Ecohydrological Analysis in Watersheds of Mountain Areas of S ã o Paulo State Coastal, Brazil

: In the Brazilian state of S ã o Paulo, the coastal municipalities have watersheds in mountains with active relief evolution (Serra do Mar). The coastal regions are more vulnerable to ﬂooding and landslides. A large number of people live on the slopes of Serra do Mar; these places are more vulnerable to landslides, which cause biodiversity loss and damage to human and natural environments. This study seeks to present an ecohydrological analysis to categorize coastal watersheds into clusters, considering the spatial characteristics of NDVI, DTM, soil depth, climate, and Soil Organic Carbon (SOC)


Introduction
The intersection of ecological and hydrological processes is the subject of ecohydrology research [1,2].Ecohydrology takes a systematic approach since there is a relationship between various components that leads to an integrated conclusion after taking into account methodological considerations that arise from the two scientific domains (hydrology and ecology) [3].The analysis of the effects of hydrological processes on the composition and operation of ecosystems is a multidisciplinary field.
The development of quantitative models to evaluate environmental conditions provides the basis for the development of ecohydrological researches.The models investigate how biological and hydrological processes interact at the watershed level [4][5][6].However, it is possible to assume that geomorphological and hydrological processes are directly connected, and that any suggestion for an ecohydrological model must take this into account [1,2,4].
The Band et al. [4] integrates ecohydrological and geomorphological processes, modelling the relationship between water, carbon, and nutrient in their interaction with the stability of rough terrains, helping to weight the levels of stability of different parts of a watershed.
The importance of proposing an ecohydrological model aimed at mountainous areas has two main objectives: (1) to understand the influence of vegetation on the stabilization of slopes and its importance in carbon sequestration; (2) understand that watershed in mountainous areas provide a diversity of ecosystem services on a regional and global scale, including quality and quantity water, regulation of the hydrological cycle and regional climate.In context of these two goals, it is essential to understand the structure and spatial distribution of the vegetation cover when analyzing how it affects landslide processes, and the territorial unit watershed is the best territorial for this [4].
A model for the State of São Paulo coast was proposed based on bibliographic literature on the issue of ecohydrology (Figure 1 and Table 1), and it uses geographic data analyzed and in a Geographic Information System (GIS) to identify areas most susceptible to landslides in the watershed.

Materials and Methods
There were four stages in the development of the ecohydrological analysis for the coast of the State of São Paulo (figure 2): The steps of research were developed in the SIG ArcGIS®, version 10.5.In the rescheduling process, performed on the Rescale by function tool, all variables were classified on a scale from 1 to 10.Then, the ecohydrological analysis considered the sum of the variables: -Normalized Difference Vegetation Index (NDVI): using LANDSAT image to 2016 available in United States Geological Survey (USGS).The NDVI is used in ecohydrological assessments in the [4][5][6] research.Because the NDVI is a satellite with a medium spatial resolution and is appropriate for research on a regional scale, we used it in our study [7][8][9].The images have been processed in ArcGIS Raster Calculator tool were used to generate NDVI images; -Digital Terrain Model (DTM): topographic maps from Brazilian Institute of Geography and Statistics (IBGE, portuguese acronym) and the Topo to Raster interpolation tool were used in ArcGIS [10]; -Soil depth: data obtained from the [11].In the present study, the following weights were assigned to each class (table 2): The soil depth is related to the slope of the terrain, an important variable in the identification of landslide susceptibility areas.We were attributed weight 10 to shallow soils that occur on slopes above 30°, predominant in the mountains (Serra do Mar) and weight 3.5 and 1 to deep and very deep soils with slopes below 15°, predominant in the coastal flat; -Climate data: data from average of Total Annual Precipitation from 2006 to 2015 (source: National Water Agency -ANA, portuguese acronym) and average of Temperature from 2006 to 2015 (source: National Institute for Space Research -INPE, portuguese acronym) interpolated using the IDW tool; -Soil Organic Carbon (SOC): We used the data available by [12].The concentration of organic matter in the soil, either by fungi or roots hyphae, increases the degree of cohesion of soil particles, which strongly depends on the mechanical behavior of its aggregates [13,14].The next step was the weighted sum based on the SOC mapping, from 0 to 5 cm, 5 to 15 cm, 15 to 30 cm using the Weighted Sum tool.The following classes of models were produced as a result of dividing this phase into equal weight assignment (value 1) for each variable entered and the sum of variables: very low, low, medium, high, and very high.
The statistical study of watersheds was the next phase, which determined the relationship between the variables SOC 0-5 cm, SOC 5-15 cm, SOC 15-30, DTM, NDVI, precipitation, temperature, soil depth, and the occurrence of landslides in watersheds.After was calculated the cluster analysis, using the packages ggcorplot and factoextra in RStudio.

