The spatial knowledge of local conditions obtained using advanced digital technologies and modelling scenarios is one of the most debated and discussed approaches to preparing and adapting urban systems to deal with future challenges [1
]. Among others, climate change will be one of the greatest threats for urban areas [3
], since the quantity, duration, and intensity of heavy rains, urban heat islands, storms, wind, and other climate-related extreme events will increase the exposure of both natural and human-made systems, with a greater impact in urban areas due to the population density [4
Despite the technological and social predictive capacity enabling us to understand where and how new phenomena will occur, it is certain that the traditional timescale, occurrence, and intensity of extreme events have changed and will continue to do so rapidly in the next few decades, according to the most recent studies [6
Therefore, the failure of the traditional determination of hazard areas requires analysis to focus on the vulnerability concept instead of risk [3
]. Considering that all cities can be potentially affected by extreme events, the vulnerability assessment includes an entire catchment, and not only how and where the catchment is prone to be impacted by certain circumstances (with a certain degree of intensity and with a predicted return time), as assumed by the traditional risk diagnosis.
To this matter, international literature demonstrates how the changing dynamics of extreme natural events and their impacts on environments and humans affect the health, the security, and the economy of cities [8
]. Among other kinds of hazards, floods can induce health losses, deaths, accidents, mental health issues, diseases, vector-borne illnesses, and waterborne diseases [9
], therefore flood susceptibility models are nowadays urgently demanded [10
This research deals with cloudburst events, as introduced by Rosenzweig et al. [12
], from the perspective of intrinsic vulnerability, as the propensity of an urban system to be flooded during a cloudburst event (e.g., the rainfall event that occurred in Copenhagen on 2 July 2011, where, in some parts of the city, over 100 mm of rain fell in one hour [13
As stated by Rosenzweig et al. [12
], there are three basic categories of knowledge to support cloudburst resilience: (1) knowledge of the current weather and future climate conditions that determine a cloudburst hazard; (2) understanding of the vulnerability of urban social, ecological, and infrastructure systems, and (3) knowledge of potential strategies for cloudburst management [12
This research investigates the second category, employing the model called “Urban Flood Risk Mitigation” of InVEST software (Integrated Valuation of Ecosystem Services and Tradeoffs, version 3.8.5) [14
], an open-source modelling suite jointly developed by Stanford University, the University of Minnesota, the Nature Conservancy, and the World Wildlife Fund. This specific Ecosystem Service (ES) model was presented and taught during the course “Introduction to the Natural Capital Approach and Urban InVEST”, organized by Stanford University during the 10th World Ecosystem Service Partnership Conference in Hanover (21 November 2019). The ES modelling output has been used to evaluate the distribution of biophysical values of run-off retention in the metropolitan catchment of Turin (Italy). According to other previous studies [15
], the vulnerability of urban areas to a cloudburst can be measured with the run-off assessment during the present predicted scenario, assuming the model returns a picture of the site-specific vulnerability in the analyzed urban catchment and that it is possible to run alternative scenarios to verify how specific nature-based or technological solutions help to improve retention performance [17
]. It is worth mentioning that flooding by cloudburst has a different dynamic from riverine or fluvial flooding. While fluvial flooding is due to rivers exceeding their capacity, cloudburst flooding (or pluvial flooding) is due to extreme precipitations. It does not depend on river dynamics, but it can represent a concurrent cause of fluvial flooding, contributing to a rapid overcharge of the water streams during a cloudburst event. This means that run-off is a key parameter that determines how vulnerable the entire system is (not only riverine areas) and where the system should be transformed and adapted to reduce run-off (not only along with river-beds). This assumption clarifies that the flooding vulnerability assessment in urban areas has its peculiarity, since it is strongly conditioned by the morphological development of buildings and their distribution in public and private spaces [11
]. The changes of natural streams during urbanization by substitution with anthropic channels limit the capacity of the system to convey floodwaters [21
]. Roads and buildings increase the flood hazards by affecting the natural hydrological cycle and making the estimation of the water discharge during a rain event much more difficult due to the complexity of the urban environment and the numerous variables that should be considered [22
]. A detailed evaluation of how the morphological characters of the urban environment affects the run-off volume can be found in several international publications [23
]. As stated by Brunetta and Salata [15
], the vulnerability assessment is the first step towards adaptation, considered as the measure of the vulnerable biophysical characters used to build a detailed spatial knowledge [26
], and learning and understanding the system dynamics [29
]. Even the slippery concept of resilience [30
] deals with adaptation toward vulnerability reduction, broadening the “physical” meaning to incorporate socio-economic structures and governance, while including pro-active and co-evolutive feedback encompassing preparedness with knowledge and physical adaptation [31
]. In practical terms, the resilience concept is operationalized at its base throughout the preliminary vulnerability measurement and identifies solutions to cope with emerging dynamics.
