Urban Development Detection Along the Transportation Corridors of the Mongolian Plateau Supported by SDGSAT-1 NTL Data

Highlights

What are the main findings?

• A method for characterizing urban development supported by multi-source remote sensing data, including nighttime light data with higher resolution.
• The City Development Index (CDI) indices of 11 cities were obtained, revealing the geographical differences of urban development on the Mongolian Plateau.

What are the implications of the main findings?

• The CDI was established to be used in urban development detection in grassland areas.
• Both SDGSAT-1 NTL data and socioeconomic data were used for urban development detection.

Abstract

Conducting urbanization monitoring at key nodes of the Mongolian Plateau holds significant importance for evaluating the economic and social development of cities within the China–Mongolia–Russia transportation corridor. These cities are sparsely distributed in the vast grassland areas, posing challenges for detecting urban changes in the transboundary regions. This study proposes a method for characterizing urban development supported by multi-source remote sensing data, including nighttime light (NTL) data with superior resolution, focusing on 11 major cities along the China-Mongolia-Russia Economic Corridor within the Mongolian Plateau. A City Development Index (CDI) is introduced, utilizing “information entropy” with the entropy weight method. The development status of 11 cities in the study area was analyzed using fractal dimensions, compactness, index of economy, index of social development, and the CDI. Results show Baotou has the highest CDI (0.82), followed by Hohhot (0.77), Ordos (0.73), and Ulanqab (0.66). In contrast, Mongolian cities exhibit significantly lower CDIs, relatively. The capital, Ulaanbaatar, has a CDI of 0.63, while no other Mongolian city exceeds 0.60, correlating with scattered populations. The results indicate that Inner Mongolia and Mongolia exhibit different geographical features in urban development. This approach provides a quantitative method for detecting and assessing urban development of the arid and semi-arid regions using NTL satellite data.

1. Introduction

Urbanization is a key area of global sustainable development, referring to population growth, spatial expansion of built-up areas, and transformation and upgrading of industrial structures. The United Nations Sustainable Development Goal 11 (SDG 11) emphasizes the importance of building inclusive, safe, resilient, and sustainable cities and human settlements [1]. Over the past few decades, rapid urbanization has occurred globally, and it has been suggested that by 2030, the urban population will reach approximately 8.6 billion [2]. Urbanization is accompanied by human-induced changes in the Earth’s landscapes [3]. Therefore, evaluating urban sustainable development is crucial, aiding in understanding the geographical changes in cities, as well as in formulating urban development policies.

Beyond urban remote sensing using daytime satellites, new nighttime light (NTL) data have been widely used in urbanization studies. NTL captures artificial light sources in cities, reflecting various social conditions indicating urban development level [4,5,6]. The Defense Meteorological Satellite Program-Operational Line Scan System (DMSP-OLS) and National Polar-Orbiting Partnership-Visible Infrared Imaging Radiometer Suite (NPP-VIIRS) have been extensively used to study urban cover and dynamics. Jin et al. used DMSP-OLS data to analyze urbanization processes and patterns in Hangzhou, eastern China [7]. Zhang et al. analyzed the spatiotemporal evolution of the Lanzhou-Xining urban agglomeration in western China using standard deviation ellipses and hotspot analysis [8]. Rafael et al. estimated the global metropolitan area using NTL data [9]. Zhang et al. revealed U.S. urban development trends using DMSP-OLS [10]. Tselios et al. applied spatial analyses based on DMSP-OLS data to study European urban clusters, functional zones, and patterns, focusing on cities such as London and Paris [11]. Gilbert et al. explored the correlations between built-up area dynamics and climate parameters in Lagos in West Africa using DMSP-OLS [12].

However, these applications still face challenges owing to the coarse spatial resolution of NTL sensors. For fine-resolution studies, multiple satellites, such as “Luojia-1”, “Jilin-1”, and SDGSAT-1, provide additional new NTL data. Among these, SDGSAT-1, launched in 2021, is the world’s first scientific satellite designed to meet the research requirements of the Sustainable Development Goals (SDGs) and support the United Nations’ 2030 SDGs agenda [13]. With a 10-m resolution, SDGSAT-1 can trace human activities in cities and provide up-to-date urban observation data. Liu et al. conducted a study on urban built-up area extraction in Shanghai, the core city of the Yangtze River Delta urban agglomeration in China, to facilitate the detailed monitoring of the SDGs [14]. Rodriguez-Antunano et al. provided methodological inspiration for the application of nighttime light remote sensing to the governance of intermediate cities [15]. Duan et al. carried out a study on the spatial distribution of urban population in the Guangdong-Hong Kong-Macao Greater Bay Area to support urban planning and SDGs evaluation [16]. Many previous studies have examined urban NTL remote sensing in developed regions. However, their application in sparsely populated areas, such as grassland areas, remains unclear. The arid/semi-arid environment of the Mongolian Plateau exhibits slow urbanization due to nomadic traditions driving seasonal population movements: winter concentration in capitals versus summer dispersion into Gers in grasslands. Different population densities, geographical conditions, and livestock systems create significant contrasts in urbanization between China and Mongolia in this region. Crucially, transportation corridor development along the China-Mongolia-Russia Economic Corridor—particularly railways and highways—has profoundly impacted regional urbanization. The corridor’s western railway connects key Sino-Mongolian cities, driving rapid urban-economic growth in grassland cities like Hohhot, Baotou, Ulaanbaatar, and Darkhan [17,18]. Evaluating their developmental status using fine-resolution NTL data coupled with the socio-economic context is still an urgent challenge.

Therefore, this study evaluated the development status of major cities along the transportation corridor of the Silk Road in the Mongolian Plateau using fine-resolution NTL data. First, built-up areas were isolated from the SDGSAT-1 NTL data. The fractal dimensions and compactness of these areas were then calculated and integrated using the temporal information entropy. Second, an indicator system was added for a more detailed assessment supported by socio-economic indicators. Given that the available SDGSAT-1 NTL data are currently limited to the period 2023–2025, we analyze the 2023 to reveal the current development status of 11 cities on the Mongolian Plateau. This study provides a new approach for the comprehensive characterization of urban development in cities in arid and semi-arid areas.