Results and Discussion
Figure 3 shows the association between the variables under study: Between the three SOC levels and between soil depth and SOC 0-5 and 5-15 cm, there is a correlation that is more than 0.75.Between the three levels of SOC, DTM, and NDVIsoil depth, SOC 15-30 cm (0.69), and NDVI (0.59)-correlations between 0.5 and 0.75 are seen.Among the variables considered, these had the highest positive correlation.
The watershed's input factors were evaluated, and it was found that 87.25% of the watershed had NDVI concentration values above 0.7, which indicated a high density of biomass and vegetative cover.
Regarding elevations, the coastal plain and the mountain region were separated by 30 m in elevation [15].93.13 percent of the watershed located in the mountains, with a high slope and landslide risk, and 6.86 percent of the sub-basins are located in the coastal plain.
About soil depth, 63.72 percent of the watershed's soils are shallow or shallower, and they are spatially concentrated in mountainous places with sloping terrain and a history of landslides.These soils are more common along the North Coast and in the Baixada Santista communities of Cubatão and São Vicente.
The range of the average temperature is 15.18 °C to 45.72 °C.The watershed in the south coast has the greatest temperatures, whereas the north coast and the Baixada Santista have milder temperatures.Lower temperatures are associated with watersheds that are greater susceptible to landslides.The distribution of precipitation included in this analysis ranged from 1193.71 mm to 3421.11 mm, with a concentration of volumes in the watersheds of Cubatão, São Vicente, Santos, and Guarujá in Baixada Santista, as well as in the north of the municipality from Ubatuba.
The SOC 0 to 5 cm has an average variation of 22.62 to 24.80 g/kg, with the Serra do Mar de Cubatão and Ilhabela with the highest concentrations of values.The concentration varies from 11.62 to 56.58 g/kg between 5 and 15 cm, with high values being found in the Cubatão and Ilhabela.Between 2.03 and 38.76 g/kg of carbon are in the soil at depths of 15 to 30 cm, with greater concentrations observed in Baixada Santista and the north coast.
The areas with a high or very high susceptibility of landslides occurring correspond to 24.06% of the total area of the São Paulo Coast for the 0 to 5 cm scenario, 27.04% for the 5 to 15 cm scenario, and 29.47% for the 15 to 30 cm scenario in the scenarios of the ecohydrological analysis (figure 4 and table 3).The distribution of watershed with the occurrence of landslide is: in the south coastal located in the municipalities of Peruíbe and Iguape, with 30 records of landslides; Baixada Santista, in the municipalities of Santos, São Vicente, Cubatão and Praia Grande with 82 records, this high number is related to a large number of occupations in risk areas; on the north coast, Caraguatatuba, Ubatuba and Ilhabela have 20 records.
In the scenarios of the ecohydrological, the north coast concentrates the largest area of occurrence of landslides, with 50%.The Baixada Santista has 42.43% of its area with the susceptibility of landslides.On the south coast, an area of 5.14% is bordered by Baixada Santista in the municipalities of Peruíbe and Iguape.The other 2.43% are located in other municipalities on the south coast.
After elaborating on the scenarios for the ecohydrological analysis, we performed a cluster analysis of the average values of the variables by watersheds (figure 5, 6 and table 2).Source: Organized by the authors.* Tw = Total watersheds; wLr = watershed with landslides.
The clusters 1 and 4 have the lowest altitudes, with an average of 88.53 m (cluster 1) and 49.45 m (cluster 4).This variable is an indicative that the watersheds are located in the coastal plain, isolated hills or low slope of Serra do Mar.What can be confirmed with the lowest SOC values, with values between 22.85 -23.84 g/kg in cluster 1 and 19.39 -20.2 g/kg in cluster 4. In addition to the concentration of very deep soils > 200 cm, default depth in coastal plain areas.Cluster 1 registers the second highest occurrence of landslides, this data is related to the occurrence of occupation on the low slope of Serra do Mar, concentrating neighborhoods with high population density, while cluster 4 watersheds in areas without human occupation.
The cluster 2 with 143 watersheds has the highest number of landslides recorded in 46 watersheds.This cluster also has the highest average SOC, between 29 and 30.7 g/kg, in addition to the highest average altitude, being 358.56 m in all watersheds of the cluster and 418.32 m in the wLr basins.The average NDVI was 0.85 in all watersheds of the cluster and 0.84 in the wLr watersheds.The climatic data show that this cluster is not the one with the highest volume of precipitation, registering 1654.50 mm in all watersheds and 1562.00 mm in the wLr watersheds.At the temperature cluster 2 recorded an average of 23.17 °C in all watersheds, and in the wLr basins an average of 22.95 °C.In terms of climate variables, cluster 2 is very similar to cluster 4. The soil depth, cluster 2 registers a value of 8.2 in all basins (including wLr), this value demonstrates that shallow soils of 50 to 100 cm, advance to the occurrence of landslides.
The cluster 3 has the second highest average SOC, with a minimum value of 25.4 g/kg in all watersheds and a maximum of 26.5 g/kg in wLr, with carbon concentrated in the first 5 cm of the soil (SOC 0-5 cm) in this class.Altitude values are 160.64 m in all watersheds and 218.60 m in watersheds with landslides.The NDVI values are high, 0.83, as well as the depth of the shallow soils of 50 to 100 cm, similar to cluster 2. As for the climatic data, cluster 3 registers the lowest values, with average precipitation of 1414.24mm and an average temperature of 22.9 °C.

