Low-Altitude, Overcooled Scree Slope: Insights into Temperature Distribution Using High-Resolution Thermal Imagery in the Romanian Carpathians

Abstract

Advective heat fluxes (chimney effect) in porous debris facilitate ground cooling on scree slopes, even at low altitudes, and promote the occurrence of sporadic permafrost. The spatial distribution of ground surface temperature on an overcooled, low-altitude scree slope in the Romanian Carpathians was analyzed using UAV-based infrared thermography in different seasons. The analysis revealed significant temperature gradients within the scree slope, with colder, forest-insulated lower sections contrasting with warmer, solar-exposed upper regions. Across all surveyed seasons, this pattern remained evident, with the strongest temperature contrasts in December and April. February exhibited the most stable temperatures, with thermal readings primarily corresponding to snow surfaces rather than exposed rock. Rock surfaces displayed greater temperature variation than vent holes. Vent holes were generally cooler than rock surfaces, particularly in warmer periods. The persistent presence of ice and low temperatures at the end of the warm season suggested the potential existence of isolated permafrost. The results confirm the chimney effect, where cold air infiltrates the lower talus, gradually warms as it ascends, and outflows at higher elevations. UAV-based thermal imagery proved effective in detecting microclimatic variability and elucidating thermal processes governing talus slopes. This study provides valuable insights into extrazonal permafrost behavior, particularly in the context of global climate change.

1. Introduction

Low-altitude cold screes, also referred to as periglacial screes, are geomorphological features characteristic of mid-latitude mountain regions below 1500 m. These landforms comprise steep accumulations of unconsolidated clastic material, including boulders, gravel, and sand, where sporadic or isolated permafrost may persist [1,2,3,4,5,6,7]. Their thermal regimes are marked by consistently low ground temperatures, pronounced thermal offsets, and active geomorphic dynamics, influencing both surface and subsurface processes [4,7,8]. The presence of permafrost at such low elevations challenges conventional periglacial models, necessitating further investigation of their thermal behavior and underlying mechanisms.

At these sites, slope evolution remains influenced by periglacial processes, including frost weathering, permafrost creep, ice segregation, and frost heaving, albeit to a lesser extent than during the Last Glacial Maximum [9]. These processes facilitate the continuous accumulation of unconsolidated debris at the base of slopes and can occasionally trigger mass movements, such as rockfalls and rock slides [10]. Certain “coarse blocky landforms”, including relict rock glaciers, scree slopes, and talus slopes, can sustain cold ground microclimates, allowing the persistence of isolated permafrost below the regional alpine permafrost boundary [7,11,12,13,14,15,16].

A key driver of these cold anomalies is the “chimney effect” [4], a mechanism in which bidirectional airflow within coarse debris is regulated by seasonal temperature gradients between the atmosphere and internal void spaces. During winter, warm air rises through the upper sections of the slope, while cold air is aspirated in the lower part of the scree. Conversely, in summer, cold and dense air accumulates in the lowermost part of the debris, while external warm air is aspirated into the upper section [11,17].

Monitoring surface temperature variations is essential for understanding periglacial processes and improving predictive models of permafrost evolution [18,19]. UAV-based infrared thermography (IRT) has emerged as a powerful tool for high-resolution thermal monitoring. Equipped with thermal sensors, UAVs offer significant advantages over traditional ground-based methods by enabling detailed assessments of complex terrain. They can capture temperature variations both vertically and horizontally, facilitating comprehensive analyses of thermal dynamics across spatial and temporal scales [20]. Their advanced navigational capabilities allow for data collection in inaccessible environments, such as scree slopes and rocky ridges, where ground-based monitoring is challenging [21,22]. Additionally, UAVs can overcome logistical challenges, making them an optimal tool for monitoring complex periglacial environments [23,24].

Despite the increasing use of UAV-based IRT in glacier monitoring [25], landslide assessment [26], and rock wall stability analysis [27,28], its application in low-altitude, cold scree environments remains limited. In particular, there is a lack of studies investigating the thermal dynamics and air circulation patterns that govern permafrost persistence in debris accumulations. Addressing this gap is crucial for refining conceptual models of periglacial processes at lower elevations and improving methodological frameworks for UAV-based IRT in such environments.

Detunata Goală has been a focal point of geomorphological research since the early 2000s due to its distinctive microclimatic and geomorphological characteristics [29]. Early studies hypothesized the presence of sporadic low-altitude permafrost based on summer ice occurrences within coarse deposits, spring water temperature, and bottom temperature of snow cover (BTS) measurements [29]. Ref. [7] investigated the microclimatic conditions at Detunata Goală, confirming the presence of permafrost through geophysical measurements. In 2022, a borehole drilled to a depth of 13 m identified a thin permafrost layer at approximately 7 m [30]. However, detailed thermal monitoring of this site remains limited, particularly regarding the role of air circulation in sustaining the overcooling of the lowermost part of the scree slope.

