Full text

Turn on search term navigation

1. Introduction

The concept of brownfields originated in the late 20th century [1], with countries such as the United States [2], the United Kingdom [3], France [4], Canada [5], EU states [6], Japan [7], and China interpreting the concept differently. However, the two defining attributes of brownfields are their existing or potential contamination and their abandoned or idle state, coupled with the need for reuse and redevelopment. In China, brownfields refer to “sites where known or potential contamination caused by human activities exists, and whose redevelopment requires risk assessment and remediation based on intended land use” [8]. Brownfields are diverse, vast in quantity, and widely distributed, encompassing abandoned mining sites, disused industrial land, idle infrastructure, landfills, war-impacted lands, cemeteries, and disaster-affected areas [9]. Brownfield regeneration is a global challenge [10]. There are at least 450,000 brownfields in the United States [11]; over 8000 brownfields in the United Kingdom [12], France [13], Canada [14], and the Czech Republic [15]; and more than 200,000 contaminated sites in China [16], forming a point–line–area network spatial pattern [16].

Brownfields are characterized by both abandonment and contamination; contribute to spatial resource wastage, environmental degradation [17], and health hazards [18]; and hinder surrounding development [19]. Brownfield regeneration could promote the functional recovery of damaged ecosystems and repurpose idle land, thereby enhancing overall human living environments. Over the past decade, numerous studies have demonstrated that the spatial characteristics of brownfields are a key determinant of their regeneration potential [20]. A full understanding of the spatial characteristics of brownfields is helpful in effectively formulating brownfield regeneration strategies [21,22,23]. Studies have found that brownfields adjacent to urban centers often have greater renewal potential [24]. Brownfields located in areas with good transportation conditions and high population density are especially suitable for regeneration [25,26,27]. Špirić identified eight site characteristics that affect brownfield regeneration potential, including the internal configuration of building complexes, original land use, proportion of open space, degree of facility preservation, and availability of infrastructure [28]. Building on spatial cognition theory, scholars have developed decision-making frameworks and prioritization strategies for renewable brownfields, such as regeneration priority evaluation methods and decision support pathways based on the DSR model [29], approaches grounded in green infrastructure network connectivity [30,31], and models that integrate economic, social, and ecological spatial attributes [32].

Industrial brownfields—sites predominantly formed through activities such as mineral extraction, smelting, and processing—are a primary focus of ecological restoration and urban renewal in China. These sites are numerous, widely distributed, and characterized by distinctive spatial patterns, complex contamination profiles, and strong connections to surrounding urban areas. Integrating their historical use, spatial attributes, and contamination characteristics, industrial brownfields can be further classified into three classes: raw-material mining brownfields, raw-material manufacturing brownfields, and non-raw-material manufacturing brownfields (Table 1). From the perspective of material flow analysis [33], these three types correspond to the three stages of mineral resource flow. Raw-material mining brownfields are abandoned sites after the extraction of natural ores. Raw-material manufacturing brownfields refer to abandoned sites where raw ores were smelted into finished mineral materials. Non-raw-material manufacturing brownfields are abandoned sites where finished mineral materials were further processed into industrial products. This refined classification enhances our understanding of the spatial distribution and morphological features of each type. It provides a basis for assessing the reuse potential of brownfields, supporting coordinated planning decisions at the urban scale.

Large-scale brownfield spatial identification is crucial for assessing their regeneration potential and supporting decision making from a regional perspective. The spatial identification of brownfields refers to determining their geographic location and delineating their boundaries. In the early 21st century, researchers proposed spatial identification methods to supplement national brownfield databases [34]. Ferrara et al. [35], Gregory et al. [36], and Huang et al. [37] identified brownfields using object-oriented methods based on remote sensing features, including multispectral information collected via Multispectral Infrared and Visible Imaging Spectrometry, aerial photography, and high-spatial-resolution SPOT-5 remote sensing imagery. Mao et al. [38], Hu et al. [39], and Sun et al. [40] employed semantic segmentation and deep learning techniques to automatically identify vacant urban land and derelict industrial sites, respectively. Konrad et al. [41] integrated a Convolutional Neural Network (CNN) with a Vision Transformer (ViT) to classify brownfields, active land, and construction areas. Ge [42] and Zhu [43] employed DInSAR (Differential Interferometric Synthetic-Aperture Radar) technology, along with its derived methods SBAS-InSAR (Small-Baseline-Subset Interferometric Synthetic-Aperture Radar) and PS-InSAR (Persistent-Scatterer Interferometric Synthetic-Aperture Radar), to detect coal mining subsidence areas within regional scopes. Cheng et al. [44], Skála et al. [45], Frickel et al. [46], Lin et al. [47], and He [48] identified brownfields using visual interpretation methods based on data such as environmental contamination records, land-use attributes, contamination-related cost data, enterprise information, and monitored soil data. However, the key data used in these methods are not open-source, which severely limits the applicability and scalability of such identification approaches. Volpe et al. [49], Stuczynski et al. [50], Li et al. [51], Song et al. [52], and Lin et al. [53] adopted visual interpretation methods based on data, including aerial imagery, CORINE Land Cover data, DF-2 remote sensing imagery, Baidu Street View, land bidding and listing information, and environmental incident reports. These methods use a variety of open-access data and are often calibrated using field surveys [46,47,52,53]. Although less efficient in identification, these methods are replicable and offer practical value for broader implementation.

Existing spatial identification methods for brownfields exhibit three major limitations. First, they often neglect the classification of brownfield types, overlooking the diversity of historical land uses and contamination profiles, which impedes coordinated regeneration in urban settings. Second, automated approaches depend heavily on ultra-high-resolution remote sensing imagery. In contrast, manual techniques often require hard-to-obtain data, such as detailed contamination reports or land transaction records. Third, the abandonment status is typically determined through labor-intensive field surveys. Although auxiliary tools like Baidu population density heat maps and Street View have been explored [51], their use still necessitates significant effort.

The aim of this research was to advance both the theoretical and the practical dimensions of urban-scale brownfield studies with a scalable and broadly applicable method. The study was structured to address the following key questions:

(1). How can multi-type industrial brownfields be identified at the city scale?

(2). What are the spatial characteristics of each industrial brownfield subtype?

(3). What factors influence the spatial characteristics of industrial brownfields?

To answer these questions, this study proposes the Target–Classification–Modification (TCM) method. Tangshan City in Hebei Province, China, was chosen as the case study due to its numerous, diverse, and urgently needed brownfield regeneration projects. The TCM method was applied in this work to analyze the spatial characteristics of industrial brownfields and the factors shaping their distribution and form.