Conclusions
As measured by the NDVI, the dense vegetation maintained constant high values for the research region since the examined area is primarily made up of Atlantic Forest, which is insufficient to reduce the incidence of landslides.
The examined results show that SOC concentration rises as depth increases.Shallow soils are more prevalent in regions where landslides are very likely to occur.Landslides in the watershed are also more common in higher elevations and on slopes.103 basins out of a total of 719 register at least one landslide, or 14.32%.
The cluster analysis classified the watersheds most susceptible to landslides as cluster 2. The concentration of the highest SOC values in these watersheds suggests that landslides may have a role in the atmospheric release of carbon dioxide.They also concentrate the highest NDVI values; however, clusters 1 and 3 also have high NDVI values due to the study area's predominant atlantic forest vegetation, while cluster 4 has lower NDVI values due to its predominance of watersheds in the coastal plain, which is where the urban areas, exposed soil, and sparse vegetation are concentrated.The indicator that shows how often landslides occur on the steep slopes of Serra do Mar, where relatively shallow soils prevail, is altitude above 350 m (values of 8.2).

Citation:
Folharini, S. O.; Vieira, A.; Oliveira, R. C. Ecohydrological analysis in susceptibility areas the landslides: study municipalities to São Paulo State coastal, Brazil.The 4th International Electronic Conference on Geosciences 2022, 69, x. https://doi.org/10.3390/xxxxxAcademic Editor: First name Lastname Published: date Publisher's Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Figure 4 .
Figure 4. Scenarios of the ecohydrological analysis.Source: Organized by the authors.

Figure 6 .
Figure 6.Clusters dimensions.Source: Organized by the authors.

Table 3 .
Landslide frequency to class of ocorrence susceptibility.