Cliff-dominated landscapes are globally widespread, commonly occurring in sandstone- and conglomerate-dominated tablelands as well as in red-bed terrains. Despite their ubiquity, these terrains lack automated, quantitative classification frameworks because traditional methods struggle to delineate discrete landform objects. This methodological limitation constrains comparative geomorphology in escarpment landscapes worldwide. We present the first automated, morphometric framework for the quantitative classification of Danxia landforms characterized by cliff-dominated morphology in red beds. Our four-step workflow integrates: (i) cliff detection via an effective-gradient criterion combined with local-slope constraints; (ii) delineation of scenic landform units through reversed-DSM watershed segmentation; (iii) extraction of core landforms (residual hills) using base-plane and median-reference-surface intersections; and (iv) rule-based classification employing planform size and shape, and profile prominence indices. Applied to Danxiashan UNESCO Global Geopark, the framework extracted 351 core landforms and classified them into six morphotypes: mesas, ramiforms, walls, ridges, columns, and stumps. The classification achieves high overall accuracy (OA = 94.0%) against reference data visually interpreted from high-resolution 3D models. Notably, ramiform landforms—highly irregular, branching residuals—emerge as the dominant morphotype (41.3%), representing a transitional evolutionary stage between intact mesas and smaller residual hills. The observed morphologic sequence (mesa → ramiform → walls/ridges/columns/stumps) suggests that landscape evolution in these humid, densely jointed terrains is driven primarily by fracture-guided internal dissection rather than simple marginal scarp retreat. This framework provides a transferable geomorphometric workflow for quantitative classification of cliff-dominated landforms and may support geomorphic diversity assessment and comparative studies of Danxia geomorphology across different regions.

Headwater catchments, comprising hillslopes, valley heads, and colluvial valleys, are critical sources of water, sediments, and nutrients for downstream river networks. As mass movements dominate these landscapes, topographic relief has long been recognized as a primary factor controlling geomorphic processes and the spatial arrangement of geomorphic units in headwater catchments. However, how the transition from hillslope- to channel-dominated processes varies with relief remains debated. Particularly, the ways in which valley head location—defined as the uppermost tip of an unchanneled valley—and the lower boundaries of headwater catchments—defined by the downstream limit of colluvial dominance—scale with relief are not fully understood. To address these questions, we analyzed a drainage basin in South Korea that spans a wide range of relief. We delineated valley heads and headwater catchment extents, and quantified the likelihoods of hillslope materials reaching both valley heads and downstream river networks. Our results reveal that valley heads shift downslope with increasing relief and that the lower boundaries of colluvial channels extend downstream, enlarging headwater catchments. Likewise, both likelihoods of sediment connectivity increase with relief. These findings demonstrate that both valley head position and headwater catchment extent are governed by relief through its effects on hillslope material supply, colluvial valley morphology, and debris flow transport. Notably, the observed positive correlation between valley head source area and relief challenges the stream power-based channel initiation model, which exhibits a negative slope-area relationship. This study advances understanding of geomorphic processes in headwater landscapes and provides insights for managing mountainous environments.

<p>In mountainous areas in the Republic of Korea, landslides occur primarily due to heavy rainfall in the summer monsoon season. While these landslides tend to occur seasonally in summer, the rainfalls that have triggered landslides are complex and varied. Temporal prediction of landslides based on the variables of rainfall events mainly uses two variables: intensity-duration or cumulative event rainfall-duration. However, it does not consider the antecedent rainfall, another critical factor that could characterize various complex rainfalls regarding landslide occurrences. Here, we first attempted to determine critical rainfall variables and their threshold values for landslide occurring using the decision tree method necessary to consider multiple rainfall variables simultaneously. We then classified landslide-triggering rainfall based on the identified critical variables using the K-means clustering method. We chose as the study area Chuncheon in the middle of the Korean Peninsula, an eroded granite basin surrounded by schist and gneiss mountains, since it has not been affected hardly by earthquakes and thus is suitable for studying rainfall-induced landslides. According to the decision tree analysis, cumulative rainfall and 5-day antecedent rainfalls were determined as critical variables, implying that considering antecedent and cumulative rainfall simultaneously is significant for landslide prediction. The K-means clustering analysis classified landslide-triggering rainfalls into four types: 1) low cumulative rainfall (198.6 ± 90.9 mm) with high antecedent rainfall for seven days prior to the landslide, 2) medium cumulative rainfall (308.3 ± 81.1 mm) with a gradual increase in antecedent rainfall for four weeks, 3) high cumulative rainfall (534.5 ± 85.7 mm) with low antecedent rainfall for four weeks, and 4) high cumulative rainfall (538.4 ± 59.8 mm) with a gradual decrease in antecedent rainfall for four weeks. In particular, the high cumulative rainfall after gradually decreased antecedent rainfall caused the most frequent landslides. Our results suggest that the threshold of cumulative rainfall varies with the antecedent rainfall pattern and that antecedent rainfall data of at least four weeks have meaningful information in forecasting and preparedness for landslide occurrence.</p>

