Introduction

Net Primary Productivity (NPP) refers to the portion of organic carbon fixed by vegetation after deducting its own respiratory consumption, which not only directly reflects the production capacity of vegetation communities under natural conditions but also serves as a key indicator for assessing ecosystem health status1,2. With the intensification of impact on ecosystems causing from global climate change and human activities, NPP research has expanded from the early field of plant physiological ecology to multiple fields such as global carbon cycling, ecosystem service assessment, and regional sustainable development.

Early research on NPP mainly focused on the field of plant physiological ecology, used to evaluate the capacity of natural and agricultural production, and was subsequently used to analyze the impact of human activities on ecosystems3. NPP is an important indicator for measuring vegetation communities4; the higher the NPP, the richer the biodiversity, then the healthier the vegetation communities, the more abundant the biomass, and the stronger the ability to respond to environmental changes5. From the perspective of global climate change, NPP serves as a core indicator to reflect the capacity of carbon sequestration6,7. Currently, many scholars use models to study regional vegetation NPP. Compared with the traditional observation methods, model estimation not only greatly improves the efficiency of vegetation NPP estimating, but also meets the demand for large-scale regional NPP observation8. Models used to estimate vegetation NPP can be divided into climate-related models, eco-physiological process models, and light use efficiency models9. Among them, the CASA(Carnegie- Ames-Stanford Approach) model is the most representative. In recent years, scholars such as Geng YX et al.10, Sun H et al.11, Wang S et al.12, and others have conducted extensive research on the dynamic changes of regional NPP using the CASA model or improved CASA models, proving that the CASA model has wide applicability in estimating regional NPP.

Significant progress has been made in NPP estimation methods, multi-scale driving factor analysis, and human-land coupling effects. For instance, Yan F et al.13found that NPP was positively correlated with precipitation in their study of the spatial distribution pattern and driving mechanisms of NPP in the Taihang-Yanshan mountain area; Yu Z et al.14found that land use, temperature, and precipitation were factors affecting NPP in their study of the main factors affecting grassland growth in Inner Mongolia, among them the precipitation having the greater impact on NPP; Naeem S et al. (2020)15 found that climatic factors had the greater impact on vegetation coverage than human activities, and precipitation had the greater impact on NPP than temperature; Sinan W et al.16found that climate change and human activities were factors affecting NPP.

The Loess Plateau is an ecologically vulnerable area in China and a core area for soil and water conservation17. The restoration of vegetation and enhancement of carbon sink capacity in the Loess Plateau are key to the national strategy of “Ecological Protection and High-Quality Development of the Yellow River Basin.” The Lanzhou section of the mainstream, located in the upper reaches of the Yellow River, is a very important ecological protection barrier and water supply area in the northwest China. The temporal and spatial differentiation of vegetation NPP in this area directly affects the ecological security and carbon coordination status of the entire region18. Since the implementation of the vegetation greening project in the northern and southern mountain along the Lanzhou section of the Yellow River mainstream in 2000, the natural environment has been basically improved, and vegetation types have gradually diversified. The northern and southern mountain in the semi-arid area of the Lanzhou section of the Yellow River mainstream are influenced by both terrain uplift and monsoon circulation, resulting in significant vertical differentiation of climatic elements19. The geomorphological type of the northern mountain is mainly characterized by hills covered with thick loess, while the southern mountain has widespread exposure of bedrock (metamorphic sandstone), with weak water retention capacity but strong anti-erosion properties. Additionally, due to differentiation in human activities, the northern mountain is adjacent to the urban area of Lanzhou and are affected by urban expansion, while the southern mountain has stronger vegetation management and protection investment (such as irrigation and replanting) due to ecological tourism development. This further leads to significant differentiation in the spatial pattern of NPP between the northern and southern mountain. However, the differences between the northern and southern mountains have not been systematically explored, and questions such as how these huge differences affect the spatial distribution of NPP and what causes them have not been fully explained, and their impact on future vegetation construction remains unclear. Therefore, conducting NPP research is of great significance for understanding vegetation coverage changes in the northern and southern mountain along the Lanzhou section of the Yellow River mainstream and balancing ecosystem stability.

