Complex Landforms Research Articles

A Multi-Scale Feature Fusion Deep Learning Network for the Extraction of Cropland Based on Landsat Data

In the face of global population growth and climate change, the protection and rational utilization of cropland are crucial for food security and ecological balance. However, the complex topography and unique ecological environment of the Qinghai-Tibet Plateau results in a lack of high-precision cropland monitoring data. Therefore, this paper constructs a high-quality cropland dataset for the YarlungZangbo-Lhasa-Nyangqv River region of the Qinghai-Tibet Plateau and proposes an MSC-ResUNet model for cropland extraction based on Landsat data. The dataset is annotated at the pixel level, comprising 61 Landsat 8 images in 2023. The MSC-ResUNet model innovatively combines multiscale features through residual connections and multiscale skip connections, effectively capturing features ranging from low-level spatial details to high-level semantic information and further enhances performance by incorporating depthwise separable convolutions as part of the feature fusion process. Experimental results indicate that MSC-ResUNet achieves superior accuracy compared to other models, with F1 scores of 0.826 and 0.856, and MCC values of 0.816 and 0.847, in regional robustness and temporal transferability tests, respectively. Performance analysis across different months and band combinations demonstrates that the model maintains high recognition accuracy during both growing and non-growing seasons, despite the study area’s complex landforms and diverse crops.

Characterization of extreme rainfall changes and response to temperature changes in Guizhou Province, China

Different geographical zones have regional heterogeneity in underlying earth surface structure and microclimate which result in different evolution trends and their response to climate change varies in extreme rainfalls in these zones. In the Guizhou province of China, there are complex landforms, which lead to spatial redistribution of rainfall, frequent extreme rainfall, and disasters high risk of geologic disasters. Research on extreme climate in Guizhou mostly paid attention to its spatio-temporal characteristics and modeling, but lack of analysis on its characteristics of extreme rainfall variability and response to temperature changes under different subsurface conditions. This study investigated the characteristics of the extreme rainfall spatiotemporal and recurrence periods in Guizhou province and discussed the relationship between the response of extreme rainfall to temperature change. Daily rainfall data from 1990 to 2020 and 2021–2100 at 31 meteorological observation stations throughout the province were collected to calculate extreme precipitation. This research had the following results. (1) Both historical and future periods show an upward trend in extreme rainfall in Guizhou province, with a spatial distribution pattern of “high in the south and low in the north, high in the east and low in the west” and “high in the southeast and low in the northwest”, respectively; the spatial distribution of extreme rainfall under each recurrence period is consistent with the non-recurrence period. (2) Both historical and future periods show an upward trend in temperature in Guizhou province, with a spatial distribution consistent with that of the extreme rainfall in the corresponding period. (3) The change in extreme rainfall intensity with increasing temperature is almost always greater than the C–C rate for different periods and underlying earth surface structure; Extreme rainfall has a Hook response structure to temperature change, and the climate response structure shifts to the right with climate warming. The results of the study can provide a basis for decision-making on regional disaster prevention and mitigation in the context of temperature change.

An Ensemble Learning Framework for Anomaly Detection of Important Geographical Entities

Abstract. Due to the complex landforms and the limited resolution of remote sensing imagery, it is difficult to avoid the problem of incorrectly capturing geographical entities, such as buildings. Therefore, anomaly detection of important geographical entities is of great significance to ensure the authenticity and accuracy of geographical entity data. In this paper, we propose an ensemble learning framework for anomaly detection of geographical entity by aggregating the predicted labels generated by multiple deep learning models. In detail, we explore multiple change detection and semantic segmentation model and fully utilize the advantages of various deep learning neural network architectures. The proposed anomaly detection strategy of buildings has been performed on two benchmark datasets, including WHU Building change detection dataset and LEVIR building change detection dataset, the experimental results prove that the proposed method can achieve a more robust and better performance than using single change detection model in terms of quantitative performance and visual performance.

Autonomous Robots Traverse Multi-Terrain Environments via Hierarchical Reinforcement Learning with Skill Discovery

Although deep reinforcement learning (DRL) has been widely used in robotic mapless navigation tasks, most of the current research focuses on structured environments such as indoor or maze scenes, and little is targeted on outdoor environments. Unlike indoor environments and mazes, outdoor fields tend to be unstructured with complex landforms and sparse rewards for robotic navigation. The performance of most DRL-based strategies is directly affected by the design of the reward function, which greatly deteriorates its generalizability to outdoor environments. To this end, here we propose a two-stage learning paradigm based on skill discovery and hierarchical reinforcement learning (HRL) to cope with this challenge. Specifically, we implement skill discovery through a pre-training stage to acquire diverse skills with terrain-adaptive exploration strategies; then we select multiple skills using HRL to cope with more complex scenarios. We carry out the robotic multi-terrain traverse task based on a high-fidelity robotic simulation platform, Webots, and implement extensive comparative experiments and ablation studies to demonstrate the effectiveness of our approach.

