Sea Caves and Other Landforms of the Coastal Scenery on Gozo Island (Malta): Inventory and New Data on Their Formation

: Sea caves are a type of cave formed primarily by the wave action of the sea. The coastal scenery of the Gozitan coast is very interesting in that sea caves and other coastal landforms, such as sea arches, develop at the sea level. We mapped seventy-nine semi-submerged sea caves opening at the sea level, ﬁve completely submerged sea caves, seven sea arches, one sea stack, and one shelter around the coast of Gozo, mainly in the Western and Eastern parts of the island, due to favorable lithological and topographical conditions. Additionally, we surveyed the topography of the emerged part of nine sea caves using the iPhone build-in LiDAR sensor, and eight sea caves in the submerged part using SCUBA equipment. This inventory represents the most detailed example of a database of coastal caves and related forms in the Mediterranean, mainly sourced from a swimming survey along the entire island. Thanks to the combination of outputs of the above-water emerged and submerged surveys, we deﬁned three types of semi-submerged sea caves: (i) box caves, (ii) joint caves, and (iii) complex caves. Moreover, we added a cave-like landform above the sea level on calcarenites called shelter, or a little extended notch deeply carved into the cliff. The shape mainly depends on the structural and lithological setting of sea cliffs. In the Western sector of the island, we also discovered the only sea cave in Gozo, measuring 122 m in length and 10 m in width, with its ﬂoor developing above the mean sea level. This cave base is of interest due to rounded landforms related to marine erosion. In the innermost part of the cave, there is also a beach with rounded pebble at an elevation of about 7 m asl. Considering the tectonic stability of the island, it could be possibly related to the MIS 5.5 highstand.


Introduction
Sea caves are very common landforms in areas where there is a steep sea cliff. Wave action weakens the rock masses along joints and faults, enlarging them to form hollows and sea caves [1]. Sea caves occur along all the limestone coasts of the Mediterranean Sea, and as such, various aspects, including biological, geomorphological, and exploration, have been studied. Research results are mainly confined to local papers or proceedings of karst conferences. The volume [2] on Italian marine caves and [3] on Tyrrhenian sea caves discussed and took stock of the current state of knowledge regarding these types of caves. Cave deposits, such as speleothems and other cave deposits, have successfully been used as sea level markers, also in the Mediterranean Sea [4,5].  temperatures range from 12 • C to 27 • C. Due to the central position of the Maltese Islands in the Mediterranean Sea, they are affected by long fetches mainly in the second and fourth quadrant [37]. Dominant winds come from NW [37]. The wave climate around the Maltese Islands is dominated by sea waves rather than swell waves: sea waves, generated by winds in the local area, are characterized by being steep and have relatively short-wave periods when compared to swell waves, which travel into the local area from elsewhere. Tidal oscillations, predominantly semi-diurnal, reach a maximum range of just 20.6 cm on average for spring tides and are reduced to 4.6 cm during neap tides [38].

Materials and Methods
The investigation of sea caves along the Gozitan coastline was carried out via swimming surveys in 2013 using the protocol of the Geoswim program [14,39,40]. Data collected were verified in a later campaign in May 2022. In particular, the first survey permitted the location of most of the sea caves presented in this study, while the second survey was aimed at the verification of the 2013 data as well as to perform underwater surveys in selected caves. Another additional survey in October 2022 was carried out in the cave named G32 (see Table 1) at Dimitri Point as sea level markers attributable to MIS 5.5 have been identified within the cave.
The minimum size of sea caves has not been precisely defined, but well-defined hollows at the sea level were mapped. Joints and fractures smaller than approximately 1 m in size were considered as discontinuities enlarged by marine processes.