We select the mountain around Jiuzhoutai in the northern mountain and the mountain peaked by Dajianshan in the southern mountain of Lanzhou as research objects. Based on the CASA model, monthly NPP data from 2000 to 2020 are calculated. The Sen-MK trend analysis method is used to explore the spatiotemporal differentiation characteristics of NPP. The geographical detector method is used to analyze the spatial correlation between land use types, topography, climatic factors (precipitation, sunshine duration, temperature), NDVI and NPP. The reasons for the different distribution patterns of NPP in the northern and southern mountain will be discussed. We hope to provide some useful suggestions for vegetation construction and ecological protection in the Lanzhou section of the Yellow River mainstream based our research.

Study area

The Yellow River is an important ecological corridor connecting the eastern, western, and central provincial regions of China, with a total length of approximately 5,464 km. It has significant importance for China’s economic, social, and ecological security. The Lanzhou section of the Yellow River mainstream tarts from the confluence of the Yellow River and Datong River at Village Chalu, Dachuan Town, Xigu District of Lanzhou, flows through six counties and districts including Xigu, Anning, Qilihe, Chengguan, Yuzhong, and Gaolan, ends at Dabaogou, Qingcheng Town, Yuzhong County, with a flow length of 152 km.

The study area includes part aera of Anning District, Chengguan District, Qilihe District, Xigu District, and Gaolan County of Lanzhou City, Yongjing County of Linxia Hui Autonomous Prefecture, and Lintao County of Dingxi City beginning from 1500 m contour. According to the distribution of contour lines with 50 m elevation intervals, the study areas is divided into northern and southern mountain (as shown in Fig. 1). By marking different contour lines, it can be clearly seen that the northern mountain range runs north-south, while the southern mountain range runs east-west.

Fig. 1
figure 1

Overview of the northern and southern mountain along the Lanzhou section of the Yellow River mainstream.

Full size image

Data processing and research methods

Data sources and processing

All data in this study are unified to the coordinate system WGS_1984_UTM_Zone_48N, and the spatial resolution is resampled of 30 m×30 m in.tiif format. All figures are drawn using Origin 2022 software.

Normalized difference vegetation index (NDVI)

The vegetation NDVI data in this paper comes from the Resource and Environment Data Cloud Platform of the Chinese Academy of Sciences (https://www.resdc.cn).

Data processing flow: using bilinear interpolation method for resampling; using the Google Earth Engine (GEE) platform to perform time series fusion of MODIS and Landsat data, eliminating cloud coverage and spatiotemporal discontinuity issues; the 3σ criterion is used to remove anomalies in meteorological and remote sensing data, and the Space-Time CoKriging (STCOK) method20 is used to fill in missing data.

Meteorological data

Surface meteorological station data is obtained from the National Scientific Meteorological Data Center website: monthly precipitation (Pre), temperature (Tem), and sunshine duration (SD) data from 2000 to 2020 form the Lanzhou station (ID: 52889) and the Yuzhong station (ID: 52983) of the China Meteorological Administration.

To make the acquired climate data closer to the true values, the Inverse Distance Weighting (IDW) method21 is used for data correction of climatic factors. The basic principle of IDW is to estimate unknown points based on data from known observation points through distance weighting. Specifically, the closer the observation point to the target point, the greater weight and greater influence the observation point has on the estimated value of the target point. This method has obvious advantages in processing meteorological data with strong spatial heterogeneity, improving the spatial accuracy of meteorological data. The method first searches for the distance between known stations and interpolation stations, calculates the square of the reciprocal of the distance, and uses this weight as the main parameter for spatial interpolation calculation22.

Remote sensing data

  1. (1)

    Digital Elevation Model (DEM) data comes from the Geospatial Data Cloud Platform (http://www.gscloud.cn), obtaining factors such as slope and elevation in the study area. The data type is raster data with a resolution of Tif (30 m×30 m).

  2. (2)

    Land cover data produced by Yang and Huang23. The land use types obtained for this study area include 7 categories: cropland, forestland, shrubland, grassland, water bodies, unused land, construction land.

  3. (3)

    Vector data for the study area is sourced from the Resource and Environmental Science Data Center of the Chinese Academy of Sciences (http://www.resdc.cn).

Research methods

CASA model

This study uses the CASA model to calculate NPP referring to the research method of Potter SC et al.24. This model is a process model based on light utilization efficiency, which estimates NPP by inputting meteorological data (monthly rainfall, monthly average temperature, and monthly solar radiation) and monthly average NDVI.

