Journal of Geography  & Natural Disasters
Open Access

Cloud-Based Application for Spatial and Statistical Analysis about Environmental Fragility and Land Cover

Luiz Fernando de Novaes Vianna^* and Fabio Martinho Zambonim ^*Correspondence: Luiz Fernando de Novaes Vianna, Department of Biology, Environmental Resources and Hydrometeorology Information Center of Santa Catarina, Florianopoli, Brazil, Email:

Author info »

Introduction

Mainly occurring in tropical forests (Atlantic Rainforest and the Amazon), deforestation is a worldwide environmental threat, particularly considering the associated factors of climate change and loss of biodiversity [1]. The Brazilian Annual Land-Use and Land-Cover Mapping Project (MapBiomas) developed a free and open-source platform to monitor Brazilian land cover changes since 1985 [2]. With the MapBiomas platform (https://plataforma. brasil.mapbiomas.org/), it is possible to analyze annual land cover change and quantify the rate of natural vegetation loss. It is also possible to determine land cover changes. However, land cover changes are not homogeneous given a contiguous mosaic of landscapes, and the impact of change is also different, depending on physical and anthropic aspects of landscapes. Therefore, to increase the analytic power of land-use and land-cover LULC, we created the Environmental Fragility Mapping System (EFMS) linking LULC change analysis to real-time physical characteristics of the landscape. Developed from Google Earth Engine (GEE) and based on MapBiomas datasets [2,3], landsat Normalized Difference Vegetation Index (NDVI), road and hydrographic networks, and terrain metrics, EFMS is a customizable application to generate raster datasets, charts, and tabular data, embracing the environmental fragility concept [4-10].

Environmental Fragility (EF) analysis originates from ecodynamics and consists of Potential Environmental Fragility (PEF), representing physical attributes of the landscape (e.g., terrain metrics, hydrologic networks, climate and soil) and Emergent Environmental Fragility (EEF), arising from the effects of Land-Use And Land-Cover (LULC) changes on PEF [4,8]. EF analysis is superior to LULC change analysis by its flexibility in evaluating multiple scenarios for different aspects of environmental challenges, such as conservation of natural ecosystems, prioritization of fragments for landscape restoration, soil erosion, natural hazards, and monitoring EF changes [5,9-13].

PFI represents physical fragility of the environment caused by the factors of relief and hydrography. LCFI represents land-cover fragility by LULC classes of MapBiomas, Landsat Normalized Difference Vegetation Index (NDVI) and road networks. EFI is the result of combination between PFI and LCFI.

To implement the concept of EF in EFMS, we combined the spatial analysis power of Google Earth Engine (GEE) with a multi-criteria approach, using the Analytic Hierarchy Process (AHP) to assign weights to MapBiomas LULC classes considering the degree of environmental fragility [3,14]. We also adopted AHP to assign default weights for spatial data through aggregating the knowledge of a multidisciplinary research team. Notably, EFMS enables the user to change all default weights and create other parametrizations, depending on need; in other words, it is customizable.

EFMS can perform spatial analysis and generate maps and data for environmental fragility analysis for multiple objectives, such as legal compliance, ecological corridor design, environmental recovery planning, soil loss, and landslide and flood fragility mapping.

In this article we describe the environmental fragility analysis method implemented in EFMS, provide the code in Java Script and present a case study applied to the municipality of Alfredo Wagner, in Santa Catarina, Brazil. The objective of the case study was to demonstrate the potential of EMFS to generate fragility scenarios for selecting priority areas for environmental recovery.

Materials and Methods

EFMS was implemented in GEE using JavaScript [3]. Two versions of EFMS code were written, one in English and other in Brazilian Portuguese. The English version of EFMS is available at https://ee-vianna.projects.earthengine.app/view/efi as shown in Figure 1, while the Portuguese version can be found at https:// ee-vianna.projects.earthengine.app/view/ife. From both links, the user manual of EFMS is also available, but only in Brazilian Portuguese. The user manual (https://doi.org/10.5281/ zenodo.10033619) details the method and presents some practical examples for using the EFMS to generate maps and export charts and tabular data [15,16].

Figure 1: English version of EFMS. The layout is divided into three parts, two side panels and map. The main panel (left) contains a brief description of the system, link to the user manual, tools for parameterizing the Priority Area (PA), and fragility index calculation models (PFI, LCFI and EFI). The panel on the right side shows the legends and graphs of area per mapped class. The map shows all layers generated. In this example, the 2022 LULC layer of Santa Catarina is drawn.