Soil erosion and environmental degradation in Nepal: A Review of Environmental Policies

Soil erosion is a natural process that occurs more frequently in mountainous regions in the Himalayas due to the complex landforms. Nepal also experiences severe forms of soil erosion due to its geographical complexity. This study examines the issues of soil erosion in relation to the environment and evaluates the effectiveness of existing environmental policies in Nepal, using secondary sources of data. The study reveals that Nepal has an average annual soil erosion rate of 25 tons per hectare per year, resulting in the total annual loss of 369 metric tons. This rate is higher compared to other areas around the world. The middle mountain region of Nepal has a higher mean annual soil erosion rate per hectare per year (which is of 38.4 tons) than other physiographic regions. Soil erosion is caused by various factors such as rainfall, water flow, land cover, geographical setting, and conservation practices – which leads to a decrease in agricultural productivity. Both on-site and off-site effects pose a significant risk to agriculture, development projects and the environment. It is suggested that the steps have to be taken to the environmental safeguard for reducing soil erosion and mitigate the problem through government agencies by emphasizing on effective implementation measures of the existing environmental policies, regulations, and development plans.

Spatial characteristics and cultural factors of Yi nationality traditional dwellings based on spatial syntax

The spatial layout of traditional residential houses is closely related to national and regional characteristics and culture. The traditional village of Yi nationality is selected, and the organization structure of the outside village and the inside living space is studied through the graphical and quantitative analysis method of space syntax. The primary purpose of this study is to analyze the spatial characteristics of Yi dwellings and infer the cultural factors behind them. To explore how to integrate traditional spatial characteristics into contemporary design and promote the preservation and inheritance of architectural culture. The results show that 1) the outer village space of residential houses is characterized by weak identifiability, compact and loose organizational structure, weak permeability, and weak accessibility; 2) The interior space of dwellings presents a spatial layout with the main room and courtyard as the core as a whole. The interior space of dwellings is abundant and flexible, and the accessibility of ancillary functional space is weak. 3) The cultural causes of the formation of these spatial features include complex landforms, family systems, hierarchical systems, semi-agricultural and semi-pastoral production modes, the concept of valuing family harmony, the use of regional materials, respect for nature, fire worship, etc. 4) The spatial characteristics of traditional dwellings are inherited in contemporary dwellings. Fuzzy elements can be added in villages according to family relations, and characteristic Spaces such as courtyards and fire ponds in traditional dwellings can be preserved.

Fine Resolution Mapping of Soil Organic Carbon in Croplands with Feature Selection and Machine Learning in Northeast Plain China

Unsustainable human management has negative effects on cropland soil organic carbon (SOC), causing a decrease in soil health and the emission of greenhouse gas. Due to contiguous fields, large-scale mechanized operations are widely used in the Northeast China Plain, which greatly improves production efficiency while decreasing the soil quality, especially for SOC. Therefore, an up-to-date SOC map is needed to estimate soil health after long-term cultivation to inform better land management. Using Quantile Regression Forest, a total of 396 soil samples from 132 sampling sites at three soil depth intervals and 40 environmental covariates (e.g., Landsat 8 spectral indices, and WorldClim 2 and MODIS products) selected by the Boruta feature selection algorithm were used to map the spatial distribution of SOC in the cropland of the Northeast Plain at a 90 m spatial resolution. The results showed that SOC increased overall from the southern area to the northern area, with an average of 17.34 g kg−1 in the plough layer (PL) and 13.92 g kg−1 in the compacted layer (CL). At the vertical scale, SOC decreased, with depths getting deeper. The average decrease in SOC from PL to CL was 3.41 g kg−1. Climate (i.e., average temperature, daytime and nighttime land surface temperature, and mean temperature of driest quarter) was the dominant controlling factor, followed by position (i.e., oblique geographic coordinate at 105°), and organism (i.e., the average and variance of net primary productivity in the non-crop period). The average uncertainty was 1.04 in the PL and 1.07 in the CL. The high uncertainty appeared in the area with relatively scattered fields, high altitudes, and complex landforms. This study updated the 90 m resolution cropland SOC maps at spatial and vertical scales, which clarifies the influence of mechanized operations and provides a reference for soil conservation policy-making.