2. Materials and Methods

2.1. Study Areas

This study focuses on Tangshan City in Hebei Province, North China, a century-old heavy industrial city known as the “cradle of modern industry in China”. Tangshan covers a total area of approximately 13,472 km2 and has a resident population of about 7.7 million. The city transitions from the Yanshan Mountains in the north to expansive plains in the central and southern areas, with elevation tapering southward. Tangshan is rich in mineral resources, with 51 proven mineral types within its territory, including coal, iron, gold, limestone, dolomite, petroleum, and natural gas (Figure 1).

Tangshan’s industrial evolution began in the late 19th century and spans steelmaking, petrochemicals, ceramics, and machinery manufacturing. In 2013, it was designated as a resource-based city, and urban transformation efforts were initiated. Over the past decade, the closure of numerous industrial and mining enterprises has generated a large and varied inventory of brownfields, rendering Tangshan an archetypal study area. The Tangshan Territorial Spatial Ecological Restoration Plan (2021–2035) highlights ecological degradation in mining zones. A total of 159 “Baichashan” (meaning “bare hills” in Mandarin) were documented, reflecting extensive land erosion and geomorphic destruction caused by open-pit mining. These areas require ongoing remediation [54]. Moreover, in 2021, Tangshan was selected among China’s inaugural urban renewal pilot cities. Within these zones, 35.04 km2 of former industrial land, 69.8% of the total renewal area, has been earmarked for redevelopment [55]. As such, industrial brownfields in Tangshan are critical priorities for ecological restoration, security pattern optimization, and urban renewal. This underscores the urgent need for precise spatial identification.

2.2. Framework of the TCM Method for the Spatial Identification of Brownfields

This study proposes a TCM-based framework for the spatial identification of brownfields. “T” stands for the rapid identification of potential brownfield target areas via novel data sources (e.g., national mineral-site and Gaode Map POIs). “C” represents the classification and refinement of brownfield subtypes through remote-sensing-based rules. “M” refers to the integration of multiple correction procedures to accurately determine the abandonment status and thereby modify the results.

To be specific, the TCM method for municipal-scale brownfield identification comprises two components. The first is a set of identification rules, derived from remote sensing classification criteria, visual interpretation guidelines, and site modification protocols. The second is the technical identification workflow used to implement these rules. Identification rules serve as the core basis for defining the subtypes, spatial boundaries, and abandonment status of industrial brownfields (Figure 2). Based on these rules, the workflow proceeded in three steps (Figure 3). First, target areas were positioned according to multi-source industrial and mining POIs. Second, multi-temporal remote sensing images in Google Earth Pro 7.3.6 were used to preliminarily assign the types and approximate boundaries. Finally, in ArcGIS Pro 3.3.2, the preliminary map with layers of key industrial and mining sites, current and planned land uses, and compound population density heatmaps was overlayed to correct and finalize the spatial identification of industrial brownfields.

2.3. Data Sources

Data quality plays a critical role in brownfield identification. Five categories of publicly accessible data (Table 2) were used in this study: historical and documentary records, urban land-use maps, industrial and mining POIs, high-resolution remote sensing imagery, and population density heatmaps. Important industrial and mining sites were identified using historical and documentary records. Current and planned industrial land-use information, including geographic location and boundaries, was obtained from urban land-use maps. The National Mineral Resource Databases and Gaode Map POIs were used to obtain industrial and mining POIs across the entire study area. Multi-temporal remote sensing images were accessed through Google Earth Pro 7.3.6. A composite population density heatmap was generated by extracting point data from Baidu heatmaps, performing kernel density analysis, classifying using the natural breaks method, and overlaying multi-temporal results.

2.4. Classification and Identification Rules for Industrial Brownfields

2.4.1. Remote Sensing Image Classification Rules

Spatial features are key to identifying brownfield subtypes. Variations in original site functions, industrial operation processes, and types of mineral resources result in diverse traits like boundary, facilities, colors, elevation, transportation, and texture. Based on these multidimensional spatial attributes, a set of remote sensing image classification rules for industrial brownfields were developed in this study (Table 3). Additionally, a remote sensing image case library of typical examples was established (Figure 4), laying the groundwork for the accurate identification of industrial brownfields.

2.4.2. Visual Interpretation Rules

In visual interpretation, delineating precise spatial boundaries and accurately assessing the abandonment status are paramount. Boundaries must be closed, non-overlapping, and align with clear physical demarcations such as roads, rivers, or fences. If a site is divided by a major road (e.g., an expressway or national highway), each segment is treated as a separate brownfield; however, if divided only by secondary roads but enclosed by higher grade thoroughfares, it remains a single unit. The abandonment status is inferred using three indicators: operational features, land cover changes, and elevation changes (specifically for raw-material mining sites). For example, smoke emissions from chimneys, the presence of new or expanded structures, or visible equipment in multi-temporal imagery indicates active use. Conversely, the gradual growth of irregular vegetation suggests natural recolonization and confirms abandonment. Significant elevation alterations in core extraction zones imply ongoing mining activity, while stable topography indicates abandonment (Figure 5).

2.4.3. Site Modification Rules

To validate and refine the preliminary identification, site modification was performed through three sequential overlays. First, the preliminary results were intersected with a comprehensive layer of key industrial and mining sites to capture any initially missed locations. Next, current and planned land-use layers were overlaid. Sites with planned land uses designated as industrial or mining were deemed active and thus removed. However, sites with non-industrial planned uses were retained as brownfields, even if current industrial activities were present. A composite population density heatmap was then generated and classified into seven levels using the natural breaks method (e.g., 0; 0–0.01; 0.02–0.03; 0.03–0.04; 0.05–0.07; 0.008–0.013; 0.014–0.02). Sites with overlapping areas above the two lowest density levels were considered active and excluded from the brownfield inventory.

2.5. Technical Workflow of the TCM Method

2.5.1. Target Area Localization via POI Data

To locate potential brownfield target areas, industrial and mining POI data were retrieved by querying keywords such as “coal”, “iron”, steel”, factory”, “metallurgy”, and “manufacturing” on the BIGEMAP GIS Office map server using Gaode Map POIs. Mining-related POIs were extracted from the 2020 National Mineral Resource Database. A total of 3317 industrial POIs and 2032 mining POIs were obtained. These datasets were cleaned, georeferenced, and projected in ArcGIS Pro 3.3.2. The study area was divided into 625 grids of 5 km × 5 km. Among them, 217 grids (34.7%) did not contain at least one industrial or mining POI and were designated as non-target areas. Visual interpretation was carried out inside these target grids, thereby enhancing identification efficiency (Figure 6).

