Figure 1. Standardized criteria for HAND Model values, altitude, slope, and proximity to rivers and lakes, and Normalized Difference Built-up Index (NDBI) for the Municipality of Sorocaba, State of São Paulo, Brazil.Source: prepared by the author.
This map aimed to identify priority areas for increasing green infrastructure to mitigate flooding in urban landscapes. The model was based on multicriteria evaluation and literature review.
The map production for Sorocaba was approached using a set of contributing factors arranged into two groups: susceptibility and impact. Using the Weighted Linear Combination, we combined five (5) criteria, four as susceptibility and one as impact (Figure 1).
The susceptibility was based on HAND model values (Nobre et al., 2016), altitude, slope, and proximity to rivers and lakes. The impact was the Normalized Difference Built-up Index (NDBI). Susceptibility factors are related to the city's geomorphology, indicating its proneness to floods. The impact category included aspects concerning the presence of impervious surfaces (made of concrete and asphalt), once the location of this urban infrastructure can influence flood processes (Du et al., 2015).
Although the quality of the data sources varied depending on their nature, they were all resampled to obtain the final possible resolution and allow for accurate aggregation. Thus, these criteria were standardized to 0 - 1 using the linear fuzzy membership function (Figure 1) (Malczewski et al., 2003).
We assume that all analyzed criteria are equally crucial for the model and have implemented the process of neutralizing their influence. For this, the steps involved finding the weighting values (Factor Weights) for each criterion to neutralize their influence. Thus, we apply different Factor Weights to the criteria so that the final weight of each criterion will depend on the number of criteria present in each analysis and the level of the relative influence of each spatially explicit criterion (Floridi et al., 2011). For example, the Factor Weight for the slope criterion was 0.111903596, while for the Normalized Difference Built-up Index (NDBI), it was 0.216862843 for this model, with five (5) criteria under analysis (Table 5.2).
For the validation, we used the ground truth points of floods in 2022 and 2023, which were kindly provided by the Urban Security Secretariat - Municipal Coordination of Protection and Civil Defense of Sorocaba. This record primarily focused on the Summer Plan period, which covers December to March. The data was provided on August 10, 2023, as documented in Administrative Process No. 18468/2023.
The results were organized on a continuous scale where the high-priority areas (nearly 1) were identified as the most suitable for implementing new green infrastructure.
We used the participatory technique involved consulting the Sorocaba population using an electronic form approved by the UFSCar Ethics Committee (CAAE: 53408421.9.0000.5504) to improve the research about green infrastructure spatial planning model maps focusing on flood mitigation (Access the form here: https://zcwp9yd4wxh.typeform.com/to/FaPxmMJm). In this form, we consulted the citizens about flood occurrences and their consequences in their lives. The methodology used here is "citizen science" because it involves collaboration between the population and scientists to map flooding points based on reports from those who experience the problems (Bonney et al., 2016).
Fig. 2. Priority areas for increasing green infrastructure using five (5) criteria and assessed through Equal Factor Weights (FW) and Different Factor Weights (FW) in the Municipality of Sorocaba, State of São Paulo, Brazil.
Results:
The model focusing on flood mitigation identified potential flood-prone areas characterized by low elevation, proximity to rivers, and low permeability. The Normalized Difference Built-up Inde (NDBI) data brought greater precision and sensitivity to the model, for example, not identifying riparian areas as potential areas for green infrastructure where vegetation or anthropogenic fields already exist (Figure 2).