data

Data Descriptor A New Multi-Temporal Forest Cover Classification for the Basin,

Margaret Kalacska 1,*, Oliver Lucanus 1, Leandro Sousa 2 and J. Pablo Arroyo-Mora 3

1 Applied Remote Sensing Laboratory, Department of Geography, McGill University, Montreal, QC H3A 0B9, Canada 2 Flight Research Lab, National Research Council of Canada, Ottawa ON K1A 0R6, Canada 3 Laboratório de Ictiologia de Altamira, Universidade Federal do Pará, Altamira PA 68372040, Brazil * Correspondence: [email protected]

 Received: 26 June 2019; Accepted: 30 July 2019; Published: 1 August 2019 

Abstract: We describe a new multi-temporal classification for forest/non-forest classes for a 1.3 million square kilometer area encompassing the Xingu River basin, Brazil. This region is well known for its exceptionally high biodiversity, especially in terms of the ichthyofauna, with approximately 600 known species, 10% of which are endemic to the river basin. Global and regional scale datasets do not adequately capture the rapidly changing land cover in this region. Accurate forest cover and forest cover change data are important for understanding the anthropogenic pressures on the aquatic ecosystems. We developed the new classifications with a minimum mapping unit of 0.8 ha from cloud free mosaics of Landsat TM5 and OLI 8 imagery in Google Earth Engine using a classification and regression tree (CART) aided by field photographs for the selection of training and validation points.

Keywords: Landsat imagery; land cover change; deforestation; biodiversity; conservation; Xingu river basin

1. Summary Ongoing deforestation in the tropics is a well-known threat to biodiversity [1–3]. Species with restricted ranges are particularly at risk of extinction from habitat loss [4]. Global- and regional-scale maps developed to analyze large-scale trends (e.g., [5–8]) do not necessarily capture the small-scale changes experienced by restricted range species. Therefore, these areas require the development of detailed land cover datasets that allow for a better understanding of historical and current environmental threats (e.g., deforestation, mining, etc.). In addition to their high biodiversity, Amazonian forests are also important contributors to drivers of global atmospheric circulation [9] and climate change mitigation [10]. The Xingu River basin encompasses the fourth largest catchment area of the Amazon. It is known for its high fish biodiversity, with an estimated 600 species, 10% of them endemic to this river basin with many remaining to be described [11–14]. Several ethnic groups inhabit a mosaic of indigenous lands in a large part of the basin, contributing greatly to a slower rate of deforestation than is seen in unprotected Xingu basin headwaters [15–17]. The health of the riparian ecosystems is linked to changes in the overall forest cover and has a direct impact the unique fish biota [18]. In order to assess the rapid changes that the Xingu River basin has experienced over the last four decades, we developed a forest/non-forest classification system for four time periods from 1989 to 2018 as a case study applicable elsewhere in the tropics. These datasets were produced from cloud free surface reflectance mosaics of Landsat TM5 and Landsat 8-OLI images, classified with a classification and regression tree (CART) in Google Earth Engine.

Data 2019, 4, 114; doi:10.3390/data4030114 www.mdpi.com/journal/data Data 2019, 4, x FOR PEER REVIEW 2 of 9 reflectance mosaics of Landsat TM5 and Landsat 8-OLI images, classified with a classification and regressionData 2019, 4, 114tree (CART) in Google Earth Engine. 2 of 8

2. Data Description 2. Data Description The four classifications are available for download in Geotiff format (30 m pixel size) (Figure 1). Each Theclassification four classifications is in a geographic are available projection for download (latitude/longitude) in Geotiff format with (30 mWGS84 pixel size)datum. (Figure Pixels1). representEach classification one of three is in classes: a geographic forest projection (1), water (latitude (2), or non-forest/longitude) (3). with Our WGS84 definition datum. of Pixelsforest representis “areas withone ofapproximately three classes: greater forest (1),than water 30% (2),tree or canopy non-forest cover”, (3). which Our definition is consiste ofnt forest with isthe “areas Brazilian with definitionapproximately [19]. greater The implementation than 30% tree canopy of this cover”, definition which iswas consistent based withon a thepproximately Brazilian definition 3000 field [19]. photographsThe implementation (see Section of this 3). definition The non-forest was based class on encompasses approximately all 3000 land field covers photographs and land (see uses Section that are3). notThe considered non-forest classforest encompasses or open water. all The land wa coverster class and represents land uses thatexposed are not surface considered water. forest or open water. The water class represents exposed surface water.

