Introduction

Ecosystems are vital sources of resources and services essential for human survival, advancement, and prosperity. They serve an irreplaceable function in sustaining the dynamic equilibrium of the Earth’s environment12;. About 60% of ecosystem services (ESs) are being reduced due to unsustainable land use34;, and it is predicted that degradation will worsen in the first half of the twenty-first century, seriously harming human wellbeing and drastically reducing the benefits of ESs5. To meet the growing population requirements, people have been accelerating resource development and land use6, which has caused many ecological issues, such as the loss of biodiversity, water pollution, and habitat fragmentation7,8,9, and led to changes in urban landscape patterns10. The relationship between humans and ecosystems is in increasing conflict, and there are now serious dangers to regional and global ecosystems. It is critical that diverse ESs have a stable supply11.

Changes in land use/cover (LULC) and spatial configurations are closely related to ESs9. One of the key factors influencing landscape-scale changes in ESs and their interactions is LULC12,13,14,15. By directly or indirectly influencing regional ecological patterns and processes, changes in LULC not only change the spatial distribution of ESs, but also change the ESs’ potential functions1617;. Researchers and decision-makers are increasingly incorporating concepts and indicators related to ecosystems into the creation of sustainable landscapes18,19,20. Landscape pattern is the spatial organization and integration of diverse types, sizes, and forms of landscape components that have an impact on ESs by altering the landscape’s composition and layout21,22,23,24,25. Landscape patterns are characterized in two main forms: landscape composition (e.g., patch type and size) and landscape configuration (e.g., patch size, shape, and spatial layout)2627;. The services supplied by ecosystems and the impacts on human well-being are a significant topics of research in the field of ecology2829;. Optimizing landscape composition and configuration can help to enhance ecological functions and maintain the sustainable development of ESs. Therefore, the first step in implementing multi-service landscape management is to examine the coupling relationship between landscape patterns and ESs30,31,32.

Research on LULC-based ecosystem management at the landscape scale remains underexplored1415;. While some scholars have investigated the nexus between landscape patterns and ESs213133,34,35;;, their research has focused on water-related ecosystem services3436;. The majority of research employs linear regression models and correlation analysis to clarify the linkage37,38,39. Landscape patterns and ESs exhibit dynamic shifts across temporal and spatial dimensions. Ocloo explored the relationship between ecosystem services and landscape patterns in Ghana, West Africa, with the help of correlation and regression analyses, but did not analyze the extent to which landscape pattern affects ESs from a geospatial perspective40. Ma revealed the trade-offs and synergies between ESs and landscape pattern driving mechanisms in the Beichuan River Basin from a geospatial perspective26; Zuo focused on the spatial impacts of landscape connectivity and landscape fragmentation on ESs41. The above studies focused on the spatial influence of landscape configuration on ESs. Effective territorial spatial planning mandates a profound understanding of the intricate interplay between landscape patterns and ESs across diverse geographical contexts, alongside a clear comprehension of the optimal timing and location for plan implementation. Consequently, the examination of ES responsiveness to landscape patterns, undertaken from a geographic perspective, assumes paramount significance. The strengths of this study are: (1) the introduction of a geographically weighted regression (GWR) model to spatialize the local landscape patterns driving ESs, thus supporting the differential management of local spaces; (2) the inclusion of landscape composition in the drivers as an important influence indicator to explore the extent of its explanation of ESs; (3) Harbin, a typical coldland provincial capital city, was selected as the study area to provide a reference for the spatialized management of ESs in coldland cities in China.

This study focuses on Harbin City, a prototypical cold-climate urban center in China, to explore the temporal changes in landscape patterns and ESs in the period from 2000 to 2020. Harbin City in northeast China has a unique dual identity as both an industrial and agricultural city. Harbin is a typical industrial city in China. Its ecosystem has been seriously damaged and the ecological environment is in an unsustainable negative state due to the accelerated pace of urbanization and the region’s inherently vulnerable climatic conditions characteristic of the cold zone42. Due to its abundant fertile black soil resources, Harbin also has a significant agricultural role. Agricultural land is approximately 50% of the city’s total area, making Harbin an important commercial grain production center in China.

The research aims to elucidate the magnitude and directional influence exerted by landscape patterns on ESs. There are four objectives: (1) to quantify the coupling relationship between key landscape patterns and ESs in Harbin; (2) to ascertain the extent of the landscape pattern’s impact on ESs; (3) to evaluate the localized responsiveness of spatially-distributed ESs to landscape patterns; and (4) to propose ecological optimization of the natural land space in Harbin City. This research has potential to facilitate beneficial interactions between ESs and landscape patterns in cold-climate cities amid accelerating urbanization, while providing scientific support for the sustainable governance of such urban spaces.

