Belt and Road Initiative Support Project: Remote Sensing Time Series Analysis of the Environmental Disturbance of the China-Belarus Industrial Park Construction

Note: This research project was completed with the support of the Joint Research and Cooperation Program of the Belt and Road International Science Organizations Alliance.

The China-Belarus Great Stone Industrial Park (hereinafter referred to as the China-Belarus Industrial Park) is located near Minsk, the capital of Belarus, and is a joint venture between China and Belarus. The Chinese shareholders are China National Machinery Industry Corporation (SINOMACH) and Harbin Investment Group Co., Ltd., holding 60% of the shares; the Belarusian shareholders are the Minsk Regional Government, the Minsk Municipal Government, and Belarusian Horizon Holding Group, holding 40% of the shares. The industrial park is conveniently located near Minsk International Airport, with a total land area of ​​91.5 square kilometers, and the first phase of the project covers an area of ​​8.5 square kilometers.

The Great Stone China-Belarus Industrial Park is a Belt and Road cooperation project jointly built by China and Belarus, and is also China's largest overseas industrial park. This project, which has been jointly cared for and promoted by the leaders of the two countries, has developed into a modern park with complete infrastructure, a superior business environment, and advanced management concepts after nearly ten years of development. It is also Belarus's largest investment attraction project and is known as a "pearl on the Belt and Road".

The park has multiple standardized factory buildings available for companies to lease or purchase. The design and construction of the factory buildings fully embody the concept of green development, utilizing natural light, ventilation, and energy-saving technologies to improve resource utilization while reducing environmental impact.

Introduction to Data Processing Methods

1. Long-term imagery from 2016 to 2024 was collected within the construction area of ​​the China-Belarus Industrial Park, and the NDVI (Normalized Difference Vegetation Index) and NDBI (Normalized Difference Built-up Index) were calculated. These two indices reflect the distribution of surrounding vegetation and building areas.

 NDVI= (NIR-Red)/(NIR+Red) 

NDVI is an index used to measure the health of vegetation. By comparing the reflectance in the near-infrared (NIR) and red bands, NDVI can effectively identify and distinguish green vegetation. NDVI values ​​typically range from -1 to 1.

A high value (close to 1) indicates healthy, lush vegetation.

Low values ​​(close to 0 or negative values) indicate a lack of vegetation or areas without vegetation cover such as bare soil or water bodies.

 NDBI=(SWIR-NIR)/(SWIR+NIR) 

NDBI is an index used to measure built-up areas. By comparing the reflectance of shortwave infrared (SWIR) and near-infrared (NIR) bands, NDBI can effectively identify and distinguish built-up areas. NDBI values ​​typically range from -1 to 1.

High values ​​(close to 1) indicate building areas.

Low values ​​(close to 0 or negative values) indicate non-building areas, which may be vegetation, water bodies, etc.

△Location map of the China-Belarus Great Stone Industrial Park

△ Changes in remote sensing images of the industrial park area from 2000 to 2024

Introduction to Data Processing Methods

2. Before acquiring images, clouds were filtered based on 25% cloud cover, and a cloud masking algorithm was used to remove clouds. Since the target area has a temperate continental climate with distinct seasons, cloud cover is high from November to March of the following year, resulting in numerous data gaps. Therefore, for areas with insufficient effective pixels, a method of combining monthly and quarterly averages was used to integrate the data, supplementing the pixels obscured by clouds in each image, and resampling to a 30-meter resolution.
3. After the data is processed, the raster images of each scene are integrated into a netCDF multidimensional raster or converted into vector points, and time attributes are added.
4. The data with time attributes mentioned above are aggregated into a spatiotemporal cube, which includes NDVI and NDBI numerical information.
5. Emerging spatiotemporal hotspots, local outliers, and change points are analyzed on the spatiotemporal cube.
6. Output and visualize the above results, interpret them according to the actual situation of the project, and verify them by comparing them with the original RGB image.

Results Display

1. Analysis of Emerging Spatiotemporal Hotspots

△NDVI Vegetation Index Emerging Spatiotemporal Hotspot Analysis Results

According to the analysis results, the changes in vegetation in the project area from 2016 to 2024 are as follows: The building areas in the project area are classified as persistent cold spots, continuous cold spots, and intensified cold spots.

The persistent cold spots are identified as buildings that existed before the project started and were still there as of May 2024. Comparing satellite imagery reveals that these include some major roads, substations, logistics parks, warehouses, and customs offices that existed before the project began.

The continuous cold spot refers to the land features that supplement the persistent cold spot. These are plots of land planned before the project started and were completed in May 2024.

The enhanced cold spots can be interpreted as construction sites that have been under construction intermittently until May 2024.

Scattered cold spots represent areas that, as of May 2024, still have no vacant land or low-density buildings.

The cold spots in the fluctuations represent areas that are covered by vegetation for most of the project's duration, until the monitoring period ends and the land is converted to open space or buildings.

△ NDBI Building Index Emerging Spatiotemporal Hotspot Analysis Results

According to the analysis results, the changes in buildings within the project area from 2016 to 2024 are as follows:

The building areas within the project area are classified as scattered hotspots, fluctuating hotspots, persistent hotspots, and gradually decreasing hotspots.

Continuous hotspots indicate areas that remained vacant or under construction throughout the project's duration. Fluctuating hotspots represent areas that were previously mostly vegetated but eventually became predominantly under construction. Scattered hotspots indicate areas under ongoing construction. Gradually decreasing hotspots can be interpreted as areas with gradually increasing vegetation density, possibly from landscaping, artificial cultivation, or restoration. The remaining scattered cold spots are vegetated areas.

2. Results of Local Outlier Analysis

△ Analysis results of local outliers in NDVI vegetation index

Local outlier analysis can identify statistically significant clusters and outliers in spatial and temporal environments. In this case, multiple types indicate that the temporal and spatial variation patterns of the NDVI vegetation index are complex and highly variable, without obvious significant clustering. The low-low clusters in the figure represent areas with persistently low vegetation indices, which corroborates the results of previous analysis of emerging spatiotemporal hotspots.

△ NDBI Building Index Local Outlier Analysis Results

The clustering results of the building index simultaneously identified densely built areas and densely vegetated areas. It was also noted that some outliers appeared near the main road in the northeast direction, indicating that there was some land development activity in that area.

3. Results of change point detection and analysis.

△ NDBI Building Index Change Point Detection Results

△ NDVI Building Index Change Detection Results

Based on the NDVI and NDBI change point monitoring results, it was found that the periods from March to October 2017 and from February to December 2022 were the concentrated development periods for the project, during which the changes in vegetation and building indices were the most significant. There were intermittent development processes during these periods. The change monitoring also clearly indicated the construction progress of roads within the western part of the industrial park: the southern section was completed from the end of October 2019 to the end of January 2020; the middle section was completed from February to mid-May 2020; the northern section was completed from the end of August to mid-December 2020; and the connection project between the middle and southern sections was completed from the end of March to mid-July 2021.

Overall, throughout the continuous construction of the industrial park from 2016 to the present, the changes in vegetation cover and the changes in buildings have been largely consistent in terms of time and space, with clear stages and regional divisions. This indicates that the land development and utilization have been carried out in an orderly manner according to the plan, the disturbance to the surrounding vegetation has been controlled at a low level, some vegetation cover has been preserved during land development, and artificial landscaping and vegetation restoration have been carried out during the construction process.

This also verifies that time-series remote sensing imagery can provide richer information on land change, enabling more complete and comprehensive monitoring of changes over large areas of the environment. It provides a more powerful analytical tool for environmental protection, engineering progress monitoring, and the construction of high-standard farmland.