(a) (b)

Figure 1. Cont.

DataData2019 2019, 4,, 1144, x FOR PEER REVIEW 3 of3 9 of 8

(c) (d)

FigureFigure 1. 1.Four Four classifications classificationsof of forestforest covercover for the Xingu Xingu river river basin basin for for (a ()a circa) circa 1989, 1989, (b) ( bcirca) circa 2000, 2000, (c)( circac) circa 2010, 2010, and and (d ()d circa) circa 2018 2018 from from LandsatLandsat TM5TM5 and OLI OLI 8 8 im imagery.agery. Boundaries Boundaries of ofstates, states, larger larger rivers,rivers, and and landmarks landmarks including including cities cities andand thethe JericoJericoáá rapirapidsds are are shown shown for for reference. reference.

3.3. Methods Methods TheThe forest forest cover cover classifications classifications werewere createdcreated for a 1.3M km km22 areaarea encompassing encompassing the the Xingu Xingu River River basin,basin, Brazil, Brazil, representing representing circa circa 1989,1989, 2000,2000, 2010, and 2018 2018 from from Landsat Landsat imagery imagery (TM5 (TM5 and and OLI OLI 8) 8) (Table(Table1). The1). The area area crosses crosses the the borders borders of five of five states: states: Par áPará,, Mato , Grosso, Amazonas, Amazonas, Tocantins, Tocantins, and and Goi ás (FigureGoiás1 ).(Figure The USGS1). The Landsat USGS SurfaceLandsat ReflectanceSurface Reflectance Tier 1 imagery Tier 1 imagery collection collection was queried was queried through Googlethrough Earth Google Engine Earth (Table Engine1). Clouds(Table 1). and Clouds cloud an shadowsd cloud shadows were masked were masked using the using pixel the QA pixel bands QA to createbands a cloudto create free a compositecloud free compos of the medianite of the surface median reflectance surface reflectance for each for reference each reference period. period. In order In to achieveorder to cloud achieve free cloud composites free composites for such for a large such areaa large in area the tropicsin the tropics which which is prone is prone to persistent to persistent cloud cover,cloud the cover, composites the composites represent represent periods periods spanning spanning 1985–1989, 1985–1989, 1995–2000, 1995–2000, 2005–2010, 2005–2010, and and 2017–2018. 2017– 2018. A classification and regression tree (CART) [20] was trained in Google Earth Engine for each A classification and regression tree (CART) [20] was trained in Google Earth Engine for each mosaic, mosaic, based on user-generated points split 70/30 for training and validation. based on user-generated points split 70/30 for training and validation. For the most recent mosaic, the training and validations points were selected based on For the most recent mosaic, the training and validations points were selected based on photographs photographs taken during four field expeditions in 2015–2018. The photographs were taken along taken during four field expeditions in 2015–2018. The photographs were taken along the Xingu River the Xingu River (by boat) from Vitória do Xingu, north of São Félix do Xingu (400 km straight line (bydistance boat) from to the Vit south)ória do(Figure Xingu, 2). Additional north of S ãphootographs Félix do Xinguwere taken (400 along km straight the PA-415 line road distance between to the south)Altamira (Figure and2 ).Vitória Additional do Xingu photographs and along werethe BR taken 230 (Transamazônica) along the PA-415 highway road between between Altamira Altamira and Vitandória an do unnamed Xingu and settlement along the near BR 230the (Transamazriver’s naturalônica) diversion highway to the between west, above Altamira the entrance and an unnamed to the settlementVolta Grande, near thewhere river’s the river natural drains diversion from the to Braz theilian west, shield above to thethe entranceAmazonian to lowlands. the Volta In Grande, the wheresouth, the photographs river drains were from taken the Brazilian along primary shield toroads the and Amazonian rivers (Culuene, lowlands. Suiá In Missu, the south, Darro, photographs Curuá) werein takenan area along bounded primary by roadsCanarana and rivers(SE), (Culuene,Sorriso (SW), Sui áCasteloMissu, Darro,do Sonhos Curu (NW),á) in an and area halfway bounded bydownstream Canarana (SE), the SorrisoSuiá Missu (SW), towards Castelo its do confluence Sonhos (NW), with the and Xingu halfway River downstream (NE). The thephotographs Suiá Missu towardsallowed its for confluence an understanding with the Xingu of the River landscape (NE). Thefrom photographs the satellite allowedimagery forin anorder understanding to improve of the landscape from the satellite imagery in order to improve differentiation between the forest and Data 2019, 4, 114 4 of 8 non-forest classes. Based on the landscape in the most recent imagery, these photographs were also used to visualize interpretations of the previous years’ mosaics for the manual selection of training points and validation points.