Materials and methods

Study area

Harbin is located in the northeast of the Northeast China Plain (Fig. 1), with an area of about 53,000 km2. As a regional central city in northeast China, Harbin is an important international transportation hub for the first Eurasian Continental Bridge and air corridor, is one of the core cities of the Harbin–Changchun urban agglomeration, and has a crucial role in the sustainable development of the northeast urban agglomeration. The city is one of the most important industrial cities in China, and is the most typical provincial capital city of the cold climate region in northeast China, with a total population of 9,431,700 at the end of 2021. The city experiences a mesothermal continental monsoon climate characterized by the presence of four well-defined seasons, with year-round precipitation mainly concentrated from June to September, with an average annual precipitation of 539.03 mm and an average annual temperature of 4.9 °C43. Harbin is located in the Songhua River basin and has many rivers, the most important of which are the main stream of the Songhua River and its tributaries. In addition, Harbin also has a large number of artificial reservoirs, natural lakes and ponds, which provide the city with abundant water resources and ecological landscapes. Harbin is rich in arable land resources, with fertile soil and a vast area, providing a superior basic environment for the development of agriculture. The 31 provincial and national nature reserves in the area provide a variety of crucial ecosystem services, including carbon sequestration and oxygen release, water nourishment, and climate regulation. With the growth of urbanization, the influence of people on the ecosystem has gradually increased. In this context, it is extremely important to balance the relationship between regional economic and social development and ecosystem health. The scientific justification for implementing sustainable development in Chinese cities with cold climates may be established with the aid of a thorough research of the relationship between landscape patterns and ESs.

Fig. 1
figure 1

Location of Harbin City in northeast China.

Full size image

The maps were generated by ArcGIS Pro (https://www.esri.com/enus/arcgis/products/index).

Data sources

To evaluate six important ESs in Harbin City and to investigate the geographical heterogeneity of ESs in response to landscape patterns, a variety of datasets on land LULC, meteorology, topography and geomorphology, vegetation, and socio-economics in 2000, 2010 and 2020 were integrated. Table 1 lists the data sources.

Table 1 Data sources for ESs.
Full size table

ESs assessment

Grain production in Heilongjiang Province ranks first in the country, and Harbin, as its capital city, has nearly half of the city’s total arable land area, and most of the arable land resources are nutrient-rich black soil44, its grain production in Heilongjiang Province and the country plays a very important role. In addition, the uneven distribution of water resources in the region and the severe degradation of permafrost in recent years have reduced the stability of regional water resources and disrupted ecosystem processes. Black soil has a deep humus layer and is an important agricultural soil resource in China, and the black soil area is an important commercial grain production base in China, At present, there is a serious soil erosion problem in the black soil area of Northeast China45, and the black soil layer is facing the danger of thinning, hardening, and loss, and if effective prevention and control measures are not taken, the majority of the black soil layer will be disappeared, which will seriously constrain the economy and agricultural production in the black soil area. With urbanization, the demand for carbon sink services is on the rise46, and a carbon storage crisis has been brought about by changes in land use. The deteriorating water quality of rivers and lakes and the destruction of habitat quality have been caused by the extensive cultivation of food crops and the application of large quantities of pesticides and chemical fertilizers.

Considering the characteristics of cold urban ecosystem services and data availability, six ecosystem services are assessed: food supply (FP), water yield (WY), soil conservation (SC), carbon stock (CS), water purification (WP) and habitat quality (HQ).

Food supply

About half of Harbin’s land area is used for agriculture, and food supply addresses the most basic human needs47. This study uses Net primary productivity data combined with statistical yearbooks to quantify food supply services. The calculation formula for food supply is:

$$\:\begin{array}{c}{F}_{x}=\frac{{NPP}_{x}}{NPP}\times\:F\end{array}$$
(1)

where Fx represents the crop yield on the agricultural land grid cell x; F is the total annual crop yield in Harbin City; NPPx is the NPP value of grid cell x in agricultural land; NPP is the total annual NPP of agricultural land in Harbin City.

Water yield

The water yield module of the InVEST model is based on the principle of water balance, where water yield is obtained by subtracting the amount of evapotranspiration from the amount of rainfall. The formula for water yield is:

$$\:\begin{array}{c}{Y}_{xj}=\left(1-\frac{{AET}_{xj}}{{P}_{x}}\right)\times\:{P}_{x}\end{array}$$
(2)

where Yxj is the water yield of land cover type j in grid x; AETxj is the actual evapotranspiration of land cover type j in grid x; Px is the precipitation of grid x; AETxj/Px is the ratio of the actual evapotranspiration to the precipitation. More information about the model can be available in the Supplementary Material.

Soil conservation

The difference between potential soil erosion and actual soil erosion is calculated based on the Universal Soil Loss Equation calculation method at the image element scale, and the result is the soil conservation (SC)4849;. The formula is:

$$\:\begin{array}{c}{SC}_{i}={RKLS}_{i}-{USLE}_{i}\end{array}$$
(3)
$$\:\begin{array}{c}{SC}_{i}={RKLS}_{i}-{USLE}_{i}\end{array}$$
(4)
$$\:\begin{array}{c}{RKLS}_{i}={R}_{i}\times\:{K}_{i}\times\:{LS}_{i}\end{array}$$
(5)

where SCi is the annual soil conservation; RKLSi is the potential soil erosion; USLEi is the actual soil erosion; R is the rainfall erosivity factor; K is the soil erodibility factor; L is the slope length factor; S is the steepness factor; C is the vegetation cover and crop management factor; and P is the soil and water conservation measures factor. More information about the model can be available in the Supplementary Material.

Carbon stock

The carbon storage module of the InVEST model uses above-ground biomass, below-ground biomass, soil carbon pools, and dead organic matter to quantitatively assess carbon stocks. Total carbon stock is calculated as:

$$\:\begin{array}{c}{C}_{t}={C}_{a}+{C}_{b}+{C}_{s}+{C}_{d}\end{array}$$
(6)

where Ct is the carbon stock; Ca is the aboveground fraction of the carbon stock; Cb is the belowground fraction of the carbon stock; Cs is the soil carbon stock; and Cd is the carbon stock in dead organic matter. More information about the model can be available in the Supplementary Material.

