The upper Iapa watershed is located within the Bistrița Half-window (Outer Carpathians Flysch Zone), where the Tarcău and Vrancea Nappe’s (Outer Moldavides), are exposed. Dominating the landscape are vertical or reversed anticlines and synclines, which took shape during the Upper Cretaceous to Lower Miocene period. Overall, the geological formations exhibit a distinct north-south orientation, with faulted flanks characterized by a vertical or reversed configuration (Brustur et al., 2019).

The Tarcău Nappe, which is of interest for the present study, comprises two main units: the Tarcău Sandstone Digitation, formed by the Lower Tarcău Formation (Paleocene-Lower Eocene); and the Strigoiu Scale, composed of: Putna Formations (Paleocene-Lower Eocene), Straja Formation (Lower Eocene-Middle Eocene), and Ciunget Formation (Lower Eocene-Middle Eocene). The Strigoiu Formation marked the frontal part of the Tarcău Nappe that thrusted towards east the Vrancea Nappe along the Vaduri Digitation (Guerrera et al., 2012). Within this lithological configuration, the landslide that triggered the Iapa LDF in 2010 (Figures 2A, B) occurred in the Ciunget Formation and Straja Formation deposits (Figures 3A, B), very close to the contact with the Putna Formation lithofacies, which consists of dark-greyish clays, calcarenites, and limestone with a thickness of 4-5 m on the left flank of the Iapa LDF (Figure 3C).

The Ciunget Formation (Lower Eocene-Middle Eocene) is represented by a flysch lithofacies with intercalations of microconglomerates and thin layers of red clays. The main predominantly pelitic mass consists of Tarcău sandstones, occurring in beds that reach 10–12 m in thickness. Intermittently, alongside the sandstones, there are also beds of organogenic microconglomerates, with thicknesses ranging from 0.15–0.20 m, containing green schists. Towards the upper part of the lithological succession, there are also beds of friable conglomerates, mainly composed of well-rounded white quartzite fragments, alongside fragments of metamorphic rocks embedded in a silty matrix (Grasu et al., 1988).

The Straja Formation (Lower-Middle Eocene) stands out as a lithological and structural unit marked by a diverse sandstone lithofacies. Therefore, from a petrographic standpoint, the formation exhibits four distinct lithofacies: quartz-arenites, gaizes and spongolites, siltstones, as well as green and dark-red clays (Grasu et al., 1988). The quartz-arenites, which are the main lithological component in the LDF source area, are fine greenish, glassy sandstones with a sharp fracture, form the basal turbidites components. Microscopically, these sandstones predominantly feature quartz grains and cherts, accompanied by lesser amounts of muscovite, feldspars, chloritized biotite, zircon, and pyrite. Gaizes and spongolites make up either the basal layers of specific turbidites or the upper sections in sequences of calcareous microconglomerates and sandstones. Meanwhile, siltstones may manifest as slender interlayers within red and green shale or as more substantial turbidite beds displaying parallel to convolute lamination. In both instances, the matrix is composed of clay-sericite or clay-carbonates. The clays lithofacies encompass a significant proportion, accounting for 62% of the Straja Formation column. In the study area, these clays are observed in distinct layers measuring 5–20 cm, displaying varying shades of red or green (Grasu et al., 1988). However, the stratigraphy of the Tarcău sandstones (Figure 3D) in both the Ciunget Formation and Straja Formation, with varying thicknesses and hardness (Figures 3E–G), and interspersed with siltstone and clay layers, was the main lithological factor of Iapa LDF event in the synoptic context of the summer of 2010.