This study aims to bridge the gap in the high-resolution, multi-seasonal thermal monitoring of scree slopes using UAV-based infrared thermography (IRT). To achieve this, we will generate high-resolution thermal orthomosaics from UAV-acquired imagery collected during four field campaigns (December 2023, April 2024, October 2024, and February 2025) at Detunata Goală. These datasets will be used to analyze the spatial and seasonal variability in surface temperatures across the scree slope. Additionally, we will establish methodological benchmarks for UAV-based thermal image acquisition and processing in extrazonal permafrost environments.

By integrating multi-seasonal, UAV-based IRT with ground temperature measurements, this study aims to enhance the understanding of thermal dynamics in low-altitude cold screes and establish a methodological framework for future research in similar environments.

2. Methodology and Materials

2.1. Study Area

Detunata Goală is a volcanic ridge in the Apuseni Mountains, part of the Western Romanian Carpathians, with a maximum elevation of 1158 m (Figure 1). It is primarily composed of basaltic andesite, distinguished by hexagonal columnar jointing [31]. The disintegration of these columns has resulted in two major scree deposits: a larger accumulation on the western slope and a smaller one on the southern slope [32].

The site has a temperate climate, with a mean annual temperature of 7.4 °C and an average annual precipitation of approximately 800 mm, based on records from the Roșia Montană meteorological station, located 6 km northwest of the study site. Snowfall typically occurs between late January and early March, with maximum snow depths reaching 60–70 cm [7].

Although the investigated open scree slope is relatively small (243,735 m²), it exhibits distinct geomorphological and microclimatic features. Persistent ice remains until early summer, while cold air currents emerge at the base of the slope during the warm season. These conditions are driven by the high porosity of the block accumulations, which facilitate air convection [33] and advection through the coarse deposits. Additionally, the Balch effect [34] influences airflow dynamics based on the configuration of interstitial spaces [31].

2.2. Methodology

The data acquisition workflow was based on the Structure-from-Motion (SfM) technique, employing UAV-based photogrammetry following the approach detailed by [35]. This method was specifically adapted for the capture and processing of high-resolution thermal and optical imagery, incorporating adjustments to optimize the handling of thermal data (Figure 2).

To analyze variations in surface and subsurface temperature regimes within the scree, four field campaigns were conducted. The first took place between 16 and 17 December 2023, the second between 26 and 27 April 2024, and the third between 26 and 27 October 2024, followed by the last between 14 and 15 February 2025. The October campaign focused solely on daytime thermal data to assess the effects of solar radiation on surface temperatures. In contrast, the February campaign included both daytime thermal imagery, recorded over freshly fallen snow, and nighttime thermal imagery, collected when the snow cover was reduced, to capture thermal contrasts under varying climatic conditions.

All campaigns included daytime UAV flights to collect RGB reference imagery. Additionally, nighttime thermal data acquisition at 23:00 was conducted for all surveys except October to minimize the influence of solar radiation on temperature measurements.

Temperature monitoring at Detunata Goală was conducted using a DJI Matrice 300 RTK, an industrial-grade UAV equipped with a Zenmuse H20T thermal sensor. This system ensures precise and reliable data acquisition through advanced navigation technologies, including GPS, GLONASS, BeiDou, and Galileo. To enhance spatial accuracy, image coordinate data were corrected using a high-precision RTK system, integrating the DJI RTK-2 Base Station over a known reference point. This correction process allowed for precise alignment between models generated across multiple field campaigns, ensuring consistent comparisons (Figure 3). The Zenmuse H20T sensor is equipped with a radiometric thermal camera (640 × 512 resolution), a zoom camera, and a wide-angle camera [36], enabling the capture of high-resolution thermal and optical data. This configuration was essential for mapping the spatial distribution of surface temperatures and analyzing thermal variations within the scree slope.

Flight planning was carried out using DJI Pilot 2 software, incorporating an analysis of elevation differences, vegetation height, and terrain characteristics to ensure UAV safety and optimize data acquisition. Key flight parameters, including the area of interest, flight altitude, camera settings, and image overlap, were carefully configured to maintain consistency across surveys. To ensure uniform data quality, a constant altitude of 100 m above the ground level was maintained using the UAV’s real-time terrain-follow functionality, which was guided by a 1 m resolution digital elevation model (DEM) obtained from [37] ANCPI—MNT LAKI II (geoportal.ancpi.ro). Since the original DEM was in Stereo70 (EPSG: 31700), it was reprojected to WGS84 (EPSG: 4326) for seamless integration with the DJI application. This step ensured a consistent ground sampling distance (GSD) across the study area. To enhance the accuracy of orthomosaic reconstruction, a high image overlap of 80% was applied for both forward and side overlaps. While automated image acquisition was employed for structured coverage, manual flights were also conducted during nighttime thermal mapping to ensure comprehensive data collection over the entire cold scree area.