2.5.2. Visual Interpretation for Brownfield Classification

Building on the target area localization, we employed Google Earth Pro 7.3.6’s multi-temporal, high-resolution imagery to classify and delineate industrial brownfield boundaries through visual interpretation guided by our identification rules. Imagery from October 2020, chosen for its optimal vegetation contrast and minimal cloud cover, served as the baseline, while additional scenes from 2018 to 2022 informed our assessment of the abandonment status. This process yielded 1930 industrial brownfields, including 441 raw-material mining, 707 raw-material manufacturing, and 782 non-raw-material manufacturing brownfields.

2.5.3. Identification Modification

Following the preliminary identification, a systematic modification process was conducted. We used historical records, urban land-use maps, and a compound population density heatmap to identify additional sites and exclude those still in operation. Historical documents, including the Tangshan Gazetteer and relevant publications, were referenced to locate key industrial and mining sites, such as the Tangshan Ceramics Plant, the Tangshan Jinma Steel Plant, and the Majiagou Mine. Urban land-use maps were sourced from the Tangshan Territorial Spatial Master Plan (2021–2035) and the 14th Five-Year Urban Renewal Special Planning Scheme for the Central Urban Area of Tangshan. The compound population density heatmap provided insights into real-time spatial population distribution. To account for temporal differences in activity patterns, data from 16 September 2020 (Wednesday) were selected, with heatmaps captured at 10:00 AM, 1:00 PM, and 4:00 PM. This composite heatmap was a critical reference for determining the abandonment status of sites and served as a second chance to revise the preliminary identification result with satellite imagery (Figure 7). After the modification process, the final number of identified industrial brownfields was refined to 1706 sites (Table 4 and Figure 7).

2.6. Accuracy Evaluation of the TCM Method

To evaluate the accuracy of the TCM method, we first determined an appropriate sample size using Yamane’s (1967) [56] simplified formula for proportions:

n=N1+Ne2

where N = 1706 (the total number of identified sites), and e = 0.10 (for ±10% precision). A 95% confidence level and p = 0.5 were assumed. This yielded 86 sites—very close to 5% of the population. Selecting a 5% (86 sites) sampling ratio, therefore, provided a statistically defensible balance between confidence (95%) and operational constraints such as funding and labor. With this rationale, we employed Excel’s uniform random number generator to select exactly 5% of the 1706 brownfield sites. This procedure yielded 86 locations for ground-truth visits. The final sample comprised 24 raw-material extraction sites (5.7% of that class), 17 raw-material manufacturing sites (3.0%), and 46 non-raw-material manufacturing sites (6.5%). We calculated the overall accuracy as the percentage of field validations that matched the TCM classification out of the 86 surveyed sites.

2.7. Spatial Characteristic Analysis Method for Industrial Brownfields

To analyze the spatial characteristics of multi-type industrial brownfields at the urban scale, this study proposes a method encompassing three dimensions: distribution patterns, morphological characteristics, and spatial connections. First, based on previous studies [26,57], the distribution patterns of the industrial brownfields were characterized qualitatively by describing their relationships with major built-up areas, transportation networks, and natural elements such as mineral resources and water bodies. Second, three metrics, i.e., area (A), perimeter (P), and shape dimension index (D), were selected to quantitatively analyze the spatial morphology of the industrial brownfields. The A and p values were obtained from the ArcGIS attribute table, and D was calculated as:

D=2×ln⁡P÷4ln⁡A 

where A represents the size of the occupied space, and P represents the extent of the contact surface with the surrounding environment. The index D measures shape complexity: when D = 1, the shape is square, and larger values indicate greater complexity. Finally, due to material flow correlations among the three types of industrial brownfields, their spatial associations were further qualitatively analyzed. This analysis helped explain the structure of urban industrial spatial organization in conjunction with the spatial distribution characteristics.

3. Results

3.1. Performance Evaluation of the TCM Method

Following a 5% field sampling survey, the TCM method achieved an overall accuracy of 84.9%, demonstrating its reliability in identifying the brownfield distribution across the city (Table 5). This high accuracy was largely due to the integration of abandonment status determination in both visual interpretation and identification modification, supported by multi-source data. In the visual interpretation stage, multi-temporal remote sensing images were used to assess site abandonment based on operational features, land cover changes, and elevation shifts. During site modification, historical records, urban land-use maps, and population density heatmaps were used to refine the initial identification. As a result, the preliminary count of 1930 brownfields was modified to 1706, increasing the precision by approximately 11.6%.

Precision=N(final results)N(preliminary results)

Moreover, the TCM method efficiently narrowed the analysis scope by pre-locating potential brownfields, enhancing the identification efficiency by 34.88% in Tangshan compared to traditional full-area visual interpretation methods.

3.2. Identification Results of Industrial Brownfields in Tangshan

A total of 1706 industrial brownfields were identified in Tangshan, comprising 422 raw-material mining sites (124.03 km2), 576 raw-material manufacturing sites (189.06 km2), and 708 non-raw-material manufacturing sites (82.14 km2). Spatially, the brownfields exhibit a multicentric pattern, clustering along mountain ridges, water bodies, railways, highways, and major built-up areas. The highest densities occur in the southern urban core and flanking the Qiluan Railway, while in the northern part of the city, the brownfield sizes diminish, giving way to a more scattered distribution (Figure 8).

3.3. Spatial Characteristics and Influencing Factors of Industrial Brownfields

The industrial brownfields in Tangshan displayed notable spatial differentiation in distribution patterns and morphological characteristics, with evident spatial connections among different types. The findings reveal that the primary driving forces behind these spatial characteristics are natural resource endowments, the layout and accessibility of transportation networks, and the historical organization of industrial production spaces (Table 6).

3.3.1. Spatial Distribution Patterns

The spatial distribution of brownfields reflects the interplay of regional natural geography, transportation infrastructure, and historical industrial organization. Resource endowment fundamentally constrained site selection. Raw-material mining brownfields cluster in bedrock outcrop areas (southern Yanshan foothills) and Luan River alluvial fans, forming terrain-aligned belts. However, transportation networks modulated this pattern. Sites north of the Qiluan railway are significantly larger, demonstrating how rail access synergized with resource availability to enable cost-effective, large-scale extraction.

Similarly, raw-material manufacturing brownfields showed a dual dependence. Their linear distribution along the Sha River’s left bank (Guye to Fengnan Districts) stems from the convergence of hydrological needs (e.g., ore dressing and coal washing) and historical siting of water-reliant industries. In contrast, non-raw-material manufacturing brownfields prioritized transport connectivity over resource proximity, forming “corridor-embedded” layouts along national highway G112 and Tangjin Highway. This divergence highlights how transportation infrastructure interacts with sectoral requirements. Resource-based sectors balanced resource access and transport efficiency, while non-resource sectors prioritized market logistics. Transport nodes further restructured the spatial organization, driving clustering near Fengrun District’s southern hub.