Water purification

Nitrogen pollution is the current major source of pollution flowing through the Songhua River basin50, and higher nitrogen output (NE) represents a weaker water purification capacity. In this study, NE was used as a negative indicator to quantify the WP condition. The water quality purification calculation formula is:

$$\:\begin{array}{c}{ALV}_{x}={HSS}_{x}\times\:{pol}_{x}\end{array}$$
(7)

where ALVx is the nitrogen output of raster x; HSSx is the hydrologic sensitivity score of raster x; polx is the output coefficient of raster x.

$$\:\begin{array}{c}{Q}_{xj}={H}_{j}\left(1-\left(\frac{{D}_{xy}^{Z}}{{D}_{xy}^{Z}+{k}^{z}}\right)\right)\end{array}$$
(8)

Habitat quality

The habitat quality module is a dimensionless indicator with a value from 0 to 151. The calculation formula is:

where Qxj is the habitat quality of grid cell x in land cover type j; Hj is the habitat suitability of land cover type j; \(\:{D}_{xy}^{Z}\) is the habitat stress level of grid cell x in land cover type j; k is the half-saturation coefficient, usually taken as half of the maximum value of \(\:{D}_{xy}^{Z}\); and x represents a constant. More information about the model can be available in the Supplementary Material.

Landscape indices

Landscape pattern indices were computed utilizing FRAGSTATS 4.2 to offer a comprehensive and intricate depiction of alterations in landscape patterns. The selected landscape indices, as outlined in Table 2, encompassed aspects such as regional edges, patch shape complexity, landscape aggregation, landscape fragmentation, and landscape diversity. Utilizing the moving window method for computation, it becomes possible to spatially represent the distribution pattern and the temporal evolution of landscape composition and configuration. Too small a moving window scale may lead to discontinuity in the landscape pattern image, while too large a scale may lead to loss of image details. In this study, the optimal moving window scale is determined by calculating the semi-variance function of landscape pattern index under different moving window radii. After debugging, an area of 720 m × 720 m was selected to calculate the landscape pattern of Harbin City.

Table 2 Landscape indices used in the study.
Full size table

Trade-offs and synergies

In this investigation, the trade-offs and synergies between the ESs were ascertained using Spearman’s correlation analysis5253;. Spearman correlation analysis was performed using the R language “corrplot” package. The formula is:

$$\:\begin{array}{c}{R}_{ab}=\frac{\sum\:_{i=1}^{n}\left({a}_{i}-\overline{a}\:\right)\:({b}_{i}-\overline{b}\:)}{\sqrt{\sum\:_{i=1}^{n}{\left({a}_{i}-\overline{a}\:\right)}^{2}\:\sum\:_{i=1}^{n}{({b}_{i}-\overline{b}\:)}^{2}}}\end{array}$$
(9)

where Rab is the correlation coefficient of the two ecosystem services; ai, bi are the values at the i-th sample point for ecosystem services a and b; \(\:\overline{a},\:\overline{b}\) are the mean values of ecosystem services a and b; and n is the number of samples.

$$\:\begin{array}{c}q=1-\frac{\sum\:_{h=1}^{L}{N}_{h}{\sigma\:}_{h}^{2}}{N{\sigma\:}^{2}}=1-\frac{SSW}{SST}\end{array}$$
(10)

Geographic detector

The driving force of the variability in the spatial distribution of ESs is revealed by factor detection in the usage of geographic detector (GD). Factor detection primarily involves the computation of spatial heterogeneity within the dependent variable while evaluating the capacity of various influencing factors to explain the spatial variations in ESs within the study area. This explanatory capability is quantified using the q-value, a metric ranging from 0 to 1, where a higher value signifies a stronger explanatory power. It was calculated by the formula:

where h (h = 1, 2,. . ., L) is the stratification of variable Y or factor X; Nh and N are the number of cells in stratum h and the whole region, respectively; \(\:{\sigma\:}_{h}^{2}\) and \(\:{\sigma\:}^{2}\) are the variance of the Y values in stratum h and the whole region, respectively; SSW and SST are the sum of the variances within the stratum and the total variance in the whole region, respectively.

$$\:\begin{array}{c}SSW=\sum\:_{h=1}^{L}{N}_{h}{\sigma\:}_{h}^{2},SST=N{\sigma\:}^{2}\end{array}$$
(11)

Geographically weighted regression

In order to properly account for regional variation, geographically weighted regression (GWR) creates localized coefficients by using spatial location characteristics5455;. The extent of the impact of landscape indices on ESs at various geographic areas was examined using the GWR model. The calculation formula was:

$$\:\begin{array}{c}{y}_{k}={\beta\:}_{0}\left({u}_{k},{v}_{k}\right)+{\sum\:}_{i=1}^{n}{\beta\:}_{i}\left({u}_{k},{v}_{k}\right){x}_{ki}+{c}_{k\:}\end{array}$$
(12)

where yk is the ES value; xki is the landscape index; n is the total number of spatial units in the analysis; ck is the random error term; (uk, vk) is the spatial location of sample k; β0(uk, vk) is the intercept at location k; and βi(uk, vk) is the coefficient of the i-th dependent variable of sample k.

