**1. Introduction**

Landslide risk is the possibility that a landslide occurs in a specific area and in a specific period of time, causing damages to population, buildings, infrastructure and services [1,2]. As a consequence, landslide risk is influenced by the overlapping in time and space of hazardous areas (where landslides are likely to occur) and potentially vulnerable exposed elements, resulting in an impact that could cause damages or losses. This has been traditionally translated into mathematical form by the classical equation [1]:

$$
\mathbb{R} = \mathbb{H} \cdot \mathbb{V} \cdot \mathbb{E} \tag{1}
$$

where R is the risk, H is the hazard (the probability for a dangerous event of a given intensity to happen in a certain place and time), V is the vulnerability (the degree of loss expected from the element impacted by the landslide) and E is exposition (the value of the elements exposed to the event).

Following this approach, quantitative risk analyses have been mainly published for small areas or, at the most, in regional scale applications [2–9].

A quantitative landslide risk assessment for very large areas (e.g., an entire nation) is still a very challenging objective, as it requires facing technical and scientific issues such as availability of complete, homogeneous and good quality input data of differ-

Segoni, Definition of Environmental Indicators for a Fast Estimation of Landslide Risk at National Scale. *Land* **2021**, *10*, 621. https:// doi.org/10.3390/land10060621

S.;

Caleca,

F.

**Citation:**

Academic Editors: Enrico Miccadei, Cristiano Carabella, Giorgio Paglia and Paul Aplin

Received: 7 May 2021 Accepted: 8 June 2021 Published: 9 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

ent natures (pertaining at least to the fields of geology, economy, demography and civil engineering) [10].

Italy is no exception to this, and to date the main strategies for assessing landslide risk at the national scale in Italy were based on different strategies. For instance, [11] proposed a statistical analysis on the spatial variability of recorded fatalities, to account for societal landslide risk in the whole Italian territory. Recently, [12] proposed a set of landslide risk indicators based on freely accessible data from an online governmental platform, including exposed population, number of buildings and landslide hazard zones as defined by the Italian regulation. Although representing a very complete overview of landslide risk in Italy, this approach has the drawback of presenting a spatial aggregation at the municipal level and leaving unexploited some scientific products that have a finer spatial resolution, such as landslide susceptibility maps, which have been proposed for several Italian regions [13–17] and for the whole Italian territory [18], or monitoring products of the artificialization of the territory such as soil sealing maps, which monitor the evolution of the processes of artificialization of the territory at high spatial resolution (10 m) at yearly time steps [19]. However, the approach of addressing national scale landslide risk problems with a set of simple indicators, rather than with a full QRA, seem promising and quite consolidated in landslide studies [10,20,21]. Undoubtedly, indicators are, by definition, simple means to describe and comprehend a complex phenomenon and are widely used in environmental studies by scientists and governmental agencies.

The purpose of this manuscript is to propose a new set of environmental indicators to characterize landslide risk over very wide areas and to apply it to characterize the Italian municipalities. The novelty in the proposed approach is to use advanced and highresolution thematic layers: already existing landslide susceptibility maps [18] are used to identify hazardous areas, and soil sealing maps are used as they have a high resolution and constantly updated representation of the spatial distribution of the elements at risk (soil sealing maps are released on a yearly basis to monitor the expansion of urban fabric [19,22]). At the same time, the general objective is keeping the resulting indexes easy to understand, quick to update and flexible enough to be adapted at varying spatial units. In its basic formulation, a spatially distributed Landslide Risk Index (LRI) is defined on a pixel basis at 50 m resolution. Afterwards, we show an application to the whole Italian territory, in which the LRI is aggregated at the municipal level following two different approaches, generating two additional indexes that can be used to gain useful understanding on the interferences between geomorphological slope dynamics and urban expansion, which give birth to landslide risk.

#### **2. Materials and Methods**

## *2.1. Test Site*

The study area considered for this work is the whole Italian territory (301,340 km2) (Figure 1a). Italy is a peninsula located in Southern Europe and extending into the Mediterranean Sea. It is characterized by two main mountain ranges: the Alps, to the north, which separate Italy from the rest of Europe, and the Apennines, forming the backbone of the peninsula and running from NW to SE.

**Figure 1.** (**a**) Overview of Italy; (**b**) administrative subdivision into 7904 municipalities.

The geological setting and morphological features of the Italian peninsula are the result of a still active geological process that led to the formation of the two mountain chains [23]. The Alps are the typical example of a collisional belt: it was generated during the Cretaceous period by the convergence of the Adriatic continental upper plate (Argand's African promontory) and a subducting lower plate including the Mesozoic ocean and the European passive continental margin. In the Eocene, a complete closure of the ocean marked the onset of the Adria/Europe collision. The collisional zone is represented by the Austroalpine-Penninic wedge, a fossil subduction complex, showing that even coherent fragments of light continental crust may be deeply subducted in spite of their natural buoyancy [24]. The Apennines extend from the northwest part of the peninsula to the isle of Sicily, and link the western Alps with the Magrebian chain of North Africa [25]. The Apennines are a NW–SE oriented fold-and-thrust belt formed during the Oligocene period by the closure (started during the Cretaceous period), of the Mesozoic Tethys Ocean and following the collision between the European (Corso-Sardinian block) and African plates [26,27].