To improve the analysis of temperature variations between vent holes and surrounding rock surfaces, thermal images of the largest vent holes, locally known as ”breathers”, were captured at night during all field campaigns except October. These datasets allowed for a detailed comparison of thermal differences in forested environments, where temperature dynamics may differ from those observed in the open scree. A major challenge encountered during data collection was the need for manual thermal image acquisition in dense forested areas. The presence of thick vegetation posed a risk of UAV collision and limited feasibility of automated flight plans. As a result, thermal imagery for these vent holes was captured through manual UAV operation, and in certain cases, handheld thermal imaging was employed to ensure complete data coverage. Another challenge was the inconsistent RTK signal in the forested terrain, which could affect georeferencing accuracy. To mitigate this, a pre-measured reference point was established using a Topcon Hiper-V RTK system, allowing for precise positioning of the DJI Base Station. This approach ensured accurate georeferencing of the acquired thermal imagery despite the environmental constraints posed by the dense vegetation.

To refine the accuracy and reliability of the UAV-based thermal imagery, on-site environmental measurements were conducted during each field campaign. Air temperature and humidity were recorded using a Kestrel 4000 Thermo-Hygrometer manufactured by Nielsen-Kellerman Co. (NK) in Boothwyn, PA, USA, to assess atmospheric conditions influencing thermal emissions. Ground surface temperature (GST) measurements were performed using snow probes at various depths to provide reference data for calibrating the physical parameters used in thermal image processing. These measurements were essential for thermal image calibration, ensuring accurate temperature extraction and minimizing potential deviations caused by atmospheric influences and surface emissivity variations. The collected data were later integrated into the analysis workflow to refine the interpretation of thermal anomalies and improve the reliability of the UAV-derived temperature datasets.

The data processing workflow integrated photogrammetric techniques, GIS-based methods, and statistical analysis to ensure accurate and reliable results. Thermal images captured by the Zenmuse H20T sensor were initially stored in DJI’s proprietary R-JPG format, which does not directly provide absolute temperature values. To enable quantitative thermal analysis, the R-JPG files were converted to TIFF format using ImageJ version 1.54k software with the IRImage-UAV plugin developed by [38,39,40]. This conversion allowed for the extraction of precise temperature values, ensuring consistency in data interpretation (Figure S1, Supplementary Materials).

During the conversion process, critical physical parameters measured on-site were calibrated to enhance the precision of temperature readings. These parameters included atmospheric humidity, ambient air temperature, sensor-to-surface distance, and emissivity, all of which were adjusted based on the prevailing climatic conditions during each survey. In December, the ambient temperature was −2 °C, the humidity was 80%, and the emissivity was set to 0.97. In April, the temperature increased to 5 °C, the humidity was recorded at 75%, and the emissivity was adjusted to 0.94, following the findings of [41] for massive basalt surfaces. In October, the air temperature was 7 °C, the humidity was 70%, and the emissivity remained at 0.94. In February, the emissivity values differed between the nighttime and daytime surveys. For the nighttime surveys, the emissivity was set at 0.96 due to the high humidity (90%) and air temperature of 0 °C. For the daytime survey, the emissivity was increased to 0.98 to account for the presence of fresh snow, with the humidity at 80% and air temperature at −4 °C. For the thermal surveys conducted in the forest sector, the emissivity values remained at 0.98 in both December and February, with humidity levels at 85% and air temperatures of −2 °C and −4 °C, respectively. In April, the emissivity was adjusted to 0.94, with the humidity recorded at 75% and an air temperature of 5 °C.

The converted TIFF thermal images and RGB data were subsequently processed using Agisoft Metashape Professional version 1.6.3 to generate high-resolution thermal maps and orthophotos for further analysis. To enhance the resolution and accuracy of the outputs, high-quality processing settings with aggressive filtering were applied. This ensured successful image alignment and sparse point cloud generation. Following the initial alignment, dense point clouds were generated, facilitating the creation of Digital Surface Models (DSMs) as well as thermal and RGB orthomosaics for all the field campaigns (Figure S2, Supplementary Materials). The processed dataset statistics, including image alignment accuracy, are presented in Table S1 in the Supplementary Materials. Some misalignment issues were observed, particularly in the thermal datasets, where the low contrast and grayscale nature of TIFF thermal images resulted in an insufficient number of tie points for automatic alignment.

To ensure consistent and accurate data comparison, all models were resampled to a uniform spatial resolution of 8 cm, matching the dataset with the lowest resolution. This step enabled precise coregistration of thermal and RGB datasets using ArcGIS Pro version 3.3.2 and specialized geospatial tools, ensuring consistency and reliability across analyses. Additionally, the coordinate system was reprojected from WGS 84 (EPSG: 4326) to the national Stereographic 70 system (EPSG: 31700) for compatibility with regional datasets. During the final processing stage, data from both manual and automated UAV flights were integrated to create a comprehensive and unified dataset, ensuring full spatial coverage while minimizing data gaps.

The processed models were imported into ArcGIS Pro for cartographic visualization, spatial analysis, and statistical evaluation. Local temperature statistics, including minimum, average, and maximum values, were extracted for key features such as vent holes and adjacent rock surfaces to assess thermal variations across different seasons. To differentiate between vent holes and rock surfaces, a combination of thermal imagery, RGB orthomosaics, and field observations was used. Vent holes were identified based on their distinct thermal signatures, which exhibited temperature contrasts relative to the surrounding rock surfaces. These thermal anomalies were further validated using RGB imagery, where visible openings, cracks, or depressions indicated potential air circulation pathways. Field observations provided an additional verification step, ensuring the accurate mapping of vent holes and rock surfaces across all seasonal datasets. Cross-referencing thermal and RGB data helped reduce misclassification errors, particularly in shaded areas where temperature variations alone may not reliably distinguish between surface features.