Sen-MK trend analysis

The Sen-MK trend analysis method25,26,27 is a non-parametric statistical method suitable for trend analysis of long time series data, which can effectively identify the trend and significance of NPP temporal changes. This method is insensitive to outliers and widely used in time series analysis in environmental and ecological fields unlimited by the type of data distribution. This study uses the Theil-Sen Median trend analysis method is used to reflect the trend and magnitude of NPP changes from 2000 to 2020 through the Sen slope, to evaluate the significance of the trend through the Mann-Kendall test, comprehensively to reveal the temporal evolution characteristics of NPP in the study area.

Geographical detector

The geographical detector is a statistical method used to detect spatial heterogeneity and reveal driving factors28. Its core idea is that if an independent variable has an important influence on a dependent variable, then the spatial distributions of the independent and dependent variables should have similarity. There are four types geographical detector methods: differentiation and factor detection, interaction detection, risk detection, and ecological detection29. This paper mainly uses the differentiation and factor detection and interaction detection methods to analyze the influence mechanisms of factors such as land use type, topography, climatic factors, and NDVI on NPP in the study area30.

Differentiation and factor detection mainly explores the extent to which a certain factor explains the spatial differentiation of vegetation NPP, represented by the q value. The range of q value is 0 ~ 1; the larger the q value, the stronger the explanatory power of the factor for vegetation NPP in this area; the smaller the q value, the weaker the explanatory power of the factor for vegetation NPP; if a factor’s q value is 0, it indicates that the factor has no relationship with vegetation NPP31. The calculation formula is as follows:

$$\:q=1-\frac{{\sum\:}_{k=1}^{L}{N}_{h}{\sigma\:}_{k}^{2}}{N{\sigma\:}^{2}}$$
(9)

where h = 1, …, L is the stratification of variable Y or factor X; Nₕ and N are the number of units in layer h and the entire area, respectively; σₕ² and σ² are the variances of Y values in layer h and the entire area, respectively.

Interaction detection can identify interactions between different risk factors Xs. The principle of this method is to evaluate the strength of explanation of factors X1 and X2 on vegetation NPP under their joint action, or to analyze whether these factors’ effects on Y are independent of each other32. The evaluation method includes first calculating the q values of two factors X1 and X2 on Y separately, then overlaying the two layers X1 and X2 to obtain a new layer X1∩X2, calculating its q value on Y, and thus judging the interaction between the two factors33.

Terrain position index

The terrain position index is calculated using elevation and slope values, and divided into 4 intervals. The calculation formula is as follows34:

$$\:\text{T}=\text{l}\text{o}\text{g}\left[\right(\frac{E}{\stackrel{-}{E}}+1)\times\:(\frac{S}{\stackrel{-}{S}}+1\left)\right]$$
(10)

where T represents the terrain position index, E represents the elevation value of any grid, Ē represents the average elevation within the calculation unit, S represents the slope value of any grid, and \({\stackrel{-}{S}}\) represents the average slope within the calculation unit. The terrain position index is divided into four gradients: low, medium-low, medium-high, and high, with a range of 0.2535.

Results analysis

Temporal and Spatial distribution of NPP

This paper selects 2000–2020 as the research timeline to study the spatiotemporal changes and formation mechanisms of NPP in the northern and southern mountain. The year is divided into four seasons according to the international seasonal classification standard: winter (December-February), spring (March-May), summer (June-August), and autumn (September-November). According to the actual situation of the study area, NPP grades are divided into 5 grades using the natural breaks method: the first grade is 0-135 gC·m⁻²·a⁻¹, the second grade is 135–188 gC·m⁻²·a⁻¹, the third grade is 188–261 gC·m⁻²·a⁻¹, the fourth grade is 261–363 gC·m⁻²·a⁻¹, and the fifth grade is ≥ 363 gC·m⁻²·a⁻¹.