The research assumes the definition of the areas vulnerable to cloudburst events as the first step towards the “resilience of the system”, according to the vision of “co-evolutionary resilience” [34
]. Through this vision, it is possible to identify the conditions for resilient urban planning: the preservation of the system’s constitutional framework and the ability to evolve, thus aiding the decision-making processes [35
] while identifying problems (e.g., vulnerability analysis) and determining the kind and quantity of solutions through plans and projects.
In the famous cases of Phoenix and New York City (USA) or Copenhagen (Denmark), the utilization of run-off modelling through detailed spatial maps has been virtuously employed in water management plans and projects while adapting the city to cloudbursts.
This study aims to apply a similar methodology in an Italian context, while modelling with Geographic Information System (GIS)-based software the spatial distribution of the run-off, and to understand where and how nature-based solutions can contribute to adapt the system towards a greener and healthier environment.
The model and related approach presented in the paper is tested in the metropolitan catchments of the city of Turin (IT), an Italian city of medium size that has undergone rapid growth during the 20th Century. Mainly, Turin is intrinsically vulnerable to every rain event since its urban land is characterized by a high rate of imperviousness and a compact, densely built-up settlement system [36
The manuscript is structured as follows: first is the methodology, presenting the catchment area and its distinctive characteristics; next, the functioning and input data required for the Urban Flood Mitigation model are described; and finally, the results describing the numerical/graphical outputs of the model, the discussion summarizing the evidence derived by the run-off retention service, and the conclusions, briefly presenting all relevant matters and proposing innovative and potential applications for future research, are presented. The novelty of this work consists in applying a common methodological framework for defining the area subjected to flooding, the intensity of the phenomenon, and the most suitable solutions that could substantially reduce the vulnerability, by using an open-access GIS experimental ES model. The use of a case study aims to highlight the possible planning application and how this kind of approach needs to be considered for setting future urban strategies.
2. Materials and Methods
2.1. The Turin Metropolitan Area
The catchment area includes 21 municipalities that border the city of Turin (Figure 1
), defined as the “first ring” due to their strong interaction with the main city, namely, Baldissero Torinese, Beinasco, Borgaro Torinese, Caselle Torinese, Castiglione Torinese, Collegno, Druento, Gassino Torinese, Grugliasco, Leinì, Mappano, Moncalieri, Nichelino, Orbassano, Pecetto Torinese, Pianezza, Pino Torinese, San Mauro Torinese, San Raffaele Cimena, Settimo Torinese, Torino, Venaria Reale, and Volpiano. Despite the high administrative fragmentation, the city of Turin consists physically of a unique semi-dense continuous built-up system that from the core area shapes the surrounding zones while comprising a heterogeneous land use [37
]. The average daily temperature varies from below 1.4 °C in January to 23.6 °C recorded in July, the climate is warm and temperate with significant rainfall throughout the year (about 864 mm of annual precipitation), and the driest month is January, with 38 mm, while May is the wettest month, with an average of 108 mm (https://en.climate-data.org/
accessed on 16 September 2020).
The catchment area (Figure 1
) covers 56,777 hectares, with an altitude that varies between 180 and 715 m above sea level (Turin is located at 248 m a.s.l.). About 36% of the area is hilly land (mean altitude ranging between 100 and 600 m a.s.l.), while the rest is lowland areas (mean altitude below 100 m a.s.l.), according to the regional topographic database.
The area consists of a varied heterogeneous landscape, from the suburban eastern hills with a semi-dense built-up zone to the dense and continuous built-up area of the city center and the lowland river areas of Sangone, Dora, and Stura that shape the orography of this flood-prone catchment area.
According to the Land Cover dataset (2017) of the National Superior Institute for Environmental Protection and Research (ISPRA), the land use of Turin consists of 35.41% artificial surface, 0.28% unvegetated natural areas, 23.44% trees, 0.14% shrubs, 39.86% agricultural land, and 0.87% water bodies and wetlands.
The extent of artificial surfaces with impermeable cover is relatively high in Turin, highlighting the peak of 100% of sealed surfaces (e.g., the productive sites or the densely built-up residential areas) in some urban districts. The combination of floodplain soils and high sealed surfaces makes this territory prone to flooding during any rainfall events: both the prolonged and less intense events that mostly cause riverine flooding, and the single and unpredictable cloudburst event.
The most recent event was on 17 August 2020, when more than 70 mm of rainfall fell in less than one hour, flooding the city at several sites while creating damage to infrastructures and a sewage system black-out causing dangerous situations (Figure 2
). Two metro stations were partially inundated, while for two consecutive days, the streets, canals, and sewage system were temporarily inoperative due to the high quantity of debris that flowed into the city, especially along the Po River and the hill (see local-newspaper-reported photos and videos, e.g., TorinoToday https://www.torinotoday.it/meteo/maltempo-temporale-strade-fiumi-17-agosto.html
accessed on 3 October 2020).