In tectonically stable regions, bedrock channels develop along deep and narrow valleys that are usually associated with resistant lithology, which enables the transport capacity of these channels to exceed the upstream sediment supply. However, bedrock channels are not restricted to the reaches with a high transport capacity. They are often located in reaches with a lower transport capacity in wide valleys underlain by erodible bedrock. Even though bedrock channels are distributed over a wide range of lithologies, the effects of erosional resistance on the transport capacity required for bedrock channel formation have not been fully clarified. We investigated the bedrock channels in the Seo River catchment in South Korea to determine the relationship between erosional resistance and the occurrence and morphology of bedrock channels. We found that more than half of the bedrock channel segments were in resistant regions. However, a considerable portion also occurred in erodible regions. Most of the bedrock channels in the resistant regions were steep and narrow, generating a higher unit stream power. Therefore, these channels had long stretches of exposed bedrock with actively eroded features such as well-abraded surfaces. In the erodible regions, most of the bedrock channels were wide and gentle, leading to a lower unit stream power. Thus, the exposed bedrock surface was poorly polished, and the channel morphology preserved more of the underlying bedrock's joint or bedding structures. These results suggest that erosional resistance affects the occurrence of bedrock channels by adjusting the threshold stream power required for bedrock incision. Moreover, lithology strongly regulates the bedrock channel morphology in erodible regions.

지형학 관련 답사는 지형경관이 수려하거나 지형학 이론을 모식적으로 설명할 수 있는 장소를 위주로 수행되어왔다. 사범대학의 지형답사는 학생 삶의 공간에서 지형을 인지하고 지형학 지식을 스스로 구성해 나가며 지형학 지식이 자신 삶에 필요한 것임을 느낄 수 있도록 일상생활 공간, 즉 도시의 지형경관을 인지하는 것에서 출발할 필요가 있다. 여기서는 학교 인근의 도시 지형 이해에 초점을 둔 지형답사 프로그램을 제안하였다. 서울의 주변부 관악산 도림천 상류 유역을 대상으로 진행된 이 프로그램은 인간에 의한 변형이 적은 지역과 도시화된 지역을 비교하도록 구성되었고 학생이 도시화 이전의 지형경관을 복원하는 활동을 포함하고 있다. 또한 답사는 일상의 지형경관을 이해하는데 주안점을 두었으며, 100년 이내의 지형학적 현상과 인위적인 배수 체계를 관찰하는 장소를 포함했다. 참가 학생은 도시의 지형경관을 이해하고 나아가 도시 환경문제와 자연재해를 지형학적 맥락에서 해석하게 되었다. 따라서 이 프로그램은 지리교육의 지형 단원 목표를 달성하는데 효율적인 지형교육 수단으로 판단된다.

The mountain catchments where deep valley networks develop continuously are the main places for agriculture and forestry activities. Understanding the geomorphological characteristics of the valley is primarily needed for managing the natural landscapes in mountain catchments. With respect to the characterization of valley geomorphology, delineating the valley bottom is the most fundamental step. However, previous studies on valley bottom extraction focused on wide and unconfined valleys; they excluded colluvial valleys in headwater catchments. Here, we evaluated the suitability of two methods widely used to extract valley bottoms, specifically applying to a mountainous catchment where colluvial valleys dominate. Our results demonstrated that the Valley Confinement Algorithm (VCA) that delineates valley bottoms by adjusting flood height was more suitable than the Geomorphons-based method that extracts valley bottoms according to the topographic patterns corresponding to valley bottom morphologies. It is mainly because the Geomoprhons-based method should be applied differently to each valley type. In contrast, the VCA applies consistently to the whole area of the study catchment. The extracted valley bottom width is proportional to the power of the drainage area in this study area. In particular, the valley bottom width of fluvial valley increase more rapidly than that of colluvial valley.

