Next Article in Journal
VE-LIOM: A Versatile and Efficient LiDAR-Inertial Odometry and Mapping System
Previous Article in Journal
Research on Jianghan Plain Water System Dynamics and Influences with Multiple Landsat Satellites
Previous Article in Special Issue
Three-Dimensional Rockslide Analysis Using Unmanned Aerial Vehicle and LiDAR: The Castrocucco Case Study, Southern Italy
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 16
Issue 15
10.3390/rs16152771
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Comparison of Landforms and Processes Detection Using Multisource Remote Sensing Data: The Case Study of the Palinuro Pine Grove (Cilento, Vallo di Diano and Alburni National Park, Southern Italy)

by
Mario Valiante
1,*,
Alessandro Di Benedetto
1 and
Aniello Aloia
2
1
Department of Civil Engineering, University of Salerno, 84084 Fisciano, Italy
2
Cilento, Vallo di Diano and Alburni National Park Authority, 84078 Vallo della Lucania, Italy
*
Author to whom correspondence should be addressed.
Remote Sens. 2024, 16(15), 2771; https://doi.org/10.3390/rs16152771 (registering DOI)
Submission received: 5 June 2024 / Revised: 18 July 2024 / Accepted: 23 July 2024 / Published: 29 July 2024
(This article belongs to the Special Issue Aerial and Drone LiDAR Data for Geomorphological Mapping, Landform Extraction and Landscape Evolution)
Browse Figures
Review Reports Versions Notes

Abstract

:
The automated recognition of landforms holds significant importance within the framework of digital geomorphological mapping, serving as a pivotal focal point for research and practical applications alike. Over the last decade, various methods have been developed to achieve this goal, ranging from grid-based to object-based approaches, covering a range from supervised to completely unsupervised techniques. Furthermore, the vast majority of the methods mentioned depend on Digital Elevation Models (DEMs) as their primary input, highlighting the crucial significance of meticulous preparation and rigorous quality assessment of these datasets. In this study, we compare the outcomes of grid-based methods for landforms extraction and surficial process type assessment, leveraging various DEMs as input data. Initially, we employed a photogrammetric Digital Terrain Model (DTM) generated at a regional scale, along with two LiDAR datasets. The first dataset originates from an airborne survey conducted by the national government approximately a decade ago, while the second dataset was acquired by UAV as part of this study’s framework. The results highlight how the higher resolution and level of detail of the LiDAR datasets allow the recognition of a higher number of features at higher scales; but, in contrast, generally, a high level of detail corresponds with a higher risk of noise within the dataset, mostly due to unwanted natural features or anthropogenic disturbance. Utilizing these datasets for generating geomorphological maps harbors significant potential in the framework of natural hazard assessment, particularly concerning phenomena associated with geo-hydrological processes.

1. Introduction

In the field of digital geomorphological mapping, many efforts have been put into the automated recognition of landforms and processes starting from grid-based data such as DTMs (Digital Terrain Models) using GIS-based analyses. Several methods have been proposed in recent times exploiting different techniques. Simpler methods rely on grid-based analyses, using a DTM and its common derivatives such as slope, aspect, curvature, etc., or various types of statistics such as neighborhood analyses [1,2,3,4,5,6,7,8,9,10], which deliver an outcome based solely on cell-by-cell values. Other methods are based instead on data segmentation following the principles of Object-Based Image Analysis (OBIA), stacking DTMs and their derivatives [11,12,13,14,15,16,17,18], RGB or multi-spectral imagery [19,20,21,22], or both [23,24,25]. Other techniques, such as change detection, also include the time factor into the analysis and are based on the recognition of changes rather than on the processing of a time-static set of data [26,27,28]. Moreover, in the last decade, all of these techniques have been increasingly gaining the support of a plethora of artificial intelligence algorithms [26,29,30].
Other than landforms, research has been focused also on the analysis of processes acting on the earth’s surface using geomorphometric techniques. Common fields of application can be related to natural hazards, such as landslide activity and susceptibility [9,31,32,33,34,35,36,37,38,39,40,41,42,43], landscape evolution modelling and tectonic geomorphology [44,45,46,47], or geomorphic transport law and surficial processes types [10,48,49,50,51], among many other fields.
For a successful geomorphological analysis, it is crucial to first generate a numerical Digital Elevation Model (DEM), which provides a detailed and reliable representation of the ground surface object of a particular study. When it comes to data collection, if the area is relatively small (in the order of tens of hectares) and well-lit by the sun, a Terrestrial Laser Scanner (TLS), based on LiDAR (Light Detection and Ranging) technology, can be effectively used. The TLS offers both high precision and high resolution [52]. However, for larger areas, the TLS method can be time-consuming and costly, both in terms of field measurements and data processing. In such cases, an Aerial Laser Scanner (ALS) deployed from a helicopter, aircraft, or UAV (Unmanned Aerial Vehicle) might be a more suitable choice [53]. The data collected using the LiDAR (Light Detection and Ranging) technique, which results in a point cloud that is interpolated to construct the DEM, are highly versatile. It can be used for a variety of applications, including detailed geomorphometric analyses, landslide studies, basin analysis, and road mapping [27,54,55,56,57]. It is worth noting that local authorities are increasingly commissioning periodic topographical surveys of their jurisdictions. This trend is providing scientists with access to up-to-date data at no cost. Such data, in some cases, consist of raw point clouds acquired through different techniques, each with varying characteristics and resolutions. Generally, to transition from the original point cloud to the DTM, it must undergo filtering to exclude non-ground objects (such as vegetation and artifacts), and then be interpolated onto a grid. Both filtering and interpolation processes require testing, the parameters to be set change depending on the type of morphology and coverage, and the accuracy of the resulting DTM must be verified [58].
Another crucial consideration is that increasing the resolution of data, such as with LiDAR data acquired from UAVs, is not always beneficial for subsequent geomorphological analyses. In fact, higher resolution can introduce additional sources of uncertainty in some interpretations [59].
This phenomenon can be attributed to the fact that many filtering algorithms, for the removal of vegetation, were originally designed for LiDAR data from aerial platforms, which typically have much lower resolutions than the current LiDAR data obtained from UAVs. Given this context, it becomes essential to make careful choices regarding filtering and terrain modeling parameters. By doing so, we can fully leverage the data while minimizing the emergence of additional uncertainties.
The work hereby presented aims at the assessment of geomorphometric performance of datasets having different characteristics in terms of source, interpolation procedure, and resolution. Such goal has been achieved through the computation of landforms and processes domains using a modified and updated version of the procedures presented in [10], with a final focus on such landforms and processes related to flooding hazard.
A test site from Southern Italy, along the Cilento coastline, has been chosen because it recently experienced floodings, threatening the population, local activities, and infrastructures. The area is a pine grove located to the south of the harbour of Palinuro, nestled within the bay created by the cape bearing the same name (Figure 1). The general geological settings of the Palinuro cape are that of a Jurassic limestone relief covered with plio–pleistocene sandy-conglomeratic coastal deposits with variable thickness. The entire study area falls within the borders of the Cilento, Vallo di Diano and Alburni National Park, UNESCO Global Geopark and World Heritage site. In October 2022, after heavy rainfall, several debris flows to floods (sensu [60,61]) occurred along the channels in the study area. Fortunately, there were no casualties because it happened during low season and only minor damages were recorded to local activities and the harbour; however, if the same phenomena would have been occurred during high season, the risk would have been much higher because of the high tourist influx.

2. Materials and Methods

Analyses have been performed on three datasets, each derived from a distinct source and featuring different resolutions in order to achieve a multi-scalar comparison of the approaches:
  • 5 m resolution DTM derived from a photogrammetric survey of the regional authority.
  • 1 m resolution DTM derived from LiDAR data of the Ministry.
  • 20 cm resolution DTM from UAV LiDAR produced by the authors for this study.
For the recognition of landforms, a grid-based neighborhood analysis has been used, while for the estimation of process types acting on the hillslope, Slope–Area plots have been exploited.