Surveys of Sea Caves
Regarding the surveys of sea caves, we present data collected during the 2013 swim survey of the island of Gozo [39], as well as additional punctual surveys both of the submerged (Figure 2A,B) and emergent parts of sea caves ( Figure 2C,D).
The 2013 survey followed the Geoswim protocol described by [14,15]. Geoswim consists of swim surveys of selected long sectors of rocky coast, mainly sloping coasts and plunging cliffs [22], with detailed observations of lateral changes in coastal landforms, ecological features of the coast, and mapping of prominent geo-objects.
Instruments, such as GPS, cameras, CTD sensors, echosounder, etc., were fixed onto the instrumental-supported raft (ISR, Figure 2). Two cameras were set in waterproof housings, above and below the ISR, and allowed the collection of ongoing videos and time-lapse images of the coastline being surveyed.
Regarding the DAQ (data acquisition) system, each device was autonomous with respect to the other. The synchronization of data collected during each surveying day was carried out with respect to the Coordinate Universal Time (UTC) and the comparison with GPS coordinates [14,15].
The surveys followed a predetermined route at approximately 1 m to 5 m from the coastline. Small changes in the original routes were caused by nuances in the local topography, and due to the orientation of some of the sea caves [14,15].
Observations were generally reported via radio to the support boat and written in the field booklet, which also includes other navigational information.
The survey covered semi-submerged sea caves, but some submerged caves, noticeable from the surface, were also reported and studied. SCUBA dives were conducted to collect data at eight sea caves using open-circuit dive equipment and underwater cameras.
Morphometric data were acquired in the field using manual instruments for measurements, such as an invar rod or an acoustic distance meter, and also terrestrial photogrammetry ( Figure 3) [14,15]. Table 1. List of the sea caves surveyed in Gozo (Malta). The names have been reported when available. The location name refers to the local place names of the sea cliffs. Morphometric parameters which were measured are the following: total length (L), depth at the entrance (D), height at the entrance (H), and width at the entrance (W). All the morphometric parameters are given in meters. The codes of sea caves are reported following the sequence of survey. The table also reports the location of sea arches and sea stacks as part of the coastal scenery.

Surveys of Sea Caves
Regarding the surveys of sea caves, we present data collected during the 2013 swim survey of the island of Gozo [39], as well as additional punctual surveys both of the submerged (Figure 2A,B) and emergent parts of sea caves ( Figure 2C,D). The 2013 survey followed the Geoswim protocol described by [14,15]. Geoswim consists of swim surveys of selected long sectors of rocky coast, mainly sloping coasts and plunging cliffs [22], with detailed observations of lateral changes in coastal landforms, ecological features of the coast, and mapping of prominent geo-objects.
Instruments, such as GPS, cameras, CTD sensors, echosounder, etc., were fixed onto the instrumental-supported raft (ISR, Figure 2). Two cameras were set in waterproof housings, above and below the ISR, and allowed the collection of ongoing videos and timelapse images of the coastline being surveyed.
Regarding the DAQ (data acquisition) system, each device was autonomous with respect to the other. The synchronization of data collected during each surveying day was carried out with respect to the Coordinate Universal Time (UTC) and the comparison with GPS coordinates [14,15].
The surveys followed a predetermined route at approximately 1 m to 5 m from the coastline. Small changes in the original routes were caused by nuances in the local topography, and due to the orientation of some of the sea caves [14,15].
Observations were generally reported via radio to the support boat and written in the field booklet, which also includes other navigational information.
The survey covered semi-submerged sea caves, but some submerged caves, noticeable from the surface, were also reported and studied. SCUBA dives were conducted to collect data at eight sea caves using open-circuit dive equipment and underwater cameras.
Morphometric data were acquired in the field using manual instruments for measurements, such as an invar rod or an acoustic distance meter, and also terrestrial photogrammetry ( Figure 3) [14,15]. (C) sea cave#G33, developed on a joint but evolving for roof collapse. When available, the aforementioned models were used to measure the width and height of the cave entrances.

Terrestrial Photogrammetry
We performed photographic acquisition to model the inner part of sea cave#G32, and to reconstruct its shape. This approach was also used to obtain the morphometric parameters of the entrances of most of the caves around the island. These high-resolution (HR) images were taken at various elevations roughly perpendicular to the cave walls using a Figure 3. Example of 3D model of sea cave entrances. Sea caves at Gozo: (A) Sea cave#G74 developed along a joint parallel to the coastline; (B) sea cave#G43, developed from collapse of blocks from the roof; (C) sea cave#G33, developed on a joint but evolving for roof collapse. When available, the aforementioned models were used to measure the width and height of the cave entrances.