The project was developed for Santa Catarina State, Brazil as shown in Figure 1, However, the code is open-source and can be adopted worldwide. To do this, it is necessary to overwrite the Areas of Interest (AOI), the LULC map, the hydrography and the OpenStreetMap (OSM) road maps with values of the new area for the same parameters. The system is spatial scale sensitive. Larger spatial scales of input data generate more accurate outputs. For the Santa Catarina State project, the EFMS maximum scale is 1:50,000, the highest scale for Landsat products [17,18].

The method for calculating the indexes can be divided into three steps as shown in Figure 2, including generating terrain metrics, NDVI, and distance maps based on raw data (Step 1), min-max scaling the raster datasets to 0-1 (Step 2), and defining weights for LULC classes and for min-max scaled raster and applying map algebra to calculate indexes (Step 3).

Figure 2: Workflow of EFMS method.

To run EFMS, the user first selects the Area of Interest (AOI), defines the parameters for mapping the Priority Areas (PA), selects the year of analysis and then weights the land cover classes and variables that constitute the indexes. The system executes all calculations for indexes each time the "Calculate EFI" button is pressed. The processing speed is proportional to the AOI. Large areas slow down the processing.

Raw data

We used the Shuttle Radar Topographic Mission (SRTM) Digital Elevation Model (DEM), Landsat 5, 7 and 8, and LULC data from the Brazilian Annual LULC Mapping Project (MapBiomas Collection 8; https://brasil.mapbiomas.org/colecoes-mapbiomas/) from 1985 to today [19]. The 30 m spatial resolution of raw data defines the output spatial resolution of indexes. We also used the 1:10,000 hydrography network and the Open Street Map (https://www.openstreetmap.org/) road network for Santa Catarina State to calculate Euclidean distance [20].

Step 1

In the first step, two terrain metrics, NDVI, two Euclidean distances and the PA mask raster datasets are generated. The system clips the data for the selected AOI. From SRTM-DEM, the system generates slope and Topographic Position Index (TPI) raster datasets [21]. From Landsat images, it generates the median NDVI, using Landsat 5 data for years between 1985 and 2011, Landsat 7 data for 2012, and Landsat 8 data for years beyond 2012. From the MapBiomas raster dataset, the system selects the year and clips the LULC raster dataset. From the hydrography network and road network maps, the system calculates two Euclidean distance raster datasets. The system also generates the PA mask raster dataset according to values defined by the user. By default, the values for altitude are 0-1900 m, the range found in the state of Santa Catarina. The parametrization of elevation enables the user to calculate the indexes by range of altitude. The default parametrization values are also 0-30 m for distance of hydrography, 0º-45º for slope and 0.6-1 for TPI. These are the values that best represent the Permanent Preservation Areas (PPA), as defined by the Brazilian National Forest Code [15]. However, no parametrization values are fixed, allowing the user to customize masks. At the end of the first step, seven variables are generated and added as eight layers to the map: Altitude, slope, TPI, distance from hydrography, distance from roads, NDVI, land cover and land cover within PA.

Step 2

In this step, the raster datasets are min-max scaled for 0-1, according to the environmental fragility of each variable. Min-max scaling is linear and considers the minimum and maximum values of each variable within the AOI, which makes the index sensitive to the scale defined by the AOI. Min-max scaling is calculated as Equation 1.

In scaled variables, values close to 0 represent lower environmental fragility, while values close to 1 indicate higher fragility. The variables Slope and TPI are directly proportional to fragility. This means that higher values of slope or TPI correlate with higher environmental fragility. The variables NDVI, distance from hydrography and distance from roads are inversely proportional to environmental fragility. Lower values of NDVI and distances from hydrography correlate with higher environmental fragility. After min-max scaling, NDVI and distances are inverted, multiplying by -1 and adding 1. This is necessary to adjust values for the adopted scale.

Also in this step, MapBiomas land cover classes are weighted. The weight of each class is directly proportional to environmental fragility. The default weights were defined by a multidisciplinary team of researchers and experts, applying the Analytic Hierarchy Process (AHP) [14]. In Table 1, we show the weight of each LULC class. The natural LULC classes were considered as the lowest environmental fragility and weighted to zero.

Table 1: Weight of MapBiomas LULC classes. Values defined through AHP [14].

