The Lima metropolitan area faces very frequent critical problems regarding the illegal occupation of unused land and land trafficking. The most recent ones occurred in April 2021, when two large areas on the outskirts of Lima were informally occupied by citizens (Fig. 1). The people who participated declared to the local news that they were affected by the pandemic and could not afford their rents any longer. The Morro Solar, located in the Chorrillos district, was first occupied on 5 April 2021. Then, the Lomo de Corvina sector, located in the Villa El Salvador district, was occupied on 12 April 2021. According to the local news, about 5000 and 3000 people occupied the Morro Solar and Lomo de Corvina, respectively. During this sudden occupation, lots were delimited with chalk, and makeshift shelters were built with wooden sticks and plastic sheets (see Fig. 1e and f). It is worth mentioning that the government declared the Morro Solar an intangible area in 1977 and a historical monument in 1986. The police, in coordination with the local authorities, removed the inhabitants that participated in the invasions on 14 and 28 April in the Morro Solar and Lomo de Corvina, respectively, and the makeshift shelters were removed several days later.

The main concern this study addresses is that the occupied areas are unsafe against natural hazards. Luque and Núñes (2011) reported a geological evaluation of Inty Llacta, an informal settlement located close to the recently invaded area in the Morro Solar. The report referred to identified rockfalls and slope failure hazards, representing a severe danger to the inhabitants and buildings. Likewise, Medina et al. (2013) reported a technical evaluation of 29 informal settlements located in the Lomo de Corvina sector and pointed out that potential slope failure may occur in the areas referred to. It was also mentioned that the western part of the Lomo de Corvina, where the recent invasion was located, is the most critical area and is not suitable for urban development.

Figure 1Location of the study area. (a) Location of metropolitan Lima (blue rectangle) within Peru. (b) Location of the study area (blue rectangle) within metropolitan Lima, which is partly located in the districts of Chorrillos and Villa el Salvador. (c) Seismic microzonation of the study area. The green rectangle denotes the location of inset (i), the Morro Solar. The blue rectangle denotes the location of the inset (ii), the Lomo de Corvina. Basemap of insets: MAXAR imagery. (d) Color composite of synthetic aperture radar (SAR) backscattering intensity images. Red band: image recorded on 14 April 2021; green and blue band: image recorded on 3 December 2020. (e) Photograph of the squatter settlement in the Morro Solar recorded by Gestion (2021). (f) Photograph of the squatter settlement in Lomo de Corvina (modified from Inga, 2021a).

Regarding seismic hazards, seismic microzonation studies have been performed in metropolitan Lima, which included the implementation of field tests in the city's urban areas. Within the framework of these studies, the underlying soil deposits that conform to the capital city were classified into different zones depending on their mechanic and dynamic behavior. Thus, Zone I corresponds to the stiffest soil of the city and Zones IV and V to the most inadequate regions for being populated due to their particularly bad soil conditions. In the case of the Chorrillos district, it was divided into four zones, and the areas in the Morro Solar are categorized as Zone IV mainly because of the steep slopes that might fail under the effect of strong motions (CISMID, 2010). Furthermore, Lomo de Corvina is also classified as Zone IV due to the existence of deep eolian sand deposits and the largest fundamental vibration period found in Lima, which is slightly larger than 1 s (CISMID, 2011). Recent studies have shown that Lomo de Corvina might evidence important values of amplification factors due to the generation and interference of surface waves along the slope (Gonzales et al., 2019). It is worth mentioning that the collapse of the light makeshifts built during the recent invasions may not have represented an effective danger condition to the inhabitants. However, non-engineering masonry houses could have been constructed in the short term if the inhabitants were not removed. Note, however, that the sole action of removing shelters will exacerbate the need for residence.

Figure 2(a) Simplified scheme of the elevation profile at the line $\stackrel{\mathrm{‾}}{\mathrm{AE}}$ in Fig. 1c, inset (ii). (b) Diagram of the radar pulse, I_a, backscattered from the ground. (c) Diagram of the radar pulses backscattered from the ground, I_a, and the makeshift shelter, I_b. (d) Red lines: SAR backscattering intensity time series at some representative points of the recently invaded land in the Morro Solar. The sudden increment from 14 April denotes the change in the backscatter mechanism from the mechanism shown in (b) to that shown in (c). Blue lines: SAR backscattering intensity time series at some points of an old informal settlement located close to the recently invaded land.

Download

Figure 3Sentinel-1 SAR-based map of the recent informal settlement in the Morro Solar (a) and Lomo de Corvina (b). Basemaps: MAXAR imagery.