<p>Valleys associated with convergent topography are common features on the earth’s surface. Valley bottoms along which most rivers flow have been considered geomorphologically important since stream channels adjust their forms within them and, consequently, the width of the valley bottom determines the planform and type of channel reach. The headwater streams comprise transitional and colluvial channels and account for nearly 70 to 80 percent of the river network. They play a role in the movement or retention of sediments and nutrients transferred from surrounding hillslopes, and the provision of habitats for the various aquatic and riparian organisms in mountainous regions. Thus, accurate delineation of the valley bottoms, including colluvial valley portions, is fundamental for managing mountain rivers. In this study, we applied widely used methods for extracting valley bottoms into an intermontane catchment where colluvial valleys dominate (i.e., the Huengjeong basin in South Korea), verified and compared their results to suggest a practical procedure for applying these methods to the uppermost basins in a mountainous region. We compared two methods: the Valley Confinement Algorithm (VCA) which extracts valley bottoms based on flood-prone areas estimation, and the Geomorphon-based method using geomorphic units matching with the topography of valley bottoms. Our results show that the valley bottoms delineated from the VCA with flooding factor 1, smaller than the value (i.e., 3) used for extracting downstream unconfined valley bottoms, characterize well the valley bottoms in the study area. However, the valley bottom width becomes wider where low order tributary flows over flat areas such as river terraces. For the Geomorphon-based method, it turned out that the geomorphic units based on a Geomorphon calculated using a single set of parameter values (i.e., analysis radius and flatness threshold) did not properly capture both headwater colluvial valley and downstream fluvial valleys. Therefore, we extracted valley bottoms for colluvial and fluvial valley bottoms, respectively: Colluvial valley bottoms could be delineated by a set of geomorphic units (valley, hollow, slope, and spur) based on a Geomorphon calculated using a 50 m analysis radius and 0.5-degree flatness threshold; The fluvial valley bottoms could be delineated by a set of geomorphic units (valleys, depressions, foot slopes, and flats) based on 125 m analysis radius and 5-degree flatness threshold. While choosing the geomorphic units corresponding to valley bottoms for the Geomorphon-based method is highly subjective and time-consuming, the VCA could easily detect flood-prone areas below river terraces by only adjusting flooding depth. Thus, the VCA is thought to be suitable for extracting valley bottoms from colluvial valley-dominated areas.  </p>

<p>Headwater catchments, each of which consists of hillslopes, valley heads, and colluvial channels, make up a major portion of a drainage basin, supplying sediment, water, and nutrients downstream. In a headwater catchment, valley heads where hillslope diffusive transport transits to fluvial transport play an important role in channel initiation. Headwater catchments where mass movements are dominated are sensitive to human activities. Human activities in headwater catchments, such as logging and crop cultivation, change the rate of hillslope erosion, thereby increasing sediment inputs and leading to channel form change and stream habitat destruction. In recent years, such human activities have increased significantly in headwater catchments. As such, delineations of the extent of headwater catchment and valley head initiation become increasingly important for watershed protection and management. Previous studies have shown that the area of the headwater catchment ranges from 10<sup>4</sup> to 10<sup>6</sup> m<sup>2</sup>, but little is known about what factors affect its range. The evolution of headwater catchment topography is mainly determined by surface processes such as landslides and runoff. The rates of these processes vary depending on the hillslope gradient closely related to topographic relief. To understand the impacts of relief on the valley head initiation and the extent of headwater catchment, we analyzed the slope-area relations of the Seo River drainage in South Korea. Firstly, we found that the upslope area at the valley head shows a weak positive correlation with relief. This finding seems to be associated with hillslope material input to fill valley heads. Steep hillslopes in a high relief region could induce more hillslope material supply, consequently filling valley heads. Such abundant flux into valley heads probably enhances the hillslope length and makes valley head initiation downstream. Secondly, the upslope area of the headwater catchment, which is set by the downstream limit of the colluvial channel increased exponentially with relief. This exponential correlation would be related to the length of debris flow-dominated channel. In high relief regions where the channel slope is steeper, debris flows scour for a further distance, resulting longer colluvial channels. These results reveal the importance of relief as controls on valley head initiation and headwater catchment extent.</p>