To evaluate the spatial variability in thermal dynamics influenced by the chimney effect, the scree was divided into two distinct sections: the lower scree, located near the forest edge, where permafrost retention is more likely [42], and the open upper scree, where the chimney effect is expected to generate warmer surface temperatures [43]. This segmentation allowed for a detailed seasonal comparison of temperature patterns and provided insights into how the chimney effect influences thermal conditions within the scree. For statistical analysis, vent holes and rock surfaces were mapped across all datasets. In the December dataset, 482 vent hole points and 533 rock surface points were recorded. In April, the dataset included 495 vent hole points and 335 rock surface points. The October dataset consisted of 431 vent hole points and 310 rock surface points. The February dataset was analyzed separately for nighttime and daytime conditions. During the nighttime survey, 351 vent hole points and 350 rock surface points were mapped, whereas in the daytime survey, 271 vent hole points and 210 rock surface points were identified. These mapped features were used to assess temperature variations between the two surface types under different seasonal and diurnal conditions, providing a comprehensive understanding of thermal behavior across the study site.

A detailed summary of the hardware and software components utilized in data processing is presented in Table S2 in the Supplementary Materials. The selected setup was designed to efficiently manage the computational demands associated with processing high-resolution thermal and optical imagery using Structure-from-Motion (SfM) techniques in Agisoft Metashape Professional. Additionally, the system was optimized to support the subsequent spatial analyses and statistical evaluations conducted in ArcGIS Pro.

Furthermore, thermal images captured within the forest sector were analyzed using the DJI Thermal Analysis Tool v3.3.2 (https://www.dji.com/global/downloads/softwares/dji-dtat3, accessed on 1 November 2024, [44]). This software allowed for a detailed examination of individual thermal images, incorporating previously calibrated physical parameters to ensure accurate temperature assessment. In December, a total of 16 images were analyzed, while in April, 37 images were included. No thermal images were collected in October for this sector, whereas in February, 12 images were examined. This assessment provided a comparative evaluation of temperature variations within specific vent holes, offering insights into their thermal dynamics across seasons.

This methodological approach generated comprehensive spatial and thermal datasets, enabling a detailed analysis of the scree slope’s thermal behavior and microclimatic characteristics.

3. Results

3.1. UAV-Derived Thermal and RGB Maps

The UAV-derived thermal orthomosaics reveal the spatial distribution of surface temperatures across the scree slope throughout different seasons. Each dataset (December 2023, April 2024, October 2024, and February 2025) highlights distinct temperature patterns, with notable differences between the upper and lower scree sections, vent holes, surrounding rock surfaces, and snow-covered areas (Figure 4, Figure 5, Figure 6 and Figure 7). Complementary RGB orthomosaics offer a visual reference for surface conditions, illustrating variations in vegetation, exposed rock, and snow cover.

The December 2023 dataset displays distinct thermal contrasts across the scree slope, with colder temperatures concentrated in the lower section and progressively increasing values toward the upper part (Figure 4). The RGB orthomosaic indicates the presence of snow patches, primarily at lower elevations. The thermal data further emphasize a well-defined gradient, showing a temperature difference of 10.3 °C across the slope. In contrast, the April dataset exhibits higher surface temperatures than those recorded in December, with a wider temperature range from −1.9 °C to 10 °C. The highest temperatures are observed in the upper scree section, while cooler values are concentrated near the forest boundary (Figure 5). The RGB orthomosaic reveals fully exposed rock surfaces with no remaining snow cover, along with denser vegetation along the scree perimeter.

The October dataset, collected under daytime and solar-exposed conditions, records the highest surface temperatures, reaching up to 53 °C in the upper scree section. The thermal orthomosaic reveals a distinct temperature gradient, with cooler values in the lower part and warmer temperatures in the middle and upper sections of the scree slope (Figure 6). The RGB orthomosaic further differentiates between shaded and sun-exposed areas, offering additional context for the observed thermal distribution.

The February 2025 dataset includes both nighttime and daytime thermal orthomosaics, enabling a comparative analysis of temperature variations under different conditions (Figure 7). The nighttime survey captures a temperature range from −5.51 °C to 8.51 °C, whereas the daytime thermal map records lower values, ranging from −9.49 °C to 3.14 °C. The RGB orthomosaic reveals a newly deposited snow layer, which appears more extensive in the daytime dataset compared to the nighttime survey.

3.2. Seasonal Thermal Patterns of the Open Scree

Figure 8 presents the distribution of temperature points corresponding to vent holes and rock surfaces on the thermal orthomosaic from the December 2023 survey, distinguishing between the upper and lower sections of the scree slope. Colder temperatures (−14 °C to −8 °C) are primarily concentrated in the lower scree section, while higher values (−4 °C to −2 °C) are recorded in the upper section. Intermediate temperature ranges (−7.99 °C to −4 °C) are distributed across the transition zone, illustrating a distinct thermal gradient.