2. Materials and Methods

2.1. Study Area

The Mongolian Plateau serves as a core area for the China-Mongolia-Russia Economic Corridor passing through several major cities. In China, the cities under this corridor include Hohhot, Baotou, Ordos, Ulaanqab, and Erenhot, while in Mongolia they include Ulaanbaatar, Zamyn-Uud, Saynshand, Choir, Darkhan, and Sukhbaatar. We selected these 11 cities as the sample for an urban development study in the Mongolian Plateau (Figure 1).

2.2. Data

The SDGSAT-1 satellite is equipped with several advanced instruments, including a Thermal Infrared Spectrometer, a Glimmer Imaging Unit (GIU), and a Multispectral Imager for Littoral Areas. This instrument can acquire NTL images with 40-m multispectral, 10-m panchromatic, and 300-km-wide coverage. Eleven scenes of SDGSAT-1 NTL data were used for urban monitoring for 11 cities, spanning March to July 2023.

The ESRI Land Use Land Cover (LULC) product with 10 m resolution was selected for comparison because it has clear urban classification in the same year. This study incorporated socioeconomic data for cities, mainly obtained from OpenStreetMap and World Data Bank. The details of the data used are presented in

Table 1. Details of the Data.

Dataset	Spatial Resolution	Data Source
SDGSAT-1 nighttime light data	10-m panchromatic 2023	International Research Center of Big Data for Sustainable Development Goals
ESRI Land Use Land Cover (LULC) product	10-m LULC data, 2023	Environmental Systems Research Institute, Inc. (Redlands, CA, USA, https://livingatlas.arcgis.com/landcoverexplorer. (accessed on 13 June 2026))
Road vector data (OSM)	Vector Data 2023	OpenStreetMap (https://www.openstreetmap.org. (accessed on 13 June 2026))
POI vector data (OSM)	Vector Data 2023	OpenStreetMap (https://www.openstreetmap.org)
Population Distribution Data	Statistical Data 2023	World Pop (https://www.worldpop.org. (accessed on 13 June 2026))
Gross Domestic Product (GDP)	Statistical Data 2023	Kummu, Kosonen. and Masoumzadeh. 2025. (https://doi.org/10.5281/zenodo.10976733. (accessed on 13 June 2026))
China (Inner Mongolia) and Mongolia Statistics Data	Statistical Data 2023	World Data Bank official website (https://databank.worldbank.org/. (accessed on 15 June 2026))

.

2.3. Technical Roadmap

A technical roadmap for this study is shown in Figure 2. Built-up areas were first isolated based on SDGSAT-1 NTL data. It was used to obtain the fractal dimensions and compactness of cities. Considering economic and social development, an indicator system was introduced reflecting the economic and social development level. Information entropy, as defined in Section 3.3, was then used in order to comprehensively establish the City Development Index (CDI). The results were then verified by comparing them with the ESRI LULC data.

2.4. Isolating Urban Built-Up Areas Using Nighttime Light Data

The panchromatic band data of SDGSAT-1 NTL includes the spatial extent of urban and suburban administrative areas. Preprocessing was performed to eliminate noise using an algorithm developed by Liu et al. specifically for SDGSAT-1 NTL data [19]. Subsequently, SDGSAT-1 NTL data were converted into DN values. The typical range of the built-up areas was extracted using the threshold method. The regions in which the DN values exceeded the empirical threshold value (25) were masked, which was determined through visual interpretation by comparing NTL data with existing ESRI land cover data and Google Earth remote sensing imagery. The raster data were converted into polygons and merged. A morphological optimization algorithm was applied in order to finalize the isolation of urban built-up areas.

The fractal dimension and compactness of urban built-up areas were calculated based on the range extracted from the NTL data. Kernel density analysis was employed to calculate the density relationship between the NTL within the city and its surrounding neighborhood. The formula is as follows:

$$f\left(x\right)={\displaystyle \sum _{i=1}^n}{\displaystyle \frac{1}{r^2}}k({\displaystyle \frac{\left|x-c_i\right|}{r}})$$

(1)

where, f(x) represents the kernel density function value at x, r is the search radius, and k is the spatial weight function, which is the DN value from the SDGSAT-1 NTL images. The term |x−c_i| denotes the distance between the ith point around x and x itself, and n is the total number of points within a distance less than or equal to r.

The fractal dimension P indicates the capacity of the urban built-up area to fill space and the complexity of its irregular boundaries. The formula is as follows:

$$P={\displaystyle \frac{2\mathrm{l}\mathrm{n}(C/4)}{\ln S}}$$

(2)

where, S and C represent the area and perimeter of the urban built-up area, respectively.

Compactness is a crucial indicator that reflects the spatial shape of a city and is used to describe the degree of spatial concentration. It is given by Formula (3).

$$Q={\displaystyle \frac{2\sqrt{S\pi }}{C}}$$

(3)

where Q represents compactness and S and C denote the area and perimeter of the urban built-up area, respectively.

The urban built-up areas extracted from the NTL data were defined as the NTL-derived built-up area (S_N). Simultaneously, the built-up area was extracted from the ESRI Land Use Land Cover (LULC) product—a refined dataset primarily based on traditional daytime remote sensing—and defined as the daytime remote sensing-derived built-up area (S_D). Consequently, a new factor, the Urban Built-up Area Activity Ratio (M_S), was defined, where M_S = S_N/S_D. The M_S has a value range of 0 to 1. A standard deviation ellipse is commonly used to measure trends of sets of points or areas. This method involves calculating the standard distances in the x- and y-directions that then define the axes of an ellipse that encompasses the distribution of all elements. The mean center is used as the origin, and the standard deviations of the x- and y-coordinates define the axes. The formula used is that of the centroid (or mean center):

$$\left(X,Y\right)=\left({\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n}{k}_i{x}_i}{{\displaystyle {\sum }_{i=1}^n}{k}_i}},{\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n}{k}_i{y}_i}{{\displaystyle {\sum }_{i=1}^n}{k}_i}}\right)$$

(4)

where, (x_i, y_i) represents the coordinates of the ith light grid, and ki represents the NTL DN value of the ith grid element. The rotation angle θ is given by:

$$\tan \theta ={\displaystyle \frac{A+\sqrt{A^2+B^2}}{B}}$$

(5)

where

$$A={\displaystyle \sum _{i=1}^n}{(x_i-\overline{X})}^2-{\displaystyle \sum _{i=1}^n}{(y_i-\overline{Y})}^2$$

