The environmental criteria, that were selected through the literature review (Tables 1 and 2), represent important characteristics of landscapes under urban sprawl, in the context of the prioritizing areas to obtain forest functional connectivity (Curiel-Esparza et al., 2015; Lakicev et. al., 2016; Vettorazzi and Valente, 2016; Santos, et al., 2018, and Mello et al., 2018).
The topographic criteria have effects on drainage capacity (Loritz et al., 2019; Hojati and Mokarram, 2016), forest composition and structure (Jucker et al., 2018), habitat reduction and isolation (Zhang, et al, 2018) and, consequently, on species patterns and richness (Li et al., 2019, Russo et al. 2016, and Keeley, Beier, and Gagnon, 2016), which are essential conditions for measuring forest connectivity.
The distance from drainage network criterion indirectly brings the concept and the importance that riparian forests have in maintaining forest connectivity (La Fuente, et al., 2018). Still, the question of the flow among forest patches, that is compromised by the expansion of urban infrastructure, which produces barriers in the landscape (McRae et al., 2012), habitat loss (Alamgir et al., 2019), and changes in animal movement (Dickie et al., 2019).
As we mention in Table 2, the criteria have different importance for the functional connectivity and forest conservation, representing a part of the process to obtain the main objective. Importance that has been highlighted in the literature review. (Ayram et al., 2016, Fernádez and Morales, 2016, Unda and Etter, 2019 and Rincón et la., 2019).
This way, the criteria selection based on literature review brings robustness to the study, considering that the set was previous evaluated in other researches. Also eliminates the subjectivity that we bring to the project when experts opinion is required, considering the restricted number of people who contributes to the process (Silveira et al., 2014, Valente et. al., 2017 and Mello et al., 2018).
In the case of criteria selected for the study area in the GBBR-SP, Brazil, they are not spatially correlated and no have information overlap (Figs. 4 and 5). Thus, they can be used to prioritize areas to functional connectivity.
Relating the main characteristics bring for the criteria, we obtain through the slope that 91% of the study area showed declivity at most 20% (mean value of 11.83%; ± 5.17) (Table 4 and Fig. 4A). This predominance of declivity value associated with the low variation of its standard deviation supports the definition of a spatial pattern for the criterion, which was identified by LISA index (I = 0.457) (Fig. 5A). Kane and John (2018) cited that topographical factors determine land-use decisions by local communities, considering if the relief allows greater accessibility, there is an increase in deforestation (Bax, Franscesconi and Quintero, 2016) and a critical potential for natural regeneration (Molin et al., 2018).
Authors as Adams, Barnard and Loomis (2014) complement that tree growth and forest productivity can be affected by slope, as well as for the TWI and aspect due to influences on runoff and wind exposure. Thus, these factors allow the elaboration of forest connectivity models based on relief, as the species have different habitat requirements (Czarnecka, Rysiak, and Chabudzinski, 2017).
Relating to TWI criterion, we obtained approximately 62.1% of the total area varying from 5–12, including soils classified as well-drained (TWI of 4–5), moderately drained (TWI of 5–7), and poorly drained soils (TWI of 7–12) (Li et al., 2006). Approximately 43% of this total is associated with the moderately drained soils, contributing to classification of this areas as low-low by LISA Index (Fig. 5B), since they are showed a scattered distribution through the landscape.
According to our results, regions associated with the 5–12 TWI range are occupied by forest fragments, showing an average of 7.08 and a low standard deviation (± 2.27) (Table 4).
This way, TWI was characterized as a heterogeneous criterion, having Moran Index (I) of 0.190) as also mentioned by Da Silva, Santos and Oka-Fiori (2019).
Similarly, the aspect also was defined as heterogeneous, having an I value of 0.292, however with a standard deviation of ± 70.67, (Table 4). This criterion behavior occurs in our study area, even it characterized by has a topographic divider of water flow and, having 65.3% of its area associated with East-West slopes facing (Fig. 4C). These facings are associated with forest patches groups (LISA index, Fig. 5C), which are the most regular patches of the landscape placed in faces with the lowest aspect values (CCA analysis, Fig. 8). This can be explained by the predominance of anthropic land-use on the east landscape face (Victor et al., 2004).
Aspect reveals a tendency to coming toward the slope values because they came from DEM. However, we cannot establish a statistical and significant pattern between them. Adding the fact, that the criterion does not present high spatial correlation, it cannot be considered adequate for a model of prioritization areas, aiming the functional connectivity. This justified by the uncertainties in connectivity analysis, that the aspect can bring to the model.
According to our results, the aspect is the only environmental criterion not adequate for prioritizing areas to obtain forest functional connectivity, in landscape subject to the urban sprawl.
In the same way that the slope and TWI, the distances from forest patches and drainage network criteria are essential to model of the prioritization of areas subject to urban sprawl (Fig. 4D).
According to these maps the average distance of forest patches from the water courses was 199.27 m, with 56% of them at most 200 m. Considering that 36% of the total forest patches is between 200m and 400 m from the river and ± 126.48m as maximum value of deviation standard (Table 4), we obtained that the majority patches (including the 56%) is from 325.75 m from a watercourse.
The distances from the drainage network criterion supported these results. Firstly, because the watercourses are adequately distributed throughout the study area (I = 0.477) (Figs. 4 and 5D), that resulted in a high spatial autocorrelation with a value superior to the topographic criteria. Second, because the autocorrelation supported the clusters structuration in regions, where there were the minor distances of the patches from watercourses (LL in LISA), as well as there were the highest (HH) distances (Fig. 5D).