In documenting the direct relationship between climatic conditions and landslide processes that occurred in the upper Iapa watershed, we utilized both general data characterizing the climate on the eastern flank of the Goșmanu Mountains (Apostol, 2004) and local data recorded at the Luminiș pluviometric station (46°48′01″N/26°28′53″E) from the lower Iapa watershed (Cojoc et al., 2015). Therefore, the general climatic characteristics include annual mean temperatures ranging between 5°C (Nechit, 450 m a.s.l.) and 8°C (Săvinești, 280 m a.s.l.) at lower altitudes in the region, and between 2.5°C and 3°C at the meteorological station located above 1,500 m a.s.l. (e.g., Ceahlău Mts.). In both cases, the annual thermal characteristics are associated with a higher value of humidity. The mean precipitation amounts range between 600 mm (Moldavian Subcarpathians) and 900–1,000 mm (upper Iapa watershed). The snow cover duration ranges between 80 and 82 days, but in the case of upper part of the mountains valleys, like the Iapa, Calu and Nechit rivers, the snow cover duration can exceed 90 days due to shelter conditions. Overall, this type of climate is characteristic of mountains with moderate altitudes and corresponds to the mixed forest zone in the Eastern Carpathians (Cheval et al., 2014; Cojoc et al., 2015).

Regarding the precipitation regime in the trigger zone of the Iapa LDF, the data recorded at the Luminiș pluviometric station (300 m a.s.l.) between 1971 and 2021 indicate that the mean precipitation amount is 703.5 mm, with June (108.5 mm) and July (107 mm) being the wettest months of the year (Supplementary Table S1). However, the annual rainfall amount increases with altitude, especially from east to west, where in the study area, these values frequently exceed 1,000 mm. This phenomenon also occurred in 2010, when according to data recorded in Iapa wathershed, 12 km downstream of the Iapa LDF, the annual precipitation amount was 1,007 mm (Cojoc et al., 2015; Romanescu et al., 2018a). Moreover, only during the June-August interval of 2010, when the LDF occurs, the cumulative precipitation amount accounts for 70% of the total precipitation recorded during the entire year. However, this pluviometric regime indicates direct control of climatic conditions on geohazard manifestations (Mihu-Pintilie, 2018a), as well as an indirect causality between the occurrence of landslides and debris flow triggered by extreme rainfall in areas without forest vegetation (Lesenciuc et al., 2017).

The Iapa River, along with the Calu River to the north and the Nechit River to the south, are the main tributaries of the Bistrița River from the northeastern flank of the Goșmanu Massif. All three watercourses flow from west to east, perpendicular to the general orientation of geological strata (Brustur et al., 2019). Consequently, the entire hydrographic network is concentrated on the main valley with a few short and low-discharge tributaries. Also, due to the north-south orientation of the geological structures, the main watercourses are segmented by numerous lithological thresholds (e.g., Duraș waterfall on the Iapa River; Brustur et al., 2019).

In this hydro-morphological configuration, the lowest point in the Iapa watershed (76.01 km²) lies at the confluence of the Bistrița River (264 m a.s.l.), while the highest point is situated in the western part of the basin at Murgoci Peak (1,293 m a.s.l.). The Iapa watershed exhibits a slightly asymmetrical catchment basin, characterized by a parallel hydrographic network type (Jung et al., 2019). The left slope, covering a width of 1–2 km, lacks permanent tributaries. In contrast, the right slope, extending 3–4 km, is drained by several significant streams, including Mânza, Măniș, and Mălina. However, in the area where the LDF occurs, specifically on the southern slope of the Drumul Chinezilor ridge, the hydrographic network consists solely of temporary streams (e.g., Limpedea brook), all of which are collected by the Iapa River. Consequently, the Făgețel Lake, formed by the LDF dam obstructing a 300 m long-section of the Iapa Valley (see A2, B2 and C2 in Figure 2) at 20.5 km upstream of the confluence with the Bistrița River, lacks any other permanent tributary.

Concerning the flow rate data of the Iapa watercourse, groundwater contributes with 40% to the annual discharge (Cojoc et al., 2015). Therefore, from 1950 to 2021, the multi-annual discharge at the Luminiș gauging station averaged 0.63 m³/s, with a minimum mean discharge rate of 0.2 m³/s in 2013 and a maximum mean discharge rate of 1.7 m³/s in 1970 (Supplementary Table S2). However, climatic conditions, which control the rest of 60% of the flow rate and are often influenced by extreme rainfall events, make the maximum discharge data more representative for the present study. In this context, the highest discharge values recorded on the Iapa River at the Luminiș gauge station occurred on 8 June 1969 (75.3 m³/s), 28 May 1972 (65.8 m³/s), 12 July 2005 (58.8 m³/s), and 2 June 2016 (68.6 m³/s). In 2010, the maximum flow rate was recorded on 26 June, reaching 40 m³/s. However, a detailed analysis of this flash floods event and its correlation with LDF and Făgețel Lake formation will be detailed in next sections.