2.1. Available Data

The first dataset, a 5 m resolution DTM, is part of the open-source geo-spatial data within the National Topographic Database (NTDB) derived from the Regional Technical Numerical Cartography, produced at the scale of 1:5000. The NTDB, produced by the Campania region’s Public Administrations for the purposes of planning and land management according to the Content Specifications of the Catalogue of Spatial Data, annexed to the Italian Ministerial Decree of 10 November 2011 “Technical rules for the definition of the specifications of the Geotopographic database content”, following the General Guidelines stated by the INSPIRE Directive (Directive 2007/2/EC of 14 March 2007). Specifications are available at the following link: https://sit2.regione.campania.it/notizia/database-topografico (accessed on 1 July 2024) (Figure 2a). In particular, this DTM results from a regional aerophotogrammetric survey performed in 2004.
The second dataset is derived from airborne LiDAR data acquired from 2010 to 2014 under commission of the Italian Ministry of Environment, Land and Sea Protection (“Ministero dell’Ambiente e della Tutela del Territorio e del Mare”—MATTM), as part of the Not-Ordinary Plan of Remote Sensing (“Piano Straordinario di Telerilevamento”—PST-A) (Law 179 of 31/07/2002, art. 27), and is provided on demand (Figure 2b). The LiDAR data on the area were acquired on April 2013 using an Optech ALTM 3100 laser scanner by Optech, mounted on an aircraft flying at an altitude between 1500 and 1800 m AGL (height Above Ground Level). The scan frequency has been set as equal to 100 kHz and the direction of flight is parallel to the coastline; the maximum scan angle used is 25. The LiDAR data were processed using TerraScan software (version 12). The specifications of the data stated by the authority are as follows: point density greater than 1.5 points/m2, planimetric accuracy (2σ) of 30 cm, and altimetric accuracy (1σ) of 15 cm.
DTMs from LiDAR provided by the Ministry exhibit an undesirable “squaring” effect. This occurs because the coordinate system (φ, λ) used to interpolate the data is not an isothermal coordinate system on the ellipsoid. In other words, the spacing between points is not uniform across the entire model. As a result, the representation of terrain features can appear blocky or squared off. Researchers have explored this issue, and a de-tailed discussion can be found in [62]. Therefore, to obtain a 3D representation of the land and the built environment that would provide an accurate rendering of geomorphological and anthropic elements with their typical shape, an automated tool in ArcMap environment has been used to perform the entire optimization process [62].

2.2. New Acquisition

2.2.1. UAV LiDAR Survey

The LiDAR survey was carried out over the entire area of about 40 Ha with a DJI Matrice 300 RTK UAV (DJI Italian section, online store, Italy), with a quadcopter having a weight of about 9 kg, and was equipped with the DJI Zenmuse L1 LiDAR sensor (DJI Italian section, online store, Italy). Table 1 lists the main characteristics of the LiDAR sensor used. The survey was conducted on 25 March 2023.
The LiDAR survey was carried out by a flight plan, which was designed taking into account the morphological characteristics of the area and the requirements of the cartographic outputs to be produced, including the density of the point cloud (points/m2). A total of three flight plans were carried out, one with a nadiral sensor and swipes along the northeast–northwest direction, parallel to the slope, and two along the main channels to the east and west of the area.
The flight height was set as equal to 70 m, and the side overlap of individual LiDAR swipes was set as equal to 20%. The point acquisition frequency was set as equal to the maximum frequency. The UAV is equipped with GNSS (global navigation satellite system) receivers to acquire, in relative positioning in the nRTK (network Real Time Kinematic) mode, the trajectory directly in the current National Reference System UTM33/RDN2008 (EPSG: 7792), with an ellipsoidal height. In post-processing, the ellipsoidal heights were converted to orthometric heights; the geoid model used is ITALGEO2005. Transformations between the different reference systems were carried out using the grids provided by the Italian Military Geographic Institute (“Istituto Geografico Militare”—IGM) (in *.GK2 format), covering the extent of one element of the map of Italy at the scale of 1:50,000 (Figure 3).
No Ground Control Points (GCPs) were used to improve the georeferencing of the point cloud, nor Check Points (CPs) to assess the quality of georeferencing, since the accuracy of the survey from UAV in the nRTK mode largely guarantees the accuracies needed for our purposes [63]. In addition, the area is covered by vegetation and any artificial targets would not have been visible.

2.2.2. DTM Interpolation

The three acquired LiDAR point clouds, one for each plan of flight, were aligned with Atlascan software (version 1.3.0 Teledyne Optech). Figure 3a shows the top view of the point cloud, and Figure 3b shows an excerpt of the density map (points/m2). The average density is about 800 points/m2, with a maximum value of about 6000 points/m2.
The point cloud obtained from a LiDAR survey contains all the features present on the terrain, and, consequently, it is necessary to perform filtering of the data before subjecting it to interpolation processes. Such filtering consists of the semi-automatic elimination of points not belonging to the ground surface. Only points belonging to the bare ground, without vegetation, buildings, or other objects, will be used to generate the DTM.
Multiscale Curvature Classification (MCC) software (version 2.1) was used for the ground filter. The MCC algorithm is an iterative multiscale algorithm for classifying lidar returns using a spline-based technique for data interpolation and smoothing (thin plate splines, TPS) [64]. There are two parameters that the user must set: the scale parameter λ, which defines the cell resolution and sampling interval (density) of the point cloud and is a function of the size of objects on the ground, and the curvature threshold, t. The results are highly dependent on their correct balance. In this study, the parameters suggested by Barbarella et al. in 2019 [58] were used.
The points belonging to the ground were then subjected to spatial interpolation processes in order to generate the DTM. Terrain modeling was completed with a TIN (Triangulated Irregular Network) according to Delaunay’s criterion, which allows us to represent the terrain surface more effectively than the GRID model, assuming the input data are free of outliers.
The last step of the process consists of the conversion of the 3D TIN into a two-dimensional image (raster), where each pixel is associated with an elevation value. The choice of cell size should align with the representation scale, which, in turn, must be congruent with the accuracy and density of the input data. The interpolation method employed is linear; each triangle (face) within the TIN defines a plane in space, passing through the three vertices of the triangle itself. Using this interpolating plane, the elevation values of the corresponding raster cell are computed. Generally, as the resolution increases, the output raster will more accurately represent the TIN surface. In our case, a grid step of 20 cm was used.
Figure 4 shows the shaded relief map of the DTM with contour lines with an equidistance of 1 m superimposed.

2.3. Geomorphometric Analyses

Geomorphometric analyses have been carried out in GIS environments using Free and Open-Source Software, such as Grass GIS (version 8.3.2) as a primary tool for the storage and elaboration of grid data, and the QGIS (version 3.34.8 LTR) for visualization and layout exports. Further statistical analyses have been carried out using the R software (version 4.4.1) within the RStudio environment (version 2024.04.2+764). The workflow outlined in the next sections has been implemented for all three datasets.

2.3.1. Landform Recognition

Landform recognition has been performed using an original grid-based neighborhood analysis based on the Deviation from Mean elevation (DEV) [4,65]:
DEV = (z0 − zmean)/SD
DIFF = z0 − zmean; TPI = int(z0 − zmean) + 0.5
where z0 is the elevation of a given DTM cell, zmean is the mean elevation of its neighborhood, and SD is the standard deviation of the elevation values in the same neighborhood.
DEV has been preferred to other methods based on the Difference from Mean elevation (DIFF), such as the raw DIFF itself [10,65] or its discretization, which is the Topographic Position Index (TPI) [2,66], because it has been proven to be more effective in such cases where the elevation difference among adjacent DTM cells is very low [4]. This configuration holds true for flat landscapes at any level of detail, but in our case, even if we are on a coastal hillslope, using very high-resolution data (1 m and 20 cm) drastically decreases the relative height difference between the adjacent cells. For the sake of comparison, the same method has been used also for the 5 m DTM.
To overcome the subjective choice of the neighborhood size, the neighborhood analysis hereby presented has been performed iteratively using 11 circular neighborhoods with radiuses progressively increasing following powers of 2, ranging from 20 to 210 DTM cells in order to have the same amount of cells at any radius size, regardless of the effective DTM resolution. This step produced 11 DEV datasets for each DTM. Then, each stack of 11 datasets has been aggregated computing the average value cell-by-cell, resulting into 3 mean DEV datasets, one for each DTM (Figure 5).
Generally, as for the DIFF, the positive values of DEV represent convex morphologies, while the negative values of DEV represent concave morphologies. Values around zero are characteristic of flat morphologies, being that horizontal or variously dipping, such as plain slopes.
After these computations, the mean DEV datasets have been classified into elementary landforms with the integration of slope data, following the classification procedure as described in Dramis et al., 2022 [10], for DIFF values (Figure 5).