(6)

and

$$B=2{\displaystyle \sum _{i=1}^n}(x_i-\overline{X})(y_i-\overline{Y})$$

(7)

The standard deviation along the x- and the y-axes are, respectively,

$${\sigma }_x=\sqrt{2}\sqrt{{\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n}{[\left(x_i-\overline{X}\right)\cos \theta -(y_i-\overline{Y}\left)\sin \theta \right]}^2}{n}}}$$

(8)

and

$${\sigma }_y=\sqrt{2}\sqrt{{\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n}{[\left(x_i-\overline{X}\right)\sin \theta +(y_i-\overline{Y}\left)\cos \theta \right]}^2}{n}}}$$

(9)

2.5. Validation for Urban Built-Up Areas

To objectively evaluate the reliability of the urban built-up area results extracted from NTL, this study employs the confusion matrix as the fundamental tool for accuracy validation. The confusion matrix is a 2 × 2 square matrix, with rows representing the ground-truth classes (built-up area based on ESRI LULC data) and columns representing the predicted classes (built-up area based on NTL data). Each entry in the matrix corresponds to the number of pixels classified into the respective row–column combination. The sum of diagonal elements gives the total number of correctly classified pixels, while off-diagonal entries indicate misclassification errors. Based on this, the following basic statistics are defined:

True Positives (TP): pixels truly belonging to built-up area (LULC) and correctly predicted as built-up area (NTL);

False Positives (FP): pixels not belonging to built-up area (LULC) but incorrectly predicted as built-up area (NTL);

False Negatives (FN): pixels truly belonging to built-up area (LULC) but misclassified as non-built-up area (NTL);

True Negatives (TN): pixels that are neither truly nor predicted built-up area.

Based on these definitions, four commonly used metrics are calculated to comprehensively assess classification performance:

Precision: measures the proportion of pixels predicted as a built-up area that belong to that class, indicating the model’s exactness. The formula is as follows:

$${\mathrm{Precision}}_i={\displaystyle \frac{TP}{TP+FP}}$$

(10)

Recall: measures the proportion of actual built-up area pixels that are correctly predicted, indicating the model’s completeness. The formula is as follows:

$${\mathrm{Recall}}_i={\displaystyle \frac{TP}{TP+FN}}$$

(11)

F1 Score: the harmonic mean of Precision and Recall, balancing both metrics and mitigating evaluation bias caused by class imbalance. The formula is as follows:

$${ \mathrm{F}1}_i=2\times {\displaystyle \frac{\mathrm{Precision}\times \mathrm{Recall}}{\mathrm{Precision}+\mathrm{Recall}}}$$

(12)

Overall Accuracy (OA): The ratio of correctly classified pixels across all classes to the total number of validation pixels, reflecting the global classification performance. The formula is as follows:

$$\mathrm{OA}={\displaystyle \frac{TP}{N_{\mathrm{total}}}}$$

(13)

where $N_{\mathrm{total}}$ is the total number of validation pixels.

2.6. CDI and Weighting Method Based on Information Entropy

The fractal dimension and the compactness of the urban built-up area were extracted from the NTL data. Considering the differences in economic and social development levels among cities along the main transportation corridors of the Mongolian Plateau, we built the indicator system for economy and social development of the urban area from socioeconomic data (

Table 2. Indicator system for economy and social development of urban area.

System Level	Indicator Level	Indicator Descriptions	Properties
Economy Development	GDP per capita (USD)	GDP/Population	+
	Population	Population inside city administrative boundaries	+
	Saved at a bank or similar financial institution (% of population)	The percentage of respondents who report saving or setting aside any money at a bank or similar financial institution in the past year.	+
	Gini index	The extent to which the distribution of income among individuals or households within an economy deviates from a perfectly equal distribution.	-
	Poverty headcount ratio at $3.00 a day (% of population)	Poverty headcount ratio at $3.00 a day is the percentage of the population living on less than $3.00 a day at purchasing power adjusted prices.	-
	GDP (annual % growth)	The percentage change over each previous year of the constant price series in United States dollars.	+
	Total Revenue and Grants (USD)	Dataset: Macro Poverty Outlook. Refer to: https://www.worldbank.org/en/publication/macro-poverty-outlook. (accessed on 14 July 2026)	+
	Gross Fixed Investment as percentage of GDP	Dataset: Macro Poverty Outlook. Refer to: https://www.worldbank.org/en/publication/macro-poverty-outlook. (accessed on 14 July 2026)	+
	Capital and Financial Account Balance (USD)	Dataset: Macro Poverty Outlook. Refer to: https://www.worldbank.org/en/publication/macro-poverty-outlook. (accessed on 14 July 2026)	+
	General Government Debt Stock (% of GDP)	Dataset: Macro Poverty Outlook. Refer to: https://www.worldbank.org/en/publication/macro-poverty-outlook. (accessed on 14 July 2026)	-
	Private Consumption (% of GDP)	Dataset: Macro Poverty Outlook. Refer to: https://www.worldbank.org/en/publication/macro-poverty-outlook. (accessed on 14 July 2026)	+
	Inflation Rate	Dataset: Macro Poverty Outlook. Refer to: https://www.worldbank.org/en/publication/macro-poverty-outlook. (accessed on 14 July 2026)	-
	Population growth (annual %)	Annual population growth rate for year t is the exponential rate of growth of midyear population from year t-1 to t, expressed as a percentage.	+
Social Development	Number of traffic facilities (Count)	Number of traffic facilities (Count) based on OSM traffic facilities vector data	+
	Number of roads (Count)	Number of roads (Count) based on OSM roads vector data	+
	Number of motorways (highways) (Count)	Number of motorways (highways) (Count) based on OSM roads vector data	+
	Number of hospitals (Count)	Number of hospitals (Count) based on OSM POI vector data	+
	Number of public facilities (Count)	Number of public facilities (Count) based on OSM POI vector data	+
	Number of commercial facilities (Count)	Number of commercial facilities (Count) based on OSM POI vector data	+
	Number of railway (Count)	Number of railway (Count) based on OSM railway vector data	+
	Life expectancy at birth, total (years)	The number of years a newborn infant would live if prevailing patterns of mortality at the time of its birth were to stay the same throughout its life.	+
	Level of water stress: freshwater withdrawal as a proportion of available freshwater resources (% of freshwater resources)	Freshwater withdrawal as a proportion of available freshwater resources is the ratio between total freshwater withdrawn by all major sectors and total renewable freshwater resources, after considering environmental water requirements.	-
	Greenhouse gas (GHG) emissions caused by the transportation sector (Tonnes of CO2-equivalent)	Source: Climate Watch (World Resources Institute), World Resources Institute (WRI)	-
	Percentage of individuals using the internet (ITU) (% of population)	Proportion of individuals who used the Internet from any location in the last three months. Access can be via a fixed or mobile network.	+
	Global Cybersecurity Index—Overall Score (ITU GCI)	The Global Cybersecurity Index (GCI) measures the commitment of countries to cybersecurity along five pillars: (i) Legal Measures, (ii) Technical Measures, (iii) Organizational Measures, (iv) Capacity Development, and (v) Cooperation.	+
	Used a mobile phone or the internet to pay bills, population 25+, cumulative (% of population)	The percentage of respondents who report using a mobile phone or the internet to pay bills in the past year.	+
	Educational attainment, at least Master’s or equivalent, population 25+, cumulative (% of population)	The percentage of population ages 25 and over that attained or completed Master’s or equivalent.	+
	Physicians (per 1000 people) (Count)	Generalist and specialist medical practitioners.	+
	Domestic general government health expenditure (% of GDP)	Public expenditure on health from domestic sources as a share of the economy as measured by GDP.	+
	Human Capital Index Plus (HCI+): Health Pillar Score	The HCI Plus Health pillar captures productivity effects of surviving to working age and achieving adequate early-life physical growth.	+
	Human Capital Index Plus (HCI+): Education Pillar Score	The HCI Plus Education pillar aggregates human capital accumulated during formal schooling.	+
	Universal Health Coverage (UHC) service coverage index	A composite index representing coverage of essential health services based on 14 tracer indicators in the areas of reproductive, maternal, newborn, and child health, infectious diseases, noncommunicable diseases, and service capacity and access.	+
	Unemployment Rate	Unemployment refers to the share of the labor force that is without work but available for and seeking employment.	-
	GCI 4.0: Electrification rate (% of population)	The percentage of population that has access to electricity. Refer to: World Economic Forum (WEF)	+