3.3.2. Spatial Morphological Characteristics

Different types of brownfields exhibit significant variations in spatial morphological characteristics such as area (A), perimeter (P), and shape dimension index (D) (Figure 9), indicating differences in land-use scale and the degree of interaction with surrounding environments. Both raw-material mining brownfields and raw-material manufacturing brownfields displayed large-scale spatial characteristics, with 192 (45.5%) and 291 (50.5%) sites exceeding 10 hectares. The largest single site covers more than 700 hectares. There is a strong correlation between site perimeter and area, with 165 (39.1%) and 209 (36.3%) sites having perimeter lengths greater than 2 km. This reflects how natural constraints (e.g., terrain forcing dispersed layouts) and industrial logic (e.g., continuous production requiring multi-unit structures) jointly shaped the expansion patterns, creating complex boundaries that challenge redevelopment coordination. In contrast, non-raw-material manufacturing brownfields appeared generally smaller in size and more compact. In total, 371 (52.4%) sites fell within the 0–5 ha range, and 583 sites (82.3%) showed perimeter lengths under 2 km. This regularity results from standardized industrial planning and integration with urban built-up areas, factors that systematically outweigh terrain influences.

The shape dimension index (D) values further revealed these interactions. Raw-material mining brownfields, where rugged terrain dominated over planning, showed the highest irregularity (82.9%, D > 1.1). Raw-material manufacturing sites exhibited moderate irregularity (69.6%, D > 1.1), as production unit expansion created patchwork layouts within topographical limits. Non-raw-material manufacturing sites displayed regularity (57.8%, D >1.1) due to standardized urban planning superseding natural factors.

3.3.3. Spatial Association Characteristics

Spatial correlation underscores pronounced systemic synergies between resources, transport, and industrial organization. Raw-material manufacturing brownfields are concentrated in the central industrial belt, including the southern fringe of the urban core, the southern edge of the Guye District, and the Qiluan Railway corridors. Proximity to mining sites minimized the logistics costs, while rail/port access enabled scale economies. The Caofeidian Port region exemplifies this dynamic, as manufacturing clusters adjacent to deep-water berths capitalize on port infrastructure to aggregate high-throughput raw-material industries. Conversely, non-raw-material manufacturing brownfields are interwoven within the central and southern built-up areas in a “factory in front and city behind” configuration where market accessibility and intensive urban land-use supersede resource ties. Their nested association with raw-material manufacturing sites, particularly in the city’s southern districts, reflects historical industrial zoning that segregated, yet co-located, complementary functions. These patterns crystallize Tangshan’s dual locational logic: resource–transport–industrial symbiosis for material sectors and market–urban planning integration for non-material sectors. Together, these approaches formed a spatial structure where natural advantages were amplified by infrastructure and institutionalized through historical production organization.

4. Discussion

4.1. Factors Influencing Identification Accuracy

The main factors that affect identification results are the temporal coverage and the spatial coverage of the data. In terms of temporal coverage, the initial identification relied on multi-temporal remote sensing imagery from 2018 to 2022. The modification stage used multi-temporal population density heatmaps captured on Wednesday, 16 September 2020, at 10:00 a.m., 1:00 p.m., and 4:00 p.m. These time slices were selected as representative samples. Expanding the temporal scope of the imagery and increasing the frequency of heatmap observations could further improve accuracy.

The spatial coverage of our datasets spans the entire Tangshan municipal area, including historical records, industrial POI data, high-resolution remote sensing imagery, and population heatmaps. However, the heatmaps were generated from Baidu heat point data with a spatial resolution of 200 m, resulting in sparse data coverage in suburban and rural areas. This limitation likely reduced the identification accuracy of non-raw-material manufacturing brownfields, the abandonment status of which heavily depends on population density signals. Moreover, the urban land-use maps employed for correction were confined to the central urban area, constraining their effectiveness for refining brownfield identification across Tangshan’s full administrative boundary.

4.2. A Comparison with Existing Methods

The TCM method for the spatial identification of brownfields allows for advances compared to conventional approaches by incorporating three sequential steps, namely, target-area localization, visual interpretation, and classification, in addition to site modification. In contrast, similar identification workflows typically commence only with imagery analysis and post-classification refinement. The existing brownfield identification methods involve processing and analyzing images or datasets over the entire study region, incurring substantial time, computational resource, and labor costs. However, TCM’s use of national mineral datasets and Gaode Map POIs for initial target-area localization rapidly delineates candidate brownfield zones, thereby constraining the scope of subsequent visual interpretation and markedly improving manual efficiency.

Additionally, the TCM method distinguishes itself in three additional dimensions: the diverse identification objects, the well-defined classification rules, and the techniques for determining the abandonment status, as described below.

(1). Most existing methods treat brownfields as a homogeneous category, obscuring critical differences in site characteristics and hindering tailored regeneration strategies. In contrast, the TCM method distinctly classifies industrial brownfields into three categories, thereby revealing subtype-specific spatial distribution, morphological patterns, and interrelationships. This granularity provides a nuanced foundation for targeted reuse and redevelopment planning.

(2). Similar brownfield identification methods often rely on environmental contamination records, corporate filings, or land transaction logs. While accurate, these data sources are limited in accessibility and temporal lag. The TCM method instead anchors its subtype differentiation in remote sensing classification rules. Remote sensing imagery delivers a comprehensive, up-to-date view of land-use status, is readily obtainable, and ensures replicable and reliable identification outcomes. This establishes a solid platform for subsequent spatial analyses and assessments of the redevelopment potential.

(3). Typical workflows defer abandonment assessment to post-classification field surveys, which are labor-intensive and time-consuming. Song et al. [52] used Baidu Street View for preliminary judgment, improving modification efficiency. In contrast, the TCM method embeds abandonment determination both during visual interpretation and throughout the site modification phase, utilizing multi-source data to improve accuracy. On this basis, validation via a stratified field-survey sample yielded an overall accuracy of 84.9%, exceeding 90% for raw-material mining sites.