2.3.2. Process Domains

Process domains acting on the hillslope have been evaluated exploiting Slope–Area plots [10,49,50,51,67]. Slope–Area plots are based on the analysis of the graphical plot of drainage area values versus slope values; different gradients and curvatures of the plot denote different types of processes acting on the slope, ranging from diffusive processes acting on ridges to channelized fluvial processes within valleys. A summary of the investigated process is provided in Table 2. The scatter plot is produced by plotting values for each cell of the domain. For our study, the domain has been identified as the set of the watersheds of the channels of interest, removing areas with a strong anthropogenic remodeling at the foot of the slope to minimize noise influence (Figure 2). Drainage area values have been computed using the Multiple flow direction algorithm [68], while slope values are expressed in percentages. From the scatter plot of cell values, data are aggregated computing the average value for each bin spanning a quarter of a logarithmic cycle.
After plotting drainage area and slope binned values, thresholds can be identified along the plot based on the gradient and the curvature of the plot itself, dividing the plot into 6 process domains, based on [10], rather than the usual 4 domains, which are commonly reported in the literature [49,50,51]. The first region (Ia) is characterized by diffusive aerial processes acting on ridges where, usually, slope values increase along with the drainage area. The first threshold marks the boundary between the first and the second region (Ib), which represents diffusive aerial processes acting on the shoulders and backslopes and is represented by a peak in the Slope–Area plot, thus an inverse gradient; in fact, in this region, slope values decrease with increasing the drainage area. The second threshold is marked by a maximum point of curvature, which can be positive or negative depending on the morphology of the hillslope and is positioned at the first maximum or minimum point of the 2nd-order derivative of the plot, after the first threshold has been identified. In this region (II), representing the beginning of concentrated erosional flows within hollows or headwaters, slope values will decrease with a different rate than the previous values, or they can even increase again in such cases where this region of the plot is strongly concave. The third threshold is represented by an inflexion of the plot inverting its concavity or convexity, and it is marked by a zero point on the 2nd-order derivative plot. In this region (IIIa), there is another break in the slope values, which will decrease at a slower pace, or will start to decrease again in case of a previous inversion. This region represents the transition between concentrated and proper channelized flows, usually with an erosive behavior. An inversion of the slope rate marks the fourth threshold with the next region (IIIb), and it is represented by a zero point of the 1st-order derivative plot. Here, slope values will start to increase again along with the drainage area, representing depositional zones on the lower portions of the hillslopes. This region represents both erosional and depositional behavior of channelized flows. The beginning of the last region (IV) is marked by a linear and rapid decrease of slope values with increasing drainage areas representing the typical profile of proper fluvial processes. However, this region is difficult to describe within the small catchment characterized by impulsive and turbulent discharge, such as the channels object of this study.

3. Results

3.1. Landforms Analysis

Starting from the input data, consisting of three DTM having different resolutions, elementary landforms have been classified based on the mean DEV values discussed in the previous section (Figure 6). From a geomorphological perspective, the study area comprises a hillslope along the northern coast of the Palinuro cape. Such a structure is reflected in the output of the classification phase. All the DTMs highlight the gradual transition from summit convex landforms, such as ridges and crests, along the southern portion of the study area, which correspond to a portion of the main ridge of the Palinuro cape. Such entities are well represented by convex morphologies detected by the grid-based analysis, and they are represented by landforms 1, 2, 3, 4, 5, and 6 (refer to Figure 5). On the 5 m DTM, transitions between landforms are much sharper and convex morphologies are only detected along the main ridge, while increasing the resolution of the grid leads to fuzzier transitions and to the detection of convex shapes also in other portions of the relief; in particular, small hills at the foot of the main slope are well detected using the 20 cm dataset. In terms of coverage, the set of convex morphologies extend for about 148,500 m2 on the 5 m DTM, about 102,100 m2 on the 1 m DTM, and about 70,000 m2 on the 20 cm DTM.
Downslope, there is the transition to flat landforms such as middle slopes or shelves, represented by landforms 7 and 8; the latter constitutes the most representative landforms at any resolution in terms of unique landforms (Figure 6d). As for previous morphologies, boundaries between flat morphologies and the others are sharper on the 5 m dataset, and the flat region is almost continuous. Instead, increasing the resolution of the dataset allows us to distinguish more features within the middle section of the relief. Flat morphologies extend for about 79,000 m2 on the 5 m dataset, about 122,000 m2 on the 1 m dataset, and about 142,800 m2 on the 20 cm dataset.
Concave morphologies group landforms 9–13, and they are located on the lower portion of the slope on the 5 m grid. This area corresponds to the lower section of the relief reaching the sea level, which is crossed by the main road connecting the Palinuro village to the harbour, and where several tourist establishments are located. On the 5 m grid, only the lower portions of the slope are classified as concave landforms, while, when increasing the resolution of the dataset, channels are highlighted uphill along the relief, up to the point that, looking at results from the 20 cm dataset, channels reach the summit area of the relief. Concave morphologies extend about 48,550 m2 on the 5 m dataset, 51,200 m2 on the 1 m dataset, and 40,160 m2 on the 20 cm dataset. However, if values related only to strongly concave morphologies are isolated, which could be related to channelized landforms, on the 5 m DTM they extend for about 500 m2, 1981 m2 on the 1 m DTM, and 3225 m2 on the 20 cm DTM.

3.2. Process Domains Analysis

Slope–Area plots for each dataset of the analyzed domain are shown in Figure 7. The plot produced with the 5 m grid does not feature a prominent maximum point in the first section of the plot, as most data from the literature suggest, instead it is characterized by a small initial peak, followed by the maximum point at the center of the plot. Such a deformation could be related to the fact that ridges are under-represented within the watershed area, mostly because in the study area ridges are narrow; in fact, our watershed is bordered by cliffs. Instead, the plot built from the 20 cm dataset exhibits a double peak in the first section of the plot. However, looking at the drainage area values at the first peak, it corresponds roughly to 0.1 m2. Such a small value is likely due to the very high resolution of the dataset and probably related to noise. Therefore, the next peak, which corresponds also to the maximum point of the plot, is chosen as threshold between regions Ia and Ib. All the plots do not reach the fluvial threshold as they do not exhibit a linear decreasing trend for more than two points, as could be expected from such small catchments.
Process domains have been visualized on a map classifying drainage area datasets with thresholds derived from the analysis of the plots (Figure 8). On the 5 m dataset, most of the area falls within the IIIa region, denoting the prevalence of erosive surficial processes transitioning from unchanneled to channeled settings. Some channeled processes are also highlighted (region IIIb). On the 1 m grid, the scenario is much more detailed. Open slopes are now classified as diffusive (Ib) to concentrated (II) processes, and channeled settings are more branched, retracing the actual drainage network. The 20 cm dataset highlights, again, more details. More areas are classified as Ia, as more convex morphologies can be recognized along the slope. Channels are more branched, and they are traced further uphill, as is the case for the landforms. Moreover, at this resolution, anthropogenic features also come into play, as paths along the pine grove influence the flow direction. All the datasets highlight, to a different extent, the erosional–depositional behavior of the channels within the study area, with channel heads progressively moving uphill, increasing the dataset resolution.