Table 1. Landsat image collections used for the classifications through Google Earth Engine.

Period Image Collection 1985–1989 LANDSAT/LT05/C01/T1_SR 1995–2000 LANDSAT/LT05/C01/T1_SR 2005–2010 LANDSAT/LT05/C01/T1_SR 2017–2018 LANDSAT/LC08/C01/T1_SR

After classification, a minimum mapping unit (MMU) of 90 m 90 m (0.8 ha) was applied. × Based on the field expeditions, this MMU is reasonable for capturing the majority of the forest fragments in the region. First, a standard sieve routine was applied in ENVI 5.2 to remove isolated pixels. Only pixel groupings of six or more were retained, followed by a clump routine to improve the spatial coherence (e.g., fill in the holes left by the sieve operation). With a kernel of 3 3 pixels, × the routine first performs a dilation, followed by an erosion. The end result is a clumping of adjacent similarly classified pixels. Confusion matrices for the four datasets (Tables2–5) indicate a high accuracy for all three classes. The ground truth data allowed for selecting training points that represent the expected variability of the classes (Figure2). Nevertheless, these confusion matrices represent the maximum accuracy, and with additional ground truth data, classification errors may be found in areas with rapid land cover change (e.g., secondary growth in abandoned clearings).

Table 2. Confusion matrix for the forest cover classification (around 1989). Sample size for training = 413.

Reference Forest Non-forest Water Classification Forest 56 2 0 Non-forest 0 66 0 Water 0 1 52

Table 3. Confusion matrix for the forest cover classification (around 2000). Sample size for training = 390.

Reference Forest Non-forest Water Classification Forest 53 0 0 Non-forest 2 60 0 Water 0 0 52

Table 4. Confusion matrix for the forest cover classification (around 2010). Sample size for training = 415.

Reference Forest Non-forest Water Classification Forest 57 1 0 Non-forest 4 63 0 Water 0 1 52

Table 5. Confusion matrix for the forest cover classification (around 2018). Sample size for training = 583.

Reference Forest Non-forest Water Classification Forest 62 5 0 Non-forest 0 81 0 Water 0 0 58 Data 2019, 4, 114 5 of 8

Data 2019, 4, x FOR PEER REVIEW 5 of 9

Figure 2. Cont.