4.3. Limitations

Despite its strengths, the TCM method has some limitations that warrant further research. First, the brownfield boundaries delineated by this method are the administrative or management zones rather than the true contamination extents, as pollutants can migrate via soil, groundwater, or air beyond the visible surface limits. Thus, incorporating soil contamination surveys or monitoring data is necessary to more accurately map pollution footprints. Second, the method focuses on above-ground brownfields and is less applicable to subsidence-related sites. Radar-based remote sensing techniques should be adopted for mining subsidence brownfields [42,43] using threshold values of subsidence rates to define spatial boundaries. Third, there are limitations on the spatial coverage of heatmaps and urban land-use maps, especially in suburban areas, which could be strengthened by incorporating mobile data. Fourth, although the 5% sampling ratio (n = 86) aligns with statistically derived thresholds [56] and operational feasibility, we acknowledge that the uniform random sampling approach may carry risks of sampling bias during the TCM accuracy evaluation process. Future studies could further mitigate this bias by adopting proportional stratified sampling or by intentionally oversampling under-represented categories to bolster the statistical reliability. Finally, although the TCM approach improves the accuracy and efficiency of identification using multi-source POI data, explicit rule sets, and sequential modification steps, it remains a largely manual workflow, incurring substantial time costs and vulnerability to subjectivity and human error. In the future, the integration of drone photography with high-resolution satellite remote sensing imagery could further improve the identification accuracy. Additionally, with the accumulation of a large number of brownfield samples, labeled data can be used to train machine learning classification models, paving the way for automatic brownfield identification.

5. Conclusions

This study proposes a “Target–Classification–Modification” method for brownfield spatial identification. Compared with conventional approaches, which typically analyze the entire study area and rely solely on image classification and post hoc refinement, the TCM method delivers greater typological detail and has higher efficiency and improved accuracy. In Tangshan City, three subtypes of 1706 industrial brownfields were identified, and their spatial characteristics were further analyzed. The results revealed notable spatial differentiation in their distribution patterns and morphological characteristics, as well as spatial connections among the different types. Moreover, the identification efficiency improved by 34.7% through target areas’ pre-localization, while precision increased by 11.6% through a multi-approach site modification process. The overall identification accuracy reached 84.9%, as validated with field sampling. Although this empirical case study was conducted in Tangshan, the TCM method is both modular and data-driven, such that it is readily transferable to other urban contexts. It shows strong potential for large-scale, multi-type brownfield identification. It also provides valuable support for informed regeneration planning. These findings contribute both theoretical advancements and practical methodologies for municipal-scale brownfield identification.

With the advancement of artificial intelligence and the proliferation of open-access spatial datasets, machine learning approaches for automated brownfield identification are beginning to appear. However, their dependence on costly, hard-to-obtain data hinders their scalability and widespread adoption. Similarly, object-oriented image analysis techniques demand extensive parameter tuning. While they offer high specificity, their transferability across different urban contexts is limited. Future work will extend the application of the TCM method to additional cities to assess its generalizability. As the repository of labeled brownfield samples grows, we will investigate the integration of machine learning classifiers for automated detection. This hybrid strategy combines the rule-based precision of the TCM method with the speed and adaptability of data-driven models. The goal is to enable a more accurate and efficient brownfield mapping. This, in turn, is expected to support research, planning, and policy development at the municipal scale.

Author Contributions

Conceptualization, Q.F. and X.Z.; methodology, Q.F., J.Z. and M.C.; software, Q.F. and J.Z.; investigation, Q.F., J.Z., Y.H. and Z.L.; resources, Q.F. and Z.L.; data curation, Q.F. and J.Z.; spatial analysis, Q.F. and J.Z.; writing—original draft preparation, Q.F., J.Z. and Z.L.; writing—review and editing, J.Z. and X.Z.; visualization, Q.F., J.Z. and Y.H.; supervision, X.Z. All authors have read and agreed to the published version of the manuscript.

 Data Availability Statement

The data presented in this study are available on request from the corresponding author.

 Conflicts of Interest

The authors declare no conflicts of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1 Geographic location of Tangshan.

Figure 2 Rules for the spatial identification method of industrial brownfields.

Figure 3 Technical procedures of the “Target–Classification–Modification” (TCM) method for the spatial identification of industrial brownfields.

Figure 4 Typical remote sensing image case library of industrial and mining brownfields.

Figure 5 Examples of visual interpretation rules.

Figure 6 Spatial distribution of industrial and mining POIs, and non-target areas in Tangshan.

Figure 7 Identification and refinement of industrial brownfields in Tangshan.

Figure 8 Spatial distribution of industrial brownfields in Tangshan by type. (A) Tangshan Brownfields Distribution Map. (a) Raw-material mining brownfield. (b) Raw-material manufacturing brownfield. (c) Non-raw-material manufacturing brownfield. (A1) Zoomed-in view of high-density industrial clusters. Note: (a1), (b1), (c1) are field validation photos; (a2), (b2), (c2) are Google Earth planimetric maps.

Figure 9 Spatial morphological characteristics of industrial brownfields in Tangshan.

Types and characteristics of industrial brownfields.

Brownfield Type Historical Use Common Site Examples Spatial Characteristics Contamination Features Redevelopment Challenges and Potential
Raw-material mining brownfields Open-pit mining for mineral extraction Mining pits, waste dumps, tailing ponds Irregular terrain, surface degradation Heavy Terrain reshaping, ecological restoration
Raw-material manufacturing brownfields Processing of raw ores, including washing, crushing, smelting, and sintering Coal washing plants, steel mills, cement plants, coking plants Flat terrain, large industrial buildings and structures Heavy Contamination remediation, industrial heritage utilization
Non-raw-material manufacturing brownfields Further processing of finished mineral materials Steel pipe factories, machinery plants, construction material plants Flat terrain, large-scale factory buildings Light Functional space replacement

Data sources for spatial identification of industrial brownfields.

Data Type Dataset Description Source
Historical and documentary records - Important industrial and mining projects, shutdown and remediation status of tailing ponds Local government official websites, books, and literature
Urban land-use maps - Current land-use and planning maps Local government official websites
Industrial and mining POIs National Mineral Resources Database (2020 edition) Mineral sites with latitude, longitude, mineral types, and utilization status, in SHP format National Geological Archives of China (NGA), http://dcc.cgs.gov.cn, accessed on 26 December 2020
Gaode Map POIs Names, latitude, and longitude, in KML format BIGEMAP
High-resolution remote sensing images Google Earth Pro remote sensing imagery Multi-temporal, meter-scale, high-resolution remote sensing imagery Google Earth Pro
Population density heatmaps Baidu heatmap Heatmap point data with a 200 m resolution, in SHP format https://www.baidu.com/, accessed on 16 September 2020

Classification and spatial characteristics of industrial brownfields.