Results

Landscape pattern changes from 2000 to 2020

Changes in land use

In the LULC composition of Harbin, agricultural land is the dominant category, with approximately 50% of the total area from 2000 to 2020 (Figs. 2 and 3). This was closely followed by woodland with 35% of the area and grassland with 8% of the area. A notable 796.67 km² of agricultural land was converted to woodland, representing 31.76% of the overall agricultural land that changed use (Fig. 3). The conversion of woodland to construction land constituted a notable portion, amounting to 30.81% of the total agricultural land conversion. As compared to the period 2000–2010, the area converted out was significantly higher during 2010–2020. The primary transformation of woodland predominantly involved its conversion into grassland, representing a substantial 67% of the total area transitioned from woodland. In turn, grassland saw conversions into various land uses, including woodland (832.77 km²), agricultural land (463.11 km²), and construction land (144.65 km²). Wetland was predominantly converted into water bodies, making up 69.41% of the total converted area, while water body area was largely repurposed for agricultural use (81.19 km²). Construction land was mainly converted to agricultural land, which made up 83.24% of the total area transferred out of it.

Agricultural land area continued to fall from 2000 to 2020, but wetland and forest areas displayed a trend of first reduction and subsequent increase. The impact of ecological protection policies was discernible, as despite the continuous growth in built-up areas, the area of woodland ecosystems and wetland ecosystems did not decrease. This trend indicates a synergistic advancement of economic development in concert with ecological conservation. This trajectory underscores the realization of a synergistic nexus between economic development and ecological conservation in Harbin.

Fig. 2
figure 2

LULC of Harbin City from 2000 to 2020.

Full size image

The maps were generated by ArcGIS Pro (https://www.esri.com/enus/arcgis/products/index).

Fig. 3
figure 3

LULC transition matrix from 2000 to 2020.

Full size image

AL: Agricultural land; WoL Woodland GL Grassland WeL Wetland WB Water body CL Construction land.

The maps were generated by Origin 2022 (https://www.originlab.com/).

Changes in landscape pattern at the class level

Changes in the level of landscape pattern index types in Harbin City from 2000 to 2020 are shown in Table 3. From 2000 to 2010 the most notable reductions in agricultural land patches were in NP (12.90%) and PD (12.89%), while the most notable increase was in SPLIT (0.78%). More significant decreases in woodland patches were in IJI (3.91%) and LPI (3.17%), while the most significant increases were in NP (4.11%) and PD (4.08%). The SPLIT and LPI of grassland patches decreased most significantly during the 10-year period, at 27.07% and 21.43%, respectively. From 2010 to 2020 agricultural land patches showed the biggest increase in SPLIT (9.87%) and the largest fall in LPI (5.46%) over the landscape pattern change trend compared to the previous ten years. While NP and PD dramatically decreased by 11.41% and 11.39%, respectively, in woodland patches, LPI grew significantly (6.28%). In contrast to the considerable declines in grassland PLAND, ED, TE, NP and PD, the grassland patch LPI increased by 6.61%.

Overall, from 2000 to 2010 the fragmentation and intricacy of agricultural land and grassland landscapes decreased, while woodland landscapes experienced increased fragmentation and evolved into more complex shapes. From 2010 to 2020, fragmentation and complexity in agricultural, woodland and grassland landscapes all diminished.

Table 3 Changing rates (%) in the landscape indices at the class level from 2000 to 2020.
Full size table

Changes in landscape pattern at the landscape level

Table 4 delineates the variations in Harbin’s landscape metrics at the macro level from 2000 to 2020. From 2000 to 2010, SPLIT (1.18%), TE (1.14%), LSI (1.13%) and ED (1.11%) had a pronounced uptrend, while IJI had the most substantial decline (4.83%), followed by PD (2.62%) and NP (2.59%). From 2010 to 2020, the sharpest increase was in SPLIT (6.61%), followed by SHEI (4.78%) and SHDI (4.77%), while in contrast, NP (6.75%), PD (6.72%) and LPI (5.46%) had a significant decrease.

Cumulatively, from 2000 to 2020, NP (9.16%) and PD (9.17%) had the most profound reduction, while SPLIT (7.87%) had a notable increase. The comprehensive fragmentation of Harbin’s landscape markedly diminished. There was a decrease in human disturbance and landscape intricacy, a diversification of landscape types, and an increase in spatial heterogeneity.

Table 4 Changing rate (%) in the landscape indices at the landscape level from 2000 to 2020.
Full size table

Analysis of the changes in the landscape index under a moving window

Appendix Figure S1 provides a visual representation of the temporal and spatial shifts in landscape pattern indices within Harbin City spanning the years 2000 to 2020. To simplify the content, detailed analysis of the changes of LPI, CONTAG, PD, SPLIT and SHDI is highlighted (Fig. 4). Over the 20 years, LPI in the western built-up region of the city increased, characterized by a growing area of high values and intensifying human disturbances. However, the northeastern and southeastern parts saw a decline in LPI. CONTAG in the built-up region progressively decreased, while in contrast, the northwest had an increase, the northeast a significant decline, and the southeast an increase. PD prominently declined in the city’s western and northeastern regions, but increased in the central and southeastern sections. SPLIT had pronounced spatial heterogeneity, with high values decreasing significantly in the western part of the city and increasing significantly in the central and southern parts of the city. Last, SHDI generally trended upwards, with the western urban areas having a minor decline and the eastern portion a significant rise.

