For this study, we used geological and geophysical maps of Guinea at a scale of 1:500,000 produced by Galperov G., combined with GPS surveys. ArcGIS was used for satellite image processing (Landsat 8 OLI/TIRS image from January 29, 2025, downloaded from USGS at https://earthexplorer.usgs.gov/; the sky was clear during this period). Geomatica and RockWorks 17 software were used to extract the lineaments, and Google Earth Pro was used to delimit the study area. The drilling data comes from institutional sources (SEG, SNAPE, UNICEF) and private drilling operations. These data enable a comprehensive analysis of the lineaments. The geotechnical data comes from the results of various geotechnical studies conducted by the Conakry Geoscience Agency.

The methodology adopted in this work consists of six steps: data acquisition—image processing—automatic extraction of lineaments—analysis and validation of lineaments—statistical and geostatistical analysis of lineaments—mapping of subsidence areas. It is based mainly on the use of USGS Landsat 8 imagery and geotechnical data. Researchers such as [45], [23], [2], [31], [35] and [25] used the same image acquisition methods for their study.

The combination of several spectral bands (false colors 7-6-4) in ArcGIS made it possible to create a composite image for detecting linear geological structures by emphasizing spectral and textural contrasts. This method is more effective than pixel by pixel analysis ([13]; [26]; [32]; [28]). One approach has been validated by the work of [14]. It uses the Image Analysis algorithm to generate GeoTIFF files highlighting linear structures associated with lineaments using infrared/red coupling. The Landsat 8 image, acquired under clear skies, did not require extensive atmospheric corrections. We therefore performed a visual check of its color composite directly in Google Earth. This was previously calibrated on the same date as the raw image acquisition date. From there, a visual comparison identified landscape discrepancies and general rendering imperfections. Several digital image processing techniques, including standard color composites, were used by [8] to map linear structures. The automatic detection algorithm in Geomatica was used to identify lineaments. [36], [42] and [1], applied the same method to generate vector data, which was then validated and analyzed in GIS software. The validation, inspired by the methods of [26] and [17], consisted of a visual and systematic comparison of the detected lineaments with road and hydrographic networks and drilling data to verify their geological relevance in an urban context. The final results were exported to Excel and RockWorks 17 for statistical analysis, which consisted of characterizing the lineaments (length, orientation) using distribution laws. Geostatistics was used to view the spatial organization of the lineaments via a variogram adjusted to a theoretical experimental model. After converting the vector data (lineaments) into point data in ArcGIS 10.3, we used kriging for spatial interpolation of the data on a map background. The variogram model used made it possible to obtain a map showing areas of low mechanical strength (areas with high interstitial void density, areas favorable or unfavorable for water). These areas are where two or more lineaments intersect. The lineament results obtained are then superimposed on the geotechnical test data to locate areas at risk of subsidence in the municipality of Ratoma. The geotechnical parameters used for this purpose are water content, particle size distribution, and Atterberg limits. Grain size, in particular the fine particle content, is the important factor that directly influences the Atterberg limits of a soil, which in turn determines the soil’s subsidence character ([18]; [21]; [3]; [47]). The influence of particle size distribution on Atterberg limits can be explained by physicochemical principles related to the surface of the grains. These electrically charged sheet-like surfaces enable water molecules to be strongly adsorbed and retained. This interaction between water and fine particles gives the soil its cohesion and plasticity, explaining the high Atterberg limits ([4]; [33]; [10]; [29]).

The relationship used in this article to calculate the liquidity index is the one used in the work of [3]. Based on soil mechanics research, we compared our results with the conditions of [15], which link the degree of saturation Sr0 of a soil to its liquidity index on the one hand, and its liquidity index to its settlement potential on the other. In 1966, Jaroslav Fedaalso proposed in his work a collapse index based on natural water content, degree of saturation, plasticity limit, and plasticity index. The definitive identification in this research is based on the liquidity index. The particle size distribution was interpreted by separating the pass and reject fractions using a 0.08 mm sieve.

Finally, a multi-criteria approach was used to map areas at risk of subsidence. The method is based on the Analytic Hierarchy Process (AHP) developed by [41] in 1970 and [7]. This structured technique breaks down the problem into a hierarchy, allowing criteria to be compared and weighted in pairs ([5]). The process includes identifying and developing decision criteria, classifying and standardizing criteria, and weighting and aggregating these criteria. This generates a summary map that facilitates informed decision-making. Despite the results obtained in this research, the use of only 11 samples for the entire municipality remains insufficient for reliable spatialization of geotechnical information.

Landsat 8, launched in 2013 by NASA and the USGS, is a satellite in the Landsat program designed for environmental monitoring, mapping, and natural resource management. It offers a spatial resolution of 15 m (panchromatic), 30 m (visible, NIR, SWIR), and 100 m (thermal). With a spectral resolution of 11 bands and a radiometric resolution of 16 bits, it allows for better distinction of spectral variations. Thanks to its radiometric depth and the program’s historical archives, Landsat 8 is suitable for long-term studies. Landsat 8 products already undergo automatic geometric corrections by the USGS.

Automatic lineament extraction was performed using Geomatica. The configuration of the parameters (thresholding, contrast threshold, minimum length, and fileting) underwent three iterations in order to optimize the process and achieve the result shown in Figure 1.

Figure 1. Mapping of lineaments in the study area.

Validation consisted of superimposing the extracted lineaments on the hydrographic and road networks. Those that coincide with these surface features are selected and removed from the lineament network. Spatial analysis of the lineaments using high-yield drilling data reveals that all of these drill holes are located near a major lineament or in an area with a high density of lineaments, unlike low-yield drill holes, which are located further away.