Karst landform classification considering surface flow characteristics derived from digital elevation models

AbstractKarst landforms are one of the most typical geographical units with a specific physical process on the earth's surface. The classification of karst landforms is an important aspect for understanding their landform processes and mechanisms. However, influenced by various interior and external forces, karst landforms have an extremely complex surface morphology, increasing the difficulty of their automatic classification. In this study, we considered hydrological features as an important factor in characterizing karst landforms and proposed a method that considers surface flow for karst landform classification. In this method, terrain was reversed for hydrological analysis to achieve the landform units. Then, the watershed boundary of the reversed terrain is extracted by hydrological analysis. The boundary of the karst landform unit is determined by erasing the plain area from the watershed boundary. Thereafter, the graph theory segmentation method is employed to merge the landform units belonging to the same karst landform entity. The proposed approach is validated and applied in two sample karst areas, Fenglin and Fengcong, located in Guilin, China, using digital elevation model data with 30 m spatial resolution. In addition, a comparative analysis is conducted to evaluate the accuracy of the proposed method. The results demonstrated that the typical karst landform units of Fenglin and Fengcong can be effectively classified. The overall classification accuracy is 94.44%. The proposed method produced more reasonable and accurate boundaries compared with the contour tree and terrain feature point methods. Furthermore, the classification results indicate various landform development stages of the karst landform process in the study area. The proposed method considering surface flow characteristics can be further extended to other landform types with highly complex landforms.

Mangrove ecosystem properties regulate high water levels in a river delta

Abstract. Intertidal wetlands, such as mangroves in the tropics, are increasingly recognized for their role in nature-based mitigation of coastal flood risks. Yet it is still poorly understood how effective they are at attenuating the propagation of extreme sea levels through large (order of 100 km2) estuarine or deltaic systems, with complex geometry formed by networks of branching channels intertwined with mangrove and intertidal flat areas. Here, we present a delta-scale hydrodynamic modelling study, aiming to explicitly account for these complex landforms, for the case of the Guayas delta (Ecuador), the largest estuarine system on the Pacific coast of Latin America. Despite coping with data scarcity, our model accurately reproduces the observed propagation of high water levels during a spring tide. Further, based on a model sensitivity analysis, we show that high water levels are most sensitive to the mangrove platform elevation and degree of channelization but to a much lesser extent to vegetation-induced friction. Mangroves with a lower surface elevation, lower vegetation density, and higher degree of channelization all favour a more efficient flooding of the mangroves and therefore more effectively attenuate the high water levels in the deltaic channels. Our findings indicate that vast areas of channelized mangrove forests, rather than densely vegetated forests, are most effective for nature-based flood risk mitigation in a river delta.

Quantifying spit growth and its hydrodynamic drivers in wind-dominated lake environments

Many sand spits are morphodynamically complex landforms, that are either analysed with complex and expensive computational models or at a conceptual level. Therefore, most case studies on spits in different environments are descriptive. A novel method based on the use of polar coordinates was devised to quantitatively analyse spit morphodynamics in a non-tidal, wind-dominated lake environment, using the Marker Wadden islands in Lake Markermeer, the Netherlands, as a case study. A high-resolution morphological data set allowed for the quantification of sedimentation processes around two spits, in two distinctive depth zones. Spit-platform growth is governed by alongshore currents that transport sediment over the spit-platform into deeper waters; the size of the spit-platform in turn affects the growth of the spit around the mean water level. Insight in this complex interplay of processes is crucial to understand spit behaviour in low-energy lake environments. At the Marker Wadden the submerged spit-platform grows during high energy wind events while the emerged spit part grows under mild to moderate energy conditions. With this new method we can quantitatively explore the role of different wave and flow conditions and predict spit growth direction in non-tidal, wind-dominated environments, beyond the level of conceptual descriptions.