In summary, the western sector of the city exhibited heightened levels of human disturbance activities, concomitant with a marked decline in landscape aggregation. Conversely, the southeastern and northern sectors of the city witnessed a decline in fragmentation, concomitant with an upsurge in landscape diversity and aggregation.

Fig. 4
figure 4

Spatial distributions of the partial landscape indices at the landscape level from 2000 to 2020.

Full size image

LPI: Largest patch index; CONTAG: Contagion index; PD: Patch density; SPLIT: Splitting index; SHDI: Shannon’s diversity index.

The maps were generated by Fragstats 4.2.681 (https://fragstats.org/index.php/downloads).

Changes in ESs from 2000 to 2020

Spatial-temporal variations of ESs

From 2000 to 2020, there were significant spatial and temporal variations in ESs (Fig. 5). FP increases significantly in 2000–2010 and decreases slightly in 2010–2020. This may be related to the development of modern agricultural technology. Spatially, FP is mainly concentrated in the western, central and southeastern corners of the city, and FP in the central and southeastern corners is slightly higher than that in the western part of the city. WY showed a continuous increase, with the high value area mainly concentrated in the eastern part of the city and the densely populated distribution area, and the low value area concentrated in the western part of the city and along the Songhua River. This may be related to the increase in precipitation.SC, CS and NE generally show a decreasing and then increasing trend. SC and CS show a spatial distribution pattern of “low in the west and high in the east”, while NE shows a distribution pattern of “high in the west and low in the east”. From 2000 to 2020, the overall HQ shows a rising trend, but due to the continuous expansion of construction land, the HQ in the western part of the city decreases, and the western built-up area is obviously located in the HQ low value area. The reason may be that the eastern part of the city has the ecological barrier areas of Xiaoxinganling and Zhangguancaililing, which have high soil conservation, carbon storage and habitat quality.

Fig. 5
figure 5

Changes in ESs from 2000 to 2020.

Full size image

FP: food supply; WY: water yield; SC: soil conservation; CS: carbon stock; WP: water purification; HQ: habitat quality.

The maps were generated by InVEST 3.12.1 (https://naturalcapitalproject.stanford.edu/software/invest).

Trade-offs and synergies between pairs of ESs

In the 15 pairs of correlation among the six ESs analyzed, six had positive relationships, while the remaining nine were negatively correlated (Fig. 6). Notably, each ecosystem service had significant intercorrelations across the study years, barring a non-significant association between WY and SC in 2000. Synergistic relationships were evident among FP, WY and NE. However, these services demonstrated trade-offs with SC, CS and HQ. SC, CS and HQ had mutual synergies. Areas with less human activity, higher vegetation cover, or more undulating topography tend to have higher SC, CS, and HQ. In contrast, areas with more human activities, lower vegetation, or flatter topographies tend to have higher FP, WY and NE.

Fig. 6
figure 6

Trade-offs and synergies between pairs of ESs from 2000 to 2020.

Full size image

FP: Food supply; WY: Water yield; SC: Soil conservation; CS: Carbon stock; WP: Water purification; HQ: Habitat quality.

The maps were generated by R 4.3.1 (https://www.r-project.org/).

Impacts of landscape pattern changes on ESs

The influence of landscape pattern indices on ESs was examined using the GD package in the R language (Fig. 7). The driving power of other landscape pattern indices on ESs passed the significance test, with the exception of NP’s explanatory power on WY, which failed at the p = 0.01 level. The proportions of agricultural land (PLAND_al) and woodland (PLAND_wl) were chosen as the influencing factors for factor detection due to the predominance of these types of land in the urban expanse. For FP, the chief determinants were PLAND_al and PLAND_wl, whereas LPI and AI predominantly influenced WY. SC was chiefly impacted by PLAND_wl, LPI and AI. Notably, PLAND_al and PLAND_wl had substantial explanatory power for CS, NE and HQ. It is noteworthy that the explanatory power of PD was more pronounced for ESs in 2000, with a tapering effect in subsequent years.

In summary, among the landscape indices, PLAND_al, PLAND_wl, LPI and AI emerged as the most influential on ESs, with landscape composition proving more influential on ESs than landscape configuration.

Fig. 7
figure 7

Contributions (q values) of the landscape indices to ESs.

Full size image

FP: Food supply; WY: Water yield; SC: Soil conservation; CS: Carbon stock; WP: Water purification; HQ: Habitat quality; PLAND_al: Percentage of landscape of agricultural land; PLAND_wl: Percentage of landscape of woodland; ED: Edge density; LPI: Largest patch index; TE: Total edge; LSI: Landscape shape index; AI: Aggregation index; CONTAG: Contagion index; NP: Number of patches; PD: Patch density; SPLIT: Splitting index; DIVISION: Landscape division index; SHDI: Shannon’s diversity index; SHEI: Shannon’s evenness index.

The maps were generated by Origin 2022 (https://www.originlab.com/).

Spatial interactions among landscape patterns and ESs

Five landscape indices—PLAND_al, PLAND_wl, LPI, CONTAG and PD—were selected to elucidate pairs and mitigate multicollinearity concerns. The influence of the landscape pattern on ESs was scrutinized using the GWR model. As presented in Fig. 8, the model’s Local R2, coupled with standard residual values predominantly in the − 2.5 to 2.5 range, supported the model’s robust fit and the reliability of the findings. Additionally, each landscape pattern index’s regression coefficients from 2000 to 2020 were calculated (Figs. 9, 10, 11, 12 and 13).