3.3. Data Comparison

In order to verify consistencies between the landform detection and the process analysis, results have been compared using the bivariate technique of the Frequency Ratio (FR) [69,70], a widespread methodology exploited for the evaluation of natural hazards’ susceptibility [40,71,72]. The focus was to ascertain whether any landform has a particular type of associated process, or if any type of process is focused within any particular landform. To this end, FR has been computed first comparing the spatial frequency of each landform within a given process (fLiPi) with the spatial frequency of process itself (fPi) (FRPvL), then comparing the spatial frequency of each process within any landform (fPiLi) with the spatial frequency of the landforms themselves (fLi) (FRLvP):
FRPvL = fLiPi/fPi = (landformi frequency within the processi area/landformi frequency)/(processi frequency/total area)
FRLvP = fPiLi/fLi = (processi frequency within the landformi area/processi frequency)/(landformi frequency/total area)
Then, based on the score returned by the FR analysis, the spatial correlation between concave landforms related to channels’ morphologies and associated processes, has been evaluated using the Area Under Curve (AUC) for each grid dataset.
FR results highlight different relations based on the dataset used for computations. Overall, increasing the resolution of the dataset results in a clearer relation between processes and landforms and vice versa (Figure 9). Looking at results from the 5 m DTM, relations between landforms and processes are not so evident; one noticeable relation regarding process IIIb could be inferred from both graphs, in which such a process is strongly related with concave landforms and not at all with convex landforms, as could be expected (Figure 9a). Moving on, with higher resolution datasets, relations become more evident. Convex landforms show FR > 1 with diffusive processes, like Ia and Ib, which decrease below 1 as we proceed to concentrated and channeled processes, while concave landforms have FR < 1 with diffusive processes, which increase as soon we reach transient and channeled processes (IIIa and IIIb). Flat morphologies are almost neutral, having FR ~ 1 from diffusive to transient processes, and FR < 1 with channeled processes (Figure 9c,e). The same relations can be observed when also looking at the process-oriented graphs. Process Ib (diffusive processes on ridges) has FR > 1 with convex morphologies, which goes below the unit moving to concave morphologies. This is more evident on the 20 cm dataset. Process Ib (diffuse processes on hillslopes) has FR < 1 with convex morphologies, and FR ~ 1 with other landforms. The same goes for process II (concentrated processes), showing no particular positive relation with any landforms. Transient and channeled processes (IIIa and IIb), instead, have a strong relation (FR >> 1) with concave morphologies, and no relation with convex landforms (FR < 1), as could be expected (Figure 9b,d,f).
To better ascertain whether there is a relation between grid-based landforms and processes evaluating using Slope–Area plots, the AUC has been computed for concave landforms, as they performed better with the FR score (Figure 10). Looking at the results, it would appear that the 5 m DTM performed better among the different datasets, with a good result regarding landform 13 (AUC = 0.89) and moderate to low performances regarding the other landforms. The other datasets had a bad performance with an almost low score (AUC < 0.7) in all the investigated landforms, with the 20 cm dataset performing worse than the 1 m one.

4. Discussions

The landforms recognition analysis delivered different results based on the dataset on which it was performed. Overall, moving from coarser to finer resolutions of the DTM gives results compatible with different mapping scales. The 5 m dataset provides smoother landforms that highlight, for example, the main ridge of Palinuro cape, along with cliffs and the main hillslopes. The 1 m dataset starts to reveal many more details, as boundaries between landforms become fuzzier, concave landforms start to reveal channels along the hillslopes, and anthropogenic features and modifications to the natural landscapes are highlighted, such as road cuts, building areas, and so on. Further increasing the resolution of the dataset also increases the complexity of the landforms’ distribution. Channels are highlighted almost up to the main ridge, and the morphology of the main hillslope becomes much more complex with a lot of convex shapes highlighting minor landforms, such as boulders, lobes of surficial deposits, land fillings, paths, etc. Also, minor hills are highlighted at the foot of the main slope. From a geomorphological mapping perspective, increasing the resolution of the dataset helps to recognize features with much more detail and, focusing on channels morphologies, it is possible to map such landforms up to the channel head (Figure 11). Besides boundaries between landforms becoming fuzzier when increasing the resolution, there is also the matter of spatial continuity. When analyzing the distribution of concave morphologies on the finer dataset, even though they can be easily reconstructed, they are not continuous in terms of cells adjacency; in fact, there are a lot of breaks along the landforms in which, for example, flat landforms have been classified. Such irregularities could be related to the filtering necessary to remove the intense vegetation of the pine grove, which may have introduced interpolations and approximations disturbing the outcome of the grid-based classification procedure. To overcome such issues related to fuzzy boundaries and/or small groups of isolated cells, other types of classifications procedures could possibly have been more effective, such as, for example, object-based segmentation.
As for landforms, a process analysis also gave different outcomes with different datasets. The coarser grid analysis gives an outcome in which most of the watersheds are interested by surficial runout processes transitioning from concentrated to channeled, confining diffusive runout only to the main ridge of the Palinuro cape, and to the minor ridges of the small hills at the foot of the slope. As the resolution increases, the scenarios in process domains undergo significant changes, leading to a more complex outcome. The drainage system is much more detailed and minor channels can be detected, highlighting processes transitioning from concentrated and erosive, as in the channel heads, to erosive/depositional processes proper of channel courses. Also, areas affected by diffusive processes on the main hillslope are detected between the channels. Results from the 20 cm dataset add complexity to the surficial processes dynamics. Minor channels are put into evidence, and the interference of anthropogenic modifications are evident. Here, interruptions in the runout process caused by paths can be seen, and the deviation of the surficial flow along the same paths has also been detected (Figure 12). However, besides the differences, all the processes analyses conducted on the different datasets reveal the presence of different channels having erosive and depositional potential with different levels of detail at a different scale of analysis.
The statistical analysis of the datasets revealed a contradictory output compared to the spatial mapping of landforms and process domains. FR outputs highlight increasing relations between processes and landforms as the resolution of the dataset increases. Convex landforms, characterizing ridges, saddles, spurs, crests, shoulders, etc., are related to diffusive aerial processes, resulting in FR > 1 in all datasets, with some exception in the 5 m dataset (landform 2 and 4), while concave landforms, describing channels, foot slopes, valleys, etc., are related (FR > 1 to >>1) to channeled processes, both erosive and depositional. From the achieved results, it could be argued that increasing the resolution of the dataset results in a better correlation between landforms and process domains.
However, such a relation is not so evident if we consider process domains as predictive factors for the landforms recognition. The application of the ROC/AUC methodology on the relation between surficial processes and concave morphologies resulted in a decreasing correlation with increasing the resolution of the dataset, ranging from moderate to low correlations using the 5 m dataset, to low, or almost casual correlations using the finer datasets. Such results could be explained if we consider the data population. Concave landforms, such as landforms 12 and 13, are represented by few cells in the 5 m dataset (832 + 20 cells), while their number increases in finer datasets, such as in the 1 m dataset (16,996 + 1981 cells) and in the 20 cm dataset (315,233 + 80,640 cells). The finer the resolution, the higher the detail of the classified landforms, resulting in the interposition of grid cells representing convex or flat morphologies within channel areas, apparently contradicting the outcome of process domains’ analysis. Working with a very high-resolution dataset could be misleading in such an analysis, and results from automatic landforms’ classification should be carefully contextualized within the geomorphological context of the studied domain. For instance, on the 20 cm dataset, a convex morphology within a channel could be related to the presence of boulders, which are totally explainable in channels within rocky slopes, but comparing it with a channelized process domain gives a negative outcome if we use the channelized process as a classification factor for concave morphologies. In the same way, the channel’s flanks could be classified as flat landforms at high resolutions, while being inside a channelized process domain.