High-elevation steep escarpments in the passive continental margins are known to have resulted from long-lasting erosion following rifting-related marginal uplift. However, recent studies have revealed that rifting-related processes alone do not explain the high-elevation marginal escarpment. Particularly, post-rift rejuvenation could contribute to the growth of the relief of the long-lived escarpment. A steep escarpment rising above 1,700 m occurred along the eastern margin of the Korean Peninsula. It is known to be a compartment of the passive continental margin, which has been tectonically stable since the extension of a back-arc sea (i.e., the East Sea). Nevertheless, coastal terraces indicative of enhanced tectonic uplift are observed along the eastern coastline. Thus, the steep escarpment has been tectonically reactivated since the Late Quaternary. Knickzone is a geomorphic marker that can decipher the role of tectonics in landscape evolution. This study examined and classified the knickzones of the catchments draining the marginal escarpment, and simulated the response of river profiles to an enhanced tectonic forcing to geomorphologically assess the post-rift tectonic perturbation and understand its effects on the evolution of the marginal escarpment. Firstly, the knickzone analysis identified 52 knickzones, of which 15 were slope-break knickzones, usually interpreted as being transient. However, all slope-break knickzones were not transient but were stationary since they were associated with spatial heterogeneity in erosional resistance. Secondly, river profile evolution modeling illustrated that an abrupt slope break of a transient knickzone developed by an enhanced uplift rate is diffused rapidly during its migration under the condition of a steep marginal escarpment with high channel concavity. Additionally, the tilted uplift pattern induced a newly adjusted river profile of a lower channel concavity, consequently decreasing the steepness of the marginal escarpment. In conclusion, the marginal escarpment in the study area would have been being adjusted since the post-rift tectonic enhancement without the development of typical slope-break transient knickzones, reducing its steepness. These finding suggest that the slope-break knickzone should be used carefully in assessing landscape transience of the marginal escarpment.

Steep and narrow escarpments develop along the eastern margin of the Korean Peninsula. They are compartments of a passive continental margin and thus have been considered tectonically stable. In contrast to the traditional notion, geomorphic markers indicative of the enhanced tectonic uplift since the Late Quaternary (i.e., coastal terraces at several different altitudes) have been observed along the eastern coastal areas of the peninsula. Therefore, the steep escarpments in the eastern margin are assumed to be tectonically reactivated. However, the spatial magnitude and timing of the reactivation and how the escarpments have responded to the reactivation have not been well studied. Knickzone is a typical geomorphic marker, which has long been utilized for deciphering the history and distribution of tectonics. Here, we examined the knickzones of the marginal escarpments, where transient knickzones are likely to be observed, in order to understand the spatial pattern of the Late Quaternary reactivation and its effects on the evolution of the marginal escarpments. We used SRTM 1 arc-second DEMs, satellite images with fine resolution, and geological maps to identify and classify knickzones. We also conducted field surveys for the verification of the identified knickzones. As a result of the knickzone analysis, 46 knickzones were identified in the study catchments. Their mean length and gradient are 461 m and 0.19 m/m, respectively. Most knickzones are at relatively high altitudes (i.e., median elevation 532 m) and thus are placed far from the coast. According to the classification of the identified knickzones, they are formed mainly due to varying rock types (11) or changes in lithologic features of the same rock type (e.g., weathering degree of rocks) (31). Few of them are associated with the accumulation of coarse sediments at a channel junction (3) and meander neck cut-off (1). This result implies that all identified knickzones in the study catchments are stationary rather than transient. Consequently, it postulates that the Late Quaternary tectonic forcing was insufficient to generate any transient knickzone. Otherwise, potential transient knickzones due to the reactivation might have disappeared rapidly during their upstream migration, which seems highly relevant to the high concavity of the stream profiles in the drainage basins of the escarpments. Additionally, the result suggests that transient knickzone is not a good indicator for interpreting the responses of the marginal escarpments to the reactivation during the Late Quaternary.