). These indicators serve as the basis for constructing two separate indices—an economic index I-economy (based on 13 indicators) and a social development index I-society (based on 21 indicators). The criteria for selecting the 34 indicators were informed by existing urban evaluation studies and research assessing Sustainable Development Goal 11 [20,21,22].

To eliminate differences in magnitude among the evaluation indicators, the data for each indicator need to be standardized. The standardization formulas are as follows:

$${x^{\prime }}_{ij}={\displaystyle \frac{x_j-x_{min}}{x_{max}-x_{min}}}$$

(14)

$${x^{\prime }}_{ij}={\displaystyle \frac{x_{max}-x_j}{x_{max}-x_{min}}}$$

(15)

where, the standardized value x′_ij is derived from the original data value x_j for the jth indicator, with x_max representing the maximum value and xmin representing the minimum value of the jth indicator. The Formulas (14) and (15) are used for the standardization of positive (+) and negative (-) indicators, respectively. The 34 standardized indicators were processed in two separate sets: the 13 economic indicators were summed with equal weighting and normalized to the 0–1 range to produce the I-economy, while the 21 social indicators were subjected to the same procedure to generate the I-society.

The fractal dimension and compactness of the urban built-up area, I-economy, and I-society are selected to establish a City Development Index (CDI). The entropy weighting method was used to assign weights to the four indicators. The information entropy value reflects the degree of uncertainty or randomness in the distribution of the indicator levels. Higher entropy values indicate greater uncertainty, whereas lower values indicate greater predictability. The information entropy H_j for the jth evaluation indicator is calculated using the following formula:

$$H_j=-k{\displaystyle \sum _{j=1}^m}{p}_{ij} ln p_{ij}$$

(16)

where,

$$k={\displaystyle \frac{1}{ln m}}$$

and m denotes the total number of levels for the jth indicator; p_ij is the proportion of the jth indicator’s ith level to the total number of observations.

The formula for calculating the weight of each evaluation indicator using the entropy weight method is as follows:

$$W_i={\displaystyle \frac{1-H_i}{j-\sum H_i}}$$

(17)

After calculating the weights of the four evaluation indicators, normalized fractal dimensions (P_n), normalized compactness (Q_n), I-economy and I-society of the 11 cities were integrated to form a composite index. The resulting composite measure was designated as CDI. Finally, the CDI can be expressed as follows:

$$CDI=0.262 P_n+0.256 Q_n+0.181 I_{economy}+0.301 I_{society}$$

(18)

where, the City Development Index (CDI) ranges from 0 to 1.

3. Results

3.1. Urban Spatial Distribution in Nighttime Light Images

The original RGB and DN value images of NTLs for the 11 cities, including Hohhot, Baotou, Ordos, Ulaanqab, Erenhot, Ulaanbaatar, Zamyn-Uud, Saynshand, Choir, Darkhan, and Sukhbaatar, are shown in Figure 3 and Figure 4. The spatial distribution of the urban built-up areas in Hohhot exhibits a clear pattern of central aggregation (Figure 3a). In Baotou, the urban built-up areas are distributed predominantly along the east-west axis, with higher concentrations around the city center (Figure 3b). In Ordos, urban built-up areas were distributed predominantly along the southwest-northeast axis (Figure 3c). The spatial distribution of urban built-up areas in Ulaanqab, Erenhot, Zamyn-Uud, Saynshand, and Choir also exhibited a clear central aggregation pattern (Figure 3d–h). Ulaanbaatar has multiple dispersed built-up areas, with scattered cores tending to cluster around the city center (Figure 3i). The spatial distribution of the urban built-up areas in Darkhan and Sukhbaatar was influenced by the terrain, following the direction of the river and valleys through the plains (Figure 3j,k).

3.2. Results of I-Economy and I-Society

The I-economy and I-society results reflect the comprehensive developmental differences among the cities in economy and social development (

Table 3. I-economy and I-society of cities along major transportation corridors in the Mongolian Plateau.