The vegetation within the upper Iapa watershed consists of 95% mixed forests, and two distinct zones are observed: the spruce zone (Picea abies, Picea excelsa) and the beech zone (Fagus sylvatica), both occasionally associated with herbaceous vegetation. The spruce zone constitutes over 50% of areas situated above 1,000 m a.s.l., particularly in the high and middle sections of the main ridges. Consequently, spruce stands in these zones are either pure or nearly pure, characterized by dense forest masses, with a sparse or underdeveloped herbaceous or shrub layer. At lower elevations (850–900 m a.s.l.), the spruce mixes with fir (Abies alba) and beech. Notably, in the area of the landslide source of the Iapa debris flow (875–800 m a.s.l.), pine (Pinus sylvestris) also appears. At altitudes ranging from 450 m a.s.l. to 600 m a.s.l., corresponding to the elevation of Făgețel Lake (615 m a.s.l.), the emergence of the beech zone is influenced by continental and foehn-like climatic factors, albeit covering a smaller surface area. In some areas, the beech forests at these elevations also incorporate specimens of elm (Ulmus montana) and sycamore (Acer pseudoplantanus). Within the shrubbery along the Iapa watercourse, noteworthy species include rowan (Sorbus aucuparia), hazel (Corylus avellana), spindle (Euonymus europaea), red elderberry (Sambucus racemosa), and black elderberry (Sambucus nigra). However, in the water mass accumulated behind the Iapa LDF dam, of the tree species mentioned earlier, currently, only the trunks of spruce are still visible above the water’s surface.

Overall, being an intensively forested area, the main and only industrial activity in the region is the exploitation and valorization of timber. Thus, satellite images reveal numerous areas where the forest has been clear-cut in longitudinal strips towards the valleys of the Iapa, Manzu, and Maniș rivers, extending almost to the alignment of the main ridges. The most heavily exploited area was identified in the upper sector of the Iapa River, specifically on the eastern slope of the Murgoci Peak (1,293 m a.s.l.), the entire left slope of the Mânzu stream (Mânzei ridge), and in the study area, formerly known as the Limpedea logging zone or Drumul Chinezilor ridge. In this context, being an area with limited accessibility and steep slopes, heavy forest tracks was used for timber extraction, forming paths that have turned into deep cutting rigoles, gully and landslide scarps. However, only in the investigated area, the density of the road and paths network used for timber exploitation is 4.5 km/km². As we will see in the upcoming subsections, the exploitation paths network and the deforested areas was one of the main factors that triggered the Iapa LDF in the summer of 2010.

The field survey campaign took place in October 2023 and covered an area of approximately 15 ha on foot and 20 ha through UAV mapping, including the landslide source area, the debris flow paths, the fan area (LDF dam), and the water surface of the impoundment. The primary objective was to investigate the triggering in the crown area, track the flow-slide evolution in terms of both erosion and accumulation, and assess the damage in the flooding area caused by LDF dam formation. All data, including pictures, waypoints, and topographical measurements, were subsequently geolocated using a accurate GPS. Regarding the aerial measurements, the use of UAV techniques was possible in such a forested area due to the fact that the 2010 debris flow-slide event created a scar in the forest vegetation. However, after the rainfall-induced debris flow-slide event in 2010, the broken trees were collected by the forest administrators. Currently, the forest growing on the surface of the LDF is young and dispersed.

The photogrammetric data acquisition stage consisted of using a high-performance DJI Phantom 4 Pro drone with a full-frame digital camera with a 20-megapixel CMOS sensor. The flight plan was designed to perform successive flights with 70% overlap between photos. Flights were conducted at a relative altitude of 110 m above the slope where the Iapa LDF occurred, correlated with mountain topographic variation, in order to obtain a ground sampling distance (GDS) of 2 cm. The photogrammetric area surveyed is 20 ha, of which the affected area by debris flow-slide is 13.5 ha and by flood approximately 2.5 ha. The photos obtained are automatically georeferenced by the drone using the on-board RTK system to record precise geographic coordinates for each photo. A total of 571 photos were obtained, which were processed using automated procedures with appropriate spatial alignment algorithms. In Figure 4A the Projection Center Point (PCPs) coordinates of each UAV photo are indicated.