Data 2019, 4, 114 6 of 8 Data 2019, 4, x FOR PEER REVIEW 6 of 9

Figure 2. Examples of field photographs used to train the visual interpretation of the imagery for selecting classification and regression tree (CART) classification training and validation points. Figure* indicates 2. Examplesthe Northern of field photographs zone between used Vit toória train do the Xingu visual and interpretati Sao Felixon do of Xingu,the imagery ** indicates for the selectingSouthern classification sector (south and ofregression Castelo dotree Sonhos). (CART) 1:classification clearing of training Arapuj áandIsland validation across points. from Altamira*, * indicates2: partially the Northern deciduous zone forest, between sandy Vitória beach, anddo Xingu rocks*, and 3: largeSao patchFelix do of clearedXingu, forest*,** indicates 4: intact the forest*, Southern5: homestead sector (south in Amazonia of Castelo lowlands*, do Sonhos). 6: small1: clea settlement*,ring of Arapujá 7: small Island household*, across from 8:Altamira*, larger settlement*, 2: partially9: large-scale deciduous deforestation*, forest, sandy 10:beach, deforestation and rocks*, with 3: large secondary patch of growth*, cleared forest*, 11: large-scale 4: intact forest*, deforestation*, 5: homestead12: burned in land Amazonia prior to lowlands*, deforestation*, 6: small 13: settlem homestead*,ent*, 7: small 14: large-scale household* deforestation*,, 8: larger settlement*, 15: large-scale 9: large-scaledeforestation*, deforestation*, 16: aerial view10: deforestation north from thewith Jerico secondaryá rapids*, growth*, 17: aerial 11: large-scale view from deforestation*, the Xadá rapids with 12: deforestationburned land prior on the to deforestation*, unprotected side*, 13: homest 18: aerialead*, view14: large-scale of intact deforestation*, forest in protected 15: large-scale area*, 19: aerial deforestation*,view of intact 16: forest aerial at view the Iririnorth rapids*, from the 20: Jericoá small-scale rapids*, clearing 17: aerial at the view Jerico fromá rapids*, the Xadá 21: rapids intact forest withalong deforestation the Culuene on the river**, unprotected 22: exposed side*, 18: soil aerial for agriculture**,view of intact fo 23:rest cattle in protected herd**, area*, 24: Belo 19: aerial Monte dam*, view25: of isolated intact forest forest at patch the Iriri in rapids*, corn field**, 20: small-sca 26: largele clearing corn field**, at the 27: Jericoá plantation**, rapids*, 21: 28: intact extensive forest pasture along the Culuene river**, 22: exposed soil for agriculture**, 23: cattle herd**, 24: Belo Monte dam*, land**, 29: extensive cotton field**, 30: recently cut forest**, 31: pasture with forest patch **, 32: pasture 25: isolated forest patch in corn field**, 26: large corn field**, 27: plantation**, 28: extensive pasture with isolated trees**, 33: cornfield with forest patch**.

Data 2019, 4, 114 7 of 8

4. Limitations Users must be aware that due to the pixel size of the satellite imagery, smaller bodies of water and narrow rivers/streams could not be classified (Figure3). The smaller tributaries where the water is often narrower than a single pixel and may have extensive vegetation covering the banks are not included in the classifications. The MMU of ~1 ha also means that very small forest and non-forest patches will have been sieved out of the classifications. Furthermore, because each classification is representative of the median reflectance for the time period, small temporal scale changes in land cover Datamay 2019 not, 4 be, x FOR represented. PEER REVIEW 8 of 9

FigureFigure 3. 3. OpenOpen rivers rivers wider wider than than a a single single pixel pixel (i.e., (i.e., >> 3030 m) m) such such as as in in the the photograph photograph on on the the left left are are includedincluded in in the the classifications. classifications. Us Usersers are are cautioned, cautioned, however, however, in in examining examining the the surface surface water water class class for for narrownarrow rivers/streams rivers/streams ((<< 3030 m),m), especiallyespecially those those with with dense dense overgrowth, overgrowth, such such as as in in the the photograph photograph on onthe the right. right. The The classifications classifications underestimate underestimat thee the area area for for these these smaller smaller rivers rivers/streams./streams. 5. Conclusions 5. Conclusions We describe a new multi-temporal classification for the Xingu River basin, Brazil. This dataset We describe a new multi-temporal classification for the Xingu River basin, Brazil. This dataset was created through the classification of Landsat TM5 and OLI 8 imagery in Google Earth Engine. was created through the classification of Landsat TM5 and OLI 8 imagery in Google Earth Engine. The Landsat archive allowed for consistency in the classifications over the four periods. The purpose The Landsat archive allowed for consistency in the classifications over the four periods. The purpose of creating these localized classifications is to address the need for accurate forest cover change data of creating these localized classifications is to address the need for accurate forest cover change data in a region with high biodiversity as well as rapid land cover change due to anthropogenic factors in a region with high biodiversity as well as rapid land cover change due to anthropogenic factors including deforestation, hydro-electric dam impoundment, mega-scale agriculture, extensive cattle including deforestation, hydro-electric dam impoundment, mega-scale agriculture, extensive cattle ranching, and population growth. The inclusion of field photographs in the training of the CART ranching, and population growth. The inclusion of field photographs in the training of the CART models allowed for the local variability of the classes to be taken into account, an aspect difficult to do models allowed for the local variability of the classes to be taken into account, an aspect difficult to on a global scale. This classification does not rely on other reference datasets, and as such, potential do on a global scale. This classification does not rely on other reference datasets, and as such, potential errors in other data sets are not propagated further. For future studies, and to address the limitations errors in other data sets are not propagated further. For future studies, and to address the limitations of the spatial resolution of these data (e.g., underestimation of small water bodies), other global optical of the spatial resolution of these data (e.g., underestimation of small water bodies), other global satellite image collections (e.g., Sentinel-2) or radar data from archived (e.g., RADARSAT-2) or new optical satellite image collections (e.g., Sentinel-2) or radar data from archived (e.g., RADARSAT-2) data (e.g., RADARSAT Constellation Mission) should be investigated. or new data (e.g., RADARSAT Constellation Mission) should be investigated.