PLAND_al negatively affected SC and HQ in more than 50% of the area, while positive effects dominated for other ESs. The significant positive effects of PLAND_al on FP were mainly distributed in the southern part of the city. Significant positive effects of PLAND_al on WY were mainly concentrated along the Songhua River in a banded manner, while the opposite was true for CS, where the Songhua River coast was the concentrated area of significant negative impacts. Significant negative impacts of PLAND_al on SC were mainly concentrated on the east side of the city, significant positive impacts on HQ were mainly concentrated on the west side of the city, and positive impacts on NE were distributed over almost the entire area of the city. PLAND_wl had positive impacts on FP, SC, CS, and HQ in more than 60% of the area, and negative impacts were distributed over a smaller area. PLAND_wl had negative impacts on WY and NE in more than 60% of the area, and the positive impacts were mainly distributed in the central part of the city, near the Songhua River. LPI mainly had positive effects on SC, CS, and HQ, and negative effects on FP, WY, and NE. The significant positive effects of LPI on FP were mainly concentrated in the west side of the city and along the Songhua River, while most of the other areas had negative effects, which were mainly concentrated in the ecological barrier areas of Xiaoxing’anling and Zhang Guangcailing. There was no clear distribution pattern for the spatial impact of LPI on NE and HQ. CONTAG had a negative effect on SC and HQ in more than 50% of the region, and a positive effect on FP, CS, and NE. More than 50% of the region had a positive effect on WY in 2000, and more than half of the region had a negative effect in 2010–2020. The significant negative effects of CONTAG on SC were mainly concentrated in the Xiaoxinganling and Zhangguancailing ecological barrier areas, and the significant positive and negative effects on WY and CS were cross-distributed along the Songhua River. The positive effects of PD on FP and HQ had an enhanced trend, with more than 50% of the area positively affected FP in 2000, and more than 60% positively affecting HQ in 2000, and increased to 60% and 70% of the area in 2020, while the negative effects on the other ESs dominated. The spatial impacts of PD on each ESs were similar to the distribution of impact coefficients of CONTAG.

Fig. 8
figure 8

GWR model fitting results for ESs.

Full size image

FP: Food supply; WY: Water yield; SC: Soil conservation; CS: Carbon stock; WP: Water purification; HQ: Habitat quality.

The maps were generated by ArcGIS Pro (https://www.esri.com/enus/arcgis/products/index).

Fig. 9
figure 9

The quantitative effects of PLAND_al in depicting ESs through GWR.

Full size image

FP: Food supply; WY: Water yield; SC: Soil conservation; CS: Carbon stock; WP: Water purification; HQ: Habitat quality; PLAND_al: Percentage of landscape of agricultural land.

The maps were generated by ArcGIS Pro (https://www.esri.com/enus/arcgis/products/index).

Fig. 10
figure 10

The quantitative effects of PLAND_wl in depicting ESs through GWR.

Full size image

FP: Food supply; WY: Water yield; SC: Soil conservation; CS: Carbon stock; WP: Water purification; HQ: Habitat quality; PLAND_wl: Percentage of landscape of woodland.

The maps were generated by ArcGIS Pro (https://www.esri.com/enus/arcgis/products/index).

Fig. 11
figure 11

The quantitative effects of LPI in depicting ESs through GWR.

Full size image

FP: Food supply; WY: Water yield; SC: Soil conservation; CS: Carbon stock; WP: Water purification; HQ: Habitat quality; LPI: Largest patch index.

The maps were generated by ArcGIS Pro (https://www.esri.com/enus/arcgis/products/index).

Fig. 12
figure 12

The quantitative effects of CONTAG in depicting ESs through GWR.

Full size image

FP: Food supply; WY: Water yield; SC: Soil conservation; CS: Carbon stock; WP: Water purification; HQ: Habitat quality; CONTAG: Contagion index.

The maps were generated by ArcGIS Pro (https://www.esri.com/enus/arcgis/products/index).

Fig. 13
figure 13

The quantitative effects of PD in depicting ESs through GWR.

Full size image

FP: Food supply; WY: Water yield; SC: Soil conservation; CS: Carbon stock; WP: Water purification; HQ: Habitat quality; PD: Patch density.

The maps were generated by ArcGIS Pro (https://www.esri.com/enus/arcgis/products/index).

Discussion

Model validation

In order to verify the simulation results, we calculated the total annual WY in Harbin and compared it with the total water resources in Harbin recorded in the Heilongjiang Water Resources Bulletin, and the error was controlled within 10%. The sand transport calculated by the model was compared with the sand transport modulus in the China River Sediment Bulletin to verify the reliability of the SC results, and the error was controlled within 3%, which showed that the SC was well simulated. Comparison of NE with the nitrogen output concentration monitored in the field in other studies50 showed that the results were closer, indicating that the simulation was reliable.

The results of the fit of the different models are shown in Table 5. Compared to geographically weighted regression (GWR), adjusted R2 of ordinary least squares (OLS) was lower and it’s AICc was higher, indicating that spatial non-stationarity was not taken into account and led to significant estimation bias.GWR takes into account the spatial heterogeneity and localized features, and the relationship between ecosystem services and the various drivers was identified through the use of GWR that resulted in better performance.