The dense point cloud, mesh, texture, tiled model and digital surface model (DSM) algorithms implemented in Agisoft Metashape Pro v.1.6.5 software were used to produce the orthomosaic (2 cm spatial resolution) (Figure 2C) and UAV-derived DSM (10 cm spatial resolution) of the Iapa LDF area. Spatial analyses were performed using Esri ArcGIS v.10.2. The DSM was resampled to a spatial cell size of 1 m². In the final stage of data processing, to capture all topographic details within and in the vicinity of the LDF body, the final DEM was filtered to exclude vegetation (e.g., trees, shrubs) or any other point elevation that does not relate to the land surface (Figure 4A).

In order to obtain the thickness of erosion and deposition map derived from the difference of two consecutive DEMs (DOD) (De Falco et al., 2023), an elevation model before the Iapa LDF event was needed for comparison with the UAV-derived DEM obtained for 2023. Therefore, the best option for extracting altitudinal information was the topographic maps at a 1:5,000 scale, considering the low resolution of other available elevation models like SRTM or ASTER (Khasanov, 2020) in the investigated region. In this context, two topographic maps corresponding to the study area were acquired from the National Agency for Cadastre and Land Registration of Neamț County, Romania (Table 2). The contours with 2.5 m equidistance were manually digitized, along with other point elevation information (e.g., mountain peks, elevation of hydrographic network), using the old maps as a background. Based on this shapefile elevation data, a DEM with a 1 m cell size, similar to the resolution of the UAV-derived DEM, was generated using the Topo to Raster interpolation tool in ArcGIS 10.2 (Figures 4B, C).

Regarding the DOD methodology, which is a widely adopted tool for volumetric data in geomorphic change detection (GCD) (De Falco et al., 2023), especially in areas affected by landslides (Jaboyedoff et al., 2012) and debris flows (Bull et al., 2010; De Long et al., 2012), it was employed to identify the primary areas of erosion and accumulation in the case of the Iapa LDF. The difference between the 2023 UAV-derived DEM and the 1975 old maps-derived DEM was computed using the Raster Calculator tool in ArcGIS 10.2 (Figure 4D). The resulting raster was used to obtain morphometric values for the detached, transported, and deposited terrigenous material. Additionally, it facilitated the accurate delineation of the morphological features of the Iapa LDF. Furthermore, by utilizing successive Google Earth satellite views from 2010 to the present (Figure 4C), along with DOD data, we managed to identify other episodes of debris flow-slide that occurred in the crown area of the Iapa LDF after the 2010 event.

The delineation of the Iapa LDF features for hydrological and geomorphological mapping was conducted through field descriptions correlated with the geodetic data (e.g., UAV-derived orthomosaic and DEM). The measurements improve the estimate of the shape, surface area, and volume of both the initial and recently reactivated debris flow-slide in the study area. Moreover, since the present case study will be included together with other examples in an upcoming European landslide dams database, the delineation of geomorphological features to obtain accurate morphometrical values was achieved according to a specific data-collection template (Mihu-Pintilie and Dufresne, 2023) (Figure 4E). The summary of this data-collection includes informations related with: 1) Landslide triggered debris-flow data: type, trigger, total volume, deposit area, length, crown elevation, toe elevation, total drop height, total runout length, horizontal runout length in 2D and 3D, travel angle, distance scarp to river (Cruden and Varnes, 1996; Hungr et al., 2014); 2) LDF dam data: dam type (Costa and Schuster, 1988), dam volume, dam material composition, dam height, dam-surface lowest elevation, dam length, dam width, downstream and upstream slope of the dam (Fan et al., 2020); 3) Făgețel Lake data: elevation of lake surface, area, length, mean and maximum width, major axis, small axis, form coefficient, perimeter length, sinuosity coefficient, volume and volume coefficient, mean and maximum depth (Mihu-Pintilie, 2018b; Romanescu et al., 2018b).