Despite the evidence of the problem, there is no strategic plan or program that addresses this threat by proposing solutions that can solve it or limit the damage. The approach adopted is still focused on identifying risky areas by the interaction between settlements and riverine-flood-prone areas. At the same time, there is no overall strategy that involves the city in all its complexity selecting the most appropriate solutions to reduce its vulnerability.
2.2. Model Input
The Urban Flood Mitigation model has been recently added to the software InVEST (version 3.8.5) [14
], released by the Natural Capital Project. This model represents one of the first tools designed explicitly for mapping urban ES, since this suite of tools was originally designed to map ES while not considering the influence of the built-up infrastructures of the city on ES provision.
The model aims to verify the capacity of urban catchments to limit the run-off process, avoiding potential urban flooding by a cloudburst. It assumes that flood-prone areas are those where a coupled interaction between the kind of permeable/impermeable materials of artificial surfaces (e.g., urban areas) and the quantity of soil drainage (depending on water conductivity) generates the amount of water that accumulates by surface run-off during a cloudburst, resulting in temporary flooding. The model simplifies the process of run-off discharge by considering that the water on the impervious surface moves directly to the area next to it, contributing directly to a surface flow accumulation. Nevertheless, in dense urban catchments, the building roofs, terraces, and other horizontal or vertical surfaces that capture the rainwater do not directly contribute to the soil run-off. Instead, the water is temporarily withheld in various paths (e.g., water pipes of the rainwater sewer system) and contributes to the total discharge sometime later. Specific literature on this dynamic demonstrates how the biophysical quantification of the run-off in the built environment exposed to torrential hazards can be significantly difficult to estimate [20
], since the quantity, quality, and surface of buildings, the sewer systems, the presence of dust in the ground, and the dryness of the soil can affect the discharge volume during cloudburst [11
Therefore, the results of this study should be evaluated bearing in mind that the InVEST model uses an empirical simplification by calculating the discharge volume in urban areas using a plain bidimensional model based on the run-off curve number. The run-off curve number is a parameter that assumes that where there is a highly sealed surface and the soil has low conductivity, the run-off process will be the highest (e.g., high run-off means that the exact amount of rainfall is not retained by soil and flows in another part of the city). On the other hand, when there are fewer sealed surfaces and soils with high conductivity, the potential retention is more elevated.
This model considers the potential of porous green areas to reduce the run-off process by absorbing water, slowing surface flows, and creating space for water collection (in floodplains or basins). The output calculates the run-off reduction (i.e., the amount of run-off retained per pixel compared to the storm volume) and, for each watershed, gives a potential estimate of the economic damage by overlaying information on flood extent potential and built infrastructure (see Figure 3
). In this case, the model has been employed without economic evaluation due to limited data and knowledge necessary for running the model correctly.
The inputs required are:
Watershed Vector delineating areas of interest;
Depth of rainfall in mm (of a single cloudburst event);
Land Cover Map;
Soils Hydrological Group Raster;
The biophysical value corresponding to each of the land use classes in the Land Cover Map.
Since run-off retention depends on the interaction between soil and land cover, the incorporation of the following two databases is crucial:
Data on land use (or land cover map) used to set the run-off curve number (RCN). RCN represents the hydrologic soil capacity and indicates the run-off potential: the higher the RCN, the higher the run-off potential [39
Data on hydraulic conductivity was used to establish soil drainage properties to verify the quantitative measure of a saturated soil’s ability to transmit water when subjected to a hydraulic gradient [40
A Table with RCN A, B, C, and D was associated with the classes (Table 1
) on the permeability of urban soil. Therefore, the land use classification was entirely built around the USDA classes employing the regional topographic geodatabase (BDTRE) available at https://www.geoportale.piemonte.it/cms/bdtre/bdtre-2
(accessed on 10 June 2020). The data on urban green areas were extracted with a thematic overlap between BDTRE and the Land Cover Piedmont (2010), which classifies the urban areas according to the Corine Land Cover thematic legend (available at https://land.copernicus.eu/pan-european/corine-land-cover
, accessed on 10 June 2020). The final layer allows the urban districts, infrastructure system, open spaces, and buildings to be identified.