AuthorAuthor Contributions: Contributions: O.L.,O.L., L.S., L.S., M. M.K. K. and and JPAM J.P.A.-M. collected collected the thefield field data. data. M.K. M.K. carried carried out the out analysis. the analysis. All authorsAll authors contributed contributed to the to writin the writingg and andediting editing of the of manuscript. the manuscript. Funding: Natural Sciences and Engineering Research Council of Canada, and Discovery Grant Program to Funding:Kalacska andNatural CNPq Sciences Universal and Project Engineering #486376 Research/2013-3 to Coun Sousa.cil of Canada, and Discovery Grant Program to Kalacska and CNPq Universal Project #486376/2013-3 to Sousa. Acknowledgments: We thank Márcio Vieira, Damilton Rodrigues de Costa (Dani) and Antônio Silva de Oliveira Acknowledgments:(Toninho) for their help We thank in the Márcio field. WeVieira, also Damilton thank two Rodrigues anonymous de Costa reviewers (Dani) for and their Antônio comments Silva de to improveOliveira (Toninho)our manuscript. for their help in the field. We also thank two anonymous reviewers for their comments to improve ourConflicts manuscript. of Interest: The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to Conflictspublish the of results.Interest: The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

References

1. Dale, V.H.; Pearson, S.M.; Offerman, H.L.; Oneill, R.V. Relating patterns of land-use change to faunal biodiversity in the Central Amazon. Conserv. Biol. 1994, 8, 1027–1036, doi:10.1046/j.1523- 1739.1994.08041027.x. 2. Turner, I.M. Species loss in fragments of tropical rain forest: A review of the evidence. J. Appl. Ecol. 1996, 33, 200–209, doi:10.2307/2404743. 3. Westman, W.E.; Strong, L.L.; Wilcox, B.A. Tropical deforestation and species endangerment: The role of remote sensing. Landsc. Ecol. 1989, 3, 97–109, doi:10.1007/bf00131173. 4. Pimm, S.L. The dodo went extinct (and other ecological myths). Ann. Mo. Bot. Gard. 2002, 89, 190–198, doi:10.2307/3298563.