As presented in the introduction section, during the summer of 2010, significant heavy rainfall impacted the northeastern part of the Goșmanu Massif, leading to several flash floods along the Iapa Valley. The highest precipitation amounts were recorded in three intervals: 16–22 May (115.8 mm), 22–27 June (195.8 mm), and 25–27 July (70.4 mm). Overall, from May to August 2010, the cumulative precipitation amounted to 599.1 mm, which represents 70% of the total precipitation recorded throughout the entire year (Figure 5A). These weather conditions also induced historically high discharge rates on the Iapa River as a response to rainfall events: 16.4 m³/s on 22 May, 40 m³/s on 26 June, and 12.6 m³/s on 27 July (Figure 5B). According to locals, during these periods, the Iapa River flooded several courtyards and damaged several bridges and roads in settlements downstream of the Iapa LDF location, especially in Negulești and Luminiș.

However, based on discussions with stakeholders, including the foresters whose ranger office house is located 8 km downstream from the LDF dam, it could not be precisely determined which of the three rainfall events triggered the debris flow-slide event. Under these circumstances, we attempted to correlate satellite imagery with the possible weather interval when the mass-movement occurred. Therefore, due to the fact that the Iapa LDF event left a scar in the forest vegetation, the best option was to investigate successive Landsat 7 ETM+ Normalized Difference Moisture Index (NDMI) images with a resolution of 50 m per pixel (Figure 5C). In practice, NDMI is generally used to determine vegetation water content (Ochtyra et al., 2020), which, in the case of the present study, can indicate when forest vegetation was removed from the area affected by the debris flow-slide. In this framework, the analysis of NDMI rasters indicates that the Iapa LDF occurred between 03 July and 20 August, and most probably the rainfall event from 27 July was the one that triggered the LDF. Furthermore, even though the most severe meteorological conditions of 2010 were not recorded at that time, the initiation of landslide processes can be attributed to the 27 July event, although the decisive factor was the large quantities of precipitation that saturated and weighed down the southern slope of Drumul Chinezilor ridge throughout the entire summer season.

Interestingly, it should be noted that during the period of 25–27 July 2010, when it is also considered that Făgețel Lake was formed, the discharge data on the Iapa Valley did not show a proportional increase directly corresponding to the amount of precipitation that fell, as was the case with previous rainfall events. In this context, we consider that the blocking of the Iapa Valley by the LDF fan caused the accumulation of water behind the dam, creating a delay of several hours between the high quantities of precipitation and the discharge response. This delay most likely corresponds to the filling duration of the newly formed lacustrine depression, estimated to be between 2 and 4 h. After the watercourse resumed its flow over the dam surface, the discharge data recorded at the Luminiș gauge station showed a progressive increase, reaching a maximum of 12.6 m³/s.

The difference of DEMs (DOD) methodology used for GCD across the Iapa LDF area was primarily applied to obtain volumetric data and secondarily to confirm the field survey observations and measurements. In this regard, the difference between the two rasters of 2023 UAV-derived DEM and the 1975 old maps-derived DEM indicates a total volume of 782,500 m³ of displaced materials, of which more than 450,000 m³ represents the volume of the Făgețel Lake dam. The erosion and deposition thickness values range between −28 m in the source area and +30 m in the deposition area (Făgețel Lake dam). The mean thickness value is approximately ±3.4 m, suggesting that, apart from the crown area and the obstructed river sector where the erosion and accumulation amplitudes are high, the rest of the Iapa LDF exhibits a relatively shallow depth configuration (Figure 6A).

The cross section locations of the field observations and UAV-derived GPS measurements used to validate the DOD outcomes are also indicated in Figure 6B. They show values ranging between 15 m and 25 m of erosion in the landslide-triggering area and on the debris flow-slide flanks, while erosion values ranging between 2 m and 4 m are observed in the main channel of the flow-sliding area. Regarding the deposition values, they are more representative in the LDF dam area, where the flow-slide accumulation reported thickness exceeding 25 m between the valley floor and the top of the obstructed dam. However, when comparing the field measurement results with the DOD values, the errors do not exceed 1 m, especially in the deposition area, and this is a consequence of the fact that direct measurements could not capture the entire flow-sliding plane at all analyzed points. In the erosional area, the errors were minimal (less than 0.5 m), including here also the debris flow events post-2010, which are quite visible in the LDF source area.