Bedrock river is rock-bound, its bed and banks are composed mainly of in-place bedrock. Bedrock channel reaches, commonly short and intermittent, often occur where transport capacity exceeds bedload sediment flux. Despite the abundant research on the typical patterns of alluvial channel reaches, the distribution of bedrock channels has not been well studied. Rock type may affect the occurrence of bedrock channels because the strength, joint density, and erosion process of bedrock vary depending on the rock type. Previous studies have viewed the bedrock channel occurrence in the aspect of the excessive sediment transport capacity, but the influence of lithology has not been considered in the literature. To understand the influence of lithology on bedrock channel occurrence in a drainage basin-scale, we investigated the distribution of bedrock channels in relation to varying lithology and unit stream power along the Seogang River in South Korea. We used satellite images with high resolution for the identification of bedrock channel reaches and then verified them through field surveys. Geological maps and 1 arc-second SRTM DEMs were used to analyze lithological effects and calculate unit stream power. As a result of the analysis, we identified 94 bedrock channels in the studied river, varying depending on lithologies. The frequency of bedrock channels in granitic gneiss areas (0.73/km) is much higher than those in the other rock type areas (granite areas, 0.57/km; limestone areas, 0.16/km). In the more frequent granitic gneiss areas, the bedrock channels are steepened (average channel slope: 0.0074 m/m) and narrow (average channel width: 65 m) and mainly reside within steepened and narrow (average valley width: 123 m) rock-bound valleys so that their occurrence is mainly associated with high unit stream power. In contrast, the bedrock channels over the other lithologies are wider (89 m) and lower-gradient (0.0056 m/m) and occur along flat and broad valleys (391 m). Consequently, the bedrock channels in the studied river were divided into two types: confined and unconfined bedrock channels. The confined bedrock channels are within the steepened and narrow valleys composed of resistant granitic gneiss and show the evidence for recent bedrock incision processes. However, the unconfined bedrock channels are mainly within the broad and flat valleys of weak saprolites and limestone with high joint density have lower unit stream power and don't show any marker for bedrock incision. In conclusion, high-relief landscape mainly composed of more resistant rocks generates steep and narrow valleys, which leads to the formation of continuous and actively incising bedrock channels. However, low-relief landscape underlain by non-resistant rocks shows wider and lower-gradient channels, with intermittent bedrock channels due to locally more resistant rock bodies.

Along the eastern margin of the Korean Peninsula, a coastal mountain range spanning over 800 km with summits above 1500 m faces the East Sea (or Sea of Japan), the back-arc sea behind the Japanese Islands. Two contrasting hypotheses exist regarding the tectonic history of this coastal mountain range: long-lasting and progressive uplifts from the Early Tertiary to the Late Quaternary, and a short and intensive uplift during the Early Miocene. However, to date, no consensus has been reached. Here, we studied the spatial distribution of knickzones to understand the formation period and development pattern of this coastal mountain range. We extracted the knickzones in a drainage basin from digital elevation models, and investigated whether or not they are transient knickzones induced by the development of the coastal mountain range. We found that all identified knickzones were stationary, which was verified by slope-area and chi-elevation analyses. This implies that sufficient time has passed for all transient knickzones relevant to the growth of the mountain range to migrate up to the catchment boundary and disappear. We then calculated the time spent for the migration of transient knickzones from the outlet to their stream heads to be at least 5.1 to 10.6 Myr. Therefore, our results suggest that the current form of the coastal mountain range had been built at least before 5.1 Myr ago and has reached a quasi-equilibrium state up to the present, thus invalidating the prevailing hypothesis of the long-lasting and progressive development until the Late Quaternary.

Although most natural streams have concave-upward longitudinal profiles, there are conflicting perspectives about whether an alluvial channel is less concave than a bedrock channel. Alluvial channels can be classified as coarse-bed and fine-bed channels, depending on the bed grain size. Both are transport limited, but the threshold of motion differs greatly. Whereas a coarse-bed alluvial channel can be claimed to be as concave as a bedrock channel, we claim that a fine-bed channel is distinguishably less concave. We derive the concavity index of a fine-bed alluvial channel using the power-law relationships emergent at a steady-state river network. For known ranges of the scaling parameters, our formulation informs a range of concavity index as 0.07 +/- 0.09 for a fine-bed alluvial channel. Our analyses of previous laboratory experiments and real fine-bed alluvial channels in the midwestern U.S. and northern Europe also support our conclusions, i.e., small profile concavity of steady-state fine-bed alluvial channels.