Table 5 Comparison of the fitting measures for different models.
Full size table

Impacts of land use on ESs

This research examined landscape pattern alterations and their implications on ESs in Harbin City over the two decades from 2000 to 2020. The results showed that urbanization and industrialization dominated in Harbin City from 2000 to 2010, with less attention to ecological protection. However, from 2010 to 2020, the city began to focus on the protection of natural ecosystems, and the results were more obvious. Over the two decades from 2000 to 2020, agricultural land consistently reduced in area. Notably, woodland and wetland areas initially experienced a minor decline, but later had a substantial resurgence. This trajectory enhanced SC, CS and HQ, aligning with findings from prior research456;. Such dynamics can be attributed to the ESs provided by wetlands, including biodiversity conservation and societal benefits5758;. The city’s woodlands, predominantly situated within the ecological barrier zones of the Xiaoxing’an Mountains and the Zhangguangcai Range, have dense vegetation and varied terrains, strengthening their SC and CS capacities. However, woodland’s proliferation negatively impacts WY. Given the elevated evapotranspiration rates in woodlands5960;, the WY, under comparable climatic conditions, tends to be lower. Yet, intriguingly, the period from 2000 to 2020 had a marked increase in WY concurrent with an expansion of woodland areas. There are two reasons: the construction land area is expanding dramatically, and the increase in built-up area leads to an increase in impervious cover, which in turn increases WY34; and as the most typical provincial capital city in the cold region in northeast China, Harbin is sensitive to climate change and has frequent extreme disaster weather61. The rainfall increased significantly from 2000 to 2020, which led to the continuous increase of WY. Apart from the influence of landscape configurations, climatic variables emerge as critical drivers shaping ESs. Human-induced activities predominantly alter land use patterns and intensities, instigating profound transformations in ecosystem structures. In contrast, natural variables, especially climatic shifts, influence biological growth and dynamics, subsequently impacting ESs6263;.

Heilongjiang Province steadily pushed forward the structural reform of the agricultural supply side from 2000 to 2020. As the capital city of its province, Harbin City, through strategic measures, including policy direction, exemplary demonstrations, and the impetus from new managerial entities, the province adjusted its food crop planting structure. The province effectively recalibrated both the regional layout and variety structure. While the agricultural land area contracted during this period, enhancements in agricultural technology and policy-driven initiatives spurred a significant increase in FP from 2000 to 2010, which then stabilized in the subsequent decade. Urbanization introduces new pollution sources with the potential to penetrate aquifers and degrade groundwater quality, posing significant challenges to urban development64,65,66. Construction lands are key pollution contributors34. In Harbin, agricultural land comprises nearly half of the city’s total area. Excessive use of pesticides and fertilizers, coupled with household waste discharges, has tainted both soil and water quality. From 2000 to 2010, the NE diminished due to advancements in agricultural technology, a decline in agricultural land area, and a moderated growth of construction land. However, from 2010 to 2020, accelerated expansion of construction land and increased human interference led to an increase in NE.

Impacts of landscape patterns on ESs

ESs were affected by landscape patterns in different ways according on the time era. According to the results of GD and GWR models, it can be seen that PLAND_al and PLAND_wl as key landscape components and LPI and AI as key landscape configurations have a stronger explanatory power for each ESs, and the explanatory power of landscape components for ESs is generally higher than that of landscape configurations.The degree of explanation of ESs by the PD in the year 2000 is significantly higher than that of the PD in the years 2010 and 2020. The results of the study are consistent with the findings of Ma, Zhang, Abdollahi and other experts2631347273;;;;. The main reason may be that the LPI can be seen as an indicator of the intensity of human disturbance, while AI signifies the connectivity among landscape patches. Enhanced inter-patch connections foster material transfer and ecological processes, ultimately augmenting ESs35. In this study, two representative indicators, PLAND_al and PLAND_wl, were selected to explore the effect of landscape composition on ESs. The results of the study showed that PLAND_al had a positive impact on FP, WY, CS and NE. The higher the proportion of agricultural land, the more likely it is to produce more food, and FP and CS will also increase. At the same time, because the vegetation coverage of agricultural land is lower than that of woodland, the evapotranspiration is lower, so WY is higher. However, due to extensive cultivation of food crops and application of large quantities of pesticides and fertilizers, NE will increase and WP, SC and HQ will decrease. PLAND_wl had a negative effect on WY and NE because woodland has higher vegetation cover and evapotranspiration, which has an intercepting effect on pollutants, thus WY and NE are lower and have a good function of water quality purification. CONTAG reflects the degree of aggregation between landscape patches68, which has a negative effect on HQ, which presents an opposite result to that of Ma26and a consistent result to that of Zuo41. CONTAG has a positive effect on CS and NE, which is is consistent with the results of previous studies34. The main reason may be that the stronger the landscape aggregation and the better the connectivity, the easier it is to lose soil and water, the less the SC and the more the NE, and therefore the water quality purification capacity is reduced, leading to the reduction of habitat quality. The degree of landscape patch fragmentation and variety is reflected in the PD69. The higher the PD, the stronger the landscape heterogeneity, the more landscape patches, and the more rich habitat types. Patches made up of various landscapes compared to a single landscape reduced NE by reducing the negative effects of pollutants on water quality70. The results suggest that landscape patterns have complex effects on ESs, and changing a particular landscape pattern may have both positive and negative effects on ESs, which should be rationally planned for specific areas.