The annual variation results of NPP are shown in Fig. 2. The average NPP value is 118.69 gC·m⁻²·a⁻¹in the northern mountain, and 202.01 gC·m⁻²·a⁻¹ in the southern mountain. From a total amount perspective, the overall NPP level of the southern mountain is significantly higher than that of the northern mountain. In terms of temporal change trend, the annual average NPP of both the northern and southern mountain shows an overall fluctuating upward trend. However, the slope of the trend line for the southern mountain (4.86) is higher than that of the northern mountain (4.27), and the goodness of fit for the changes in the southern mountain is generally higher than that of the northern mountain, indicating that the growth rate of NPP in the southern mountain is faster and has the stronger temporal correlation and stability.

Fig. 2
figure 2

Annual variation trends of NPP in the northern and southern mountain from 2000–2020.

Full size image

As shown in Fig. 3, the NPP range is between 0 and 363 gC·m⁻²·a⁻¹, with the highest proportion falling in the first grade, which accounting for 86.9% of the total NPP content in the northern mountain, indicating that the northern mountain has stronger biological carbon sequestration capacity in low NPP grade interval. While, the NPP range is between 0-≥363 gC·m⁻²·a⁻¹, with the highest proportion falling in third grade which accounting for 32.52% of its total NPP content in the southern mountain, indicating that the southern mountain has stronger biological carbon sequestration capacity in medium NPP grade interval. According these results, the overall health and richness of vegetation communities in the southern mountain are better than those in the northern mountain.

Fig. 3
figure 3

 Average NPP proportion in the northern and southern mountain from 2000–2020.

Full size image

From a seasonal difference perspective shown in Fig. 4, the NPP in all four seasons in both the northern and southern mountain shows an upward trend, but there are significant differences: in spring, the growth rate and fitting degree of the southern mountain are slightly higher than those of the northern mountain; in summer, the southern mountain has a larger growth slope, showing more vigorous summer productivity; in autumn, the goodness of fit of the southern mountain is relatively high among the four seasons (R²=0.339), indicating that autumn NPP changes has a stronger temporal correlation; the fitting degree of winter is significantly higher than that of the other seasons, indicating that the winter growth trend is the most stable. It could be concluded that the goodness of fit of NPP in four seasons of the southern mountain is generally higher than that of the northern mountain, and the seasonal patterns of vegetation growth are more distinct.

Fig. 4
figure 4

Seasonal variation trends of NPP in the northern and southern mountain from 2000–2020 (A is spring, B is summer, C is autumn, D is winter).

Full size image

Based on the pixel-level MK trend analysis results, the spatial differentiation characteristics of NPP in the northern and southern mountain from 2000 to 2020 are shown in Figs. 5 and 6. The proportion of areas with extremely significant increases in both the northern and southern mountain is higher than 60%, and the proportion of areas with decreases is very small, indicating that the overall vegetation growth conditions in both the northern and southern mountain has been improved, and the spatial changes in NPP increases in the southern mountain (95.71%) are more significant than that in the northern mountain(94.28%).

Fig. 5
figure 5

 Significance test of NPP change trends in the northern and southern mountain from 2000–2020.

Full size image
Fig. 6
figure 6

 Proportion of NPP change areas in the northern and southern mountain from 2000–2020.

Full size image

Driving mechanism analysis

As shown in Fig. 7, with the changing of time, the NPP of the low gradient shows a flat or declining trend both in the northern and southern mountain while the NPP of the other gradients shows an upward trend in both the northern and southern mountain. NPP is improving in both mountain areas, with the southern mountain having more high-grade terrain position indices and high NPP values than the northern mountain. It could be concluded that as the terrain position index gradient increases, the NPP value also increases.

Fig. 7
figure 7

Distribution of NPP among terrain position index gradients in the northern and southern mountain from 2000–2020 (A is the low gradient, B is the medium-low gradient, C is the medium-high gradient, D is the high gradient) (Note: In the horizontal axis, N represents the northern mountain, S represents the southern mountain).

Full size image

As shown in Fig. 8, precipitation has a positive correlation with NPP both in the northern and southern mountain, while sunshine duration and temperature have negative correlations with NPP. Among these, precipitation has the highest fitting degree (R2 = 0.6148 in the southern mountain, R2 = 0.6943 in the northern mountain), followed by sunshine duration (R2 = 0.3586 in the southern mountain, R2 = 0.2242 in the northern mountain), and temperature has the lowest correlation with NPP (R2 = 0.0736 in the southern mountain, R2 = 0.0117 in the northern mountain). It could be concluded that precipitation is the main factor impacted NPP among climatic factors.