Type Boundary Facilities Colors Elevation Transportation Textures
Raw-material mining brownfields Clear and irregular No special facilities Coal: black; hematite: brownish yellow; copper: yellowish red; limestone: grayish white Elevation variation, terraced or stepped terrain Rugged roads Coal, iron, and copper: a few large sites; limestone: smaller
Raw-material manufacturingbrownfields Clear and regular with walls Industrial structures, like blast furnaces and conveyor belts Cement plant: grayish white; steel plant: brownish; coke plant: black Relatively flat Well-structured internal roads; railways Large-scale clustered industrial buildings connected by pipelines or railways
Non-raw-material manufacturingbrownfields Clear and regular with walls Storage of raw materials and miscellaneous items Blue or red corrugated steel sheets Relatively flat Well-structured internal roads Dense building clusters surpassing residential buildings in size

Number of brownfields before and after modification using different data sources.

Data Number of Brownfields
Before Process After
Sites Deleted Sites Added
Historical records 1930 —— 13 1943
Urban land-use maps 1930 47 26 1909
Population density heatmap 1930 239 —— 1691
Combination 1930 256 32 1706

Identification accuracy of industrial brownfield subtypes.

Type Sample Number Sampling Rate Actual Number Accuracy
Raw-material mining brownfields 24 5.7% 22 91.7%
Raw-material manufacturing brownfields 17 3.0% 15 88.2%
Non-raw-material manufacturing brownfields 46 6.5% 36 80.4%
In total 86 5.0% 73 84.9%

Types and spatial characteristics of industrial brownfields.

Type Spatial Distribution Spatial Morphology
Raw-material mining brownfields Clustered along mountain ranges and railways [Image omitted. Please see PDF.] Scale ranges from 0.2 to 730 hectares; irregular shape [Image omitted. Please see PDF.]
Raw-material manufacturing brownfields Adjacent to urban built-up areas; distributed along rivers [Image omitted. Please see PDF.] Scale ranges from 0.3 to 2000 hectares; relatively regular shape [Image omitted. Please see PDF.]
Non-raw-material manufacturing brownfields Scattered within urban built-up areas; distributed along roads [Image omitted. Please see PDF.] Scale ranges from 0.1 to 320 hectares; highly regular shape [Image omitted. Please see PDF.]

1. Zheng, X. Landscape Strategies for Brownfield Regeneration Based on the Concept of “Brown Earth-Work”. Doctoral Thesis; Tsinghua University: Beijing, China, 2014.

2. United States Congress. Small Business Liability Relief and Brownfields Revitalization Act. Public Law 107–118. 2002; Available online: https://www.congress.gov/107/plaws/publ118/PLAW-107publ118.pdf (accessed on 6 April 2025).

3. Ministry of Housing, Communities and Local Government. UK. National Planning Policy Framework Annex 2: Glossary. 2012; Available online: https://www.gov.uk/guidance/national-planning-policy-framework/annex-2-glossary#prev-dev-land (accessed on 28 April 2025).

4. Oliver, L.; Ferber, U.; Grimski, D.; Millar, K.; Nathanail, P. The scale and nature of European brownfields. Proceedings of the CABERNET 2005-International Conference on Managing Urban Land LQM Ltd.; Belfast, UK, 13–15 April 2005.

5. Environment and Climate Change Canada, Canada. Pollution and Waste Management. 2025; Available online: https://www.canada.ca/en/services/environment/pollution-waste-management/contaminated-sites.html (accessed on 28 April 2025).

6. CABERNET Coordination Team University of Nottingham. Sustainable Brownfield Regeneration: CABERNET Network Report. 2006; Available online: https://www.yumpu.com/en/document/read/38906007/sustainable-brownfield-regeneration-cabernet-network-report (accessed on 6 April 2025).

7. Japan: Expert Studying Group for Countermeasures against Brownfields. Current status of the Brownfields Issue in Japan Interim Report. 2007; Available online: https://www.env.go.jp/en/water/soil/brownfields/interin-rep0703.pdf (accessed on 6 April 2025).

8. China Architecture & Building PressThe Architectural Society of China. Architectural Design Dataset; China Architecture & Building Press: Beijing, China, 2017; 509.

9. Zheng, X. Multidimensional Brownfields. World Archit.; 2021; 04, pp. 26-30+129. [DOI: https://dx.doi.org/10.16414/j.wa.2021.04.004]

10. Kirkwood, N.; Xiao, L. Looking Broadly at Brownfields. Chin. Landsc. Archit.; 2015; 31, pp. 5-9.

11. EPA, United States. Brownfields. 2025; Available online: https://www.epa.gov/brownfields/about (accessed on 28 April 2025).

12. Ministry of Housing, UK, Communities and Local Government. Dataset: Brownfield Land. 2025; Available online: https://www.planning.data.gov.uk/dataset/brownfield-land# (accessed on 28 April 2025).

13. Ministère de la Transition écologique, de la Biodiversité, de la Forêt, de la Mer et de la Pêche, France. Carte Interactive. 2025; Available online: https://www.georisques.gouv.fr/cartes-interactives#/ (accessed on 28 April 2025).

14. Treasury Board of Canada Secretariat, Canada. Find Sites by Classification. 2025; Available online: https://www.tbs-sct.gc.ca/fcsi-rscf/classification-eng.aspx (accessed on 28 April 2025).

15. Ministerstvo životního prostředí, Czech Republic. Systém evidence kontaminovaných míst. 2025; Available online: https://www.sekm.cz/portal/areasource/map_search_public/ (accessed on 28 April 2025).

16. Li, X.; Jiao, W.; Xiao, R.; Chen, W.; Liu, W. Contaminated sites in China: Countermeasures of provincial governments. J. Clean. Prod.; 2017; 147, pp. 485-496. [DOI: https://dx.doi.org/10.1016/j.jclepro.2017.01.107]

17. Bai, Z.; Zhou, W.; Wang, J.; Zhao, Z.; Cao, Y.; Zhou, Y. Overall Protection, Systematic Restoration and Comprehensive Management of Land Space. China Land Sci.; 2019; 33, pp. 1-11.

18. Zhao, Y. Research on the Current Situation of Coal Mine Wastewater Pollution and Analysis of Treatment Plans. Create Living; 2019; 11, pp. 168-170.