Step 3

In the last step, the system calculates PFI, LCFI and EFI. PFI is calculated by the weighted average among the scaled datasets of slope, TPI and distance from hydrography. By default, the adopted weights are 0.4 for distance from hydrography, 0.35 for slope and 0.25 for TPI. LCFI is calculated by the weighted average among the scaled datasets of MapBiomas LULC, NDVI and distance from roads. By default, the adopted weights are 0.4 for the MapBiomas LULC, 0.35 for NDVI and 0.25 for distance from roads. EFI is the IFP multiplied by the IFSC. The result is a raster dataset with values between 0 and 1 where 0 is the lowest fragility and 1 the highest. The scale of fragility is represented in Table 2.

Table 2: Weight of variables, scale, and colors of PFI, LCFI and EFI.

EFMS parametrization for selecting environmental recovery priority areas within Permanent Preservation Areas (PPA), as defined by the Brazilian national forest code

For selecting environmental recovery priority areas within Permanent Preservation Areas (PPA), we use the EFMS default parametrization once it was implemented according to Brazilian National Forest Code [15]. The only parametrizations were to select the city of Alfredo Wagner as AOI and generate charts and tables for the year 2022.

EFMS parametrization for designing ecological corridors

For selecting environmental recovery priority areas for designing ecological corridors, we change the values of PA to create connective corridors based on hydrography, slope, and topographic position. We change the values of distance from hydrography to 0-50 m, slope to 30º-90º and TPI to 0.5-1 to select areas within 50 m of rivers, areas with more than 30º slope, and areas with upper slopes and hilltops. We also changed the weight of each physical variable to 0.35 for distance from hydrography and slope and 0.3 for TPI. Finally, we changed the weight of forest plantation LULC class from 0.05 to 0.23, the same weight as that of other agriculture and monoculture classes as shown in Figure 3.

Figure 3: EFMS weight customization to identify priority areas to create ecological corridors for environmental recovery in Alfredo Wagner, SC. Note: Redefine weights of land cover classes: Values must between 0 to 1: Non observed.

Results and Discussion

According to MapBiomas (https://plataforma.brasil.mapbiomas.org/), from 1985 to 2022, the state of Santa Catarina lost 999,412ha of natural habitat to farming, while 468,289 ha of farming land were recovered to forest or non-forest natural formation as shown in Table 3.

Table 3: Area (ha) of LULC class transition from 1985 to 2022 in Santa Catarina.

This dynamic shows that Santa Catarina is still losing natural areas to anthropic activities, mainly farming. To reverse the loss of natural areas in Santa Catarina, it is necessary to adopt environmental recovery policies. The focus of new policies for environmental recovery should be based on territorial planning rather than legal or socioeconomic mechanisms [22]. New policies should also consider all natural LULC classes as environmentally fragile and important to maintenance and recovery, not only forest formations [23]. In this effort, EFMS was designed to aid scientists, managers, politicians, and stakeholders in identifying and selecting priority sites and then calculating the area to be recovered.

The municipality of Alfredo Wagner is a representative example of ecodynamics in Santa Catarina. Alfredo Wagner is environmentally heterogeneous with altitude ranging from 400 to 1,200 m above sea level. The original vegetation of Atlantic Rainforest, Araucária Forest, Grasslands and Faxinal Forests was replaced by pasture, temporary crops and forest plantation (non-native species, i.e., Pinnus elliotti) in the 20th century [24].

According to MapBiomas (https://plataforma.brasil.mapbiomas.org/), from 1985 to 2022, the municipality lost 6,238 ha of forest formation and 3,387 of grassland for pasture, agriculture, and forest plantation. In the same period, 1,121 ha of pasture, 2,238 ha of agriculture and 229 ha of forest plantation were recovered to forest formation, and 279 ha of pasture, 40 ha of agriculture and 5 ha of forest plantation were recovered to grassland as shown in Figure 4.

Figure 4: Sankey chart of and use/land cover (level 2) changes in Alfredo Wagner from 1985 to 2022. Note: MapBiomas (https://plataforma. brasil.mapbiomas.org/).

To illustrate the practical application of EFMS, we selected and Prioritized Areas (PA) for environmental restoration in Alfredo Wagner, considering two scenarios: (i) To identify priority sites for environmental recovery in rural areas to be in compliance with the legal code that defines PPA and (ii) To identify PA to create Ecological Corridors (EC) for environmental recovery.