Fig. 8
figure 8

The correlation analysis of climatic factors (precipitation, sunshine duration, temperature) and NPP in the northern and southern mountain from 2000–2020 (A is the northern mountain, B is the southern mountain).

Full size image

As shown in Fig. 9, there is a positive correlation between NDVI and NPP, and the correlation is extremely high both in the northern and southern mountain. The fitting degree of the southern mountain (R2 = 0.8989) is better than that of the northern mountain (R2 = 0.8903), indicating that the interannual variation pattern of NDVI is more significant, and the regularity and continuity of vegetation improvement are stronger in the southern mountain.

Fig. 9
figure 9

Trend fitting of annual average NDVI and NPP in the northern and southern mountain from 2000–2020.

Full size image

The land use type is dominated with grassland and has no shrubland existing in the northern mountain, while forestland and shrubland are the main land use types in the southern mountain. What need to clearly described are that all of the forestland is planted by human completely in the northern mountain while larger part of forestland and shrublands are natural in the southern mountain. As shown in Fig. 10, forestland and shrubland are the dominant land use types in high-grade NPP, while other land types produce lower NPP. According to the overall situation of changes of NPP and land use types from 2000 to 2020 timeline, both the northern and southern mountain show an increase in high-grade NPP. Construction land and unused land are continuously increasing in the northern mountain, while forest and shrub are continuously increasing in the southern mountain. The change in land use types is one of the reasons why NPP in the southern mountain is better than that in the northern mountain.

Fig. 10
figure 10

Changing of land use types and different NPP grades in the northern and southern mountain from 2000–2020 (A is cropland, B is forestland, C is shrubland, D is grassland, E is water bodies, F is unused land, G is construction land) (Note: In the horizontal axis, N represents the northern mountain, S represents the southern mountain).

Full size image

Geographical detector results analysis

Five years (2000, 2005, 2010, 2015, 2020) were selected to analyze the relationship between six factors—land use type (LULC), precipitation (PRE), elevation (DEM), normalized difference vegetation index (NDVI), sunshine duration (SD), and temperature (TEM)—and NPP using the geographical detector. The results show that all factors have a significant impact on NPP (p < 0.01). Overall, in both the northern and southern mountain, NDVI has a decisive effect on NPP, and the degree of influence of various factors on NPP varies with time. The dominant influencing factors for NPP in the northern mountain, ranked by average q value from large to small, are NDVI (0.466094), LULC (0.108239), DEM (0.102159), SD (0.08064), TEM (0.09176), and PRE (0.042584). The dominant influencing factors for NPP in the southern mountain, ranked by average q value from large to small, are NDVI (0.654938), SD (0.257304), DEM (0.25114), LULC (0.20971), PRE (0.172703), and TEM (0.152891).

The factor interaction detection results are shown in Figs. 11. The explanatory power of interactions between various influencing factors is significantly higher than that of single factors. The interactions of NDVI and temperature with other factors all have a relatively large impact on NPP, with the explanatory power of NDVI with other factors being the highest, and the interactive effect of NDVI and temperature having the greatest overall impact on NPP.

Fig. 11
figure 11

Interaction detector analysis results of the relationship between six factors and NPP. (A is the northern mountain, B is the southern mountain; 1 is 2000, 2 is 2005, 3 is 2010, 4 is 2015, 5 is 2020; (LULC- land use type), PRE- precipitation, DEM -elevation, NDVI- normalized difference vegetation index, SD- sunshine duration, and TEM- temperature).

Full size image

Discussion

Comprehensive analysis of natural factors in NPP differentiation formation

The formation of NPP differences between the northern and southern mountain is influenced by a combination of various natural factors: topographic factors are one of the important factors affecting the spatial distribution of NPP. Our research result shows that the proportion of medium-high terrain position index in the southern mountain is greater than that in the northern mountain, and there is a positive correlation between terrain position index and NPP—the higher the terrain position index, the higher the NPP. This finding is consistent with the research results of Laamrani et al. (2014)35. The mountain range in the southern mountain trends east-west horizontally, while the mountain range in the northern mountain trends north-south, which further affects the spatial distribution of NPP. As the grade of the terrain position index increases, the NPP value also increases, this pattern that is verified in both the northern and southern mountain.