Tectonically inactive since the middle Miocene, the Korean Peninsula is generally considered a typical passive continental margin with asymmetric relief with only low levels of seismic activity. Recent reports of uplift rates as high as 150–350mmka−1 from geomorphic markers using numerical dating methods (e.g. radiocarbon, optically stimulated luminescence (OSL), and cosmogenic nuclides), however, cast doubt on the tectonic quiescence of Korea. Thus, we evaluate the geomorphic status of eastern Korea using denudation rates at geomorphic timescales of 103–106years, which are equivalent to the rates previously reported at local points. To infer regional denudation rates, we measured catchment-wide denudation rates (CWDR) in 13 watersheds using in situ cosmogenic 10Be and 14C analyses of riverine sediment samples in eastern Korea. These CWDR data suggest that the mean denudation rate during the past 5–14ka centers around ~74.7±25.4mm·ka−1, which is consistent with the long-term, geologic exhumation rate we derive of ~74±10.1mm·ka−1 since the middle Miocene. However, our CWDR data are ~2–3 times lower than rates of coastal uplift and river incision of 150–350mm·ka−1 derived from data on marine and fluvial terraces, respectively. In addition to isotopic analyses, we quantified topographic indices such as elevation, relief, and slope, and then compared these indices with CWDRs to examine morphological changes that might be related to tectonic disturbances since the beginning of late Pleistocene. The relationship between CWDRs and topographic indices indicates that most of the study area rests in a near steady-state condition. However, in contrast to the smooth concave longitudinal profile expected for a steady-state landscape, topographic analyses reveal the presence of major knickzones (confirmed by field reconnaissance). These knickzones may have been produced by changes in base level caused by tectonic events, lithologic variations or sea-level changes. Our 1-D modeling of knickpoint propagation suggests that tectonic disturbance is the factor most likely to produce knickzones in the study area, raising the possibility that erosional signals caused by late-Pleistocene tectonic disturbance affected channel gradients, but that the knickpoints have not yet propagated through the entire catchment area. Our results suggest that the eastern part of the Korean Peninsula is approaching, but has not completely reached, a geomorphic steady-state, and that transient conditions have been induced by reactivation of the rift-flank margin no later than the late Pleistocene, causing tilting of the central Korean Peninsula to the west at a rate of 7.0×10−10m/m per year over the past ~125ka.

The processes involved in the development of high-altitude, low-relief areas (HLAs) are still poorly understood. Although cosmogenic nuclides have provided insights into the evolution of HLAs interpreted as paleo-surfaces, most studies focus on estimating how slowly they erode and thereby their relative stability. To understand actual development processes of HLAs, we applied several techniques of cosmogenic nuclides in the Daegwanryeong Plateau, a well-known HLA in the Korean Peninsula. Our denudation data from strath terraces, riverine sediments, soils, and tors provide the following conclusions: (1) bedrock incision rate in the plateau (similar to 127m Myr(-1)) is controlled by the incision rate of the western part of the Korean Peninsula, and is similar to the catchment-wide denudation rate of the plateau (similar to 93m Myr(-1)); (2) the soil production function we observed shows weak depth dependency that may result from highly weathered bedrock coupled with frequent frost action driven by alpine climate; (3) a discrepancy between the soil production and catchment-wide denudation rates implies morphological disequilibrium in the plateau; (4) the tors once regarded as fossil landforms of the Tertiary do not reflect Tertiary processes; and (5) when compared with those of global paleo-surfaces (<20m Myr(-1)), our rapid denudation rates suggest that the plateau cannot have maintained its probable initial paleo landscape, and thus is not a paleo-surface. Our data contribute to understanding the surface processes of actively eroding upland landscapes as well as call into question conventional interpretations of supposed paleo-surfaces around the world. Copyright (c) 2015 John Wiley & Sons, Ltd.

Morphometric features observed from a stream longitudinal profile (SLP) reflect channel responses to lithological variation and changes in uplift or climate; therefore, they constitute essential indicators in the studies for the dynamics between tectonics, climate, and surface processes. The widespread availability of digital elevation models (DEMs) and their processing enable semi-automatic extraction of SLPs as well as additional stream profile parameters, thus reducing the time spent for extracting them and simultaneously allowing regional-scale studies of SLPs. However, careful consideration is required to extract SLPs directly from a DEM, because the DEM must be altered by depression filling process to ensure the continuity of flows across it. Such alteration inevitably introduces distortions to the SLP, such as stair steps, bias of elevation values, and inaccurate stream paths. This paper proposes a new algorithm, called maximum depth tracing algorithm (MDTA), to extract more accurate SLPs using depression-unfilled DEMs. The MDTA supposes that depressions in DEMs are not necessarily artifacts to be removed, and that elevation values within them are useful to represent more accurately the real landscape. To ensure the continuity of flows even across the unfilled DEM, the MDTA first determines the outlet of each depression and then reverses flow directions of the cells on the line of maximum depth within each depression, beginning from the outlet and toward the sink. It also calculates flow accumulation without disruption across the unfilled DEM. Comparative analysis with the profiles extracted by the hydrologic functions implemented in the ArcGIS™ was performed to illustrate the benefits from the MDTA. It shows that the MDTA provides more accurate stream paths on depression areas, and consequently reduces distortions of the SLPs derived from the paths, such as exaggerated elevation values and negatively biased slopes that are commonly observed in the SLPs built using the ArcGIS™. The algorithm proposed here, therefore, could aid all the studies requiring more reliable stream paths and SLPs from DEMs.