Cities	I-Economy	I-Society
Hohhot	0.85	1.00
Baotou	0.90	0.93
Ordos	1.00	0.84
Ulanqab	0.64	0.65
Erenhot	0.30	0.15
Zamyn-Uud	0.06	0.09
Saynshand	0.04	0.36
Choir	0.05	0.43
Ulaanbaatar	0.37	0.90
Darkhan	0.03	0.51
Sukhbaatar	0.00	0.00

). Among the cities in Inner Mongolia, Ordos stands out with the highest I-economy. By contrast, Erenhot, a border city, has a low I-economy of only 0.30. In the Mongolian region, where the I-economy levels are generally low across the various cities, Ulaanbaatar is the most prominent city, with an I-economy of 0.37. Border cities Sukhbaatar exhibits the lowest I-economy values. Other cities in Mongolia also exhibited low I-economy values under 0.10.

For I-society, among the cities in Inner Mongolia, Hohhot stands out with the highest I-society. By contrast, Erenhot has a low I-society of only 0.15. In the Mongolian region, Ulaanbaatar is the most prominent city, with a high I-society of 0.90. Border cities, such as Sukhbaatar, also exhibit the lowest I-society values. Darkhan has a relatively high I-society of 0.51. Other cities in Mongolia also exhibited low I-society values under 0.45.

3.3. Validation for Urban Built-Up Areas and Analysis of Urban Morphology

The standard deviation ellipses of cities along the main transportation corridors of the Mongolian Plateau were calculated from the extracted urban built-up area boundaries. Figure 5 illustrates the urban built-up area boundaries and standard deviation ellipses for the 11 cities. The ratio of the major axis to the minor axis of the ellipse is calculated.

As shown in Figure 6, we selected Hohhot, Ordos, and Ulaanbaatar as the samples for validation. Three validation areas in each city were selected; consequently (totally, we have 9 validation areas). In Hohhot, we have 1,110,748 pixels for validation; in Ordos, we have 625,419 pixels for validation; in Ulaanbaatar, we have 511,776 pixels for validation. We calculated the confusion matrices for the three validation sample cities and subsequently computed four accuracy indicators. The Precision values are 0.88, 0.72, and 0.98, respectively; the Recall values are 0.85, 0.96, and 0.58; the Overall Accuracy values are 0.79, 0.72, and 0.63; and the F1-Score values are 0.86, 0.83, and 0.73. The average values of Precision, Recall, Overall Accuracy, and F1-Score are 0.86, 0.60, 0.71, and 0.81, respectively. Through validation, we confirmed that the built-up area extracted from the NTL data accurately captured the urban built-up areas, thereby ensuring the reliability of subsequent analyses.

Urban built-up area boundaries were extracted from NTL. Based on these boundaries, the fractal dimensions, compactness, and urban built-up area activity ratios (Ms) of cities along the main transportation corridors of the Mongolian Plateau were calculated. The fractal dimensions and compactness of cities in 2023 are listed in Table 3. Ms can better reflect the vitality of urban built-up areas and help identify issues such as “ghost cities” and “hollow villages,” which are the challenges faced by many cities during their development process.

The fractal dimension describes the complexity of the boundary between the urban and non-built-up areas. A higher fractal dimension indicates a more irregular and fragmented boundary in the built-up area. Compactness reflects the spatial clustering in urban built-up areas, with higher compactness suggesting that built-up areas are more concentrated around their spatial center.

As shown in

Table 4. City development parameters along major transportation corridors in the Mongolian Plateau in 2023.

Cities	Fractal Dimension	Compactness	Urban Built-Up Area Activity Ratio
Hohhot	1.300	0.047	0.91
Baotou	1.398	0.021	0.87
Ordos	1.293	0.020	0.86
Ulanqab	1.369	0.033	0.85
Erenhot	1.287	0.081	0.95
Zamyn-Uud	1.428	0.044	0.96
Saynshand	1.412	0.045	0.94
Choir	1.420	0.052	0.92
Ulaanbaatar	1.467	0.012	0.79
Darkhan	1.397	0.038	0.99
Sukhbaatar	1.383	0.057	0.96

, among the cities in Inner Mongolia, Baotou had the highest fractal dimension (1.398) in 2023, indicating a relatively complex urban boundary and a trend toward built-up area expansion. By contrast, the smaller border city of Erenhot has the lowest fractal dimension (1.287), reflecting its primary role as a port city with limited additional functional zones and well-planned port areas. Notably, Ordos, another major city, has a fractal dimension of 1.293, similar to Erenhot, suggesting effective planning and management of its built-up area boundary.

In Mongolia, Ulaanbaatar exhibited the highest fractal dimension (1.467) in 2023, which was also the highest among 11 cities on the Mongolian Plateau. This reflects both the urban expansion trend and the somewhat disordered planning of built-up area boundaries. Overall, cities in Mongolia generally showed slightly higher fractal dimensions than those in Inner Mongolia, which may be attributed to differences in urban planning approaches. Ulaanbaatar had the lowest compactness value (0.012) among the 11 cities, indicating a spatially dispersed distribution of built-up areas, which is consistent with the presence of scattered ger districts and temporary settlements. In contrast, Erenhot has the highest compactness (0.081), aligning with its actual condition as a port city where built-up areas are concentrated around the port administrative zone.

3.4. Results of City Development Index

The CDI results reflect the comprehensive developmental differences among the cities in terms of urban built-up area expansion, economy, and social development. These results provide valuable insights for assessing the sustainable development of cities along the main transportation corridors of the Mongolian Plateau (

Table 5. The City Development Index (CDI) for cities along major transportation corridors of the Mongolian Plateau in 2023.

Cities	The City Development Index (CDI)
Hohhot	0.77
Baotou	0.82
Ordos	0.73
Ulanqab	0.66
Erenhot	0.58
Zamyn-Uud	0.43
Saynshand	0.50
Choir	0.55
Ulaanbaatar	0.63
Darkhan	0.52
Sukhbaatar	0.42

).

Based on the CDI results of 2023, an inspection revealed that Baotou stands out within the Inner Mongolian region of China, with the highest CDI of 0.82. Hohhot achieved a CDI of 0.77, and Ordos achieved a high CDI of 0.73. Ulanqab has undergone significant development in recent years, achieving a relatively high CDI value of 0.66. By contrast, Erenhot, a border city, has a low CDI of 0.58. In the Mongolian region, where the CDI levels are generally low across the various cities, Ulaanbaatar is the most prominent city, with a CDI of 0.63. Choir follows with a CDI of 0.55. Border cities Zamyn-Uud and Sukhbaatar exhibit low CDI values of 0.43 and 0.42, respectively. Other cities in Mongolia also exhibited low CDIs, with Saynshand at 0.50 and Darkhan at 0.52.