Figure 2. Validation of lineaments in relation to road networks in the study area.

3.3.1. Statistical Analysis

Due to the complexity of the study area (urbanized area) and the objective of the study, the detailed map of lineaments obtained after various treatments shows high densities of lineaments of varying sizes. These lineaments vary from a few meters to less than 1 km for minor lineaments with a total length of 307,414.61 m, an average of 444.88 for a total number of 691; then from 1 to 2 km for medium-sized lineaments with a total length of 281,107.62 m, an average of 1398.55 for a total number of 201. Finally, the elements with a minimum length greater than 2 km are the major lineaments, with a total length of 359,256.62 m, an average length of 3018.96 m, and a total number of 119. This gives a total of 1011 lineaments, of which 68% are minor lineaments, 20% are medium lineaments, and 12% are major lineaments. The major lineaments include approximately 6% exceeding 5 km and 17% measuring between 3 and 5 km, while only 2 minor lineaments reach approximately 0,8 km. The longest lineament (17,715 km) crosses the study area from northwest to northeast. This characterization of lineaments is very important for geological and geomorphological analysis in order to assess their significance and spatial extent.

The fracture density map shown in Figure 3 expresses the number of lineaments per unit area. It reveals areas of high concentration (17.01 to 28.33 km/km²) and low concentration (0 to 11.33 km/km²) of lineaments in the study area.

Figure 3. Map of the density of Ratoma lineam.

Figure 4 illustrates the interpolation variogram of underground lineaments, modeling fractures in all directions. The high values of this distribution of unstable areas represent 48.61% of the study area, while the low values represent 18.05%.

The orientation of the lineaments is analyzed by the rosettes illustrated in Figure 5. The lineaments are grouped into 12 orientation classes, based on angular intervals of 15˚. The distribution of fractures on the directional rosettes is relatively homogeneous for minor lineaments as shown in Figure 5(a), but irregular for major lineaments on Figure 5(c). However, all lineament families have frequencies exceeding 80%, with variations: 30% - 80% for minor lineaments, 16% - 80% for medium lineaments, and 8% - 83% for major lineaments. The layout of the lineament rosettes reveals two dominant orientation classes: NW-SE for major lineaments and NE-SW for medium lineaments. Each class also has a secondary

Figure 4. Map of the variogram model using kriging.

(a) (b) (c)

Figure 5. Diagram of directional rosettes. (a) Rosette of minor lineaments; (b) Rosette of medium-sized lineaments; (c) Rosette of major lineaments.

orientation, which is the opposite of the dominant one.

3.3.2. Geostatistical Analysis

Geostatistics is discussed here to show the spatial distribution of fractures and produce a model of fractured networks based on 2D cartographic analysis. Figure 4 shows the result of the geostatistical analysis of the lineaments. This map, which is the result of the spatial distribution of spherical experimental variograms, shows that the most fractured areas (those with low mechanical resistance) are scattered throughout the study area, with values ranging from 0.21 to 0.99. The map produced illustrates the distribution and spatial variability of these lineaments. It highlights geological discontinuities, such as groundwater circulation channels (lineaments) or shear zones. This spatial distribution shows the heterogeneity and extent of the lineaments, indicating a large-scale influence of linear geological structures.

Figure 4 (lineament density) and Figure 5 (void density), although different, complement each other. The first explains the geological behavior of linear structures, while the second characterizes their geotechnical properties. Both figures provide a comprehensive understanding of the phenomenon, incorporating both structural and mechanical factors affecting the soil. The synthesis of these two pieces of information is essential for a rigorous risk assessment and appropriate development planning.

Table 1. Geotechnical characteristics of Ratoma soils.

Source: Conakry Geoscience Agency.

The comparison between the natural water content of a soil and its Atterberg limits provides an initial qualitative indication of its potential for settlement. Among the 11 samples analyzed in Table 1, only one (Elamba) has a natural water content almost equivalent to its liquidity limit (W = 48.3 and WL = 48) and a saturation degree close to 100%. This sample was taken at a depth of between 6.5 and 7 meters in the soil. The liquidity index (LI) is the indicator used to predict the behavior of a soil. LI is presented in Table 2.

Table 2. Identification of subsidence-prone areas in Ratoma.

Table 2 shows that the Elamba sample has a liquidity index (LI) of 1.013 for a depth of 6.5 to 7 meters. This value, which is greater than or equal to 1, is consistent with the fact that its water content (W) is almost identical to its liquidity limit (WL). On the other hand, samples with an LI of less than 0, regardless of the depth at which they were taken, present a low risk of subsidence.

In this study, five criteria were used to identify areas at high risk of subsidence: slope, altitude, fracture density, mechanical resistance, and soil liquidity index. The combination of these topographical, geological, and geotechnical elements made it possible to create a map of susceptibility to subsidence, following the methodology described in the work of [38].

The subsidence map shown in Figure 6 was produced by combining these five criteria, which are divided into three factors (topographical, geological, and geotechnical). It shows five classes with an uneven distribution.

Figure 6. Map of areas at risk of subsidence.

The validation of the map of areas at risk of subsidence took into account the location of the 11 samples. The Lambanyi sample (Elamba, where IL > 1) confirmed its position, as did the Nongo Taady 1 sample (ENongTa1, where 0 ≤ IL < 1), which proved to be decisive. The final distribution indicates that 49.06% of the area is at risk of subsidence, 34.24% is at low risk, and 16.7% is at moderate risk. This categorization provides a clear representation of the vulnerability of the area.