Introduction

The rapid growth of global economy has led to an increase in energy demand1. As one of foremost fossil fuels, coal has witnessed a historic milestone in 2023, surpassing a staggering 8.5 billion tons in global consumption for the first time (source: International Energy Agency, IEA). However, coal mining, especially surface coal mining, has significantly disturbed the natural vegetation and soils. This destructive practice not only leads to large areas of land degradation, but also has a severe impact on the regional carbon balance2,3. To achieve the targets for absolute carbon reduction set by the United Nations Framework Convention on Climate Change (UNFCCC) and the Paris Agreement, it is essential to consistently enhance the carbon sequestration capacity of ecosystems in mining areas and implement significant ecological protection and restoration projects4,5,6. Vegetation plays a crucial role as a carbon sink in the carbon cycle of terrestrial ecosystems7,8,9. Its carbon sequestration capacity is a core objective of ecological protection and restoration in mining areas10,11,12. Therefore, quantifying the impacts of surface coal mining and restoration activities on carbon sequestration in vegetation (VCS) can provide essential data to achieve ecological balance in the coal industry13,14.

The carbon sinks of vegetation in surface coal mining areas typically undergo three stages of “natural vegetation-mining-restoration”. Surface mining completely removes vegetation and soils, resulting in the conversion of the original vegetation into carbon source sites such as mine pits or industrial sitesS1 online). At this resolution, the overall vegetation cover can be effectively captured and analyzed for broad trends in vegetation change within the study area. Consequently, although the 30 m resolution may not be able to meet all the needs in some details, it still provides an overall understanding of the vegetation condition in the Shendong Coal Base and serves as a valuable reference and data support for assessing the loss of vegetation carbon sinks in the region. As remote sensing and methods for calculating VCS continue to evolve, future studies will increasingly rely on higher resolution VCS data. This allows for the capture of more subtle vegetation characteristics and changes, as well as the assessment of the dynamics of vegetation carbon sinks in mining areas.

Impact of coal mining on VCS

Currently, the development of China's coal resources has shifted towards the western region, which has a fragile ecological environment. This development is mainly concentrated in four provinces: Shanxi, Shaanxi, Mongolia, and ** and implementing policies related to coal mining and ecological restoration.

Materials and Methods

Study area and data

The Shendong coal base is located at the junction of Inner Mongolia, Shaanxi, and Shanxi in China. It belongs to the Yellow River Basin and has geographical coordinates ranging from 38°42′-40°06′N and 109°41′E—111°36′E (see Supplementary Fig. S1). The Shendong coal base is one of China’s 14 major coal bases, with a total area of 18,393.7 km2 and proven coal reserves of 223.6 billion tonnes. This region has an ecologically fragile environment and a semi-arid continental monsoon climate, with natural and restored vegetation is dominated by grasslands and sparse shrubs. The region receives an average annual precipitation of approximately 386.82 mm, while the average annual temperature is around 7.36 °C. Since 2007, the Shendong coal base has undergone large-scale mining, resulting in the loss of 400.08 km2 of vegetation as of 202149. This has significantly impacted the local ecological environment.

The spatio-temporal data on vegetation disturbance from 2001 to 2021 were sourced from a previous study conducted by the authors at the Shendong coal base using an automatic method (Auto-VDR) for identifying vegetation destruction and restoration of various open-pit mines45. Images from 2012 were excluded because of the poor data quality. The data had a spatial resolution of 30 m. To ensure more accurate data for analysis, we manually inspected the recognition errors based on remotely sensed imagery. For the missed and misidentified regions shown in Supplementary Fig. S2, we reviewed all Landsat and GF imageries in turn for the years 2002–2022. The initial data were corrected because surface mining can cause significant changes in feature types, and the destruction time and restoration time were easily identifiable in the imagery. Based on this, the destruction and restoration areas for 2022 have been added. The accuracies for vegetation destruction time and restoration time were 0.94 and 0.92 after pre-processing, respectively (see Supplementary Fig. S3-4 online).

The annual net primary productivity (NPP) of vegetation in the Shendong coal base from 2001 to 2022 was calculated using the global MODIS NPP product MOD17A3HGF v061 with 500 m spatial resolution, which was acquired from the National Aeronautics and Space Administration (NASA) (https://lpdaac.usgs.gov/). This NPP product was estimated using the BIOME-BGC (BioGeochemical Cycles) model50. The annual NPP is derived from the sum of all 8-day Net Photosynthesis (PSN) products (MOD17A2H) from the given year51.

Scale transformation of NPP

The spatio-temporal data on vegetation disturbance were obtained based on the maximum NDVI data during the growing season (July–September). Therefore, the area identified as destroyed does not necessarily indicate that it was utterly devoid of vegetation throughout the entire year. Instead, it was detected as devoid of vegetation during the growing season. Vegetation may have been present in the destroyed area, accumulating fixed carbon in the months before its destruction, resulting in a lower annual NPP value known as the background NPP (NPPbg). The spatial resolution of the NPP data used in this study is 500 m, while the spatio-temporal data on vegetation disturbance has a resolution of 30 m. In some cases, vegetated and destroyed areas may exist within a 500 × 500 m area, as shown in Fig. 4a,b. Therefore, the NPP of this pixel comprises both the NPP of vegetation and the NPPbg of the destroyed area. Before conducting statistical analyses, it is necessary to scale transformation of the existing NPP to obtain the NPP of the vegetated area at a resolution of 30 m. The processing is illustrated in Fig. 4.