Scenario 1: To identify priority sites for environmental recovery in rural areas to be in compliance with the legal code that defines PPA

The Brazilian National Forest Code defines PPA based on proximity to hydrography (riparian areas), terrain slope and topographic position [15]. According to the Forest Code, PPA should be covered by natural vegetation in 100% of its area [15]. In the first objective, we used the legal definition of PPA to set the default values for distance from hydrography (0-30 m), slope (45º-90º) and TPI (0.6-1). The resultant default values generate PA close to those defined by legislation, but the spatial resolution of 30 m and the 1:50,000 scale of the raw data were unsuited for assessing PPA in larger scales.

The result shows that Alfredo Wagner has more than 30% of territory with very high and high EFI and close to 40% with intermediate EFI as shown in Figure 5 and Table 4. This represents 50,844 hectares that need environmental recovery. However, in those areas are the urban center of Alfredo Wagner and a lot of farms. To select the areas to be recovered, we assessed the environmental fragility of the PPA defined by the forest code.

Figure 5: 2022 EFI map and area chart of Alfredo Wagner, SC. Note: EFMS (https://ee-vianna.projects.earthengine.app/view/efi).

Table 4: Area of EFI classes in Alfredo Wagner (case 1).

The percentage of very high and high EFI within the PPA is higher than 43% as shown in Figure 6 and Table 4. To better explore the results, it is possible to overlay EFI maps and Google´s high-resolution images and apply transparent levels of drawing as shown in Figure 6. These areas are concentrated in valleys and primarily represent riparian regions that should be prioritized for environmental recovery as shown in Figure 6.

Figure 6: 2022 EFI map and area chart of Alfredo Wagner’s PPA, SC, to identify priority sites for environmental recovery in rural areas in order to comply with the Brazilian national forest code. In the detailed area, we overlayed the EFI and high-resolution satellite images to show three highly fragile environmental areas: (1): Deforested area with high slope; (2): Agricultural activity on hilltops, and (3): Construction of a small hydroelectric plant in a riparian area with a high slope. Note: EFMS (https://ee-vianna.projects.earthengine.app/view/efi).

Alfredo Wagner has 6,056 ha of high and very high EFI within the PPA as shown in Table 4. To recovery all PPA with high and very high EFI using an agroforest design with mean density of 603 plants/ha are necessary 3,651,768 seedlings [25].

In some sites, such as those indicated in the detailed satellite image in Fig. 6, it is possible to identify deforested high slope 1): Agricultural activities on hilltops; 2): And riparian areas; 3): According to the Brazilian Forest Code, Permanent Preservation Areas (PPA) must be forested or covered by natural vegetation [15]. In some cases, they can be used for agricultural activities, but in accordance with soil and water conservation criteria. However, most of unforested PPA areas must be completely recovered, and priority can be given to those with higher environmental fragility.

Scenario 2: To identify priority areas to create Ecological Corridors (EC) for environmental recovery

Ecological corridors can preserve functionality and guarantee the species mobility since they provide links of connectivity within a heterogeneous territory [16]. Unlike PPAs, ecological corridors are not defined by law, with rigid boundaries, and must be designed according to landscape criteria.

In the second scenario, we reduced the weight of distance from hydrography by distributing the weight of physical variables comprising PFI. This resulted in reducing the importance of riparian areas, as defined by Brazilian law, by considering both relief aspects and riparian areas as equally important areas for EC. At the same time, by increasing the weight of the LULC class termed as forest plantation, we defined it as a monoculture that contributes to biodiversity lost. With this new parametrization, EC should contribute to biodiversity to a degree exceeding that of legislation.

Considering this scenario, the areas with high and very high EFI in Alfredo Wagner total 14,669 ha as shown in Figure 7 and Table 5.

Figure 7: 2022 EFI map and area chart of Alfredo Wagner, SC, to identify priority areas to create EC for environmental recovery. Note: EFMS (https://ee-vianna.projects.earthengine.app/view/efi).

Table 5: Area of EFI classes in Alfredo Wagner (scenario 2).

To create an EC network in Alfredo Wagner, we would start prioritizing 8,197 ha of high and very high EFI for natural vegetation recovery as shown in Figure 8 and Table 5. Observing the detailed area of EFI overlayed with high-resolution satellite image as shown in Figure 8, it is possible to perceive that environmental fragility does not represent vegetation coverage, but rather the integration of LULC and potential environmental fragility by relief and hydrography. Some areas with intermediate or low EFI can be covered by natural vegetation as shown in Figures 1 and 8. On the other hand, some areas with high EFI may appear to be covered in natural vegetation, but are, instead, covered by non-native planted forests as shown in Figures 2 and 8.