Data 2019, 4, 114 8 of 8

References

1. Dale, V.H.; Pearson, S.M.; Offerman, H.L.; Oneill, R.V. Relating patterns of land-use change to faunal biodiversity in the Central Amazon. Conserv. Biol. 1994, 8, 1027–1036. [CrossRef] 2. Turner, I.M. Species loss in fragments of tropical rain forest: A review of the evidence. J. Appl. Ecol. 1996, 33, 200–209. [CrossRef] 3. Westman, W.E.; Strong, L.L.; Wilcox, B.A. Tropical deforestation and species endangerment: The role of remote sensing. Landsc. Ecol. 1989, 3, 97–109. [CrossRef] 4. Pimm, S.L. The dodo went extinct (and other ecological myths). Ann. Mo. Bot. Gard. 2002, 89, 190–198. [CrossRef] 5. De Almeida, C.A.; Coutinho, A.C.; Esquerdo, J.C.D.; Adami, M.; Venturieri, A.; Diniz, C.G.; Dessay, N.; Durieux, L.; Gomes, A.R. High spatial resolution land use and land cover mapping of the Brazilian Legal Amazon in 2008 using Landsat-5/TM and MODIS data. Acta Amaz. 2016, 46, 291–302. [CrossRef] 6. DeFries, R.S.; Townshend, J.R.G.; Hansen, M.C. Continuous fields of vegetation characteristics at the global scale at 1-km resolution. J. Geophys. Res. Atmos. 1999, 104, 16911–16923. [CrossRef] 7. Hansen, M.C.; Potapov, P.V.; Moore, R.; Hancher, M.; Turubanova, S.A.; Tyukavina, A.; Thau, D.; Stehman, S.V.; Goetz, S.J.; Loveland, T.R.; et al. High-Resolution Global Maps of 21st-Century Forest Cover Change. Science 2013, 342, 850–853. [CrossRef][PubMed] 8. Shimada, M.; Itoh, T.; Motooka, T.; Watanabe, M.; Shiraishi, T.; Thapa, R.; Lucas, R. New global forest/non-forest maps from ALOS PALSAR data (2007–2010). Remote Sens. Environ. 2014, 155, 13–31. [CrossRef] 9. Malhi, Y.; Roberts, J.T.; Betts, R.A.; Killeen, T.J.; Li, W.H.; Nobre, C.A. Climate change, deforestation, and the fate of the Amazon. Science 2008, 319, 169–172. [CrossRef][PubMed] 10. Pereira, J.C.; Viola, E. Catastrophic Climate Risk and Brazilian Amazonian Politics and Policies: A New Research Agenda. Glob. Environ. Politics 2019, 19, 93–103. [CrossRef] 11. Barbosa, T.A.P.; Benone, N.L.; Begot, T.O.R.; Goncalves, A.; Sousa, L.; Giarrizzo, T.; Juen, L.; Montag, L.F.A. Effect of waterfalls and the flood pulse on the structure of fish assemblages of the middle Xingu River in the eastern Amazon basin. Braz. J. Biol. 2015, 75, S78–S94. [CrossRef][PubMed] 12. Camargo, M.; Giarrizzo, T.; Isaac, V. Review of the geographic distribution of fish fauna of the Xingu river basin, Brazil. Ecotropica 2004, 10, 123–147. 13. Dagosta, F.C.P.; De Pinna, M. The fishes of the Amazon: Distribution and biogeographical patterns, with a comprehensive list of species. Bull. Am. Mus. Nat. Hist. 2019, 431, 163. [CrossRef] 14. Fitzgerald, D.B.; Perez, M.H.S.; Sousa, L.M.; Goncalves, A.P.; Py-Daniel, L.R.; Lujan, N.K.; Zuanon, J.; Winemiller, K.O.; Lundberg, J.G. Diversity and community structure of rapids-dwelling fishes of the Xingu River: Implications for conservation amid large-scale hydroelectric development. Biol. Conserv. 2018, 222, 104–112. [CrossRef] 15. Da Silva, J.M.C.; Rylands, A.B.; Da Fonseca, G.A.B. The fate of the Amazonian areas of endemism. Conserv. Biol. 2005, 19, 689–694. [CrossRef] 16. Schwartzman, S.; Boas, A.V.; Ono, K.Y.; Fonseca, M.G.; Doblas, J.; Zimmerman, B.; Junqueira, P.; Jerozolimski, A.; Salazar, M.; Junqueira, R.P.; et al. The natural and social history of the indigenous lands and protected areas corridor of the Xingu River basin. Philos. Trans. R. Society B-Biol. Sci. 2013, 368. [CrossRef][PubMed] 17. Garcia, E.; Ramos, F.S.V.; Mallmann, G.M.; Fonseca, F. Costs, Benefits and Challenges of Sustainable Livestock Intensification in a Major Deforestation Frontier in the Brazilian Amazon. Sustainability 2017, 9, 158. [CrossRef] 18. Frederico, R.G.; Olden, J.D.; Zuanon, J. Climate change sensitivity of threatened, and largely unprotected, Amazonian fishes. Aquat. Conserv. Mar. Freshw. Ecosyst. 2016, 26, 91–102. [CrossRef] 19. Gilbert, N. Forest definition comes under fire. Nat. News 2009.[CrossRef] 20. Breiman, L.; Friedman, J.; Stone, C.J.; Olshen, R.A. Classification and Regression Trees; Taylor and Francis Group: Boca Raton, FL, USA, 1984.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).