5. Conclusions

Automatic landforms’ recognition and process domain analysis have been conducted by applying grid-based techniques on different DTMs on the study area of the pine grove of the Palinuro cape, in the natural protected area of the Cilento, Vallo di Diano and Alburni National Park. The datasets have different resolutions and are derived from different sources; a 5 m resolution photogrammetric dataset has been compared with data coming from LiDAR surveys, both airborne and UAV based, having 1 m and 20 cm resolution.
For the identification of elementary landforms, an original classification procedure based on DEV values, instead of the more common DIFF, has been applied. From a spatial mapping perspective, the coarse resolution DTM was useful to define landforms at a lower scale of detail, highlighting the main features of the relief, while failing in the recognition of landforms at a detailed scale. Analyses carried out on the 20 cm dataset highlighted that increasing the resolution of the dataset provides much more detailed landforms, but also introduced noise that could be related to filtering procedures, or geomorphological features, which should be carefully addressed through field analyses and surveys.
Regarding the filtering of LiDAR data from UAV, the process to remove points that do not belong to the terrain, which precedes the generation of Digital Terrain Models (DTMs), is crucial. The choice of parameters significantly influences the resulting terrain model and, consequently, the analyses performed on it. The parameters used in the MCC Filter (λ = 0.55 and t = 0.095) differ slightly from the default parameters recommended in the literature. This variation is due to the irregular terrain morphology and the presence of maquis vegetation on steep slopes, which further complicates the filtering process.
In heavily vegetated areas, the DTM derived from LiDAR data is more accurate compared to the one generated from photogrammetric models. This accuracy difference arises because, in areas with bare ground adjacent to tall vegetation, the reliability of photogrammetric cloud points is lower than that of LiDAR cloud points. However, in areas without vegetation, the DTM from photogrammetry can be equally accurate. Therefore, the choice of data source is crucial, as it directly impacts subsequent analyses. Overall, LiDAR data tend to perform better, especially in vegetated areas.
The results also highlight a general relationship between the types of landforms and surficial process domains, which usually increase with higher resolution datasets, but, at the same time, process domains cannot be exploited as a standalone classification parameter for landforms analysis, especially considering very high-resolution datasets (grid size < 1 m), as fine, detailed landforms could be contradictory regarding the general type of process domain acting in a specific section of the relief.
The temporal offset between the datasets is almost a decade between each other, spanning over a 20-year period considering all the datasets (2004, 2010–2013, and 2023). A direct comparison of the three datasets may be misleading due to varying 3D resolutions and the dense vegetation in the area, leading to datasets that have undergone filtering operations and interpolations, as previously mentioned. Overall, from a geomorphological perspective, also due to the vegetation coverage, erosive rates are not to be expected to be very high. Channels in the study area are mostly ephemeral, having a modest discharge only during the wet season, and diffusive erosion is prevented by a dense underwood vegetation (Figure 11). Major erosive processes are represented by paroxysmal flows, such as the ones of October 2022, producing, at most, an incision along the main channels and the mobilization of metric to sub-metric limestone blocks from the underlying bedrock. Such episodes cannot be addressed in our current datasets because of both resolution (if we consider the 5 m DTM), and the processing techniques (1 m and 20 cm DTMs).

Author Contributions

Conceptualization, M.V., A.D.B. and A.A.; methodology, M.V. and A.D.B.; software, M.V. and A.D.B.; validation, M.V. and A.A.; formal analysis, M.V. and A.D.B.; investigation, M.V. and A.D.B.; data curation, M.V. and A.D.B.; writing—original draft preparation, M.V. and A.D.B.; writing—review and editing, M.V., A.D.B. and A.A.; supervision, A.A.; project administration, M.V. and A.A.; and funding acquisition, M.V. All authors have read and agreed to the published version of the manuscript.

Funding

The UAV survey and fieldwork are supported by C.U.G.RI. (Inter-University Research Center for the Prediction and Prevention of Major Hazards)—University of Salerno, in the framework of the research agreement with the Cilento, Vallo di Diano and Alburni National Park Authority.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors on request.