Implications for ecosystem management

Landscape units are physical land spaces that carry objective geographic differentiation laws and subjective spatial resource perceptions, which help to identify the correlation characteristics of ESs among multiple levels 78; 79, and investigate in depth the connection between ESs and landscape patterns, which helps to carry out urban ecological restoration planning. In light of the fact that different landscape patterns have varying effects on various ESs, multiple adaptive ecosystem management techniques should be employed for managing different ESs, as this study has demonstrated. In future urban planning, more attention should be paid to landscape composition and configuration to effectively improve the ESs. With the development and implementation of a series of ecological protection policies in Harbin City in recent years, such as special action to combat the destruction of wetland resources and the return of farmland to woodland, the benign development of natural ecosystems has been promoted to a large extent. It has played a good role in the protection of land resources, which has led to the conversion of a large amount of agricultural land to woodland, improving four ESs but not FP and WY.

However, due to the complex interactions between ESs, it is difficult to maximize each ES by changing the landscape pattern. In the future spatial planning of land, policy makers should consider this intricate relationship and develop each ES in a balanced manner to maximize t.he comprehensive ecological benefits. WY will also be reduced, so it is possible to increase the proportion of woodland near agricultural land to minimize some WY, thus enhancing the WP. However, it is difficult to carry out this work in practice. With the delineation of the “three zones and three lines” in recent years, the state strictly protects agricultural land, and the phenomenon of “returning woodland to farmland” has occurred, and it is difficult to shift agricultural land to woodland, which requires local policy makers to formulate a more practicable plan, to avoid contradicting national policy, and at the same time, to balance the development of the local ESs and enhance human wellbeing.

At the same time, because different districts and counties implement policies with different strengths and focuses, it is difficult for the development direction of landscape patterns to be consistent, and therefore the impacts on ESs are not the same, which has far-reaching impacts on human wellbeing. In this study, we used the GWR model to spatially characterize the impacts of different landscape patterns on ESs, which provides a scientific reference for policy makers in each region’s ecological planning(Figs. 9, 10, 11, 12 and 13). As a typical industrial city, the ecological management of Harbin’s habitat is crucial. According to the results of the GWR study, in order to improve the ecological quality of the built-up area in the western part of the city, the HQ of the densely populated area can be increased by suppressing the rapid development of impervious pavement and increasing CONTAG. Due to the prominent water pollution problems in Harbin, the proportion of cultivated land around water bodies can be reduced thereby reducing the inflow of nitrogen, phosphorus and other pollutants into the water bodies. In recent years, the precipitation in Harbin City has gradually increased, and the WY has shown a drastic increase trend, but the increase in WY does not necessarily mean an increase in human benefits, and is even prone to urban flooding disasters. Coupled with the relatively small proportion of green space in the built-up area, low biodiversity, and insufficient rainfall and flood storage capacity, the human environment in Harbin City has been seriously jeopardized. In the vicinity of the built-up area, WY shows a decreasing trend with the increase of CONTAG, so WY can be reduced in this area by increasing the spreading degree and connectivity of the landscape and constructing a network of blue and green spaces.

Limitations of this study and future work

This study was conducted for ESs and landscape patterns, and the InVEST model was chosen to evaluate WY, SC, CS, NE and HQ. Although it has the advantages of simplicity and easy access to data, it does have some uncertainties. Since some parameters are difficult to obtain, they need to be estimated according to the characteristics of the study area73. The more precise the parameters are, the more the uncertainty of ESs decreases74. For example, soil loss increases linearly with increasing C and P values, and the WY model is more sensitive to Kc21. Therefore, how to obtain the precise parameters is important in running the model. In addition, the WY model fails to take the complex topography into account in the water production module, and for areas with intensive human activities, policies such as water diversion and transfer, as well as the upstream water inflow, will have a great impact on the actual water resources, which are not taken into account in the WY model. The estimation of NE dynamics is frequently characterized by a lack of consideration for the intricate biochemical processes governing nutrient transformations. This omission consequently engenders a degree of uncertainty within the InVEST model75. This study upheld the model’s reliability through a process of ongoing parameter adjustments, complemented by thorough comparisons with existing studies and authoritative datasets.

Due to the intricacy of the findings, this study did not spatialize specific policies and clearly distinguish the many policy implementation regions and specific implementation policies, even though it made some findings about the ecological planning of territorial space. Future studies should emphasize planning the development direction of different regions, balancing the trade-offs and synergistic relationships between ESs, and optimizing the landscape pattern for human wellbeing. As well as landscape pattern, other factors also affect ESs. Thus the driving mechanism of ESs with additional influencing elements needs to be studied in the future.

Conclusion

Based on the InVEST model, this study evaluated the geographical and temporal variations of six ESs in Harbin City from 2000 to 2020, and the impact of the landscape pattern on ESs was examined using the GD and GWR models. Overall, the degree of landscape fragmentation in the study area decreased significantly over 20 years, the degree of human interference and landscape complexity declined, landscape types tended to be richer, and spatial heterogeneity gradually increased. The effects of PLAND_al, PLAND_wl, LPI and AI on the ESs were higher, and PLAND_al had a positive effect on FP, WY, CS and NE, while PLAND_wl had a positive effect on FP, SC, CS and HQ, LPI had a favorable impact on SC, CS and HQ, CONTAG had a positive effect on FP, CS and NE, and PD had a positive effect on FP and HQ. This study investigated the spatiotemporal coupling effects between ESs and landscape patterns, giving local policymakers in urban ecological planning a more thorough viewpoint. However, due to the complexity of the results, specific policies were not spatialized. Future studies should focus on planning the development direction of different regions to balance ESs and promote human wellbeing.