Terrestrial Photogrammetry
We performed photographic acquisition to model the inner part of sea cave#G32, and to reconstruct its shape. This approach was also used to obtain the morphometric parameters of the entrances of most of the caves around the island. These high-resolution (HR) images were taken at various elevations roughly perpendicular to the cave walls using a GoPro Hero 7 Black camera. A total of 1522 images were acquired and processed separately in a specialized software used to produce the dense point cloud and 3D model of the cave, according to the DP technique. The first part of DP is known as the SfM technique, and involves the importation of images of each outcrop, and camera alignment.
Collected images were imported in Agisoft Metashape TM (Agisoft LLC, St. Petersburg, Russia). The latter is a robust software package widely used for the DP technique. The above-cited software includes SfM algorithms that allow the user to align the images and calculate their relative position to each other during the camera alignment phase. The outcome of the camera alignment process was a sparse point cloud of the investigated cave. The second part of DP is known as MVS reconstruction, which involves dense point cloud production, mesh production, and the generation of a 3D textured model (Figure 4). The low-density point cloud is thickened by increasing the number of points, thus generating a dense point cloud. The final processing steps included the production of a 3D surface (mesh) and the generation of 3D textured models. cloud production, mesh production, and the generation of a 3D textured model ( Figure  4). The low-density point cloud is thickened by increasing the number of points, thus generating a dense point cloud. The final processing steps included the production of a 3D surface (mesh) and the generation of 3D textured models. In a cave environment, global navigation satellite system (GNSS) signals are blocked, so it was not possible to measure the coordinates of the markers to scale the 3D model. Additionally, not having a total station available, it was decided to scale the 3D model within the photogrammetric software by inserting 50 cm scalebars on each side of each of the ten markers positioned along the cave. Moreover, multiple measurements in situ and further checks of the accuracy of the 3D model were made using the point cloud generated via mobile phone (MP) LiDAR as a reference [40].  In a cave environment, global navigation satellite system (GNSS) signals are blocked, so it was not possible to measure the coordinates of the markers to scale the 3D model. Additionally, not having a total station available, it was decided to scale the 3D model within the photogrammetric software by inserting 50 cm scalebars on each side of each of the ten markers positioned along the cave. Moreover, multiple measurements in situ and further checks of the accuracy of the 3D model were made using the point cloud generated via mobile phone (MP) LiDAR as a reference [40].

iPhone LiDAR Survey
The MP survey aimed to build the 3D model of the innermost part of the G32 cave. We performed two MP surveys using an iPhone 12 Pro Max TM (Apple Inc., Cupertino, CA, USA) in May and October 2022. These HR images were captured at various elevations roughly perpendicular to the sea cliff. The iPhone 12 Pro Max TM weighs 0.226 kg, measuring 160.8 × 78.1 × 7.4 mm 3 , and it is equipped with three rear cameras, including a 12 MP wide-angle, f/1.6, 12 MP ultra-wide-angle, f/2.4, and 12 MP telephoto, f/2.0 with 2/2.5× optical and 10× digital zoom capacity. The 12 MP wide-angle camera was used to capture the images. Our acquisition was conducted within the overhead environment extant in caves, resulting in a very inaccurate GPS signal. The iPhone 12 Pro Max TM uses a fast stabilization procedure that provides local coordinates within a few tens of seconds. Tavani et al. [41] and Corradetti et al. [42] observed that the accuracy of this model is lower than about 10 m. The acquisition was carried out in automatic mode.
In the cave#G32 (Table 1), we used the MP mounted on a 1.7 m-long selfie stick to extend the LiDAR scanning range that is limited to 5 m. Image acquisition was performed from an average distance of one meter from the cliff face. This distance was selected to guarantee a balance between image quality and the number of images taken. The camera orientation was varied to provide complete and uniform acquisition of the vertical joints that characterize the cliff. Since the cave has heights above 15 m, we integrated the MP LiDAR data with those obtained from the photogrammetric survey (made with the GoPro) to improve the accuracy of the final 3D model.