Figure 9 illustrates the distribution of the mapped temperature points for vent holes and rock surfaces on the thermal orthomosaic from the April survey, differentiating between the upper and lower sections of the scree slope. The temperature values exhibit significant variation, with vent holes consistently showing lower temperatures compared to the surrounding rock surfaces. This trend extends to points located outside the model perimeter. Colder temperatures (−4 °C to −2 °C) are concentrated in the lower scree section, while progressively warmer values (6 °C to 10 °C) are recorded in the upper section. The transition zone contains temperature values ranging from −1.99 °C to 6 °C. Notably, rock surfaces display a more pronounced temperature gradient between the lower and upper scree sections, with cooler values near the forest boundary and higher temperatures in the upper slope.

Figure 10 illustrates the distribution of the mapped temperature points for vent holes and rock surfaces on the thermal orthomosaic from the October survey, differentiating between the upper and lower sections of the scree slope. The temperature values span a wide range, with cooler temperatures (2.46 °C to 9 °C) concentrated in the lower scree section, while progressively higher values (15 °C to 53 °C) are recorded in the upper section. The transition zone contains temperatures ranging from 9.01 °C to 15 °C. The spatial temperature distribution follows a distinct pattern, with vent holes primarily exhibiting temperatures between 3 °C and 11 °C, while rock surfaces display a broader range, with most values between 9 °C and 25 °C, particularly in the upper scree section. The highest temperatures, exceeding 40 °C, are recorded in the uppermost part of the scree, highlighting the pronounced thermal contrast across the slope.

The February 2025 dataset (Figure 11) provides both nighttime and daytime comparisons. At night, vent hole temperatures range from −3 °C to 5 °C, while rock surfaces vary from −8 °C to 8 °C. The lower scree records the coldest temperatures, particularly in depressions. During the daytime, fresh snow influences surface temperatures. Rock surfaces range from −9.49 °C to 3.14 °C, while vent holes, covered by deeper snow, maintain temperatures between −8 °C and −2 °C. Snow coverage reduces temperature fluctuations over vent holes, while rock surfaces show greater variability.

Figure 12 presents a seasonal comparison of vent hole and rock surface temperatures across all datasets, allowing for a more detailed evaluation of thermal contrasts. In December, vent holes in the lower scree register median temperatures of −8.82 °C, while rock surfaces reach −11.02 °C. The temperature distribution in December shows moderate variability, with a standard deviation of 2.39 °C for vent holes and 2.17 °C for rock surfaces. The differences between vent holes and rock surfaces indicate that vent holes retain slightly higher temperatures during winter months, while rock surfaces experience more pronounced cooling.

In April, vent holes warm to 1.68 °C, and rock surfaces reach 3.61 °C, showing an overall increase in temperature across both surface types. The broader temperature range recorded in April highlights the influence of increased solar radiation. The transition from winter to spring results in a more pronounced difference between the upper and lower scree sections, with vent holes warming at a slower rate compared to rock surfaces. The standard deviation values of 1.72 °C for vent holes and 2.03 °C for rock surfaces reflect greater thermal fluctuations, particularly in areas with higher solar exposure.

In October, the upper scree rock surface records the highest median temperature at 33.9 °C, while vent holes remain at lower values. The significant increase in rock surface temperatures compared to other seasons is primarily associated with strong daytime solar heating. The thermal differences between vent holes and rock surfaces in October are the most pronounced, as indicated by the highest recorded standard deviations among all datasets. Rock surfaces in the lower scree section exhibit a median of 9.39 °C, while vent holes maintain values around 4.9 °C, demonstrating a strong contrast in heat retention between these surface types.

The February dataset indicates that snow-covered vent holes range from −7.6 °C to −6.2 °C, while rock surfaces register −5.6 °C and −2.6 °C, depending on snow accumulation. The variability in temperature is lowest in February, as snow cover reduces the extent of surface temperature fluctuations. The standard deviation values of 0.55 °C for vent holes and 0.63 °C for rock surfaces indicate that temperature variations are more stable compared to other seasons. This is further supported by the minimal temperature differences recorded between vent holes and rock surfaces in February, where the difference between median values is 0.19 °C, significantly lower than the differences observed in April and October.

The temperature differences between vent holes and rock surfaces vary across seasons. The upper scree maintains higher temperatures than the lower scree, except in conditions where snow moderates surface temperature fluctuations. The largest contrasts occur in October and April, while the lowest values are recorded in December and February. The thermal distribution across the scree slope reflects seasonal changes in snow cover and solar exposure. The statistical values from Figure 12 provide a more quantitative assessment of these differences, reinforcing the observed seasonal trends in temperature variability between vent holes and rock surfaces.

3.3. Results from the Forest Sector Imagery

The analysis of thermal contrasts between vent holes and adjacent rock surfaces within the forest sector provides additional insights into localized thermal dynamics. This section focuses on the temperature variations observed in December 2023, April 2024, and February 2025, with an emphasis on the influence of seasonal factors such as snow cover, air circulation, and surface radiation exposure. Figure 13, Figure 14 and Figure 15 present thermal and RGB imagery highlighting the spatial distribution of temperature differences in the forest sector. A statistical comparison of temperature variations in the forest sector is summarized in

Table 1. Statistical summary of temperature measurements for vent holes and adjacent rock surfaces in the forest sector, outside the scree outline.