Acknowledgments

We would like to thank C.U.G.RI. from the Inter-University Research Center for the Prediction and Prevention of Major Hazards and its Director Domenico Guida, for sharing the LiDAR data from UAV and financial support. The authors would like to also thank the Cilento, Vallo di Diano and Alburni National Park Authority for providing administrative and logistic support for the field work in a natural protected area such as the pine grove of Palinuro cape.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. MacMillan, R.A.; MacNabb, D.H.; Jones, R.K. Automated Landform Classification Using DEMs: A Conceptual Framework for a Multi-Level, Hierarchy of Hydrologically and Geomorphologicaly Oriented Physiographic Mapping Units Background Situation Analysis. In Proceedings of the 4th International Conference on Integrating GIS and Environmental Modeling (GIS/EM4): Problems, Prospects and Research Needs, Banff, AB, Canada, 2–8 September 2000. [Google Scholar]
  2. Weiss, A.D. Topographic Position and Landforms Analysis. In Proceedings of the Poster Presentation, ESRI User Conference, San Diego, CA, USA, 9–13 July 2001. [Google Scholar]
  3. Minár, J.; Evans, I.S. Elementary Forms for Land Surface Segmentation: The Theoretical Basis of Terrain Analysis and Geomorphological Mapping. Geomorphology 2008, 95, 236–259. [Google Scholar] [CrossRef]
  4. De Reu, J.; Bourgeois, J.; Bats, M.; Zwertvaegher, A.; Gelorini, V.; De Smedt, P.; Chu, W.; Antrop, M.; De Maeyer, P.; Finke, P.; et al. Application of the Topographic Position Index to Heterogeneous Landscapes. Geomorphology 2013, 186, 39–49. [Google Scholar] [CrossRef]
  5. Lin, C.W.; Tseng, C.M.; Tseng, Y.H.; Fei, L.Y.; Hsieh, Y.C.; Tarolli, P. Recognition of Large Scale Deep-Seated Landslides in Forest Areas of Taiwan Using High Resolution Topography. J. Asian Earth Sci. 2013, 62, 389–400. [Google Scholar] [CrossRef]
  6. Samodra, G.; Chen, G.; Sartohadi, J.; Hadmoko, D.S.; Kasama, K. Automated Landform Classification in a Rockfall-Prone Area, Gunung Kelir, Java. Earth Surf. Dyn. 2014, 2, 339–348. [Google Scholar] [CrossRef]
  7. Vannametee, E.; Babel, L.V.; Hendriks, M.R.; Schuur, J.; de Jong, S.M.; Bierkens, M.F.P.; Karssenberg, D. Semi-Automated Mapping of Landforms Using Multiple Point Geostatistics. Geomorphology 2014, 221, 298–319. [Google Scholar] [CrossRef]
  8. Gruber, F.E.; Baruck, J.; Geitner, C. Algorithms vs. Surveyors: A Comparison of Automated Landform Delineations and Surveyed Topographic Positions from Soil Mapping in an Alpine Environment. Geoderma 2017, 308, 9–25. [Google Scholar] [CrossRef]
  9. Luo, W.; Liu, C.-C. Innovative Landslide Susceptibility Mapping Supported by Geomorphon and Geographical Detector Methods. Landslides 2018, 15, 465–474. [Google Scholar] [CrossRef]
  10. Dramis, F.; Guida, D.; Valiante, M. Object-Oriented Mapping as a Tool for the Assessment of Landslide Hazard in Higly Urbanized Areas. Front. Earth Sci. 2022, 10, 834078. [Google Scholar] [CrossRef]
  11. Drǎguţ, L.; Blaschke, T. Automated Classification of Landform Elements Using Object-Based Image Analysis. Geomorphology 2006, 81, 330–344. [Google Scholar] [CrossRef]
  12. van Asselen, S.; Seijmonsbergen, A.C. Expert-Driven Semi-Automated Geomorphological Mapping for a Mountainous Area Using a Laser DTM. Geomorphology 2006, 78, 309–320. [Google Scholar] [CrossRef]
  13. Dramis, F.; Guida, D.; Cestari, A. Nature and Aims of Geomorphological Mapping. In Geomorphological Mapping: Methods and Applications; Smith, M.J., Paron, P., Griffiths, J.S., Eds.; Elsevier: Amsterdam, The Netherlands, 2011; pp. 39–74. [Google Scholar]
  14. Seijmonsbergen, A.C.; Hengl, T.; Anders, N.S. Semi-Automated Identification and Extraction of Geomorphological Features Using Digital Elevetion Data. In Geomorphological Mapping: Methods and Applications; Smith, M.J., Paron, P., Griffiths, J.S., Eds.; Elsevier: Amsterdam, The Netherlands, 2011; pp. 297–335. [Google Scholar]
  15. Guida, D.; Siervo, V.; Cestari, A.; Palmieri, V. Application of “GmIS_Unisa” Geomorphological Mapping System to Regional Planning; Rendiconti Online della Società Geologica Italiana: Roma, Italy, 2012; Volume 21, pp. 1155–1157. [Google Scholar]
  16. Guida, D.; Cestari, A.; Cuomo, A.; Dramis, F.; Palmieri, V.; Siervo, V. The Salerno University Geomorphological Informative Mapping System: The Licosa Polygenetic Case Study (Cilento European Geopark, Southern Italy). In Proceedings of Geomorphometry 2015, Poznań, Poland, 22–26 June 2015; Zwoliński, Z., Jasiewicz, J., Mitasova, H., Hengl, T., Eds.; Geomorphometry for Geosciences, Adam Mickiewicz University in Poznan, Institute of Geoecology and Geoinformation, International Society for Geomorphometry: Poznań, Poland, 2015; pp. 53–56. [Google Scholar]
  17. Guida, D.; Cuomo, A.; Palmieri, V. Using Object-Based Geomorphometry for Hydro-Geomorphological Analysis in a Mediterranean Research Catchment. Hydrol. Earth Syst. Sci. 2016, 20, 3493–3509. [Google Scholar] [CrossRef]
  18. Louw, G.; van Niekerk, A. Object-Based Land Surface Segmentation Scale Optimisation: An Ill-Structured Problem. Geomorphology 2019, 327, 377–384. [Google Scholar] [CrossRef]
  19. Mezaal, M.R.; Pradhan, B. An Improved Algorithm for Identifying Shallow and Deep-Seated Landslides in Dense Tropical Forest from Airborne Laser Scanning Data. CATENA 2018, 167, 147–159. [Google Scholar] [CrossRef]
  20. Comert, R.; Avdan, U.; Gorum, T. Rapid Mapping of Forested Landslide from Ultra-High Resolution Unmanned Aerial Vehicle Data. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2018, 42, 171–176. [Google Scholar] [CrossRef]
  21. Yang, Y.; Song, S.; Yue, F.; He, W.; Shao, W.; Zhao, K.; Nie, W. Superpixel-Based Automatic Image Recognition for Landslide Deformation Areas. Eng. Geol. 2019, 259, 105166. [Google Scholar] [CrossRef]
  22. Hölbling, D.; Spiekermann, R.; Binn, A.; Betts, H. Analysing and Visualizing Spatio-Temporal Landslide Patterns. Abstr. ICA 2019, 1, 116. [Google Scholar] [CrossRef]
  23. Dabiri, Z.; Hölbling, D.; Abad, L.; Prasicek, G.; Argentin, A.-L.; Tsai, T.T. An Object-Based Approach for Monitoring the Evolution of Landslide-Dammed Lakes and Detecting Triggering Landslides in Taiwan. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2019, XLII, 103–108. [Google Scholar] [CrossRef]
  24. Rajbhandari, S.; Aryal, J.; Osborn, J.; Musk, R.; Lucieer, A. Benchmarking the Applicability of Ontology in Geographic Object-Based Image Analysis. ISPRS Int. J. Geo-Inf. 2017, 6, 386. [Google Scholar] [CrossRef]
  25. Pradhan, B.; Jebur, M.N.; Shafri, H.Z.M.; Tehrany, M.S. Data Fusion Technique Using Wavelet Transform and Taguchi Methods for Automatic Landslide Detection from Airborne Laser Scanning Data and QuickBird Satellite Imagery. IEEE Trans. Geosci. Remote Sens. 2016, 54, 1610–1622. [Google Scholar] [CrossRef]
  26. Shi, W.; Zhang, M.; Ke, H.; Fang, X.; Zhan, Z.; Chen, S. Landslide Recognition by Deep Convolutional Neural Network and Change Detection. IEEE Trans. Geosci. Remote Sens. 2021, 59, 4654–4672. [Google Scholar] [CrossRef]
  27. Okyay, U.; Telling, J.; Glennie, C.L.; Dietrich, W.E. Airborne Lidar Change Detection: An Overview of Earth Sciences Applications. Earth-Sci. Rev. 2019, 198, 102929. [Google Scholar] [CrossRef]
  28. Kariminejad, N.; Hosseinalizadeh, M.; Pourghasemi, H.R.; Tiefenbacher, J.P. Change Detection in Piping, Gully Head Forms, and Mechanisms. CATENA 2021, 206, 105550. [Google Scholar] [CrossRef]
  29. Yang, S.; Wang, Y.; Wang, P.; Mu, J.; Jiao, S.; Zhao, X.; Wang, Z.; Wang, K.; Zhu, Y. Automatic Identification of Landslides Based on Deep Learning. Appl. Sci. 2022, 12, 8153. [Google Scholar] [CrossRef]
  30. Qi, W.; Wei, M.; Yang, W.; Xu, C.; Ma, C. Automatic Mapping of Landslides by the ResU-Net. Remote Sens. 2020, 12, 2487. [Google Scholar] [CrossRef]
  31. Carrara, A.; Cardinali, M.; Detti, R.; Guzzetti, F.; Pasqui, V.; Reichenbach, P. GIS Techniques and Statistical Models in Evaluating Landslide Hazard. Earth Surf. Process. Landf. 1991, 16, 427–445. [Google Scholar] [CrossRef]