EEAGER: A Neural Network Model for Finding Beaver Complexes in Satellite and Aerial Imagery

AbstractBeavers are ecosystem engineers that create and maintain riparian wetland ecosystems in a variety of ecologic, climatic, and physical settings. Despite the large‐scale implications of ongoing beaver conservation and range expansion, relatively few landscape‐scale studies have been conducted, due in part to the significant time required to manually locate beaver dams at scale. To address this need, we developed EEAGER—an image recognition machine learning model that detects beaver complexes in aerial and satellite imagery. We developed the model in the western United States using 13,344 known beaver dam locations and 56,728 nearby locations without beaver dams. Performance assessment was performed in twelve held out evaluation polygons of known beaver occupancy but previously unmapped dam locations. These polygons represented regions similar to the training data as well as more novel landscape settings. Our model performed well overall (accuracy = 98.5%, recall = 63.03%, precision = 25.83%) in these areas, with stronger performance in regions similar to where the model had been trained. We favored recall over precision, which results in a more complete catalog of beaver dams found but also a higher incidence of false positives to be manually removed during quality control. These results have far‐reaching implications for monitoring of beaver‐based river restoration, as well as potential applications detecting other complex landforms.

Evaluation of monthly-scale soil erosion spatio-temporal dynamics and identification of their driving factors in Northeast China

Soil water erosion has caused an enormous damage to agricultural production and ecosystems worldwide. Soil erosion exhibits certain seasonal variation patterns within year under the influence of multiple factors. In this study, we calculated the multi-year average monthly-scale soil erosion modulus from 2001 to 2019 to evaluate the soil erosion risk using the revised universal soil loss equation (RUSLE) in the black soil region of Northeast China. The results showed that the average annual soil erosion modulus in this region was 8.53 t•ha−1•year−1, and the average monthly soil erosion modulus was 0.78 t•ha−1•month−1. The months of April-July and October were identified as critical periods for erosion, with erosion modulus values of 1.17, 1.89, 1.61, 1.13, and 0.93 t•ha−1•month−1 during these five months, respectively. High soil erosion was mainly distributed in southern Liaoning province and western Inner Mongolia, and these high erosion areas were mostly distributed in the regions with complex landforms, in grasslands, and in the southern temperate climatic zone. Among multiple driving factors, the topographic factor LS (L = slopelength,S = slopesteepness) was one of the dominant factors determining the spatial distribution pattern of soil erosion with 30 % of the contribution. The rainfall erosivity (R) and cover management (C) were the major factors driving significantly the intra-annual variation of soil erosion in Northeast China with 27.2 % and 28.9 % of the contribution, respectively. The intra-annual erosion varied with the rainfall and the vegetation growth, and the period of high erosion occurs in the early part of the wet season when the rainfall intensity increases and the vegetation growth is insufficient to protect the soil. Our findings highlight the need to take dynamic factors into consideration in the RUSLE model and the importance for identifying the areas at high erosion risk so as to take soil conservation actions.

Mapping evergreen forests using new phenology index, time series Sentinel-1/2 and Google Earth Engine

Evergreen forests are sensitive to climate change and their role in the exchange of carbon, water, and energy in terrestrial ecosystems is irreplaceable. The temporal and spatial variation of the evergreen forest is closely related to forest monitoring and management in China, which has a far-reaching impact on forest protection and sustainable development. However, the lack of annual fine resolution maps of evergreen forests limits our exploration of the evolution of the spatiotemporal patterns of evergreen forests. Therefore, it is necessary to timely and accurately maps evergreen forests with high spatial resolution. We used a new phenology index of NDVImax-NDVIwinter_max to identify evergreen forests at the 10-m scale. First, we mapped annual 2019 land cover types to obtain the forest mask. Second, we calculated the classification phenology index by extracting the difference between evergreen forests and deciduous forests in phenological characteristics. Finally, we extracted the evergreen forest in 2019 from the annual forest map based on the constructed phenological indicators. The kappa coefficient of our annual forest map for 2019 was 0.92, and the overall, producer and user accuracies were 97.98%, 90.09%, and 97.46%, respectively. The kappa coefficient of the annual evergreen forest map in 2019 was 0.98, and the overall, producer and user accuracies were 99.73%, 97.84%, and 99.68%, respectively. Our study shows that the new phenological index can identify and map evergreen forests on complex landforms dominated by evergreen-deciduous mixed forests, which can be applied to other regions and years in China. The results of this study have reference value for evaluating the spatial distribution and resource management of evergreen forests.