Statistic Type	Vent Holes/Rock Surfaces (December 23) (°C)	Temperature Diff (December 23) (°C)	Vent Holes/ Rock Surfaces (April 24) (°C)	Temperature Diff (April 24) (°C)	Vent Holes/ Rock Surfaces (February 25) (°C)	Temperature Diff (February 25) (°C)
Average	−6.03/−3.88	−2.15	0.08/3.86	3.78	−2.91/−1.69	−1.22
Median	−5.8/−4.25	1.59	−0.4/3.4	3.8	−1.69/−1.5	−0.19
Minimum	−9.9/−7.4	2.5	−3.6/0.2	−3.8	−4.2/−2.9	1.3
Maximum	−2.6/−0.5	3.1	4.9/9.6	4.7	−1.6/−0.4	2
StDev	2.39/2.17	−0.22	1.72/2.03	0.31	0.55/0.63	0.08

, detailing differences between vent holes and rock surfaces across the December, April, and February datasets. In December, vent holes exhibited lower average temperatures than rock surfaces, with moderate thermal differences between the two surface types. In April, the temperature gap widened, with rock surfaces experiencing the highest variability due to stronger diurnal solar heating. In February, the overall temperature difference between vent holes and rock surfaces was smaller than in December and April, indicating more uniform cooling under snow-covered conditions. The lower standard deviation in February suggests reduced variability compared to the other two periods.

In December 2023 (Figure 13), vent holes generally exhibited lower temperatures than the surrounding rock surfaces. The thermal images reveal that temperature differences varied across locations, with vent holes maintaining more stable conditions. While the adjacent rock surfaces experienced fluctuations, vent holes remained within a consistent thermal range, reinforcing their role in facilitating cold air outflow. These findings align with winter chimney circulation processes, where dense, cold air pools in lower sections and is released through vent holes. The cooling effect of vent holes is particularly evident near the forest boundary, where thermal contrasts are more pronounced.

By April 2024 (Figure 14), temperature variations between vent holes and rock surfaces increased compared to December. Rock surfaces exhibited significantly higher temperatures due to increased solar radiation exposure, while vent holes remained comparatively cooler. This seasonal shift highlights the moderating influence of the chimney circulation mechanism during spring, where vent holes continue to facilitate cool air movement while the surrounding rock surfaces warm rapidly. The increased thermal contrast underscores the role of insolation in controlling surface temperature fluctuations.

In February 2025 (Figure 15), thermal contrasts were influenced by snow cover thickness. Vent holes consistently exhibited lower temperatures than the adjacent rock surfaces, with variations depending on the depth of the overlying snow layer. In areas with substantial snow accumulation, vent hole temperatures were among the lowest recorded, whereas rock surfaces with thinner snow layers retained slightly higher values. These localized contrasts indicate that snow cover affects the surface temperature distribution by insulating the substrate. The thermal gradients visible in the imagery suggest that air circulation through vent holes remains active even under snow-covered conditions, albeit with reduced intensity compared to the snow-free periods.

Calibration of the February 2025 vent hole temperature dataset was performed by correlating UAV-derived thermal measurements with in situ ground surface temperature (GST) data (Figure 16). The adjustment process refined the distance parameter to optimize alignment between the datasets, resulting in a strong correlation (R² = 0.9423). This calibration confirmed that vent hole temperatures obtained from UAV-based thermal imagery were consistently lower than GST measurements while maintaining the same overall trend. The regression analysis provides a reference for temperature retrieval accuracy and highlights the importance of accounting for distance effects in UAV thermal surveys. Individual data points correspond to different vent hole locations, with temperatures ranging from approximately −4.1 °C to −2 °C for GST and −4 °C to −1.5 °C for the adjusted vent hole temperatures.

A detailed summary of the temperature statistics for the February 2025 forest sector dataset is presented in Table S3 in the Supplementary Materials. This includes key statistical metrics such as minimum, maximum, and average temperatures, alongside GST measurements, air temperature, and the distance parameter. The table provides a comprehensive overview of the calibration adjustments applied to ensure accuracy in temperature retrieval for vent hole measurements.

Overall, the forest sector results demonstrate that vent holes consistently maintain lower temperatures than adjacent rock surfaces across all seasons. The thermal contrasts are most pronounced in April, when increased solar radiation enhances surface heating, while in February, snow cover influences the magnitude of temperature differences. These findings reinforce the seasonal dynamics of chimney circulation in the forested scree slope, further supporting the observed patterns from the open scree section.