This section shows how the informal settlements were observed from satellite imagery. Unfortunately, the areas of interest were cloud-covered on most of the days after the invasion. From the beginning of April 2021 until 5 May, only one product from the Sentinel-2 constellation (Copernicus, 2021) recorded on 20 April was cloud-free. On the other hand, synthetic aperture radar (SAR) images are designed to pass through the clouds; thus, we focus on SAR images from the Sentinel-1 constellation. A total of 14 Sentinel-1 SAR images, recorded from 3 December 2020 to 8 May 2021, were used for the analysis. The images were taken by the vertical–vertical polarization (i.e., vertical polarization of the incident wave emitted from the antenna and vertical polarization of the backscattered wave intercepted by the antenna) from the ascending path, and the incident angles at the two study areas are about 38^∘. The SAR images were processed using the SNAP software (SNAP, 2021). Figure 1d shows a color composite of backscattered intensity recorded on different dates. The red band denotes an image recorded on 14 April 2021, and the green and blue bands denote the image recorded on 3 December 2020. Note from inset (ii) in Fig. 1c and d that the Lomo de Corvina area in the SAR image looks smaller than that from the optical image. Such geometric distortions are because of the oblique observation geometry of SAR images. Figure 2a depicts a scheme of the elevation profile corresponding to the line $\stackrel{\mathrm{‾}}{\mathrm{AE}}$ depicted in inset (ii) of Fig. 1. The profile is simplified by four segments, $\stackrel{\mathrm{‾}}{\mathrm{AB}}$, $\stackrel{\mathrm{‾}}{\mathrm{BC}}$, $\stackrel{\mathrm{‾}}{\mathrm{CD}}$, and $\stackrel{\mathrm{‾}}{\mathrm{DE}}$. In order to illustrate the geometric distortions in the SAR images, it is assumed that the range direction is parallel to the vertical plane that contains the profile. The slant range denotes the direction from which the microwave energy travels from the satellite to the ground. The microwave images are presented in ground-range format. Note that because points A and C share about the same distance to the satellite, both segments $\stackrel{\mathrm{‾}}{\mathrm{AB}}$ and $\stackrel{\mathrm{‾}}{\mathrm{CB}}$ occupy the same geographic position in the ground range ($\stackrel{\mathrm{‾}}{\mathrm{ab}}$ and $\stackrel{\mathrm{‾}}{\mathrm{cb}}$) of the SAR image. In the optical image, the profiling scheme is located over the line $\stackrel{\mathrm{‾}}{\mathrm{AE}}$ (Fig. 1c); on the other hand, the profile is located along the segment $\stackrel{\mathrm{‾}}{\mathrm{ae}}$ in the SAR image, which is significantly smaller. However, the segment that contains the informal settlement in the optical image, $\stackrel{\mathrm{‾}}{\mathrm{DE}}$, is only slightly larger than the length of $\stackrel{\mathrm{‾}}{\mathrm{de}}$. Note also that there is a shift between $\stackrel{\mathrm{‾}}{\mathrm{DE}}$ and $\stackrel{\mathrm{‾}}{\mathrm{de}}$.

Regarding the color composite SAR image shown in Fig. 1c, red tones are observed in the new informal settlements, which express an increment of the backscattering in time. Such an increment originates from the double-bounce scattering mechanism produced by the shelters. Figure 2b and c depict a scheme of the backscattering mechanism under two scenarios. When the radar pulses emitted by the satellite reach a flat area, only a small fraction of the energy is backscattered to the sensor (Fig. 2b). In the presence of an object (Fig. 2c), the radar pulses bounce off the ground towards the object, and then they are reflected from the object to the sensor. Therefore, a larger fraction of the energy is backscattered to the sensor. In the case of the new informal settlements, such objects are indeed the makeshift shelters. We exploit this pattern to map the extent of the occupied area. Figure 2d shows the time-series backscattered intensity at some representative points of the recently occupied areas (red lines) and existing built-up areas (blue lines). As expected, a clear increment is observed in the invaded area from 14 April 2021. Therefore, to map the recently occupied areas, a threshold for each pixel coordinate is computed. The threshold is set as the average plus 2 times the standard deviation of the time-series backscattering intensity until 2 April 2021. Then, if the backscatter intensity recorded on 14 April 2021 is greater than the threshold, it is assumed that the area is occupied. After thresholding, the morphological operators closing and opening were applied using a kernel size of 3×3. Then, pixel clusters were identified, and those with sizes lower than 200 pixels were filtered out. Figure 3 depicts the extent of the invaded areas in the Morro Solar and Lomo de Corvina. The black polygon shown in Fig. 3b denotes the extension of the occupied area, which is estimated from visual inspection of images and videos recorded by an unmanned aerial vehicle (UAV) (Inga, 2021b). Our results from SAR images identified 84 % of the area identified by visual inspection. Furthermore, only 67 % of the area identified from SAR images is contained within the black polygon. We believe that the main reason for the discrepancies between SAR images' results and those from visual inspection lies in the complex geometric distortions in the SAR images.

The results show that a recent squatter settlement could be identified from radar images. Considering the open access to Sentinel-1 imagery, this represents an opportunity to implement a sustainable, low-cost system to monitor informal urban growth over unsafe areas and perform hazard mitigation actions, such as relocation. Prompt measures can prevent significant losses to society. For instance, the mapped settlements are 438 000 and 265 300 m² in the Morro Solar and Lomo de Corvina, respectively. From a visual inspection of the urban settlements nearby the recently occupied areas, we estimated a total of 3051 houses in the Morro Solar and 1595 in Lomo de Corvina that would have been exposed to strong ground motion amplification and potential landslides during earthquakes.

The software SNAP is available for download at https://step.esa.int/main/download/snap-download/ (SNAP, 2021). The code for the time-series analysis is available on request to the corresponding author.

The Sentinel-1 data are available for download at https://scihub.copernicus.eu/dhus/#/home (Copernicus, 2021). The images and videos recorded by a UAV is available at: https://www.dreamstime.com/info532049_info (Inga, 2021b).

LM, FG, CG, and MD conceived and designed the study. All the authors contributed in the analysis of the data, discussion of the results, and preparation of the manuscript.

The contact author has declared that neither they nor their co-authors have any competing interests.

Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This work was partly supported by the CONCYTEC–World Bank research program “Improvement and Extension of the services of the National System of Science, Technology and Technological Innovation” 8682-PE through its executing unit PROCIENCIA (contract 038-2019 FONDECYT-BM-INC-INV), the JSPS Kakenhi 21H05001, and the TCPAI Center and the Core Research Cluster of Disaster Science at Tohoku University (a Designated National University).

This paper was edited by Daniele Giordan and reviewed by two anonymous referees.