in Green (in Portuguese) Urban Spaces in Portugal: A Niche Modeling Approach [35] Campos, M. A. A., and Uchida, T. 2002. “Influence of Using Maximum Entropy.” Ciência Florest 24 (3): Shade on the Growth of Seedlings of Three Amazon 597-607. (in Portuguese) Species.” Pesqui. Agropecuária Bras. 37 (3): 281-8. (in [25] Metz, C. E. 1986. “ROC Methodology in Radiologic Portuguese) Imaging.” Invest. Radiol. 21 (9): 720-33. [36] Nimer, E. 1979. “Climatology of Brazil.” Secretariat of [26] Marco Júnior, P., and Siqueira, M. F. 2009. “How to Planning of the Presidency of the Republic, IBGE, Determine the Potential Distribution of Species under a Department of Natural Resources and Environment. Conservationist Approach?” Megadiversidade 5 (1-2): Accessed July 4, 2017. 65-77. (in Portuguese) http://biblioteca.ibge.gov.br/visualizacao/livros/liv81099. [27] Peterson, A. T., Papeş, M., and Eaton, M. 2007. pdf. (in Portuguese) “Transferability and Model Evaluation in Ecological [37] Sousa, C. R., Rossi, L. M. B., Azevedo, C. P., and Vieira, A. Niche Modeling: A Comparison of GARP and Maxent.” H. 2003. “Paricá: Schizolobium parahyba var. amazonicum Ecography 30 (4): 550-60. (Huber ex Ducke) Barneby.” Technical Information Amazon [28] Scheldeman, X., and Van Zonneveld, M. 2010. Training Network. Acessed July 4, 2017. Manual on Spatial Analysis of Plant Diversity and https://www.infoteca.cnptia.embrapa.br/bitstream/doc/67 Distribution, 1st ed.. Rome: Bioversity International, 40. 6209/1/circtec18.pdf. (in Portuguese) [29] Oliveira, Y. M. M., and Rotta, E. 1985. “Natural [38] Moura Barros, L., Pimentel, C. R. M., Correa, M. P. F., Distribution Area of Yerba Mate (IIex paraguariensis St. and Mesquita, A. L. M. 1993. “Technical Híl.).” Presented at the Seminar on News and Forest Recommendations for the Crop of Precocious Dwarf Outlook, EMBRAPA-CNPF, Curitiba, PR. (in Cashew.” EMBRAPA, CNPAT, Fortaleza. Acessed June Portuguese) 19, 2017. http://www.ceinfo.cnpat.embrapa.br/arquivos/ [30] Rödder, D., Lötters, S., Öz, M., Bogaerts, S., artigo_2493.pdf. (in Portuguese) Eleftherakos, K., and Veith, M. 2011. “A Novel Method [39] Aguiar, F. F. A., Schaefer, S. M., Lopes, E. A., and to Calculate Climatic Niche Similarity among Species Toledo, C. B. 2002. “Production of Juçara Palm with Restricted Ranges—The Case of Terrestrial Lycian Seedlings (Euterpe edulis Mart.).” Instituto de Botânica,

410 Potential Distribution Modeling of Useful Brazilian Trees with Economic Importance

São Paulo. Acessed June 19, 2017. [41] Dos Santos, M. J. C., Rodriguez, L. C. E., and Wandelli, http://botanica.sp.gov.br/institutodebotanica/files/2011/11 E. V. 2002. “Economic Evaluation of Four Agroforestry /palmito.pdf. (in Portuguese) Systems Models in Degraded Pastures Areas in the [40] Bovi, M. L. A., Godoy Júnior, G., Nagai, V., and Cardoso, Ocidental Amazon.” Sci. For. 62: 48-61. (in Portuguese) M. 1990. “Planting Density of Heart of Palm [42] Campos, J. C. C., and Leite, H. G. 2006. Forest Intercropped with Rubber Trees.” Pesqui. Agropecuária Measurement: Questions and Answers. Viçosa: UFV, 277. Bras. 25 (7): 1023-9. (in Portuguese) (in Portuguese)