4. Discussion

The observed thermal differences in this study are largely explained by the differentiated circulation of air within the block accumulation [45]. The results indicated that air circulation within the openwork deposits is more complicated than initially considered [7], revealing significant seasonal variations influenced by factors such as land cover, slope, clasts dimensions, porosity, and the presence/absence of snow cover [46]. However, our findings suggest that chimney circulation is present during both winter and spring, facilitating the cooling of lower regions of the scree slope, such as in other similar sites [4,5,11,16]. The direction and velocity of airflow within loose sediment accumulations fluctuate mainly due to the thermal contrast between external and ground air, creating a driving pressure gradient [43]. During winter, warm air, being lighter than the colder atmospheric air, rises and escapes from the upper part of the slope (Figure 7). As a result of this movement, cold air is pulled into the lower section of the scree and can rapidly cool the substrate [43]. This process is clearly observed at the Detunata cold scree slope before the onset of snow cover when the lowest ground surface temperatures are concentrated in the lower third of the slope. Once the ground is covered by a substantial snow layer, the intensity of chimney circulation appears to weaken at the Detunata site (Figure 11). Even so, the air expelled through the funnels in the upper part of the slope is a few degrees warmer than that in the lower part (Figure 11). When the outside air temperature exceeds that of the interior, dense, cold air is gravitationally released at the base of the talus slope [4]. As a result, external warm air is diffusely drawn into the upper part of the slope [5]. This mechanism is particularly evident during the April and October campaigns at Detunata, when ground cooling is most pronounced in the lower part of the slope, as highlighted in previous studies [4,5,6,7,8,16,17,43].

Snow cover plays a significant role in periglacial environments, as its onset prevents ground cooling during the cold season due to its thermal insulating properties. However, at sites where snow thickness is relatively thin, such as Detunata, and where internal ventilation is enhanced by the high porosity of debris, the snow cover does not have a significant impact on ground cooling. According to [43], the winter phase of the chimney mechanism persists despite the presence of continuous snow cover, which can range from 1 to 3 m thick. Thanks to efficient air circulation at Detunata, a dense network of snow funnels forms during winter, facilitating continuous air exchange between the substrate and the atmosphere [7]. Furthermore, in the upper part of the slope, where warm air is expelled during winter, the outflow of warm air inhibits the accumulation of snow [7]. However, in February 2025, we noted that immediately after the fresh snow began to accumulate, the chimney circulation decreased slightly, leading to a reduction in the thermal differences between the lower and upper parts of the scree slope.

The distribution of the lowest temperatures revealed by infrared thermography aligns closely with areas exhibiting a high probability of permafrost occurrence [7,30]. Similar to other marginal periglacial sites in the Romanian Carpathians [47,48,49,50], permafrost at Detunata is relatively thin, characterized by a thick active layer, and strongly influenced by local topoclimatic conditions [7].

The temperature variations between vent holes and the surrounding rock surfaces further confirm the presence of the chimney effect. During winter, vent holes demonstrated more stable and nearly uniform temperatures compared to the adjacent rock surfaces, with colder values near the forest boundary. This suggests that vent holes serve as conduits for upward air movement, releasing relatively warmer air at the top of the talus mass. In contrast, during spring, an inverse trend was observed: vent holes were considerably cooler than the surrounding rock surfaces. This phenomenon is likely due to increased solar radiation directly heating the exposed rock surfaces during the day, while the chimney effect continues to moderate the temperature distribution within the block accumulation. These findings highlight the dynamic thermal behavior of the scree slope, influenced by seasonal changes, as observed in similar low mountain ranges in Germany [16]. They are consistent with the temperature regimes characteristics of low-altitude cold scree slopes elsewhere [43,51,52].

The results demonstrate clear seasonal and spatial variations in temperature across the scree slope, driven by the complex air circulation and distinct thermal behaviors of vent holes and rock surfaces. For vent holes, the influence of the Balch effect cannot be excluded, as it may vary depending on the configuration of interstices within the block accumulations [29]. These findings underscore the importance of seasonal thermal dynamics in shaping the microclimatic patterns of the scree slope [16].

At Detunata Goală, the chimney effect plays a crucial role in shaping the thermal dynamics within the scree slope, as described by [7]. The mechanism involves cold air being drawn into the lower parts of the talus, ascending through the porous rock debris, and warming as it rises before being expelled as warmer air at the upper parts of the talus [4]. Compared to the findings of [7], this study’s results align with the conceptual model of seasonal chimney circulation, with some key differences attributable to seasonal variations and methodological approaches.

Ref. [7] identified two distinct cooling and warming zones on the talus surface during winter. This study’s winter orthomosaic highlights lower temperatures in areas previously identified as warmer, potentially indicating temporal shifts in the chimney effect’s dynamics due to external climatic factors. Additionally, the thermal orthomosaic shows a consistently cold zone on the northeastern side of the scree slope, present in all the seasons, which further corroborates the persistent influence of the chimney effect in this region. The presence of snow cover likely contributed to lower temperature readings in this area, as snow acts as a cold surface with high reflectivity, reducing the absorption of thermal radiation. This reflective property of snow not only affects surface temperatures but also influences the overall thermal dynamics of the scree slope during the winter season.