Figure 4
figure 4

Scale transformation of NPP data. (a) The color red indicates areas where destruction has occurred and green indicates areas of vegetation. The blue circle highlights the region of 500 × 500 m pixels completely on the destroyed area. (b) An enlarged 500 × 500 m grid with vegetation and damaged area data at 30 m resolution. (c) NPP data at 500 m resolution. (d) An enlarged 500 m grid with NPP as NPPb. (e) Assign a value of 0 to the destroyed area and calculate the NPP of vegetated area using Eq. (1). (f) NPP after scale transformation.

Full size image

(i) To estimate NPPbg in the destroyed area, we randomly selected the 500 × 500 m pixels wholly destroyed within the study area (the blue area in Fig. 4c) and set their NPP as NPPbg. The background values of the destroyed area for the entire study area were obtained through Kriging spatial interpolation. These NPPbg are calculated annually.

(ii) To calculate the NPP in the vegetated area, the total area (S) and the vegetated area (Sveg) for the 500 × 500 m region were calculated separately. Subsequently, the total NPP in this region was calculated as NPPb × S. To obtain the NPP of the vegetated area at a resolution of 30 m, we subtracted the contribution of NPPbg from the fixed total NPP. The resulting NPP was then evenly distributed over the vegetated area based on its area, as shown in Fig. 4e. A value of 0 was assigned to the NPP of the destroyed area. Equation (1) shows the calculation for scale transformation. The NPP after the scale transformation is shown in Fig. 4f.

$$ NPP_{a} = \left\{ {\begin{array}{*{20}c} {\frac{{NPP_{b} \times S - NPP_{bg} \times (S - S_{veg} )}}{{S_{veg} }}{\text{, pixel is vegetation }}} \\ { \, 0{\text{ , pixel is not vegetation}}} \\ \end{array} } \right. $$
(1)

where, NPPa represents the NPP after scale transformation (g C m-2 a-1), while NPPb represents the NPP before scale transformation (Fig. 4d). S—Sveg refers to the area of destroyed area (m2).

Calculation of VCS

Research has demonstrated that 1 g of carbon in vegetation equals 2.2 g of organic matter. Based on the chemical equation for photosynthesis, vegetation absorbs 1.63 g of CO2 for every gram of accumulated organic matter52. This conversion relationship can transform NPP into VCS, as shown in Eq. (2).

$$ VCS = NPP \times 2.2 \times 1.63 $$
(2)

where VCS represents the amount of CO2 fixed by vegetation per unit area and time, which is represented by carbon sequestration in vegetation (VCS) in this paper (unit: g CO2 m-2 a-1). The coefficient of conversion from NPP to organic matter is 2.2, and the coefficient of conversion from organic matter to CO2 is 1.63.

Calculation of VCS of undisturbed state in the mining areas

A linear regression model was fitted using the pre-mining (2001-TD) VCS data of the study area, as shown in Eq. (3). The VCS of undisturbed state after the destruction time was then predicted based on the regression equation, as shown in Eq. (4), and the results formed a “Prediction line”. The predicted VCS of undisturbed state represent the original state of the VCS when the study area is assumed to be unaffected by mining activities.

$$ a = \frac{{n\sum {tVCS_{t} - \sum t } \sum {VCS_{t} } }}{{n\sum {t^{2} } - (\sum t )^{2} }}, \, b = \frac{{\sum {VCS_{t} } }}{n} - a\frac{\sum t }{n}, \, t = 2001, \cdots ,T_{D} $$
(3)
$$ VCS_{Year} = a \cdot Year + b, \, Year = T_{D} + 1, \cdots ,2022 $$
(4)

where n represents the total years involved in the regression, TD denotes the vegetation destruction time, t is the year, and VCSt denotes the VCS for the respective year.

Quantification of direct and potential changes in VCS

This paper analyzes the changes in VCS in the Shendong coal base from two perspectives (see Supplementary Fig. S4 online). The coal development activities have caused destruction and restoration of vegetation, resulting in changes in VCS. The term “direct changes in VCS” refers to changes relative to the previous year when the destruction or restoration occurred (refer to Fig. S4a), including direct decrease and direct increase. “Potential changes in the VCS”, on the other hand, are relative to the undisturbed state (Fig. S4b). To distinguish potential changes in VCS caused by mining and restoration activities, we labeled the potential changes in destroyed area as C_destroy and in restored area as C_restore. The restoration rate of vegetation (RV) is defined as the ratio of the restoration area over the destruction area, while the restoration rate of VCS (RVCS) is defined as the ratio of “the direct increase in VCS” over “the direct decrease in VCS”. All abbreviations used in this paper are summarized in Supplementary Table S2 online.

The deficit of carbon sequestration in vegetation (VCSD) from surface coal development activities was calculated using Eq. (5). VCSD is defined as the total reduction in VCS compared to the undisturbed state at the mine sites after vegetation destruction has occurred.

$$ VCSD = - (\sum\nolimits_{{{\text{t}} = 2001}}^{2022} {C\_destroy_{t} } + \sum\nolimits_{{{\text{t}} = 2001}}^{2022} {C\_restore_{t} } ) $$
(5)

where VCSD is the deficit of carbon sequestration in vegetation, C_destroyt is the potential changes in VCS in the destroyed area in year t, and C_restoret is the potential changes in VCS in the restored area in year t. If VCSD is greater than 0, it indicates that coal development activities have had a negative impact on VCS. Conversely, if VCSD is less than 0, it indicates that coal development activities have increased VCS in the mining area.