Overall, Hohhot remains the most prominent city in terms of the CDI in 2023. Except for Erenhot, cities in the Inner Mongolia region of China had relatively high CDI values. In contrast, cities in the Mongolia region generally exhibit low CDI values, highlighting the current uneven development along the China-Mongolia-Russia corridor. The underdeveloped state of cities in Mongolia is evident, with only Ulaanbaatar achieving a CDI of 0.63. This was found to be lower than the CDI levels of the four cities in Inner Mongolia, China.

Based on the CDI results, we ranked the 11 cities and classified them into four development levels. The CDI values (ranging from 0 to 1) were divided into four equal intervals: 0.75–1.0 for the high development level, 0.50–0.75 for the medium development level, 0.25–0.50 for the relatively low development level, and 0.00–0.25 for the very low development level. The classification results are broadly consistent with the actual development conditions of cities in this region. Hohhot and Baotou are categorized as high development levels, both exhibiting outstanding performance in economic, social, and green development dimensions. Zamyn-Uud and Sukhbaatar fall into the relatively low development level; both are border port cities, yet their port scales are smaller than that of Erenhot (another port city), and they have smaller populations and receive less development investment. The results also reveal an overarching geographic disparity in development between China and Mongolia within the Mongolian Plateau region: all cities on the Chinese side are at either high or medium development levels, whereas no Mongolian city reaches the high level, and the two relatively low-level cities are both located in Mongolia.

Based on the P_n, Q_n, I-economy, I-society, and CDI results, we conducted a sensitivity analysis (

Table 6. Sensitivity analysis of driving factors.

Cities	CDI	CDI (Excluding Pn Factor)	CDI (Excluding Qn Factor)	CDI (Excluding I-Economy Factor)	CDI (Excluding I-Society Factor)
Hohhot	0.77	0.70521	0.94495	0.75117	0.57866
Baotou	0.82	0.79969	0.9077	0.80618	0.68401
Ordos	0.73	0.66992	0.89212	0.66388	0.5742
Ulanqab	0.66	0.56321	0.74548	0.66314	0.58861
Erenhot	0.58	0.48055	0.43974	0.64458	0.72219
Zamyn-Uud	0.43	0.23817	0.3959	0.51413	0.54856
Saynshand	0.50	0.34662	0.49308	0.60996	0.54166
Choir	0.55	0.4087	0.52731	0.66692	0.57988
Ulaanbaatar	0.63	0.50885	0.80551	0.6961	0.46887
Darkhan	0.52	0.37754	0.54773	0.63686	0.5071
Sukhbaatar	0.42	0.2422	0.33088	0.51826	0.58238

,

Table 7. Value change in sensitivity analysis.

Cities	CDI (Excluding Pn Factor)	CDI (Excluding Qn Factor)	CDI (Excluding I-Economy Factor)	CDI (Excluding I-Society Factor)
Hohhot	−0.06479	0.17495	−0.01883	−0.19134
Baotou	−0.02031	0.0877	−0.01382	−0.13599
Ordos	−0.06008	0.16212	−0.06612	−0.1558
Ulanqab	−0.09679	0.08548	0.00314	−0.07139
Erenhot	−0.09945	−0.14026	0.06458	0.14219
Zamyn-Uud	−0.19183	−0.0341	0.08413	0.11856
Saynshand	−0.15338	−0.00692	0.10996	0.04166
Choir	−0.1413	−0.02269	0.11692	0.02988
Ulaanbaatar	−0.12115	0.17551	0.0661	−0.16113
Darkhan	−0.14246	0.02773	0.11686	−0.0129
Sukhbaatar	−0.1778	−0.08912	0.09826	0.16238

and

Table 8. Class change in sensitivity analysis.

Cities	CDI (Excluding Pn Factor)	CDI (Excluding Qn Factor)	CDI (Excluding I-Economy Factor)	CDI (Excluding I-Society Factor)
Hohhot	downgrade	unchanged	unchanged	downgrade
Baotou	unchanged	unchanged	unchanged	downgrade
Ordos	unchanged	upgrade	unchanged	downgrade
Ulanqab	unchanged	unchanged	unchanged	unchanged
Erenhot	downgrade	downgrade	unchanged	unchanged
Zamyn-Uud	downgrade	unchanged	upgrade	upgrade
Saynshand	downgrade	downgrade	unchanged	unchanged
Choir	downgrade	unchanged	unchanged	unchanged
Ulaanbaatar	unchanged	upgrade	unchanged	downgrade
Darkhan	downgrade	unchanged	unchanged	unchanged
Sukhbaatar	downgrade	unchanged	upgrade	upgrade

). We took out each of the four CDI components individually and recalculated the CDI. We recorded the value changes and the level shifts. Exclusion of P_n caused an overall underestimation of the CDI for all 11 cities. The reason is that P_n measures the capacity of the urban built-up area to fill space. Without P_n, the CDI ignores the high built-up density that exists in all cities. Exclusion of Q_n produced two opposite effects. For the two largest cities (Ordos and Ulaanbaatar), the CDI was considerably overestimated, and their development levels rose. For two smaller cities (Erenhot and Saynshand), the CDI was notably underestimated, and their levels fell. Exclusion of I-economy led to a substantial overestimation for two border-port cities (Zamyn-Uud and Sukhbaatar), and their levels increased. At the same time, it caused an underestimation for the three large cities (Hohhot, Baotou and Ordos). Exclusion of I-society caused substantial overestimation for the two port cities (Zamyn-Uud and Sukhbaatar), raising their levels. It also caused considerable underestimation for the four large cities (Hohhot, Baotou, Ordos, and Ulaanbaatar), lowering their levels.

Based on Table 7 and Table 8, we observed that removing P_n caused substantial numerical changes in the CDI for all 11 cities and resulted in the highest frequency of upward or downward shifts in urban development classification compared to the removal of other parameters. This suggests that the fractal dimension represented by P_n is the most sensitive parameter. This outcome may be attributable to the fact that the fractal dimension primarily captures the boundary complexity of all different city types, and its removal exerts the most pronounced influence on CDI results.