Climatic factors, especially precipitation, are another important factor affecting NPP. Our study found that the multi-year average precipitation in the southern mountain is greater than that in the northern mountain, which is consistent with the research results of Yan F et al. (2025)13 that NPP was positively correlated with precipitation. The northern and southern mountain are influenced by both terrain uplift and monsoon circulation, resulting in significant vertical differentiation of climatic elements, and this climatic difference directly affects vegetation growth conditions and NPP levels.

The difference in geomorphological types is also an important factor leading to different NPP distributions. The northern mountain is mainly characterized by hills covered with thick loess, while the southern mountain has widespread exposure of bedrock (metamorphic sandstone), with weak water retention capacity but strong anti-erosion properties. This geomorphological difference affects soil moisture conditions and nutrient status, which in turn affects vegetation growth and NPP levels.The research results of Michaud E et al. (2024)36 also confirmed the important influence of geomorphological types on ecological functions.

Vegetation type is a direct factor affecting NPP. Our research results show that the proportion of forest and shrub in the southern mountain is significantly bigger than that in the northern mountain, and fores and shrub are the dominant land use types for high NPP. The proportion of grass in the northern mountains is greater than that in the southern mountain, but the NPP level of grass is usually lower than that of forest and shrub. This difference in vegetation types directly leads to differences in NPP levels between the northern and southern mountain. The research results of other scholars also confirmed that changes in vegetation type will lead to changes in NPP37.

NDVI, as an important indicator of vegetation growth conditions, has a close relationship with NPP. Our study found that the goodness of fit of NDVI in the southern mountain is higher than that in the northern mountain, indicating that the continuity and regularity of vegetation improvement in the southern mountain are stronger. The geographical detector results further confirm that NDVI single factor and its interaction with other factors have a significant impact on NPP. In the study of Liu Y et al.38, interaction analysis showed that the interaction of NDVI with any other factor (such as precipitation, temperature, elevation, etc.) significantly enhances the impact on NPP, which is consistent with the results of this paper.

Differential analysis of human activity impacts

In addition to natural factors, differences in human activities are also an important reason for the different spatial distribution of NPP in the northern and southern mountain: urbanization is a key impact factor. The northern mountain is adjacent to the urban area of Lanzhou and is more affected by urban expansion. Urban construction occupies a large amount of land, reducing vegetation coverage area and lowering regional NPP levels. Sinan W et al.16 found that human activities affected NPP, which is consistent with the conclusions of this paper.

Differences in land use change trends also affect the spatial distribution of NPP. During the period from 2000 to 2020, land use types in the northern and southern mountain have undergone changes to varying degrees: in the northern mountain, the proportion of cropland has decreased, the proportions of forestland and grassland have remained basically stable, the proportion of water bodies has increased slightly, and the proportions of unused land and construction land have increased; in the southern mountain, the proportion of cropland has decreased, the proportions of grassland and water bodies have remained basically stable, and the proportions of forestland, shrubland, unused land, and construction land have increased. This difference in land use change trends has led to different changes in NPP. In particular, the increase in the proportion of forestland and shrubland in the southern mountain has played a positive role in improving regional NPP levels. Li J et al.39 proposed that regional characteristic factors and uneven spatial distribution factors were important driving factors for regional differences. The research of Ji Y et al.40 directly confirmed that the increase in forestland area was closely related to the increase in total NPP.

The difference in ecological engineering implementation effects is also an important factor. The vegetation greening project in the northern and southern mountain along the Lanzhou section of the Yellow River mainstream, which has been implemented since 2000, has achieved varying degrees of effects in the northern and southern mountain. Our results show that the vegetation restoration effect in the southern mountain is better than that in the northern mountain, which is closely related to the intensity of vegetation management and protection investment, land use changes, and natural conditions.

Systematic revelation of Multi-factor interactions

The interaction detection results of the geographical detector show that there are complex interactions between various driving factors, and these interactions often have a greater impact on NPP than single factors: the interaction of NDVI with other factors has a significant impact on NPP, especially the interaction of NDVI with land use type and NDVI with climatic factors. This indicates that the impact of vegetation growth conditions (NDVI) on NPP varies significantly under different land use types and climatic conditions. For example, the same NDVI value may produce higher NPP in forest land types and lower NPP in grassland or unused land; the same NDVI value may produce higher NPP in areas with abundant precipitation and lower NPP in arid areas. The research of Nemani et al.41 confirmed that the interaction of NDVI with climatic factors (especially temperature and precipitation) was the main mechanism driving the global increase in NPP, which is consistent with the research results of this paper.