Figure 8: 2022 EFI map and area chart of Alfredo Wagner, SC, to identify priority areas to create EC for environmental recovery. In the detailed area, we overlayed the EFI results and high-resolution satellite images to show where natural vegetation recovery should be prioritized, considering connective ecological corridors and environmental fragility. Note: EFMS (https://ee-vianna.projects.earthengine.app/view/efi).

Conclusion

This kind of analysis helps to plan for reforestation and the connectivity of EC by prioritizing areas, defining the species to be utilized on natural vegetation recovery, calculating the number of seedlings needed and designing the species composition that matches unique environmental characteristics, such as hydrography, slope, sunlight or shadows

The environmental fragility concept can be widely applied. The Alfredo Wagner, SC case study showed the potential for evaluating environmental fragility for different objectives. Combined with MapBiomas LULC change analysis, EFMS is a user-friendly tool for spatial analysis and generation of maps and data for environmental management.

Data Availability

All the raw data used by EFMS are from public datasets. The Shuttle Radar Topographic Mission (SRTM) digital elevation model (DEM) and the Landsat 5, 7 and 8 images are available from GEE datasets (https://developers.google.com/earth-engine/datasets), that can be accessed by any user with an account.

Land use land cover data from the Brazilian Annual LULC Mapping Project (MapBiomas Collection 8; https://brasil. mapbiomas.org/colecoes-mapbiomas/) from 1985 to today can be accessed by GEE dataset link (https://code.earthengine.google.com/?asset=projects/mapbiomas-workspace/public/collection8/mapbiomas_collection80_integration_v1).

Spatial and statistical data about environmental fragility and land cover, from 1985 to today, for the Santa Catarina State and other political and environmental subdivisions are available in EFMS English (https://ee-vianna.projects.earthengine.app/view/efi), or Portuguese version (https://ee-vianna.projects.earthengine.app/view/ife). Data generated by EFMS can be analyzed and viewed in the system or exported as a table (csv) or chart (png, svg). Since each analysis generates a specific dataset, EFMS itself can be considered a data repository.

Code Availability

All codes are available from Zenodo (https://doi.org/10.5281/zenodo.10018501) under Creative Commons Attribution 4.0 International. In the Zenodo repository, users will find the freely available codes in JavaScript and a user manual. AOI, hydrography and roads data are for Santa Catarina State, Brazil. To run the code for other regions, it is necessary to overwrite these data for local ones in GEE assets. We suggest using data at higher scale of 1:50,000.

Acknowledgement

This study was financed by the Fundação de Amparo à Pesquisa e Inovação do Estado de Santa Catarina (FAPESC)–Finance Code TO 2021TR001397. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author Contributions

Luiz Vianna: Conceptualization, Methodology, Software, Writing-Original draft preparation, Writing-Review and Editing, Validation. Fábio Zambonim: Validation, Writing-Review and Editing, Project administration, Funding acquisition.

Competing Interests

The author(s) declare(s) that there is no conflict of interest regarding the publication of this article.

References

1. Diaz S, Settele J, Brondízio E, Ngo HT, Guèze M, Agard J, et al. The global assessment report on biodiversity and ecosystem services. IPBES. 2019.
2. Souza Jr CM, Z Shimbo J, Rosa MR, Parente LL, A Alencar A, Rudorff BF, et al. Reconstructing three decades of land use and land cover changes in Brazilian biomes with landsat archive and earth engine. Remote Sensing. 2020;12(17):2735.

   [Crossref] [Google Scholar]

3. Gorelick N, Hancher M, Dixon M, Ilyushchenko S, Thau D, Moore R. Google earth engine: Planetary-scale geospatial analysis for everyone. Remote Sensing Env. 2017;202:18-27.

   [Crossref] [Google Scholar]

4. Tricart J. Ecodynamics. IBGE; SUPREN. 1977.

   [Google Scholar]

5. França LC, Lopes LF, Morais MS, Lisboa GD, Rocha SJ, Morais Junior VT, et al. Environmental fragility zoning using GIS and AHP modeling: Perspectives for the conservation of natural ecosystems in Brazil. Conservation. 2022;2(2):349-366.