4. Discussion

4.1. Adaptability, Method Comparison and Accuracy Validation

Monitoring urban development in the prairie regions is challenging because of the limited distribution of urban settlement areas across vast expanses. Large-area remote sensing often encounters difficulties in effectively identifying dynamic characteristics and tracking sustainable urban development. However, high-spatial-resolution NTL remote sensing data can better identify and analyze dispersed urban settlement areas, thereby facilitating the assessment of urban sustainability. Among similar datasets, SDGSAT-1 NTL data offers the highest spatial resolution. This makes it well-suited for monitoring urban sustainability in the Mongolian Plateau and in more arid and semi-arid areas across Eurasia.

Building on prior accuracy assessments, this study validates SDGSAT-1’s effectiveness for urbanization monitoring. Previous research has underscored the potential of remote sensing for examining urbanization. Park et al. analyzed urbanization in six cities in the Mongolian Plateau using 30-m Landsat data to assess spatial expansion [18]. In contrast, our study employed SDGSAT-1 NTL data with 10-m resolution. Liu et al. utilized SDGSAT-1 NTL data and a logarithmic LitBV index to extract urban built-up areas [19]. Guo et al. compared SDGSAT-1, LJ1-01, and VIIRS data, confirming SDGSAT-1’s exceptional spatial resolution [23]. Compared to other NTL products, SDGSAT-1 Glow Imagery Unit (GIU) data offers superior resolution, multispectral capabilities, and wide coverage, enabling village-scale light source analysis [24].

Similarly, we extracted grassland built-up area parameters using SDGSAT-1, validating them against ESRI LULC data with >85% accuracy for both parameters. Comparisons between the fractal dimensions and compactness of built-up areas from the ESRI LULC and SDGSAT-1 data were conducted to verify accuracy and ensure the accuracy of the urban parameters obtained from NTL data. This allows for detailed mapping of built-up areas and precise urban development assessments. Through validation, we confirmed that the urban built-up area, fractal dimension, and compactness derived from the NTL data accurately captured the characteristics of urban built-up areas, thereby ensuring the reliability of subsequent analyses. As shown in

Table 9. Comparative analysis of CDI results from SDGSAT-1 and VIIRS in 2023.

Cities	CDI Results from SDGSAT-1	CDI Results from VIIRS	Accuracy of VIIRS
Hohhot	0.77	0.59	0.76
Baotou	0.82	0.77	0.93
Ordos	0.73	0.56	0.77
Ulanqab	0.66	0.52	0.78
Erenhot	0.58	0.22	0.38
Zamyn-Uud	0.43	0.11	0.26
Saynshand	0.50	0.24	0.48
Choir	0.55	0.29	0.53
Ulaanbaatar	0.63	0.35	0.56
Darkhan	0.52	0.27	0.52
Sukhbaatar	0.42	0.15	0.36

, we calculated the CDI in 2023 based on VIIRS NTL data for comparison. In contrast to the CDI results derived from SDGSAT-1, those from VIIRS exhibit a clear underestimation of the development status across all cities, with a particularly marked underestimation observed for small and medium-sized cities such as Zamyn-Uud, Sukhbaatar, and Erenhot. The most plausible explanation is the coarser spatial resolution of VIIRS, which introduces substantial errors when delineating the built-up extents of smaller cities.

As shown in

Table 10. Results of accuracy indicators for built-up area extracted from VIIRS.

Cities	Precision (VIIRS)	Recall (VIIRS)	Overall Accuracy (VIIRS)	F1-Score (VIIRS)
Hohhot	0.71	0.69	0.66	0.69
Ordos	0.62	0.67	0.61	0.64
Ulaanbaatar	0.68	0.51	0.57	0.58
Total Average	0.67	0.62	0.61	0.64

, we calculated the result of accuracy indicators for the built-up area extracted from VIIRS for comparison. Four accuracy indicators for the same three validation sample cities and the same validation areas were calculated in Section 3.3. The Precision values for the three validation sample cities are 0.71, 0.62, and 0.68, respectively; the Recall values are 0.69, 0.67, and 0.51; the Overall Accuracy values are 0.66, 0.61, and 0.57; and the F1-Score values are 0.69, 0.64, and 0.58. The average values of Precision, Recall, Overall Accuracy, and F1-Score are 0.67, 0.62, 0.61, and 0.64, respectively. Validation results confirm that the VIIRS-based built-up area extraction yields lower accuracy than the SDGSAT-1-based extraction.

Within the framework of urban NTL remote sensing, this study has considered non-remote sensing data such as socioeconomic factors. This study uses the concept of information entropy and the entropy weight method to incorporate finer-scale urban development information [25,26,27]. This study has also used both daily and NTL remote sensing data to reflect the vitality of urban built-up areas and help identify issues such as “ghost cities” and “hollow villages,” which are challenges faced by many cities during their development process.

4.2. Anthropogenic and Natural Driving Factors of Urban Development

Understanding the factors influencing urban development requires a closer examination of human-induced constraints and natural limitations. Urban development is influenced by human and natural factors. An analysis of major cities along the transportation corridors of the Mongolian Plateau revealed that the CDI of Mongolia was low, with cities such as Hohhot and Baotou in China’s Inner Mongolia Autonomous Region having significantly higher CDIs in comparison. Ulaanbaatar, the capital and largest city of Mongolia, houses nearly 50% of the country’s population. However, approximately 60% of the population of Ulaanbaatar lives in the Ger district, which consists of informal settlements or shantytowns located on the outskirts of the city [28]. They have limited access to public transportation, healthcare, and education, thereby posing significant challenges to the livelihoods of the residents of Ulaanbaatar and urban development [29,30,31,32].

In order to comprehend the historical context of current urban layout issues, we revisited the urban planning history of Mongolia. The first comprehensive urban development plan for Ulaanbaatar was established in Moscow in 1954 and has since undergone six rounds of revisions. Before the Fifth Master Plan was formulated in 2002, Ulaanbaatar’s urban land use lacked planning and policies. This situation persisted for more than a decade during a period of social transformation [33]. The significant land privatization movement in Ulaanbaatar during the 1990s had a major impact, coupled with political and economic structural changes in Mongolia. These factors led to the chaotic expansion of the Ger areas [30,34]. These circumstances have laid the groundwork for uncontrolled urban sprawl [35]. Consequently, Ulaanbaatar now has a vertically expanding and fragmented urban structure [36].