19. Landrigan, P.J.; Fuller, R.; Acosta, N.J.; Adeyi, O.; Arnold, R.; Baldé, A.B.; Bertollini, R.; Bose-O′Reilly, S.; Boufford, J.I.; Breysse, P.N. . The Lancet Commission on pollution and health. Lancet; 2018; 391, pp. 462-512. [DOI: https://dx.doi.org/10.1016/S0140-6736(17)32345-0]

20. Novosák, J.; Hájek, O.; Nekolová, J.; Bednář, P. The spatial pattern of brownfields and characteristics of redeveloped sites in the Ostrava metropolitan area (Czech Republic). Morav. Geogr. Rep.; 2013; 21, pp. 36-45. [DOI: https://dx.doi.org/10.2478/mgr-2013-0009]

21. Holl, A. Manufacturing location and impacts of road transport infrastructure: Empirical evidence from Spain. Reg. Sci. Urban Econ.; 2004; 34, pp. 341-363. [DOI: https://dx.doi.org/10.1016/S0166-0462(03)00059-0]

22. Smith, J.P.; Li, X.; Turner II, B.L. Lots for greening: Identification of metropolitan vacant land and its potential use for cooling and agriculture in Phoenix, AZ, USA. Appl. Geogr.; 2017; 85, pp. 139-151. [DOI: https://dx.doi.org/10.1016/j.apgeog.2017.06.005]

23. Turečková, K.; Nevima, J.; Škrabal, J.; Martinát, S. Uncovering patterns of location of brownfields to facilitate their regeneration: Some remarks from the Czech Republic. Sustainability; 2018; 10, 1984. [DOI: https://dx.doi.org/10.3390/su10061984]

24. Kunc, J.; Navrátil, J.; Tonev, P.; Frantál, B.; Klusáček, P.; Martinát, S.; Havlíček, M.; Černík, J. Perception of urban renewal: Reflexions and coherences of socio-spatial patterns (Brno, Czech Republic). Geogr. Tech.; 2014; 9, pp. 66-77.

25. Abe, S.; Nakagawa, D.; Matsunaka, R.; Oba, T. Study on the factors to transform underused land focusing on the influence of railway stations in central areas of Japanese Local cities. Land Use Policy; 2014; 41, pp. 344-356. [DOI: https://dx.doi.org/10.1016/j.landusepol.2014.06.021]

26. Frantál, B.; Greer-Wootten, B.; Klusáček, P.; Krejčí, T.; Kunc, J.; Martinát, S. Exploring spatial patterns of urban brownfields regeneration: The case of Brno, Czech Republic. Cities; 2015; 44, pp. 9-18. [DOI: https://dx.doi.org/10.1016/j.cities.2014.12.007]

27. Smoļakova, A. Exploring spatial patterns of urban brownfields: The case of Daugavpils city. Proceedings of the International Scientific and Practical Conference; Rezekne, Latvia, 16 June 2017.

28. Špirić, A. Spatial criteria in urban renewal of industrial brownfield sites. Građevinar; 2015; 67, pp. 865-877.

29. Yang, H. Study of Abandoned Mine Land Regeneration under the Perspective of “Urban Renewal and Ecological Restoration”. Doctoral Thesis; China University of Mining & Technology: Beijing, China, 2018.

30. Hou, W.; Zhai, L.; Feng, S.; Walz, U. Restoration priority assessment of coal mining brownfields from the perspective of enhancing the connectivity of green infrastructure networks. J. Environ. Manag.; 2021; 277, 111289. [DOI: https://dx.doi.org/10.1016/j.jenvman.2020.111289]

31. Liao, Q.; Xu, H.; Liu, X. Regeneration of mining wasteland in view of optimization of urban green infrastructure system: A case study of Daye. Bull. Geol. Sci. Technol.; 2021; 40, pp. 214-223. [DOI: https://dx.doi.org/10.19509/j.cnki.dzkq.2021.0415]

32. Fu, Q. Spatial Patterns and Regeneration Strategies of Brownfield Clusters in Resource-Exhausted Cities, China. Doctoral Thesis; Tsinghua University: Beijing, China, 2021.

33. Brunner, P.H.; Rechberger, H. Practical Handbook of Material Flow Analysis. Int. J. Life Cycle Assess.; 2004; 9, pp. 337-338. [DOI: https://dx.doi.org/10.1007/BF02979426]

34. Haninger, K.; Ma, L.; Timmins, C. The value of brownfield remediation. J. Assoc. Environ. Resour. Econ.; 2017; 4, pp. 197-241. [DOI: https://dx.doi.org/10.1086/689743]

35. Ferrara, V.; Beriatos, E.; Brebbia, C.A. Brownfield identification: Different approaches for analysing data detected by means of remote sensing. WIT Trans. Ecol. Environ.; 2008; 107, pp. 45-54. [DOI: https://dx.doi.org/10.2495/BF080051]

36. Gregory, A.S.; Ritz, K.; McGrath, S.P.; Quinton, J.N.; Goulding, K.W.T.; Jones, R.J.A.; Harris, J.A.; Bol, R.; Wallace, P.; Pilgrim, E.S. . A review of the impacts of degradation threats on soil properties in the UK. Soil Use Manag.; 2015; 31, pp. 1-15. [DOI: https://dx.doi.org/10.1111/sum.12212]

37. Huang, D.; Liu, Q.; Liu, G.; Yan, W. Coal Gangue Yards Information Extraction Using Object-oriented Method Based on SPOT-5 Remote Sensing Images. J. Geo-Inf. Sci.; 2015; 17, pp. 369-377.

38. Mao, L.; Zheng, Z.; Meng, X.; Zhou, Y.; Zhao, P.; Yang, Z.; Long, Y. Large-scale automatic identification of urban vacant land using semantic segmentation of high-resolution remote sensing images. Landsc. Urban Plan.; 2022; 222, 104384. [DOI: https://dx.doi.org/10.1016/j.landurbplan.2022.104384]

39. Hu, X.; Zhuang, S. Large-Scale Spatial–Temporal Identification of Urban Vacant Land and Informal Green Spaces Using Semantic Segmentation. Remote Sens.; 2024; 16, 216. [DOI: https://dx.doi.org/10.3390/rs16020216]

40. Sun, Y.; Hu, H.; Han, Y.; Wang, Z.; Zheng, X. Large-Scale Automatic Identification of Industrial Vacant Land. ISPRS Int. J. Geo-Inf.; 2023; 12, 409. [DOI: https://dx.doi.org/10.3390/ijgi12100409]

41. Durrbeck, K.; Lasker, A.; Gollapalli, K.; Ghosh, M.; Sk, M.O.; Fischer, R. BrownViTNet: Hybrid CNN-Vision Transformer Model for the Classification of Brownfields in Aerial Imagery. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.; 2025; 18, pp. 8189-8202. [DOI: https://dx.doi.org/10.1109/JSTARS.2025.3541733]

42. Ge, D. Research on the Key Techniques of SAR Interferometry for Regional Land Subsidence Monitoring. Doctoral Thesis; China University of Geosciences: Beijing, China, 2013.

43. Zhu, Y. Analysis of Ground Subsidence Monitoring in Mining Areas Using InSAR with Parameter Inversion. Doctoral Thesis; Central South University: Changsha, China, 2013.