  32. Aleotti, P.; Chowdhury, R. Landslide Hazard Assessment: Summary Review and New Perspectives. Bull. Eng. Geol. Environ. 1999, 58, 21–44. [Google Scholar] [CrossRef]
  33. Chung-Jo, F.; Fabbri, A.G. Probabilistic Prediction Models for Landslide Hazard Mapping. Photogramm. Eng. Remote Sens. 1999, 65, 1389–1399. [Google Scholar]
  34. Guzzetti, F.; Carrara, A.; Cardinali, M.; Reichenbach, P. Landslide Hazard Evaluation: A Review of Current Techniques and Their Application in a Multi-Scale Study, Central Italy. Geomorphology 1999, 31, 181–216. [Google Scholar] [CrossRef]
  35. Dai, F.C.; Lee, C.F. Landslide Characteristics and Slope Instability Modeling Using GIS, Lantau Island, Hong Kong. Geomorphology 2002, 42, 213–228. [Google Scholar] [CrossRef]
  36. Ayalew, L.; Yamagishi, H. The Application of GIS-Based Logistic Regression for Landslide Susceptibility Mapping in the Kakuda-Yahiko Mountains, Central Japan. Geomorphology 2005, 65, 15–31. [Google Scholar] [CrossRef]
  37. Conoscenti, C.; Di Maggio, C.; Rotigliano, E. GIS Analysis to Assess Landslide Susceptibility in a Fluvial Basin of NW Sicily (Italy). Geomorphology 2008, 94, 325–339. [Google Scholar] [CrossRef]
  38. Piacentini, D.; Troiani, F.; Soldati, M.; Notarnicola, C.; Savelli, D.; Schneiderbauer, S.; Strada, C. Statistical Analysis for Assessing Shallow-Landslide Susceptibility in South Tyrol (South-Eastern Alps, Italy). Geomorphology 2012, 151–152, 196–206. [Google Scholar] [CrossRef]
  39. Calvello, M.; Peduto, D.; Arena, L. Combined Use of Statistical and DInSAR Data Analyses to Define the State of Activity of Slow-Moving Landslides. Landslides 2017, 14, 473–489. [Google Scholar] [CrossRef]
  40. Delchiaro, M.; Rouhi, J.; Valiante, M.; Della Seta, M.; Esposito, C.; Martino, S. Bivariate Landslide Susceptibility Analysis in the Lorestan Arc (Zagros Mountains, Iran). Ital. J. Eng. Geol. Environ. 2021, 1, 53–66. [Google Scholar] [CrossRef]
  41. Berti, M.; Corsini, A.; Daehne, A. Comparative Analysis of Surface Roughness Algorithms for the Identification of Active Landslides. Geomorphology 2013, 182, 1–18. [Google Scholar] [CrossRef]
  42. Merghadi, A.; Yunus, A.P.; Dou, J.; Whiteley, J.; ThaiPham, B.; Bui, D.T.; Avtar, R.; Abderrahmane, B. Machine Learning Methods for Landslide Susceptibility Studies: A Comparative Overview of Algorithm Performance. Earth-Sci. Rev. 2020, 207, 103225. [Google Scholar] [CrossRef]
  43. Martinello, C.; Delchiaro, M.; Iacobucci, G.; Cappadonia, C.; Rotigliano, E.; Piacentini, D. Exploring the Geomorphological Adequacy of the Landslide Susceptibility Maps: A Test for Different Types of Landslides in the Bidente River Basin (Northern Italy). CATENA 2024, 238, 107835. [Google Scholar] [CrossRef]
  44. Delchiaro, M.; Fioramonti, V.; Della Seta, M.; Cavinato, G.P.; Mattei, M. Fluvial Inverse Modeling for Inferring the Timing of Quaternary Uplift in the Simbruini Range (Central Apennines, Italy). Trans. GIS 2021, 25, 2455–2480. [Google Scholar] [CrossRef]
  45. Expósito, I.; Jiménez-Bonilla, A.; Delchiaro, M.; Conde, J.L.; Balanyá, J.; Moral-Martos, F.; Della Seta, M. Geomorphic Signature of Segmented Relief Rejuvenation in the Sierra Morena, Betic Forebulge, Spain. Earth Surf. Dyn. 2022, 10, 1017–1039. [Google Scholar] [CrossRef]
  46. Delchiaro, M.; Iacobucci, G.; Della Seta, M.; Natacha, G.; Piacentini, D.; Ruscitto, V.; Zocchi, M.; Troiani, F. A Fluvial Record of Late Quaternary Climate Changes and Tectonic Uplift along the Marche Piedmont Zone of the Apennines: New Insights from the Tesino River (Italy). Geomorphology 2023, 445, 108971. [Google Scholar] [CrossRef]
  47. Gioia, D.; Corrado, G.; Minervino Amodio, A.; Schiattarella, M. Uplift Rate Calculation Based on the Comparison between Marine Terrace Data and River Profile Analysis: A Morphotectonic Insight from the Ionian Coastal Belt of Basilicata, Italy. Geomorphology 2024, 447, 109030. [Google Scholar] [CrossRef]
  48. Dietrich, W.E.; Bellugi, D.G.; Heimsath, A.M.; Roering, J.J.; Sklar, L.S.; Stock, J.D. Geomorphic Transport Laws for Predicting Landscape Form and Dynamics. Geophys. Monogr. 2003, 135, 103–132. [Google Scholar] [CrossRef]
  49. Booth, A.M.; Roering, J.J.; Rempel, A.W. Topographic Signatures and a General Transport Law for Deep-Seated Landslides in a Landscape Evolution Model. J. Geophys. Res. Earth Surf. 2013, 118, 603–624. [Google Scholar] [CrossRef]
  50. Tseng, C.M.; Lin, C.W.; Dalla Fontana, G.; Tarolli, P. The Topographic Signature of a Major Typhoon. Earth Surf. Process. Landf. 2015, 40, 1129–1136. [Google Scholar] [CrossRef]
  51. Vergari, F.; Troiani, F.; Faulkner, H.; Del Monte, M.; Della Seta, M.; Ciccacci, S.; Fredi, P. The Use of the Slope–Area Function to Analyse Process Domains in Complex Badland Landscapes. Earth Surf. Process. Landf. 2019, 44, 273–286. [Google Scholar] [CrossRef]
  52. Barbarella, M.; Di Benedetto, A.; Fiani, M.; Guida, D.; Lugli, A. Use of DEMs Derived from TLS and HRSI Data for Landslide Feature Recognition. ISPRS Int. J. Geo-Inf. 2018, 7, 160. [Google Scholar] [CrossRef]
  53. Pirotti, F.; Guarnieri, A.; Vettore, A. State of the Art of Ground and Aerial Laser Scanning Technologies for High-Resolution Topography of the Earth Surface. Eur. J. Remote Sens. 2013, 46, 66–78. [Google Scholar] [CrossRef]
  54. Guilbert, E. Surface Network Extraction from High Resolution Digital Terrain Models. J. Spat. Inf. Sci. 2021, 22, 33–59. [Google Scholar] [CrossRef]
  55. Esposito, G.; Salvini, R.; Matano, F.; Sacchi, M.; Danzi, M.; Somma, R.; Troise, C. Multitemporal Monitoring of a Coastal Landslide through SFM-Derived Point Cloud Comparison. Photogramm. Rec. 2017, 32, 459–479. [Google Scholar] [CrossRef]
  56. Ambrosino, A.; Di Benedetto, A.; Fiani, M. LiDAR Data and HRSI to Evaluate the Mitigating Effect of Forests into Rockfall Risk Analysis Using SOM: Mt San Liberatore Case Study. Remote Sens. 2023, 15, 4523. [Google Scholar] [CrossRef]
  57. Di Benedetto, A.; Fiani, M.; Petti, L.; Repetto, E. Road Surface Modelling and Characterization from Terrestrial Lidar Data. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2023, 48, 113–120. [Google Scholar] [CrossRef]
  58. Barbarella, M.; Cuomo, A.; Di Benedetto, A.; Fiani, M.; Guida, D. Topographic Base Maps from Remote Sensing Data for Engineering Geomorphological Modelling: An Application on Coastal Mediterranean Landscape. Geosciences 2019, 9, 500. [Google Scholar] [CrossRef]
  59. Kamp, N.; Krenn, P.; Avian, M.; Sass, O. Comparability of Multi-Temporal DTMs Derived from Different LiDAR Platforms: Error Sources and Uncertainties in the Application of Geomorphic Impact Studies. Earth Surf. Process. Landf. 2023, 48, 1152–1175. [Google Scholar] [CrossRef] [PubMed]
  60. Hungr, O.; Leroueil, S.; Picarelli, L. The Varnes Classification of Landslide Types, an Update. Landslides 2014, 11, 167–194. [Google Scholar] [CrossRef]
  61. Valiante, M.; Guida, D.; Della Seta, M.; Bozzano, F. Spatiotemporal Data Management of Recurrent Debris Flow Events Using Object-Oriented Data Modelling. Ital. J. Eng. Geol. Environ. 2021, 1, 239–248. [Google Scholar] [CrossRef]
  62. Di Benedetto, A.; Fiani, M. Integration of LiDAR Data into a Regional Topographic Database for the Generation of a 3D City Model. In Geomatics for Green and Digital Transition, Proceedings of the 25th Italian Conference, ASITA 2022, Genova, Italy, 20–24 June 2022; Borgogno-Mondino, E., Zamperlin, P., Eds.; Springer International Publishing: Cham, Switzerland, 2022; pp. 193–208. [Google Scholar]
  63. Famiglietti, N.A.; Cecere, G.; Grasso, C.; Memmolo, A.; Vicari, A. A Test on the Potential of a Low Cost Unmanned Aerial Vehicle RTK/PPK Solution for Precision Positioning. Sensors 2021, 21, 3882. [Google Scholar] [CrossRef] [PubMed]
  64. Evans, J.S.; Hudak, A.T. A Multiscale Curvature Algorithm for Classifying Discrete Return LiDAR in Forested Environments. IEEE Trans. Geosci. Remote Sens. 2007, 45, 1029–1038. [Google Scholar] [CrossRef]