   [Crossref] [Google Scholar]

6. Rosa MR, Ross JL. Application of GIS in generating fragility maps. Magaz Depart Geograph. 1999;13:77-106.

   [Crossref] [Google Scholar]

7. de Jesus França LC, Piuzana Mucida D, Campos Santana R, Santos de Morais M, Rezende Gomide L, de Meneses Bateira CV. AHP approach applied to multi-criteria decisions in environmental fragility mapping. Forest. 2020;50(3).

   [Crossref] [Google Scholar]

8. Ross JL. Empirical analysis of the fragility of anthropized natural environments. Magaz Depart Geograph. 1994;8:63-74.

   [Crossref] [Google Scholar]

9. Roberts NJ, Nadim F, Kalsnes B. Quantification of vulnerability to natural hazards. Georisk. 2009;3(3):164-173.

   [Crossref] [Google Scholar]

10. Xia M, Jia K, Zhao W, Liu S, Wei X, Wang B. Spatio-temporal changes of ecological vulnerability across the Qinghai-Tibetan Plateau. Ecological Indicators. 2021;123:107274.

    [Crossref] [Google Scholar]

11. de Matos TP, de Matos VP, de Mello K, Valente RA. Protected areas and forest fragmentation: Sustainability index for prioritizing fragments for landscape restoration. Geol Ecolog Landscap. 2021;5(1):19-31.

    [Crossref] [Google Scholar]

12. Valente RA, de Mello K, Metedieri JF, Américo C. A multicriteria evaluation approach to set forest restoration priorities based on water ecosystem services. J Environ Manag. 2021;285:112049.

    [Crossref] [Google Scholar]

13. Guo L, Liu R, Men C, Wang Q, Miao Y, Shoaib M, et al. Multiscale spatiotemporal characteristics of landscape patterns, hotspots, and influencing factors for soil erosion. Scien Total Environ. 2021;779:146474.

    [Crossref] [Google Scholar]

14. Saaty RW. The analytic hierarchy process-what it is and how it is used. Mathem Model. 1987;9(3-5):161-176.

    [Crossref] [Google Scholar]

15.  Law No. 12,651, of May 25, 2012. Provides for the protection of native vegetation. Official Gazette Union. 2012.

    [Google Scholar]

16. Salomão NV, Silva LS, Fernandes GW, Piuzana D, Gonzaga AP, Machado EL. Criteria for the implementation of ecological corridors in fragmented areas in southern Brazil. Forest Scien. 2023;33:e70337.

    [Crossref] [Google Scholar]

17. Welch R, Jordan TR, Ehlers M. Comparative evaluations of the geodetic accuracy and cartographic potential of landsat-4 and landsat-5 thematic mapper image data. Photogram Engin Remote Sens. 1985;51(9):1249-1262.

    [Google Scholar]

18. Vergara OR, Cintra JP, D'alge JC. Assessment of the artographic accuracy of documents updated with orbital images and geographic information systems. Embrapa Acre. 2002.

    [Google Scholar]

19. Farr TG, Rosen PA, Caro E, Crippen R, Duren R, Hensley S, et al. The shuttle radar topography mission. Rev Geophys. 2007;45(2).

    [Crossref] [Google Scholar]

20. Aerosurvey of the state of Santa Catarina-final production report, Florianópolis. SDS. 2012.
21. Jenness J. Topographic position index. TPI Document. 2006.

    [Google Scholar]

22. Souza JM, Morgado P, Costa EM, Vianna LF. Predictive scenarios of LULC changes supporting public policies: The case of chapecó river ecological corridor, Santa Catarina/Brazil. Land. 2023;12(1):181.

    [Crossref] [Google Scholar]

23. Souza JM, Morgado P, Costa EM, Vianna LF. Modeling of Land Use and Land Cover (LULC) change based on artificial neural networks for the Chapecó river ecological corridor, Santa Catarina/Brazil. Sustainability. 2022;14(7):4038.

    [Crossref] [Google Scholar]

24. Klein RM. Phytogeographic map of the state of Santa Catarina. Herbar Barbosa Rodrig. 1978.

    [Google Scholar]

25. Hanisch AL, Vogt GA, da Cunha Marques A, Bona LC, Bosse DD. Structure and floristic composition of five waterfall areas in the North Plateau of Santa Catarina. Brazilian Fores Res. 2010;30(64):303-310.

    [Crossref] [Google Scholar]