Policy initiatives for urbanization in China differ from those in Mongolia. Government-led initiatives such as China’s Urban Agglomeration of the Yellow River Ji-Shaped Bend (UAYB) city cluster promote coordinated development. In 2020, the Central Financial and Economic Affairs Commission of China formally proposed the concept of the UAYB city cluster. This cluster includes Hohhot, Baotou, and Ulanqab, emphasizing coordinated development within the cluster [37]. This highlights the significant differences in government roles in the urbanization process among the three cities in Inner Mongolia and Ulaanbaatar. Combined with differing economic structures, these factors lead to different drivers of urban development, resulting in differences in CDI.

Natural features also play a significant role in urban expansion limitations. Urban development is constrained by various factors, including spatial planning, topography, and industrial layout. Regional natural features, such as the Altai Mountains, Selenge River and its many branches, and unique topographical characteristics, inevitably influence urban planning. In mountainous areas, slopes greater than 10 degrees and rivers are significant geographical factors that limit urban expansion [28]. Cities located in such regions face severe restrictions due to topographical factors [38]. In industrial layouts, cities also consider natural factors such as mineral distribution, education levels, and connectivity with other administrative regions. The ecological and environmental issues of the Mongolian Plateau also affect urban planning and development. For example, the frequent occurrence of sandstorms severely affects urban development in Mongolia [36].

Hohhot and Ulaanbaatar are the capital cities of the Inner Mongolia Autonomous Region and Mongolia, respectively. Both cities experienced rapid urbanization, particularly between 2000 and 2010 [29]. Ulaanbaatar is located in a valley on the Tuul River in north-central Mongolia at the foot of the Bogd Khan Mountain. It is surrounded by mountains to its north and south. This natural constraint caused Ulaanbaatar’s urban expansion to occur primarily in the east-west direction of the valley. The Tuul River hindered its expansion to the south [28]. This natural constraint distinguishes the urban development of Ulaanbaatar from that of Hohhot, located on the Tumochuan Plain. The urbanization rate in the Hohhot-Baotou-Ordos (HBO) region increased from 55.96% in 2000 to 80.99% in 2020 [39]. Therefore, there are significant differences in the natural factors affecting urban development between these two cities. These constraints are a critical reason for Ulaanbaatar’s CDI (0.63), which is lower than that of Hohhot (0.77).

Different socio-environmental contexts across regions present unique urban planning challenges. Although the living conditions of Chinese cities in the Inner Mongolia Autonomous Region have improved, fragile ecosystems remain at risk [40]. In contrast, Mongolian cities face challenges such as a single industry, low output value, unreasonable urban planning, and urban waste pollution. For example, Darkhan faces severe waste pollution in the surrounding Ger district [41]. Ulaanbaatar faces severe air pollution, with the PM_2.5 of Ulaanbaatar from January through March 2020 being 129 ± 57, 71.1 ± 32.9, and 33.1 ± 12.4 µg/m³, respectively [42]. The current research in the literature has explored the factors driving urbanization in six cities on the Mongolian Plateau from 1990 to 2015. Other studies have focused on the relationship between urbanization and development from 2000 to 2010 and their correlation with urbanization rates. The findings indicate that economic development (0.559) is the primary driver of urbanization in Inner Mongolia, whereas social goods (0.646) and economic development (0.433) exert stronger driving forces on urbanization in Mongolia [17,43].

4.3. Limitation and Future Work

This study has certain limitations, although it has clear contributions to grassland city development evaluation. Firstly, since the SDGSAT-1 satellite was launched in November 2021, the available high-quality data are currently limited to recent years. Due to the unavailability of data, this study only analyzes data for 2023 to reveal the current development status of 11 cities. VIIRS nighttime light data are similar nighttime data sources that cover long-term time series. However, the spatial resolution of VIIRS data is 500 m, which differs greatly from the 10-m resolution of SDGSAT-1 nighttime light data. This large difference indeed requires rigorous cross-sensor calibration between the two datasets. Our study has limitations in this regard. The socioeconomic indicator system in this study was developed based on empirical methods. The coverage, scientific validity, and practical effectiveness of the indicators require further verification. We note that the existing CDI mainly reflects the current level of development. However, the CDI still has limitations in measuring sustainability.

Future work will track and incorporate annual SDGSAT-1 NTL data and multi-satellite historical data to extend the temporal coverage and enhance the time-series analysis. The next study on model accuracy enhancement can be directed toward addressing blooming and background noise issues. Furthermore, the indicators comprising the CDI could be expanded in number and scope in subsequent studies (indicators such as topography or administrative planning). This will enhance the precision of the index in comprehensively reflecting urban development conditions and better supporting sustainable development. Future studies could utilize extended time-series data to further explore the spatial relevance of the China–Mongolia–Russia Corridor in the context of urban development assessment. These enable a comparative discussion of inter-city development dynamics and driver variations and uncover the spatial heterogeneity in urban development and policy planning within the corridor area.

5. Conclusions

Sustainable urbanization is a major challenge in arid and semi-arid regions. This study evaluated the development status of 11 representative cities along the major transportation corridors of the Mongolian Plateau in 2023. Daily and NTL remote sensing data were used to reflect the vitality of urban built-up areas. Information entropy was used for weighting, adding a dynamic component to the analysis that reflected the degree of uncertainty or randomness in the distribution of the indicator levels. This approach enhanced the evaluation of urban socioeconomic development level and helped establish a City Development Index, which provides a unified measure for the comprehensive assessment of urban development that combines spatial, economic, and social factors. We found that cities in Mongolia generally exhibited more complex and irregular urban boundaries than those in Inner Mongolia, whereas compactness effectively captures the spatial concentration of built-up areas, ranging from a highly dispersed pattern in Ulaanbaatar to a tightly clustered layout in Erenhot. The CDI findings indicate that cities in the Inner Mongolian region, such as Hohhot, Baotou, Ordos, and Ulanqab, exhibit the characteristics of rapid development driven by resource-dependent industries. By contrast, cities in the Mongolian region show relatively poor development and transportation conditions. The methods and findings of this study contribute to the tracking and monitoring of sustainable urban development in grassland areas and have the potential for application in other economic corridors in the Belt and Road Initiative.