Field Data
Sea caves were mapped and measured as shown in Table 1 and described in detail in the Paragraph 4.1. A separate paragraph is dedicated to sea cave#G32 because it develops almost entirely above the mean sea level and exhibits rounded potholes and channels of possible marine origin.

Mapping of the Sea Caves and Sea Arches in Gozo
A total of 79 semi-submerged sea caves around the island of Gozo were mapped, 2 of which are tunnels, because they have exits in directions opposite to their entrances. In addition, seven sea arches and five submerged caves (Table 1, Figure 5) were also mapped. Sea cave#G26, also known as the Xlendi Tunnel, is almost totally submerged. We found 64 sea caves along the plunging cliffs of the W sector, between the ria of Mġarr ix-Xini and Reqqa Point in the N side of the island, and 15 sea caves were mapped along the E sector, between Da asri, while at Reqqa Point, sea cave#G1 outcrops. In the E part of Gozo, LCL outcrops between Daleq Qorrot and Qala, the easternmost promontory of Gozo. From the latter site to Tac-Cawl rocks and Mgarr, GLO prevails, but some caves are carved on LCL rock masses (Figure 3).

Sea Cave#G32 with MIS5.5 Landforms and Deposit
Sea cave#G32 has a viable entrance with a height of 16 m (Figure 6), while the floor of the entrance is a few tens of centimeters above sea level (asl). The depth in front of the entrance rapidly increases and the area is partially rocky with sandy deposits. The total length of the cave is 122 m with a maximum width of 10 m. The height of the cave varies between 2.66 m in the innermost part and 15.5 m at the entrance ( Figure 7). Here, a large limestone rock partially obstructs the entrance ( Figure 8A) and forms a small terrace at about 15 m asl on the left side of the cave. Sea cave#G32 is a long sinuous tunnel roughly elongated in a NW/SE direction. The lower parts of the channels are occupied by seawater. The elevation of the floor increases from 0.5 m to 8.8 m asl in the innermost part. The floor of cave#G32 is generally made of well-rounded potholes as well as of rounded funnels and channels developing in the direction of the cave ( Figure 8B,C). Roughly in the middle of the cave, a 30 m long sector of the floor is occupied by collapsed blocks up to 3 m in size. Beyond these blocks, the elevation of the floor increases rapidly. In the innermost sector of the cave, a sandy-to-pebble deposit occurs. The clasts are rounded or subrounded up to half a meter in diameter ( Figure 8D,E). The deposit lies at an elevation of 8 m asl.
Living organisms occur up to about 40 m from the entrance, mainly on the floor, but also at the wall's foot. We measured morphometric parameters in 12 caves in the field and confirmed them through DP-derived outputs, while 13 caves were measured solely via photogrammetry. The longest cave is #G32, which is located at Dimitri Point. Finally, we described the sea bottom in six of the caves.

Sea Cave#G32 with MIS5.5 Landforms and Deposit
Sea cave#G32 has a viable entrance with a height of 16 m (Figure 6), while the floor of the entrance is a few tens of centimeters above sea level (asl). The depth in front of the entrance rapidly increases and the area is partially rocky with sandy deposits. The total length of the cave is 122 m with a maximum width of 10 m. The height of the cave varies between 2.66 m in the innermost part and 15.5 m at the entrance (Figure 7). Here, a large limestone rock partially obstructs the entrance ( Figure 8A) and forms a small terrace at about 15 m asl on the left side of the cave. Sea cave#G32 is a long sinuous tunnel roughly elongated in a NW/SE direction. The lower parts of the channels are occupied by seawater. The elevation of the floor increases from 0.5 m to 8.8 m asl in the innermost part. The floor of cave#G32 is generally made of well-rounded potholes as well as of rounded funnels and channels developing in the direction of the cave ( Figure 8B,C). Roughly in the middle of the cave, a 30 m long sector of the floor is occupied by collapsed blocks up to 3 m in size. Beyond these blocks, the elevation of the floor increases rapidly. In the innermost sector of the cave, a sandy-to-pebble deposit occurs. The clasts are rounded or sub-rounded up to half a meter in diameter ( Figure 8D,E). The deposit lies at an elevation of 8 m asl.
Living organisms occur up to about 40 m from the entrance, mainly on the floor, but also at the wall's foot.