Low-altitude permafrost sites, situated well below the typical altitudinal limit for permafrost occurrence, are exceptionally rare in the scientific literature [5]. The limited existing studies mainly concentrate on mapping permafrost using geophysical methods and boreholes [8], along with measuring ground temperatures via data loggers and documenting BTS conditions at the close of the cold season [7]. In some cases [11], individual thermal images captured for specific voids are used to support the cooling of the substrate. However, no previous study has employed UAV thermal imagery to map the spatial distribution of ground surface temperatures across different seasons. This approach has significant strengths, as it allows for the rapid acquisition of ground temperature data at an unprecedented spatial resolution. Despite using different methodological approaches, our findings are consistent with previous results reported at similar sites in the Swiss Alps [8], Austrian Alps [6], Central German Uplands [16], Japan [2], and the White Mountains (USA) [11], which also highlighted that the chimney effect is responsible for cooling the lower part of the scree deposits.

Operational and environmental factors also influenced data collection and interpretation. For example, during winter flights, the low ambient temperatures required preheating the UAV batteries to an operational temperature of 23–24 °C, consuming 10–15% of the battery charge. While nighttime conditions reduced the influence of solar radiation, they increased the risks of collisions and compromised anti-collision sensor functionality. External factors influencing UAV temperature readings, such as emissivity, air temperature, air humidity, wind direction and speed, as well as distance to objects, must be carefully accounted for to ensure the most accurate data analysis possible. These variables are subject to frequent temporal changes, particularly during extended flight times, and can significantly affect thermal radiation absorption and recorded temperatures. For example, during daytime flights, the influence of solar radiation can vary rapidly with shifting cloud cover, necessitating real-time adjustments to account for these changes. While the relatively small size of the study site at Detunata Goală allowed for short flight times, minimizing the impact of temporal changes in these variables, this consideration is critical for larger or more complex sites. To ensure reliable thermal orthomosaics, overlaps of at least 75% between images were maintained, and RTK signals were used to improve georeferencing accuracy. However, RTK signal reliability was reduced in forested areas. Consistent flight resolution across campaigns was critical for meaningful spatial comparisons, particularly when examining seasonal changes in the thermal dynamics of the scree slope. Additionally, a small area in the lower talus shows slightly higher temperatures than its surroundings, likely caused by the operator’s presence during data acquisition. This highlights the need for careful attention during fieldwork to minimize human-induced temperature artifacts in the dataset.

The calibration of UAV-derived thermal data is subject to multiple uncertainties related to atmospheric and surface conditions. Key factors influencing accuracy include air temperature, humidity, emissivity, and sensor-to-surface distance, all of which must be carefully measured on-site and adjusted during data processing. High humidity levels (above 85%) have been observed to overcorrect UAV thermal readings, leading to artificially lower temperatures, while low humidity (50–60%) can result in an overestimation of surface temperatures. A balanced humidity range (70–75%) has been found to provide the most reliable correction, aligning UAV readings with GST and in situ measurements. Notably, changing the humidity setting from 75% to 85% results in approximately a 1 °C decrease in recorded temperatures, while adjusting emissivity between 0.94 and 0.99 alters temperature values by only about 0.1 °C. In contrast, the sensor distance parameter plays a critical role in thermal accuracy, as improper selection can lead to temperature variations exceeding 5 °C. Determining the appropriate distance-to-subject value is essential, as even small adjustments can result in notable changes in recorded temperatures. In this study, calibration was performed by adjusting the distance parameter based on GST measurements, reinforcing the need for precise field data to refine UAV thermal imagery. These uncertainties emphasize the importance of integrating ground-based measurements for validation and employing systematic calibration procedures to enhance the reliability of UAV thermal datasets.

Integrating additional ground-based temperature sensors, such as those utilized by [7], positioned in shaded areas to avoid direct solar influence, could enhance the validation of UAV-derived thermal data. These sensors would provide reliable reference temperatures, improving the accuracy of thermal analyses across different environmental conditions. The integration of UAV thermal imagery with in situ measurements has the potential to refine the understanding of temperature dynamics within talus slopes, particularly regarding the chimney effect and its seasonal variability. This study underscores the value of UAV-based thermal surveys when combined with systematic calibration approaches and a thorough consideration of local environmental factors.

5. Conclusions

This study analyzed the seasonal thermal dynamics of the scree slope at Detunata Goală using UAV-based infrared thermography, emphasizing temperature gradients and the chimney effect’s role in air circulation. The results indicate that colder air accumulates in the lower scree, while warmer air is expelled at higher elevations, with the strongest contrasts observed in December and April and the lowest variations in February due to snow cover. Temperature differences between surface types are also evident, as rock surfaces experience more pronounced variations than vent holes, and the upper debris accumulation consistently exhibits higher temperatures than the lower scree. The largest discrepancies occur in October and April, while the coldest temperatures are recorded in December and February. Vent holes remain colder than rock surfaces, particularly in warmer months. UAV-derived, high-resolution infrared thermography effectively captured these microclimatic variations, demonstrating its utility in periglacial research. Integrating UAV thermal data with geophysical surveys, time-lapse imaging, and long-term ground surface temperature monitoring can enhance the assessment of periglacial environments. Furthermore, continued ground-based validation and careful calibration of physical parameters are crucial for improving the accuracy of thermal mapping.