44. Cheng, F.; Geertman, S.; Kuffer, M.; Zhan, Q. An integrative methodology to improve brownfield redevelopment planning in Chinese cities: A case study of Futian, Shenzhen. Comput. Environ. Urban Syst.; 2011; 35, pp. 388-398. [DOI: https://dx.doi.org/10.1016/j.compenvurbsys.2011.05.007]

45. Skála, J.; Vácha, R.; Čechmánková, J.; Horvathova, V. Various aspects of the genesis and perspectives on agricultural brownfields in the Czech Republic. Morav. Geogr. Rep.; 2013; 21, pp. 46-55. [DOI: https://dx.doi.org/10.2478/mgr-2013-0010]

46. Frickel, S.; Elliott, J.R. Tracking industrial land use conversions: A new approach for studying relict waste and urban development. Organ. Environ.; 2008; 21, pp. 128-147. [DOI: https://dx.doi.org/10.1177/1086026608317799]

47. Lin, H.; Song, Y.; Wang, S. Research on the Method of Urban Brownfield Information Identification and Brownfield Database Construction: Taking Changchun as an Example. China Land Sci.; 2016; 30, pp. 80-87+97.

48. He, T. Study on the pollution space update of Changsha-Zhuzhou-Xiangtan Urban Agglomeration. Ph.D. Thesis; Hunan Normal University: Changsha, China, 2016.

49. Volpe, L.L.; Vasques, A.R.; Lombardo, M.A.; Misra, S.C.; Revetria, R.; Sztandera, L.M.; Iliescu, M.; Zaharim, A.; Parsiani, H. The use of geoprocessing techniques to identify the brownfields railroad areas in the city of Rio Claro-Sp, Brazil. Proceedings of the 4th WSEAS International Conference on Remote Sensing; Venice, Italy, 21–23 November 2008.

50. Stuczynski, T.; Siebielec, G.; Korzeniowska-Puculek, R.; Koza, P.; Pudelko, R.; Lopatka, A.; Kowalik, M. Geographical location and key sensitivity issues of post-industrial regions in Europe. Environ. Monit. Assess.; 2009; 151, pp. 77-91. [DOI: https://dx.doi.org/10.1007/s10661-008-0251-4]

51. Li, W.; Wang, D.; Li, H.; Wang, J.; Zhu, Y.; Yang, Y. Quantifying the spatial arrangement of underutilized land in a rapidly urbanized rust belt city: The case of Changchun City. Land Use Policy; 2019; 83, pp. 113-123. [DOI: https://dx.doi.org/10.1016/j.landusepol.2019.01.032]

52. Song, Y.; Lyu, Y.; Qian, S.; Zhang, X.; Lin, H.; Wang, S. Identifying urban candidate brownfield sites using multi-source data: The case of Changchun City, China. Land Use Policy; 2022; 117, 106084. [DOI: https://dx.doi.org/10.1016/j.landusepol.2022.106084]

53. Lin, M.; Li, J.; Xie, H. Methods and Practices for Identifying Urban Brownfield Sites. World Reg. Stud.; 2013; 22, pp. 169-176.

54. Tangshan Territorial Spatial Ecological Restoration Plan (2021–2035), Public Consultation Draft. Available online: http://zygh.tangshan.gov.cn/ts/xxgk/ghtj/gtgh/10981983046046597120.html (accessed on 28 April 2024).

55. Tangshan Municipal Bureau of Housing and Urban-Rural Development. The 14th Five-Year Special Plan for Urban Renewal in the Central Urban Area of Tangshan. Available online: https://zhujianju.tangshan.gov.cn/u/cms/tszjj/202407/01110943y5bq.pdf (accessed on 28 April 2024).

56. Yamane, T. Statistics: An Introductory Analysis; 2nd ed. Harper & Row press: New York, NY, USA, 1976; 886.

57. Osman, R.; Frantál, B.; Klusáček, P.; Kunc, J.; Martinát, S. Factors affecting brownfield regeneration in post-socialist space: The case of the Czech Republic. Land Use Policy; 2015; 48, pp. 309-316. [DOI: https://dx.doi.org/10.1016/j.landusepol.2015.06.003]

Word count: 7375

Show less

© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.

Abstract

Brownfields are abundant, widely dispersed, and subject to complex contamination, resulting in waste land, ecological degradation, and barriers to economic growth. The accurate identification of brownfield sites is key to formulating effective remediation and reuse strategies. However, the heterogeneity of surface features poses significant challenges for identifying various types of brownfields across entire urban areas. To address these challenges, this study proposes a “Target–Classification–Modification” (TCM) method for brownfield identification, which was applied to Tangshan City, China. This method consists of a three-stage process: target area localization, visual interpretation and classification, and site-level modification. It leverages integrated multi-source open-access data and clear rules for subtype classification and the determination of spatial boundaries and abandonment status. The results for Tangshan show that (1) the overall accuracy of the TCM method reached 84.9%; (2) a total of 1706 brownfield sites were identified, including 422 raw-material mining sites, 576 raw-material manufacturing sites, and 708 non-raw-material manufacturing sites; (3) subtype analysis revealed distinct spatial distribution and morphological patterns, driven by resource endowments, transportation networks, and industrial space organization. The TCM method improved the identification efficiency by 34.7% through precise target-area localization. It offers well-defined criteria to distinguish different brownfield subtypes. In addition, it employs a multi-approach strategy to determine the abandonment status, further enhancing accuracy. This method is scalable and widely applicable, providing support for urban-scale brownfield research and practice.

Details

Title
“Target–Classification–Modification” Method for Spatial Identification of Brownfields: A Case Study of Tangshan City, China
Author
Fu Quanchuan 1   VIAFID ORCID Logo  ; Zhu, Jingyuan 2   VIAFID ORCID Logo  ; Zheng Xiaodi 3 ; Li, Zhengxiang 4 ; Maini, Chen 2 ; He Yuyuwei 1 

1  School of Architecture and Design, Beijing Jiaotong University, Beijing 100044, China; [email protected] (Q.F.); 
2  School of Architecture, Tsinghua University, Beijing 100084, China; [email protected] (J.Z.); 
3  School of Architecture, Tsinghua University, Beijing 100084, China; [email protected] (J.Z.);, Key Laboratory of Eco-Planning and Green Building (Tsinghua University), Ministry of Education, Beijing 100084, China 
4  School of Architecture and Urban Planning, Chongqing University, Chongqing 400045, China; [email protected] 
First page
1213
Publication year
2025
Publication date
2025
Publisher
MDPI AG
e-ISSN
2073445X
Source type
Scholarly Journal
Language of publication
English
DOI
https://doi.org/10.3390/land14061213
ProQuest document ID
3223923839
Copyright
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.