We employed two additional datasets to rank these layers in permeability/condition classes:
The National High-Resolution Land Consumption map (NHRLC) is a raster database of 10 m pixel size produced yearly (by the Italian Institute for Environmental Protection and Research) since 2015, covering the entire Italian territory and monitoring activities of institutional land take. This dataset was obtained using semi-automatic classification techniques from high-resolution satellite images to detect sealed and artificial areas and is downloadable as open access data from the SINANET portal (http://groupware.sinanet.isprambiente.it/uso-copertura-e-consumo-di-suolo/library/consumo-di-suolo
, accessed on 3 June 2020);
The NHRLC was used to calculate the permeability of urban districts, while NDVI was used to define the “condition” (poor, fair, and good) of urban open spaces according to USDA classification of RCN. For both datasets, the procedure was the same: the Land Use layer was intersected with the NHRLC or NDVI dataset. Then, the average values of imperviousness and vegetation density for each land use class were calculated. Agricultural land, natural land, and water bodies were assigned the average values found from USDA parameters.
The final biophysical table is shown below (Table 1
) and organized into 17 Land Use classes characterized by a particular run-off curve parameter for each soil group.
The second database represents the saturated hydraulic conductivity (Ksat, mm/h, Table 2
), which is defined as the soil’s ability to be vertically drained by liquids in a saturated condition. This means that soil with good conductivity allows a significant quantity of water absorption and movement in a short period (high conductivity). With a high conductivity, the water quickly reaches the aquifer, while the surface flow processes’ result is limited. Therefore, conductive soils are protective against surface erosion and ensure a better quality of surface water. On the contrary, permeable soils are not protective of groundwater because of the high conductivity. The situation is the opposite in the presence of conductivity characterized by low infiltration and high processes of surface run-off [41
Lastly, hydraulic conductivity is a function of soil porosity: the water’s movement is facilitated when pores are large and continuous, while it is more difficult when the pores are small and disconnected. The soil’s porosity is connected with soil texture: clay soils generally have a lower saturated hydraulic conductivity than the sandy and gravel soils, where the pores, less numerous but larger, facilitate the passage of significant volumes of water [15
To map soil conductivity in the catchment, we used the digital “Map of soil protection capacity” (available online at https://www.regione.piemonte.it/web/temi/agricoltura/agroambiente-meteo-suoli
, accessed on 3 June 2020). The map represents the soil’s protective capacity against the infiltration of pollutants, taking the conductivity (“Fk” values) as a proxy for protection (high conductivity corresponds to low protection and vice versa). The map is grouped into eight classes, from high to low protection.
We made a reclassification into four classes, considering the low protective capacity as the first class (Group A), assuming that all those soils that did not protect groundwater (with high Fk values) have high conductivity:
Group A (Fk 7 and Fk 8);
Group B (Fk 5 and Fk 6);
Group C (Fk 3 and Fk 4);
Group D (Fk 1 and Fk 2).
The result is presented in Figure 4
, where it can be noted how the dense built-up system of the city of Turin stands on a Group B soil, which is a good one (medium-high conductivity) while the hill (eastern part) is rooted in Groups C and D, resulting in its being extremely vulnerable to rainfall events.
The model was run considering a single rain event of 50 mm, which has been established as a threshold of cloudburst, according to Rosenzweig et al. [12
This research developed and tested the recent Flooding Risk Mitigation model of InVEST while contributing to estimating the potential biophysical effect of a cloudburst event in the metropolitan area of Turin. Flooding vulnerabilities measured in terms of the integration of surface run-off have been spatially mapped. Vulnerability has also been evaluated with the infiltration capacity, which has been spatially mapped by integrating data on land use and soil hydrological conductivity. The biophysical assessment has been analytically compared in the entire catchment, while an in-depth evaluation of the potential adaptation and planning solutions was conducted in Turin.
Results demonstrate that during a cloudburst event such as the one that happened on 17 August 2020, where more than 6.5 million cubic meters of rain fell in less than 1 h, the total volume of the run-off discharge is 3.4 million cu.m., with a measured relative run-off per hectare ranges from 230 to 430 cu.m. per hectare. Therefore, 52% of the rainfall volume (more than half) cannot be retained by the soil, thus creating a pluvial surface stream, generating severe human and public safety problems.
Run-off outputs were then· classified and superimposed with the land use plan to see where and how the adoption of site-specific NBS can reduce the vulnerability of the city. Finally, a simulation of the biophysical effects of NBS was performed to see how the integration of specific adaptation measures against cloudburst events can change the initial run-off volume. Results demonstrate how performance-based planning can aid decision-making with tangible and measurable effect at the city-scale level.
The research shows how a methodology for ES mapping could be functional to set NBS while operationalizing the ES concept for planning purposes. Moreover, the approach used in the paper can make planning decisions oriented towards the performance-based design of urban open spaces (both public and privates), for example, ensuring permeability by using porous pavements, or increasing the quality of vegetations by adding new urban trees and forestry areas, which could enhance the flooding process intercepting and abating run-off during extreme weather events.
Ensuring effective and site-specific NBS is an urgent need to include ES in the planning process, using them to define and set urban design parameters to make cities more resilient to the effects of climate changes. The study was able to provide valid knowledge support for planners and decision-makers in setting NBS using an ES-based approach.