The Profiles Based on Ridge and Valley Lines to Extract Shoulder Lines on the Loess Plateau

The shoulder line is fundamental to geomorphic evolution and erosion monitoring research on the Loess Plateau, which represents the boundary between positive terrain (intergully) and negative terrain (inner gully). The existing extraction methods mainly suffer the problems of unclear geological significance, poor landform application, and low efficiency of algorithms. This paper proposes a new loess shoulder line automatic extraction method, in which topographic feature points (ridge and valley points) were used as endpoints to generate continuous profiles, and two parameters, analysis operator size (L) and filter threshold (σ), were created for shoulder point extraction from each profile. This method can be applied to complex landforms such as the continuous shoulder lines of terraces and extracts. Herein, three typical areas on the Dongzhi Plateau were selected to assess the performance of the method, and a digital elevation model (DEM) with a resolution of 5 m was used as source data. The accuracy assessment index was the Euclidean distance offset percentage (EDOP), and the original evaluation method was improved based on Structure from Motion–Multiview Stereo (SfM-MVS) technology. The experimental results showed that the average accuracy of the proposed method in the three test areas reached 89.3%, which is higher than that of the multidirectional hill-shading and P-N methods. Via testing in different areas, it could be concluded that the extraction efficiency was less affected by the area of the test region, and the approach exhibited a suitable robustness. Simultaneously, the optimal values of parameters L and σ were examined. This study increases the possibility of accurate shoulder line extraction in the large area of the Loess Plateau.

Numerical simulation scheme of jointed rock masses using UAV photogrammetry and a disk-based discontinuous deformation analysis model

<abstract><p>The use of unmanned aerial vehicles (UAVs) for photogrammetry allows the rapid acquisition of high-resolution images of geological masses in complex landforms. However, effective analysis of the acquired image information remains a key research issue. At K158 + 837 on the Chongqing-Huaihua Railway, Baima jointed rock masses were reconstructed with high accuracy using UAV close-range photogrammetry technology, and rock discontinuities were extracted from the projected image. The proposed modeling algorithm for jointed rock masses enables the preprocessing of two-dimensional jointed rock mass slopes. Numerical simulations using the disk-based discontinuous deformation analysis method show that the discontinuity network formed by initial cutting significantly affects the subsequent crack development. Meanwhile, simulation results under different scenarios indicate the importance of the pre-reinforcement measures applied to unstable rock masses. The workflow developed based on these results can serve as a reference for the comprehensive acquisition, recognition and numerical modeling analysis of similar jointed rock masses.</p></abstract>

Risk Assessment of Flood Disaster in Sichuan Province Based on GIS

In the context of global climate change, flood disaster has become one of the significant disasters endangering human life. Sichuan Province has more annual precipitation and complex landforms, which suffer from yearly floods. To reveal the spatial distribution and spatiotemporal dynamic change characteristics, we developed a risk assessment model based on geographic information systems and natural disaster risk assessment theory. Considering both natural and human factors, precipitation, terrain, climate type, population, and GDP are selected as evaluation indexes. We Used the expert score-AHP method to obtain the annual flood disaster risk regionalization from 2017 to 2021. In the end, the results were compared with the entropy weight method in the objective weighting method. The results showed that the comprehensive flood disaster risk in Sichuan Province is higher in the east and lower in the west. The flood disaster risk in the eastern hill area, Panxi area, and northeast Sichuan Basin is generally higher. In contrast, that in the west Sichuan Plateau is lower. Based on the evaluation results, it is found that the risk assessment of flood disasters based on expert scoring-AHP method is more in line with the actual situation.

Vegetation information extraction in karst area based on UAV remote sensing in visible light band

Karst areas, with their rich species and unique ecological environment, have attracted the attention of many researchers. The unique landscape has brought a wealth of species and great difficulties for research. Karst areas have complex landforms and numerous karst caves, which not only bring a lot of manpower and time investment to the research, but also increase the risk of the research. Therefore, the common field research method is not suitable for karst areas. To solve this problem, this paper adopts the form of UAV remote sensing, uses the visible band, through the calculation of normalized vegetation index, combined with the spectral characteristics of the investigated plants, to extract the vegetation information of the study area. In this paper, the final extracted vegetation information is obtained through practice, and compared with other vegetation extraction methods. It can be seen from the results that the vegetation information extraction method proposed in this paper to the visible band UAV remote sensing has higher extraction accuracy and better effect than other existing methods. Therefore, it can be used to extract vegetation information in complex terrain areas, reduce labor cost input, save research time, and obtain more targeted regional plants for targeted research, So as to improve the efficiency and level of vegetation research in karst areas.