Observing the criterion results, we noticed that they were predominately supported by the relation with the proximity to watercourses, which is the most important region for the main objective. This way, the most adequate name for the criterion is proximity to the drainage network instead of the distances from the drainage network. The name reflects the most important region for the criterion and becomes in accordance with the literature, wherein have been considering riparian forest corridors as the most suitable environment for forest connectivity (La Fuente et al., 2018, Zimbres et al., 2018).
While the watercourses are scattered through the landscape, the main features of the distance from the low-density urban area and highways criteria showed concentrated, mainly, in its Central-to-South-west portion (Fig. 4E and Fig. 4F). Due the presence of the great urban areas and roads, the forest patches cluster of this regions presented spatial autocorrelation classified as LL (Fig. 5E; I = 0.976 and Fig. 5F; I = 0.992). It is noted in the Fig. 4E that a restricted area was created over consolidated urban agglomerations, to analyze only the effects of urban sprawl on forest remnants.
However, the features of criteria (roads and urban) and consequently theirs respectively distance maps showed an effect in the forest fragments spatialization. Around 71% of the forest patches are 2000 m from an urban area and 86% of them are 2000 m from a road.
Anthropogenic disturbance in the landscape favors generalist species that are able to explore environments, such as disturbed habitats (Magioli et al., 2019). Consequently, there is an increase in conflict factors, such running over wildlife (Abra et al., 2021), predation of farm animals (McPherson, Brown, and Downs, 2016), and animals lethal control (Blackwell, et al., 2016), being essential to define the most beneficial actions of conservation planning and implementation (Abra et al., 2021).
Conflict features that showed high spatial correlation (PCA analysis, Fig. 6), indicating that they can compose a unique criterion in future studies. Criterion that brings information related with the connectivity, since that urbans areas and road have crossed landscapes subject to urban sprawl, which have become one of the greatest threats to biodiversity conservation (Scriven et al., 2019; Madadi et al. 2017).
Kuang et al. (2014) cited that the urban expansion process and the relationship with forest fragmentation is an effect of worldwide verification, carried out on the scale of megacities, and due to the increase in the population living in urban areas (Angel et al., 2016). Thus, given that land use can occur in many different patterns, metrics that consider the spatial arrangement of urban infrastructure can better explain landscape fragmentation (Bar-Massada, Radeloff, and Stewart, 2014; Lin and Fuller, 2013).
Relating the importance to maintain the criteria distance from forest patches, our results indicate that we have forest patches supporting the native fauna and flora, especially the group formed by patches with sizes greater than 300 ha (Table 4 and Fig. 4H). As we mentioned, this group is formed by 21 patches, belong to three Protect Areas, located inside the study area, and occupy 50% of the total forest area. Even more, these forest patches are integrated with others, considering that the minor than they are scattered across the landscape (Table 3).
According to Magioli et al. (2019), the large and continuous habitats support populations with more complex trophic structures, acting as a source for biodiversity maintenance in modified habitats. Gibson et al., (2011) complemented that these habitats are essential refuges for wildlife, assuming that their similarity to natural areas (i.e, without anthropic actions).
The Brazilian environmental legislation has been encouraged the conservation of areas as patches greater than 300 ha, however, it is not enough to minimize the urban pressure effect (Romero et al., 2020). In our study areas is not only this group that has potential for connectivity in the studied landscape. We can include the group having medium size too, highlighting the importance of the criterion distances from forest patches. Considering these proximity relations among the forest patches, we can suggest for next studies the name proximity to forest cover as proposed by author as Mello et al., 2018.
Showing different behavior of other criteria, we have the NDVI. The traditional index represents the forest patches in terms of their status and/or quality (Fig. 4H) i.e., the conditions inside the remnants. Consequently, it is not a criterion that represents the conditions of the natural vegetation on the landscape, as our objective requires.
However, patches-level data are important to support the decision-makers discussion, which guided us to includes our second objective.
For the study area, the patches-level metrics showed coherent resultants, considering the classes areas, that we proposed. The largest forest remnants showed great values of perimeter and shape than the smallest, but minor values of distance among their components (Table 3).
Furthermore, the class areas supporting the evaluation of the spatial autocorrelation among metrics and after, among metrics and criteria.
The results were a high Moran index value for metrics (Fig. 7), which supported in their correspondence with the criteria since these results reflected that the external influences on the fragments did not occur randomly and that the criteria act on the landscape (with 84.1% of explanation in CCA, Fig. 8).
Different studies have been shown the potential of the ecological metrics in the definition of areas for conservation, on analysis of sprawl urban and evaluation of the landscape structure (Schindler et al., 2013, De Jesus, et al., 2015, Otero et al., 2015, Romero et al., 2018, and Lelli et al., 2019), corroborating with our results.
The CCA analysis (Fig. 8) also indicated that the reduction in the distance from low-density urban areas was associated with a reduction in the distance among forest fragments (NEAR). While the distance from the highways was associated with an increase in the perimeter of the forest patches, and consequently their respective areas. According to Pirnat and Hladink (2018), new diffuse urban areas are not subject to significant changes in terms of habitat size and this seems to happen regardless of population changes (Organization of United Nations Habitat, 2016).
Otherwise, the drier regions of the landscape, on the flatter terrains, near to the rivers, and facing to the east face were associated with our smallest and most irregular patches, that were isolated and presented the lowest NDVI values. Thus, in these landscape regions, the condition related to the urban infrastructure and topography were favorable to human occupation, as also observed by Torres, Jaeger, and Alonso, 2016.