Overall, the DOD map accurately indicates all details related to the thickness of the mass movement in terms of erosion and accumulation areas. This analysis permitted the identification of important geomorphological aspects, such as the presence of channels where major erosion occurred (C1–C7), the first slope change at the foothill (C8), the upper part of the debris flow-slide area (C9–C12), or the precise limit of the obstructive dam (C13–C15) (Figure 6B).

For a more comprehensive hydrological and geomorphological characterization of the Iapa LDF, we divide the investigated area into three landform parts: 1) Landslide triggered debris-flow area; 2) Iapa LDF dam area; 3) Lake area. All morphometric values will be recorded in an upcoming landslide dams database, alongside other representative landslide.

The location of the Iapa LDF in the flysch zone of the Eastern Carpathians, a mountainous area frequently affected by extreme rainfall events with high erosional impact in deforested locations, constitutes the main premise for triggering the recent mass-movement processes that have occurred in the upper Iapa Valley. Even though the debris flow-slide event that occurred in the summer of 2010 was extremely dynamic, a stabilized appearance of vegetated landforms in the LDF accumulation area indicates that shallow landslides and debris flows are currently active only in the source and depletion area. However, a new episode of heavy rainfall that affected the southern slope of the Drumul Chinezilor ridge can occur at any time. This hazard threat arises from the analysis of climatic conditions during the May-August period of 2010 when the time lag between the long-lasting rain culmination and the landslide triggering suggests that the debris flow-slide initiation was connected with cumulative precipitation over the 3 months in the study area. Therefore, lessons learned and future actions needed to prevent the further reactivation of the Iapa LDF suggest that local authorities should avoid clear-cutting forests on exposed slopes prone to landslide processes. Additionally, the network of forest exploitation paths should be minimized on steep slopes, as the density of these timber transportation routes has been the primary anthropogenic trigger leading to the occurrence of the Iapa LDF.

Regarding the Iapa River sector obstructed by 2010 LDF event and the formation of Făgețel Lake, although this type of landslide lakes are not unique in the landscape of the Eastern Carpathians, the uniqueness and hazardous threat arise from the dam composition and stability characteristics. Usually, the lifespan of dams formed by debris flow-slide processes is short (a few hours) due to the low cohesion of the debris matrix. Therefore, the long-term presence of Făgețel Lake and the relatively stability of the dam indicate, but only in this case, that the potential flood hazard threat was partially diminished. However, this does not guarantee that in the event of other similar occurrences, the formed dams will not fail, affecting the population located downstream. For this reason, in the watershed of the Iapa river, in addition to the area affected by debris flow-slide in 2010, other areas susceptible to landslides should also be monitored in the future. In this study, we identified three possible other areas where mass movement processes could occur on a similar or even larger scale than Iapa LDF. However, we anticipate that the outcomes of this approach will aid in enhancing the emergency response plan in the event of any other potential destructive hydro-morphological events linked to heavy rainfall in the surveyed region.

From the perspective of UAV survey methodology used in this study, both for extracting the main morphological and morphometric parameters of the Iapa LDF and for understanding the mechanism of the impoundment and formation of Lake Făgețel, there were several limitations that need to be addressed. The UAV survey conducted to obtain a DEM as accurate as possible with the topographic reality in 2023 was hindered by the forest vegetation that had spread over the area. This aspect made the generation of the DEM using DTM-derived from successive orthomosaic aerial images a time-consuming process, involving the identification and filtering of non-elevation data (e.g., tree canopy). However, in the future, we intend to use LiDAR technology to reconstruct the terrain elevation model in the area affected by the Iapa LDF, a technology that will streamline both the processing and filtering stages of geospatial data, significantly improving the accuracy of the resulting DEM. Despite these drawbacks, the use of DEM-derived from UAV survey and corrected on the ground with GPS measurements has yielded satisfactory results.