Ecogeomorphic response of a coastal dune in southern Portugal regulated by extrinsic factors

Coastal dunes are complex landforms whose morphology results from various interactions between biotic and abiotic factors. Here, we explore the longshore variability of the morphological features, plant community distribution and accumulation patterns of a dune segment (1.4 km-long) located at the downdrift end of a sandy peninsula in the Ria Formosa, Portugal. To understand the main drivers of the observed variability and the implications for dune morphological response, this information was combined with recent multidecadal shoreline evolution data. The integrated results document significant differences in dune morphology, sedimentation patterns and plant zonation, with two distinct dune configurations or states identified in close proximity. One (western sector) shows a narrower dune system, vegetation cover characterised by pioneer species with low densities, and squeezed plant zonation. Conversely, the other (eastern sector) presents a wider dune system with a new foredune, a more developed plant zonation and relatively high vegetation density. Both states could be partially explained by the recent shoreline trends and inlet shifts, with stable to retreating trends in the western sector and shoreline progradation in the eastern one. Plant zonation and accumulation patterns suggest that the dune along the retreating sector is in a cycle of inland migration, encouraged by the reduced accommodation space and the low retention capacity of the vegetation across the dune stoss. Alternatively, observations along the prograding sector suggest that the greater accommodation space and the stabilising feedback between vegetation and topography promoted the seaward progradation of the system and the development of an incipient foredune. Outcomes support the importance of biogeomorphic feedbacks for the dune configuration, but they also evidence that the role of vegetation within the feedback is primarily regulated by physical factors that ultimately promote or inhibit vegetation effects on dune topography.

Assessing DEM quality and minimizing registration error in repeated geomorphic surveys with multi‐temporal ground truths of invariant features: Application to a long‐term dataset of beach topography and nearshore bathymetry

AbstractRemotely sensed digital elevation models (DEMs) and uncertainty‐based geomorphic change detection have become very practical tools for geoscientists, including for coastal research. Through the analysis of DEMs of differences (DoDs) and the provision of DEM quality, it allows monitoring complex landforms and confidently relating the changes observed to environmental forcing conditions.With continuing remote sensing advances, and as some monitoring programmes are reaching several decades of repeated data collection, it is timely to consider approaches that enable the reconciliation of DEMs of variable and potentially unknown quality before their subsequent geomorphic analyses. In this paper, we present an original workflow whereby composite data formed by fusing available measurements over invariant features serve as multi‐temporal ground truths for assessing repeated DEMs. Results of the evaluation enable identification of DEMs of lower quality (bias and precision) and correction of registration error (horizontal and vertical bias), and thus offer the in‐built capacity for estimating and improving change detection levels afforded by the data.The workflow was applied to a freely accessible multi‐sensor: RTK‐GNSS, terrestrial laser‐scanning, drone photogrammetry and multibeam echo‐sounding dataset of high‐resolution topographic and nearshore bathymetric DEMs collected at the macrotidal pocket beach of Porsmilin (France) over the period 2003–2019. Our results show that consistently high DEM precision can be achieved in a long‐term multi‐sensor dataset, but registration errors may be present and can be minimized through co‐registration with the purpose‐built ground truths. Although the study focuses primarily on measuring height discrepancies, which is directly relevant for DoD analysis, we show that the methods can also be used for dealing with horizontal error when high‐resolution imagery is available. Finally, the detailed DEM evaluations presented, in application to a rare dataset documenting beach and shoreface change for nearly two decades, provide original insights on the performance of usual topo‐bathymetric surveying techniques.

Landform classification based on landform geospatial structure – a case study on Loess Plateau of China

ABSTRACT Landform classification, which is a key topic of geography, is of great significance to a wide range of fields including human construction, geological structure research, environmental governance, etc. Previous studies of landform classification generally paid attention to the topographic or texture information, whilst the watershed spatial structure has not been used. This study developed a new landform classification method based on watershed geospatial structure. Via abstracting the landform into the internal and marginal structure, we adopted the gully weighted complex network (GWCN) and watershed boundary profile (WBP) to simulate the watershed geospatial structure. Introducing various indices to quantitatively depict the watershed geospatial structure, we conducted the landform classification on the Northern Shaanxi of Loess Plateau with a watershed-based strategy and established the classification map. The classified landform distribution has significant spatial aggregation and clear regional boundaries. Classification accuracy reached 89% and the kappa coefficient reached 0.87%. Besides, the proposed method has a positive response to some similar and complex landforms. In general, the present study first utilized the watershed geospatial structure to conduct landform classification and is an efficient landform classification method with well accuracy and universality, offering additional insights for landform classification and mapping.