From a geomorphological point of view, Italy has a marked energy of relief: mountains are present in every Italian region and occupy more than the 35.2% of the territory. The greatest part of Italy, however, is characterized by hills, representing the 41.6% of the land surface. This juvenile morphological setting, in a still tectonically active territory, brings the consequence that landslide hazard is widespread in every part of Italy, excluding flat alluvial and coastal plains.

Landslide hazard is further exacerbated by climatic and meteorological constraints. Due to the large latitude range covered by Italy, the climate varies largely: from the cold climate of the north, EFH according to Koppen classification, typical of the highest mountain peaks, with annual precipitation higher than 2000 mm, to the Subtropical climate (BS in Koppen classification) of the southernmost coastal areas of Sicily, Apulia, Sardinia and Calabria, with long, hot, dry summers and precipitation less than 400 mm in Sicily [28]. Recently, due to the effects of climate change, periods of precipitation are becoming shorter and more intense in many parts of Italy [29,30], causing an increase in landslide activity and in the number of harmful landslide events per year [31,32].

For a full understanding of the application reported in this study, it is worth noting that Italy is subdivided into 7904 municipalities (Figure 1b), which represent the smallest administrative subdivisions of the territory and that have important responsibilities in territorial planning, urban design and risk management.

#### *2.2. Landslides in Italy: National Inventory and Existing Susceptibility Maps*

For the reasons explained above, each year hundreds to thousands of landslides affect Italy, causing victims and damages to buildings, infrastructure and cultural heritage [12,32–34]. An official landslide database exists at the national scale that is managed by ISPRA (National Institute for Research and Environmental Protection). The database is called IFFI (Italian National Landslide Inventory) and maps all known landslides (both active and inactive), mapped at the 1:10,000 scale by means of field surveys, remote sensing techniques and collection of ancillary data. According to IFFI, 620,808 landslides are present, covering about the 7.9% of the Italian territory. IFFI is openly accessible via an online platform [12] and it is acknowledged to be one of the most complete and homogeneous national-scale inventories in Europe [35–37]. IFFI is widely used as a base for landslide hazard and risk assessments at various scales [12,15,38–40].

In particular, in Italy, an overwhelming literature exists about landslide susceptibility studies. Landslide susceptibility maps (LSMs) represent, over appropriate spatial units, the spatial probability of the occurrence of landslides, and they are usually obtained by a statistical analysis of the spatial distribution of a set of predisposing factors [41]. Although LSMs do not contain temporal predictions, they are usually considered the starting point for landslide hazard and risk assessment. This is also the approach used for this work, but a literature review showed that most of the published LSMs refer to basin-scale studies [42–46]. Some examples of regional-scale susceptibility assessments also are present [13–16,47], but the use of a combination of regional maps obtained with different approaches to compose a nation-wide mosaic of landslide susceptibility would pose huge problems of consistency of the data. To our knowledge, the only LSM at the Italian scale available to be used as input data for this work is the national scale susceptibility assessment performed by [18]. The susceptibility assessment was performed separately for three different landslide typologies (rockfalls, rapid shallow slides, slow deep slides), producing three susceptibility maps at 50 m resolution. A Random Forest algorithm [48], which is a machine learning technique widely consolidated in LSM studies [49–51], was calibrated with the IFFI landslide inventory and a set of environmental variables including lithology, land cover, morphometric parameters (elevation, slope gradient, aspect, curvature), and hydrological parameters (topographic wetness index, stream power index, upslope contributing area). Overall, 196,087 sample points (50% randomly sampled inside landslides and 50% randomly sampled outside the mapped landslides) were used to train the Random Forest model and 84,641 independent points were used to quantify its accuracy in terms of AUC (area under receiver-operator characteristic curve), which is reported as 0.85.

#### *2.3. Soil Sealing in Italy*

In addition to the natural physical features (such as geological and climatic settings), anthropogenic dynamics are also deeply involved in landslide risk in Italy. On one hand, urban elements (such as buildings and infrastructure) may contribute to destabilizing slopes, acting as predisposing factors for landslide hazard. On the other hand, the ongoing expansion of urban fabric and infrastructure generates, at an alarming rate, new elements that are exposed to hazard, determining a relevant degree of landslide risk.

Since 2015, ISPRA has undertaken a nation-wide monitoring program of soil sealing. Soil sealing is the most intense form of artificial land take and it can be defined as the removal or covering of soil by buildings, constructions or other totally or partly impermeable artificial material [52]. Since then, every year, a national cartography of soil sealing is produced by remote sensing techniques and it is released as a raster map (pixel size 10\*10 m), in which the whole Italian territory is classified into two classes: sealed soil/not

sealed soil [53,54]. Sealed soil includes built-up areas, paved areas, railways, airports, ports and even reversible land consumption such as dirt roads [54]. Although all those elements are not distinguished from each other, the information conveyed by the soil sealing maps is very useful for the aim of this study because it provides useful information (updated on a yearly basis) about all anthropic elements exposed to risk, with a relatively very high spatial resolution.

To this regard, it should be stressed that, typically, urban areas in Italy are not clustered and are characterized by a peculiar diffuse pattern (referred to as "sprawl" and "sprinkling") [19]. As a consequence, other land cover/land use monitoring products (such as Corine Land Cover) are not able to adequately capture the spatial and temporal evolution of this phenomenon [54]. Moreover, the remote sensing techniques developed by ISPRA are specifically conceived and calibrated to detect the diffuse and scattered patterns of Italian urban fabric [19].