The interaction of land use type with climatic factors also has an important impact on NPP. Different land use types respond differently to climatic conditions. For example, forests may have lower sensitivity to precipitation than grasslands, but higher sensitivity to temperature than grasslands. This differentiated response leads to significant differences in NPP levels in different land use type areas under the same climatic conditions. Yu Z et al.14 concluded that land use, temperature, and precipitation all affecting NPP, which is also consistent with the research conclusions of this paper.

The interaction of topographic factors with climatic factors is equally important. Topography indirectly affects vegetation growth and NPP levels by influencing local climatic conditions, such as precipitation distribution, temperature gradient, and light intensity. For example, high-altitude areas usually have lower temperatures than low-altitude areas, which may limit vegetation growth and NPP levels; but high-altitude areas may have more precipitation than low-altitude areas, which may promote vegetation growth and NPP levels. This complex interaction leads to spatial heterogeneity of NPP.

The coupling effect of natural factors and human factors is also an important mechanism affecting NPP. Human activities affect NPP levels by changing land use types, vegetation coverage, and management methods, together with natural factors. For example, vegetation management and protection investment may produce more significant effects in areas with poorer climatic conditions, and relatively smaller effects in areas with good climatic conditions.

Future research directions

Although the temporal-spatial changing characteristics of vegetation NPP in the Lanzhou Section of the Yellow River Mainstream were analyzed in detail in this paper, but the mechanism of these changing is need to explore deeply and widely. In order to make up the insufficient of the lay of the land and terrain position index in explaining the distribution pattern of vegetation NPP in the Lanzhou Section of the Yellow River Mainstream, the slope orientation, fragmentation and continuity of terrain should be study in the future. The regulation of terrain properties to the climatic factors such as precipitation, wind direction and heat are not clear, their temporal-spatial distribution pattern and potential impacts on the distribution pattern of vegetation NPP in the Lanzhou Section of the Yellow River Mainstream should be research more.

Conclusion

This paper, the spatiotemporal differences of vegetation NPP distribution in the Lanzhou section of the Yellow River mainstream are analyzed comprehensively using NPP data obtained from the CASA model, digital elevation data, land use data, meteorological data, and NDVI data, the impact of topography, human activities, and natural factors on the temporal-spatial changing characteristics of vegetation NPP in the Lanzhou Section of the Yellow River Mainstream are explored. The conclusions are as follows: (1) Between 2000 and 2020, the overall trend of NPP in the North and South Mountain showed an increasing state, with the proportions of extremely significant growth areas in the North and South Mountain being 66.62% and 61.01%, respectively. But NPP per area is dominated with lower grades and presents a slowly increasing, especially in the North Mountain, the area with lower altitude and having stronger human activities. (2) The average NPP of the South Mountain from 2000 to 2020 was 202.01 gC·m⁻²·a⁻¹, while that of the North Mountain was 118.69 gC·m⁻²·a⁻¹. Analyses of total quantity differences, grade distribution differences, temporal variation differences, and seasonal differences all indicated that the Vegetation NPP of South Mountain was superior to that of the North Mountain; (3) This differentiation was influenced by a combination of natural factors (topographic factors, climatic factors, vegetation coverage) and human activities (land use types); (4) The geographical detector results further confirmed that NDVI was the most significant factor influencing NPP (q-value of 0.654938 for the southern mountains and 0.466094 for the northern mountains), and the interaction between NDVI and temperature had the most significant impact on NPP.

Although vegetation NPP shows an increasing trend in the Lanzhou section of the Yellow River mainstream according to our research results, but NPP per area is dominated with lower grades and presents a slowly increasing, especially in the northern mountain, the area with lower altitude and having stronger human activities. In the future, it is necessary to comprehensively consider both natural and human factors, promote ecological comprehensive governance work according to local conditions, optimize vegetation functions and coordinate ecological compensation policies to enhance regional vegetation restoration,, except put into effect the ecological restoration measures proved to be successful.