Discussion
In this section, we discuss the general characteristics of sea caves in Gozo (par. 5.1) and their statistical distribution around the island, the distribution vs. lithology, and the type of sea caves or other surveyed landforms (par 5.2). A separate paragraph (par. 5.3) is devoted to sea cave#G32, as it shows landforms that can be traced back to a paleo sea level.

Morphometric Characteristics of Sea Caves in Gozo
Sea caves are common landforms along rocky coasts. Their origin and development are due to several factors, such as wave action, karst processes, etc. [1,6,7,9] In Gozo, sea caves can be roughly divided into two main categories mainly according to the shape of the entrance ( Table 1).
The first one is the box type, with a squared shape, mainly due to successive collapse of the limestone layers at the roof. The shape of the entrance can be both horizontally enlarged ( Figure 9A) or vertically enlarged ( Figure 9B).

Sea-Cave Distribution and Lithology
The statistical analysis of the spatial distribution of sea caves along the coastline of Gozo, based on a total amount of 87 coastal landforms, or sea caves, sea arches, and sea  The second one is the joint type, which develops along vertical or sub-vertical joints in the rock. Jointing can be perpendicular to the sea cliff ( Figure 9C) or angled, up to almost parallel ( Figure 9D). However, these types represent the end members of a complex mixing of shapes. In fact, many studied sea caves have both a box and joint component simultaneously. For example, they can develop through successive collapses of the roof along a joint ( Figure 9E), or at the contact between two different geological formations ( Figure 9F). The Xlendi Tunnel is a complex cave that develops on a tectonic contact between the LCL and GLO rock masses, creating a tube that is almost completely submerged and cuts a promontory from one side to the other.
Although submerged sea caves have been reported when visible from the water surface, we did not collect any morphometric parameters. Moreover, we also reported a sea cave in Marsalforn that is carved in the upper member of the GLO rock (#G75). It can be called a shelter, since it looks like a little extended notch carved into the cliff ( Figure 9G).
Sea arches are remnant of sea cliff retreat, and possibly of paleo sea caves. We reported seven sea arches 08, but SA4, the famous Azure Window, collapsed during a severe winter storm in 2017 [43]. We also found a sea stack (SS1) near the Azure Window (#SA4) that was probably previously connected to the sea cliff ( Figure 9H).

Sea-Cave Distribution and Lithology
The statistical analysis of the spatial distribution of sea caves along the coastline of Gozo, based on a total amount of 87 coastal landforms, or sea caves, sea arches, and sea stacks showed that 43 (51%) are located on NW part of the island, 16 (19%) on the SE, and 26 (30%) on the SW side. Due to a lack of data, five sea caves were not included in this quantification.
Below the Xwejni rock, there is a shelter carved in the GLO rock mass at Xwejni Bay. It was not considered in the statistical analysis since there is only one cave of this type. Most sea caves (n = 50) are box caves (63%), while there are 20 joint caves (26%), 8 complex sea caves (10%), and 1 shelter (1%), as shown in Figure 10B. There are five submerged caves, seven sea arches, and one sea stack. Most of the sea caves developed on LCL rock masses (n = 56, 71%), followed by GLO rocks (n = 15, 19%) and only a few in UCL rocks (n = 6, 8%), and finally at the contact between LCL and GLO (n = 2, 2%). Figure 10C shows the percentages for each formation.
The types of sea caves and lithology were compared for each zone of Gozo. In the NW side of Gozo ( Figure 10A), most sea caves are located on LCL rocks and are mainly stratatype. In the SE side of Gozo, strata-type caves are also dominant but they are balanced between LCL and UCL ( Figure 10B) and are mainly in GLO. In the SW side of Gozo, most of the sea caves are located on LCL, and only one sea cave (Xlendi tunnel) has developed on the contact between LCL and GLO rock masses ( Figure 10C). No sea caves have been found on the NE side of the island, where large landslides are abundant [34].
In Gozo, most of the sea caves occur on its W side, spanning from the Northern to Southern sector, while the NE side has no sea caves due to landsliding and lithological factors. For instance, BC terrains outcrop from Marsalforn to Dahleq Qorrot Beach on the NE side of Gozo. Moreover, from Dahleq Qorrot to Hondoq, there are only six sea caves and one sea arch due to the topography of the island, which is mainly characterized by sloping coasts [22]. However, an increasing number of sea caves have developed along the remaining Southern sector towards Mgarr, comprising seven sea caves and two sea arches. No sea caves have been found between Mgarr and Mġarr ix-Xini.
Exposure appears to be the most critical factor, as most of the sea caves have developed along the NW side of the island, which is exposed to the dominant wind (Majjistral). The presence of freshwater at Dweira, as already reported by [44], and on the Southern side of the island, may have also contributed to the development of sea caves in the less-exposed sites.
The types of sea caves and lithology were compared for each zone of Gozo. In the NW side of Gozo ( Figure 10A), most sea caves are located on LCL rocks and are mainly strata-type. In the SE side of Gozo, strata-type caves are also dominant but they are balanced between LCL and UCL ( Figure 10B) and are mainly in GLO. In the SW side of Gozo, most of the sea caves are located on LCL, and only one sea cave (Xlendi tunnel) has developed on the contact between LCL and GLO rock masses ( Figure 10C). No sea caves have been found on the NE side of the island, where large landslides are abundant [34]. In Gozo, most of the sea caves occur on its W side, spanning from the Northern to Southern sector, while the NE side has no sea caves due to landsliding and lithological factors. For instance, BC terrains outcrop from Marsalforn to Dahleq Qorrot Beach on the NE side of Gozo. Moreover, from Dahleq Qorrot to Hondoq, there are only six sea caves and one sea arch due to the topography of the island, which is mainly characterized by sloping coasts [22]. However, an increasing number of sea caves have developed along