  65. Gallant, J.C.; Wilson, J.P. Primary Topographic Attributes. In Terrain Analysis: Principles and Applications; Wilson, J.P., Gallant, J.C., Eds.; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 2000; pp. 51–85. ISBN 0-471-32188-5. [Google Scholar]
  66. Tagil, S.; Jenness, J. GIS-Based Automated Landform Classification and Topographic, Landcover and Geologic Attributes of Landforms around the Yazoren Polje, Turkey. J. Appl. Sci. 2008, 8, 910–921. [Google Scholar] [CrossRef]
  67. Montgomery, D.R.; Foufoula-Georgiou, E. Channel Network Source Representation Using Digital Elevation Models. Water Resour. Res. 1993, 29, 3925–3934. [Google Scholar] [CrossRef]
  68. Qin, C.; Zhu, A.-X.; Pei, T.; Li, B.; Zhou, C.; Yang, L. An Adaptive Approach to Selecting a Flow-partition Exponent for a Multiple-flow-direction Algorithm. Int. J. Geogr. Inf. Sci. 2007, 21, 443–458. [Google Scholar] [CrossRef]
  69. Lee, S.; Talib, J.A. Probabilistic Landslide Susceptibility and Factor Effect Analysis. Environ. Geol. 2005, 47, 982–990. [Google Scholar] [CrossRef]
  70. Mondal, S.; Maiti, R. Integrating the Analytical Hierarchy Process (AHP) and the Frequency Ratio (FR) Model in Landslide Susceptibility Mapping of Shiv-Khola Watershed, Darjeeling Himalaya. Int. J. Disaster Risk Sci. 2013, 4, 200–212. [Google Scholar] [CrossRef]
  71. Abdo, H.G.; Almohamad, H.; Al Dughairi, A.A.; Al-Mutiry, M. GIS-Based Frequency Ratio and Analytic Hierarchy Process for Forest Fire Susceptibility Mapping in the Western Region of Syria. Sustainability 2022, 14, 4668. [Google Scholar] [CrossRef]
  72. Hoa, P.V.; Tuan, N.Q.; Hong, P.V.; Thao, G.T.P.; Binh, N.A. GIS-Based Modeling of Landslide Susceptibility Zonation by Integrating the Frequency Ratio and Objective–Subjective Weighting Approach: A Case Study in a Tropical Monsoon Climate Region. Front. Environ. Sci. 2023, 11, 1175567. [Google Scholar] [CrossRef]
Remotesensing 16 02771 g001
Figure 1. Geological map of the study area (excerpt from the Italian national geological map (Progetto CARG) (CRS EPSG: 7792)), the red box in the inset shows location in Southern Italy.
Figure 1. Geological map of the study area (excerpt from the Italian national geological map (Progetto CARG) (CRS EPSG: 7792)), the red box in the inset shows location in Southern Italy.
Remotesensing 16 02771 g001
Remotesensing 16 02771 g002
Figure 2. Study area: (a) 5 m DTM provided by the regional authority and (b) 1 m DTM derived from LiDAR data of the Ministry. Watersheds of interest are highlighted. (CRS EPSG: 7792).
Figure 2. Study area: (a) 5 m DTM provided by the regional authority and (b) 1 m DTM derived from LiDAR data of the Ministry. Watersheds of interest are highlighted. (CRS EPSG: 7792).
Remotesensing 16 02771 g002
Remotesensing 16 02771 g003
Figure 3. Results of the LiDAR survey: (a) top view global point cloud, in red the excerpt shown in panel (b), and (b) density map (points/m2) of the point cloud excerpt.
Figure 3. Results of the LiDAR survey: (a) top view global point cloud, in red the excerpt shown in panel (b), and (b) density map (points/m2) of the point cloud excerpt.
Remotesensing 16 02771 g003
Remotesensing 16 02771 g004
Figure 4. Results of the LiDAR survey: DTM in shaded relief with contour lines with an equidistance of 1 m (CRS EPSG: 7792).
Figure 4. Results of the LiDAR survey: DTM in shaded relief with contour lines with an equidistance of 1 m (CRS EPSG: 7792).
Remotesensing 16 02771 g004
Remotesensing 16 02771 g005
Figure 5. DEV workflow for landforms recognition (std = standard deviation).
Figure 5. DEV workflow for landforms recognition (std = standard deviation).
Remotesensing 16 02771 g005
Remotesensing 16 02771 g006
Figure 6. Results of the landforms analysis and maps: (a) 5 m DTM; (b) 1 m DTM; (c) 20 cm DTM; and (d) the spatial distribution of the landforms (CRS EPSG: 7792).
Figure 6. Results of the landforms analysis and maps: (a) 5 m DTM; (b) 1 m DTM; (c) 20 cm DTM; and (d) the spatial distribution of the landforms (CRS EPSG: 7792).
Remotesensing 16 02771 g006
Remotesensing 16 02771 g007
Figure 7. Slope–Area plots of the analyzed watershed with process domain thresholds for each dataset. Cell values of the 20 cm dataset are not represented for clarity.
Figure 7. Slope–Area plots of the analyzed watershed with process domain thresholds for each dataset. Cell values of the 20 cm dataset are not represented for clarity.
Remotesensing 16 02771 g007
Remotesensing 16 02771 g008
Figure 8. Results and maps of the process domains analysis: (a) 5 m DTM; (b) 1 m DTM; and (c) 20 cm DTM (CRS EPSG: 7792).
Figure 8. Results and maps of the process domains analysis: (a) 5 m DTM; (b) 1 m DTM; and (c) 20 cm DTM (CRS EPSG: 7792).
Remotesensing 16 02771 g008
Remotesensing 16 02771 g009
Figure 9. Frequency Ratio plots for landforms vs. processes (a,c,e), and for processes vs. landforms (b,d,f) for each dataset.
Figure 9. Frequency Ratio plots for landforms vs. processes (a,c,e), and for processes vs. landforms (b,d,f) for each dataset.
Remotesensing 16 02771 g009
Remotesensing 16 02771 g010
Figure 10. ROC/AUC curves for concave landforms for each dataset: (a) DTM 5 m; (b) DTM 1 m, and (c) DTM 20 cm.
Figure 10. ROC/AUC curves for concave landforms for each dataset: (a) DTM 5 m; (b) DTM 1 m, and (c) DTM 20 cm.
Remotesensing 16 02771 g010
Remotesensing 16 02771 g011
Figure 11. Some examples of channeled landforms from the study area (a–e) and the respective results in terms of landforms and processes for the different datasets.
Figure 11. Some examples of channeled landforms from the study area (a–e) and the respective results in terms of landforms and processes for the different datasets.
Remotesensing 16 02771 g011
Remotesensing 16 02771 g012
Figure 12. Example of interference between anthropogenic modifications (a path) and process domains: (a) 5 m DTM, (b) 1 m DTM, and (c) 20 cm DTM (CRS EPSG: 7792).
Figure 12. Example of interference between anthropogenic modifications (a path) and process domains: (a) 5 m DTM, (b) 1 m DTM, and (c) 20 cm DTM (CRS EPSG: 7792).
Remotesensing 16 02771 g012
Table 1. Characteristics of the LiDAR mounted on board the UAV.
Table 1. Characteristics of the LiDAR mounted on board the UAV.
ParameterValue
Ranging Accuracy3 cm @ 100 m
Multiple Return480,000 pts/s
FOV70.4° × 4.5°
Detection Range450 m @ 80% reflectivity
Table 2. Process domains summary.
Table 2. Process domains summary.
DomainFlow TypeProcess TypeMorphotype
IaDiffusiveAerial erosionRidges
IbDiffusiveAerial erosionShoulders and backslopes
IIConcentratedLinear erosionHollows and headwaters
IIIaConcentrated to channelizedLinear erosionChannel heads
IIIbChannelizedLinear erosion—depositionUpslope channels and streams
IVChannelizedMainly depositionFluvial channels
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Valiante, M.; Di Benedetto, A.; Aloia, A. A Comparison of Landforms and Processes Detection Using Multisource Remote Sensing Data: The Case Study of the Palinuro Pine Grove (Cilento, Vallo di Diano and Alburni National Park, Southern Italy). Remote Sens. 2024, 16, 2771. https://doi.org/10.3390/rs16152771

AMA Style

Valiante M, Di Benedetto A, Aloia A. A Comparison of Landforms and Processes Detection Using Multisource Remote Sensing Data: The Case Study of the Palinuro Pine Grove (Cilento, Vallo di Diano and Alburni National Park, Southern Italy). Remote Sensing. 2024; 16(15):2771. https://doi.org/10.3390/rs16152771

Chicago/Turabian Style

Valiante, Mario, Alessandro Di Benedetto, and Aniello Aloia. 2024. "A Comparison of Landforms and Processes Detection Using Multisource Remote Sensing Data: The Case Study of the Palinuro Pine Grove (Cilento, Vallo di Diano and Alburni National Park, Southern Italy)" Remote Sensing 16, no. 15: 2771. https://doi.org/10.3390/rs16152771

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop