Relation between relief and Badland spatial distribution in the Paleogene Pazin Basin, Croatia

ABSTRACT Badlands are specific landforms characterized by intense denudation processes. Their occurrence is mainly associated with clay-like materials and semi-arid and Mediterranean climates. This study presents the badland inventory for the Paleogene Pazin Basin located on the Istrian peninsula in Croatia. A total of 5,381 badland polygons, with a total area of 10.25 km2, were manually outlined, based on the visual interpretation of orthophotos at a scale of 1:5,000. The badlands in Istria are represented as small and isolated landforms and are exclusively associated with flysch and flysch-like materials, which cover a significant part of the study area (46%). The analysis of badland inventory shows that badlands are spatially not equally distributed but that their abundance is increasing from NW to SE part of the analyzed area. Additionally, the results of the spatial analysis indicate the positive relationship between badland occurrence and relative relief, which is presented via the relief-badland index.


Introduction
Badlands are impressive natural landforms, formed in softer sedimentary rocks and clay-rich sediment (Yang et al., 2019), typically but not exclusively in semi-arid climate zones (Gallart et al., 2002;Harvey, 2004).The impermeability of such materials, with consequent generation of high surface runoff, and lack of vegetation lead to extensive fluvial erosion, the result of which is the formation of an extremely dissected landscape (Scheidegger et al., 1968).The development of badlands is controlled by numerous factors, among which the four critical can be underlined, as follows: (i) relief vigor, (ii) a rapidly weatherable soft lithology, (iii) a climate with notable rainfall or snow-melt erosive events, and (iv) additional disturbance or restriction of the environment that influences the development of protective vegetation cover (Moreno-de Las Heras & Gallart, 2018).With such characteristics, the badlands have great potential for a whole range of exogenetic processes, including weathering, denudation, and accumulation (Gulam et al., 2014).
The dominant process on badland slopes is erosion, which conditions badland origin and sustains its existence (Bryan & Yair, 1982).Due to intensive erosion, badlands are characterized by extreme geomorphological dynamics, which enables the measurement of rapid badland relief changes in a relatively short period.This is the main reason why the majority of research topics in badland studies refer to identifying erosion processes and quantifying the erosion rates (Martínez-Murillo & Nadal-Romero, 2018).Recently, the multi-temporal photo-geological analysis was applied to reconstruct the geomorphological evolution using aerial photos, orthophotos, and satellite images (e.g.D 'Intino et al., 2020).Moreover, to assess the topographic changes, the comparison of time series digital elevation models obtained with the use of terrestrial laser scanning (e.g.Llena et al., 2020;Marsico et al., 2021;Vericat et al., 2014) and unmanned aerial vehicles (e.g.Neugirg et al., 2016;Yang et al., 2019) is increasingly applied.
To describe the badland distribution within a particular area and complete different spatial analyses, it is necessary to create a badland inventory map.Identifying and mapping badlands are most commonly performed using remote sensing techniques, by visual interpretation of satellite images, aerial photos, and orthophotos (e.g.Alonso-Sarria et al., 2011;Bianchini et al., 2016;Coratza & Parenti, 2021;Khosrowpanah et al., 2010;Liberti et al., 2009;Vergari, 2015).Still, all of those techniques are combined with inevitable field surveys.
The badland inventories provide the basic information about badland spatial distribution within a particular area, thus can improve knowledge in various badland research areas which aim to:locate the most abundant badland areas on a regional scale; -identify different badland types (e.g.Bianchini et al., 2016);analyze the relationship between the current badland extent and different environmental predisposing factors (e.g.elevation, slope, aspect, slope curvature, relief, drainage density, distance to stream, lithology) in order to identify preferable conditions of badland formation and assess the areas susceptible to soil erosion (e.g.Leipe & DeLuca, 2017;Vergari, 2015;Vijith & Dodge-Wan, 2019);analyze and measure the changes in badland areas over time, i.e. define their expansion or extinction over the years using time-series badland inventories (e.g.Torri et al., 2018).
In this paper, the badland inventory map is presented for the area of Paleogene Pazin Basin, located in Istria, which is known for the appearance of the vast majority of badlands in Croatia.However, for an area such as Istria, completing the badland inventory is quite a challenging task.While in most areas worldwide known for badlands, where they are dominating over a wide areal extent (e.g.Badlands National Park, South Dakota; North Caineville Plateau, Utah; Bardenas badlands, Spain; Cappadocia, Turkey), in Istria, badland phenomena are typically discontinuously presented as isolated and small landforms (Gulam et al., 2014).Exactly because of that, in such an environment their spatial distribution can represent the measure of the intensity of exogenetic processes within Middle Eocene clastic deposits, with which the Istrian badlands are explicitly associated.
The presented badland inventory of Paleogene Pazin Basin is used to analyze badland spatial distribution within delineated watersheds and to represent badland size characteristics.Additionally, the spatial analysis is performed to estimate the relation between badland distribution and relative relief.

Study area
Istria is the biggest Croatian peninsula located at the head of the Adriatic Sea, between the Gulf of Trieste and the Kvarner Gulf (Figure 1).From the geological point of view, according to Velić et al. (1995), three different zones can be distinguished, namely, (i) southern and western Istria presented as Jurassic-Cretaceous-Paleogene carbonate plain, (ii) eastern and north-eastern Istria as Cretaceous-Paleogene carbonate-clastic zone with overthrusting structures (Učka  The focus of this research is the Middle Eocene clastic deposits of the Paleogene Pazin Basin, consisting of transitional beds and flysch (hereinafter termed as flysch and flysch-like materials).Transitional beds are presented with clayey limestones, calcitic marls, and marls (Velić et al., 1995).The flysch complex corresponds mainly to Middle Eocene (Petrinjak et al., 2021) and can be divided into the lower unit, characterized by the rhythmical marl and carbonate sediments exchange, and the upper unit, where carbonate-siliciclastic turbidite sediments are deposited (Bergant et al., 2003).They represent a small part of a large sedimentary system, which appeared as a consequence of the geotectonic changes on the carbonate platform during the Pyrenean compressional-tangential tectonic phase, in the period from the Eocene to the Oligocene (Marinčić et al., 1996).The study area of this research presents the area of 60 delineated watersheds (896.78 km 2 ), which comprised the entire area covered with flysch and flyschlike materials within the Paleogene Pazin Basin (Figure 1).
The badlands in the Basin are commonly formed as a result of excessive erosion (Jurak & Fabić, 2000), developing typical badland landscapes (Figure 2).They are characterized by a sparse vegetation cover and an average yearly amount of denudation between 2.5 and 6.5 cm (Gulam et al., 2018).
The climate of the Istrian peninsula is quite diverse, with variations in mean annual temperatures and mean annual precipitation between the southwestern, coastal area, and the northeastern, hilly part of Istria.Therefore, the climate of the central area is classified as Continental, the western and southern coastal parts as the Mediterranean, and the eastern part as sub-Mediterranean.The area of the Pazin basin is characterized by warm summers and a mean annual precipitation of around 1,200 mm (Zaninović et al., 2008).The time-series analyses of climate factors and the amount of denudation have shown a positive correlation, with the highest rate of denudation prevailing during dry, warm periods followed by intensive rainstorms (Gulam, 2012).

Materials and methods
The input data used in this study to create badland inventory and generate derivatives for further spatial analyses are presented in Table 1.
The badland inventory was prepared through systematic visual interpretation of digital orthophotos at a scale of 1:5,000 (State Geodetic Administration of the Republic of Croatia, 2011).During the analysis of orthophotos, different coverage of badlands with vegetation was determined, which directly influences the precision of badland delineation.The lower the vegetation coverage is, the more precisely the boundary of the polygon that outlines the badland is.Accordingly, four badland categories are defined, as follows (Gulam, 2012): (i) 0%, (ii) 1-25%, (iii) 26-75%, and (iv) more than 75% of the badland area covered with vegetation (Figure 3).
To enable the analysis of badland spatial distribution in more detail, the study area is divided into smaller spatial units, i.e. watersheds.A watershed also called a drainage basin or catchment area, is a geo-hydrological unit that presents any surface area from which runoff resulting from rainfall is collected and drained through a common point (Wani & Garg, 2009).A watershed may be considered as a dynamic environmental system unit where hydrologic and geomorphic processes interact, such as precipitation, evapotranspiration, infiltration, runoff generation, erosion, transport, and sedimentation (Renschler, 2004).To extract watersheds objectively, the delineation is performed automatically using Arc Hydro Tool, as described by Baye (2020).The manual interventions are made in flat areas, where the tool cannot follow watercourses properly due to the resolution of the digital elevation model.Since flat areas do not contain any badlands nor the flysch and flysch like-materials, which are in the focus of the analyzes, those parts of watersheds were simply removed.
In order to place the watersheds in their mutual spatial relationship, the watersheds are numbered based on the spatial appearance of their centroids from NW to SE.In such a way, number one is assigned to the watershed located in the far NW part of the study area, while the larger one, number 60, is assigned to the watershed located in the far SE part (Figure 4).This allowed the badland spatial distribution to be interpreted visually within the watersheds, but in a way also to quantify the evident trend of badland spatial occurrence in the NW-SE direction.
Spatial analyses aimed to determine badland spatial distribution and its relationship with relief energy, which is considered as one of the critical parameters in badland formation.Since Istrian badlands are fully formed within flysch and flysch-like materials, spatial analyses were performed for all watersheds within which at least 5% of the area is covered with this unit.The area covered with such materials was distinguished according to basic geological maps of SFRY at a scale of 1:100,000.More precisely, four   (Gulam, 2012).
Relative relief, also known as relief energy is one of the relief characteristics determined without considering the sea level.It is described by several authors as the difference in altitude between the highest and lowest points within a defined area (Nir, 1957).This certain area can present the standard areal unit of for example 1 km 2 , the area within a drainage basin of a given order, or between an individual interfluve and the adjacent valley bottom (Ahnert, 1970).In the local context, the relative relief represents a parameter of the intensity of the development of exogenetic processes (Lozić, 1995).Since in areas with a higher relative relief, the intensity of erosion is higher, it gives an indication of the potential relief energy for soil erosion.Therefore, it is used as one of the parameters to model erosion susceptibility (e.g.Vergari, 2015;Vijith & Dodge-Wan, 2019).
In this study, the spatial variability of relative relief is analyzed in two different ways.First, the difference between the maximum and minimum altitude is calculated for the area within each watershed, which presents the relative relief at a watershed scale.It allows interpreting its general spatial variability.Second, to present the distribution of relative relief continuously within watersheds, the difference between the maximum and minimum altitude is calculated for the circle neighborhood area with a radius of 100 m around each cell of a 20 m grid, which presents the relative relief at a 20 m grid scale.It allows the detailed analysis of the relationship between relative relief and badland occurrence.
To present the results of performed spatial analyses, the badland index is used, expressed in percentages (Bartelletti et al., 2017).The badland index for the study area (BI) is calculated as the ratio between the total badland area and the part of the study area covered with flysch and flyschlike materials.The watershed-badland index (BI W ) is calculated for each analyzed watershed as the ratio between the badland area within a certain watershed and the area of that watershed covered with flysch and flysch-like materials.The relief-badland index (BI R ) is calculated for each relative relief category within the area covered with flysch and flysch-like materials as the ratio between the badland area within a certain relative relief category and the area of that category.

Badland statistics
According to the sheets of basic geological maps of SFRY, the flysch and flysch-like materials cover an area of 408.26 km 2 (46%) within the study area.In that area, a total of 5,381 badland polygons were outlined, which corresponds to an average density of 13.18 badlands per square kilometre.The badlands cover a total area of 10.25 km 2 , i.e. 2,5% of the area covered with flysch and flysch-like materials.The badlands size ranges from 21.70 m 2 to 0.08 km 2 , with an average area of 1,904.73m 2 (Table 2).Only 165 (3%) badlands are larger than 1,000 m 2 .
Considering the four badland types defined according to the vegetation cover, the number of badlands and, consequently the total badland area, are increasing from the first to the fourth category, which contains 61.64% of badlands.Conversely, going from the first to the fourth category the average badland area is decreasing from 5,218.25 m 2 to 1,310.45 m 2 (Table 2).

Badland spatial distribution
Based only on the visual inspection of the badland inventory, it is obvious that badlands are mostly concentrated in the eastern and southeastern parts of the Paleogene Pazin Basin.Moreover, it is clear that badlands are spatially not equally distributed, but the badland abundance is increasing from NW to SE.The  total badland area is the largest within watersheds 41 and 44, with an area of 1.53 and 1.45 km 2 , respectively.The badlands were not identified within twelve analyzed watersheds, located NW from watershed 28.
In order to quantify the evident trend of badland spatial distribution at the watershed scale, the BI W is presented and compared with BI (Figure 5).
When the watersheds are lined up as they spatially appear going from NW to SE, the BI W is linearly increasing (Figure 5), showing a moderate to high positive correlation (r = 0.7).It is also shown that the smallest BI W values are associated with the watersheds located in the far NW part of the Basin.Thus, for all watersheds numerated to number 31, the BI W is below the BI.For watersheds numerated with a number higher than 31, the BI W is above or near the BI line for 73% of watersheds.The largest BI W values of 11.16% and 10.68% are calculated for watersheds 46 and 59.

Badlands and relative relief
The relative relief at the watershed scale ranges from 134 m to 1,134 m (Figure 6a).When the watersheds are lined up as they spatially appear going from NW to SE, the watershed scale relative relief is linearly increasing, showing a low positive correlation (r = 0.14).The highest values, above 600 m, are calculated for three watersheds located far E and SE, namely, watersheds 41, 43, and 53.This trend coincides with the intensity of tectonic activity within the study area.Unlike the fold-and-thrust system of Ćićarija and Učka mountains in the far eastern part of Istria, western and central Istria do not show a high degree of structural disturbance and are characterized by poorly distinguished, decametric folds of small amplitudes (Marinčić & Matičec, 1991).
To present the relief energy in more detail, the relative relief at the 20 m grid scale is presented (Main Map).Calculated values of relative relief range from 0 m to 268 m.They are classified into four categories based on the natural breaks (Jenks) algorithm, as follows: 1 (<23 m), 2 (23 m-48 m), 3 (49 m-79 m), and 4 (>79 m).The area of four categories of relative relief is differently distributed within the watersheds area covered with flysch and flysch-like materials (Figure 6b).Of them, the second and third categories prevail, covering almost 80% of the analyzed area (Figure 6b).The fourth category of relative relief is absent within thirteen watersheds.Considering the badland area distribution according to each relative relief category (Figure 7), it is noticed that badlands mostly occur within the second and third categories (81% of total badland area).The smallest occurrence of badlands is detected within the first category (less than 6% of total badland area).However, if the area of each relative relief category is also taken into consideration and the relationship between badland occurrence and relative relief is presented through the BI R , it is shown that the BI R is increasing from the first to the fourth relief category.Moreover, the BI R for the third and fourth categories is above the BI line and the BI R for the first and second categories is below the BI line.
The results of this analysis show that relative relief is positively correlated with the occurrence of badlands, where denudation processes dominate.Given the fact that relative relief can be considered as an indicator of the youngest tectonic movements (Marković, 1983), it can be concluded that there is a correlation between tectonic activity and the intensity of exogenetic processes within the flysch and flysch-like materials in the study area.Uplift, by definition, tends to increase the relief, which leads to more rapid downhill transport influencing indirectly the rate of denudation (Ahnert, 1970).In most badland areas of the Mediterranean regions, the Quaternary tectonics have been quite active, leading to past and/ or present uplifting (Gallart et al., 2002).

Conclusion
Badlands in central Istria are exclusively associated with Middle Eocene clastic deposits within the Paleogene Pazin Basin, which consist of the transitional beds and flysch on the area of 408.26 km 2 .For this area, the badland inventory was created, comprising 5,381 badland polygons manually delineated by visual interpretation of digital orthophotos at a scale of 1:5,000.Outlined polygons cover 2.5% of the total area of Middle Eocene clastic deposits.
Unlike the world-famous badland sites, which dominate extremely large areas, Istrian badlands appear as isolated and small landforms.This fact enabled the analysis of the spatial distribution of Istrian badlands, which was triggered by the obvious trend of the increase of badland density from the NW to the SE part of the study area.
Given that tectonic activity has an increasing trend in a similar direction on the Istrian peninsula, it was logical to test the geomorphological parameter (relative relief) which is, among other things, highly dependent on tectonics.The increasing trend of the badland index at the level of the analyzed watershed in the NW-SE direction confirmed the hypothesis that the share of badlands in a certain area largely depends on tectonic activity, i.e. uplift.This fact provides an answer to the question of why in the area of central Istria the distribution of badlands varies.
The presented badland inventory will be used in further research to analyze the relationship between badland spatial distribution and other environmental parameters supposed to be predisposing for badland occurrence, to produce a badland susceptibility map.Furthermore, the presented data will enable the comparison with updated badland inventory derived from new orthophotos in the near future, which will allow the discussion of badland initiation or extinguishing in the area of central Istria.Additionally, future studies on Istrian badlands will be focused on measuring the erosion rates on several selected badland locations where the UAV survey is carrying out periodically since 2018.
Software ESRI ArcGIS 10.2.1 was used for: (i) visual interpretation of orthophotos; (ii) delineation of badland polygons, (iii) conduction of spatial analysis, and (iv) designing of the Main Map.The graphs are created in MS Excel 2016 and the final editing of the Main Map layout was performed using Affinity Designer 1.6.5.

and
Ćićarija mountains), and (iii) central Istria known as the Paleogene flysch basin.It dominantly consists of the Dogger-Eocene carbonate rocks with less representation of Paleogene clastic rocks, flysch, and calcareous breccia.

Figure 4 .
Figure 4.The numbering of the delineated watersheds within the study area.

Figure 5 .
Figure 5. Watershed-badland index (BI W ) for the analyzed area of 53 watersheds.The badland index for the study area (BI) is also shown.

Figure 6 .
Figure 6.Relative relief: (a) at a watershed scale; (b) at a 20 m grid scale classified into four categories, for the area covered with flysch and flysch-like materials.

Table 1 .
Input data used in this study.