Sea Cave#G32 and Relative Sea Level Change
Sea cave#G32 is the only cave on Gozo with a floor that has developed almost completely above the mean sea level. Despite this, seawater is still present on the floor, indicating that sea waves can enter the cave during storms up to a few dozen meters inside. Marine processes have significantly shaped the well-rounded channels and potholes, and the foot of the walls is also partially rounded. Moreover, the floor of the cave contains rounded-to-sub-rounded clasts up to the innermost part, where a sandy-to-pebbly deposit occurs. The clasts suggest that the deposit might be a Tyrrhenian beach deposit at an elevation of about 8 m, indicating the tectonic stability of Gozo over a long time, as well as in the Holocene, as reported by [44].
However, no datable fossil remains or even forms of rock biodegradation referable to lithodomes were found, which could be due to the low resistance of the limestones in which the cave is developed. This poor quality of rock masses suggest that these fossils were not preserved over a long time, as reported by [45,46]. The fact that rounded gravels are not cemented could raise doubts, related to the fact that the studied deposit is totally composed by uncemented clasts, and MIS5.5 deposits are usually cemented [44].
Otherwise, the altitude corrected for isostasy of the maximum transgression for MIS 5.5 at Gozo is 8.35 m, as reported by [45], and agrees to the altitude value of inner deposits.

Conclusions
A total of 84 sea caves and 8 retreat-related landforms have been investigated for the first time along the Gozitan coasts. These caves had never been inventorized and categorized for their shapes. Most of the caves were box caves (n = 50), followed by joint caves (n = 20), complex (n = 8), and shelter (n = 1) thanks to the combination of outputs of the above-water emerged and submerged surveys. All datasets were digitized and stored in a GIS, revealing that the majority of the sea caves studied developed on LCL rocks. This is likely due to the good quality of rock masses of the above-cited sedimentary rocks. Most of them (n = 41; 51%) are located on the most exposed sector (NW), indicating the significant contribution of waves and mechanical factors in their formation and shaping.
Additionally, marine landforms above the mean sea level have been reported in Gozo for the first time. Although dates and lithodomes on the cave walls are not available, the shapes suggest that the deposit found in the innermost part of the cave G32 (Table 1) may be a MIS 5.5 paleo-beach that has been preserved due to its inland location and potential protection from collapsed boulders in the central sector of the cave.

Data Availability Statement:
The sea-cave dataset is available on request from the corresponding author.