Introduction

Global climate change and land use/cover change (LUCC) have a significant impact on the production and sink processes1,2. Global warming has caused the atmospheric boundary layer to hold more water vapor, resulting in a significant increase in the intensity of extreme precipitation in many regions. This has also affected the rates of the hydrological cycle and the processes of generating runoff in these areas3. Furthermore, extensive human activities such as the construction of water projects, urbanization, and deforestation have modified the conditions of catchment substrates and the distribution of water resources on both spatial and temporal scales. These alterations have had significant effects on soil moisture, evapotranspiration, catchment production, and sink processes4. The Songnen Plain in China serves as a significant agricultural region for cereal production, including maize, rice, and soybeans5. The territory is classified as belonging to the temperate monsoon climate zone, which is characterized by significant influence from hydrothermal conditions. Consequently, the region is very susceptible to climate extremes. Furthermore, due to population growth and the rising need for socio-economic progress, the Songnen Plain has experienced significant alterations in land utilization and coverage (such as a substantial expansion of paddy fields, the implementation of water conservation infrastructure, and soil degradation), resulting in a range of water security issues6. Hence, there is an urgent need to quantitatively analyse the impacts of land use/cover change on hydrological processes in the Songnen Plain in the dual context of frequent climate extremes and food production security.

As human activities persist in disturbing land use and cover, the impact of land use change on the depth of runoff is likewise increasing7. The impact of changes in land use on the depth of runoff has become a critical issue in the study of global environmental change8. Currently, researchers have conducted numerous studies on the influence of land-use changes on the depth of runoff9. These studies primarily concentrate on three main areas: firstly, analyzing the effects of various land-use changes on runoff depth10; secondly, examining the relationship between the spatial distribution pattern of land-use and runoff depth11; and thirdly, investigating the correlation between changes in runoff depth and the corresponding socio-economic systems12. Research has been carried out to investigate the relationship between changes in land use and the depth of runoff, in order to offer theoretical backing for this study. Nevertheless, current research has primarily focused on the impact of individual land-use changes on the depth of runoff, with limited investigation into the correlation between these changes and the underlying mechanism when multiple land-use types are combined. Regarding research views, the present focus of studies on land use change and runoff depth mostly centers on examining the correlation between land use change and changes in runoff depth. However, these studies generally fail to include the influence of climate change on runoff depth during the study process. Furthermore, previous research at the scale of analysis has primarily focused on examining the influence of various land use changes on the depth of runoff from a broad perspective. However, there have been relatively few studies that have investigated the connection between these two factors from a more detailed and specific viewpoint.

Currently, researchers employ paired catchment tests13, climate resilience approaches14, statistical methods15, and hydrological modeling techniques16 to investigate the temporal and spatial variations in runoff depth within a watershed. Catchment mapping tests primarily include comparing similar catchments to analyze the impact of land use change on the depth of runoff. Nevertheless, this approach frequently has challenges in identifying catchments that share the same or similar underlying and meteorological circumstances as the catchment being studied17. The Budyko framework model, although more physically significant than the standard mathematical-statistical empirical method, has limitations in accurately modeling hydrological processes, which restricts its effectiveness in climate resilience methods18. Statistical attribution analysis methods primarily consist of multivariate linear regression and spatio-temporal geographically weighted regression model (GTWR). This methodology uses statistical analysis to investigate the correlation between runoff and environmental variables. However, it has a drawback in that it fails to consider the impact of climate change on the abrupt alteration of runoff patterns19. Hydrological simulation involves the use of various models, such as the VIC model20 and the MIKE SHE/MIKE 11 model21. These models primarily analyze the effects of climate conditions and land-use changes on runoff using hydrological scenarios (such as the runoff reduction method). However, a major limitation of this method is the inability to quantitatively investigate the relationship between changes in different land-use types and the depth of runoff22. Consequently, there is a lack of reliable techniques to effectively examine the impacts of various alterations in land use on the depth of runoff, and further investigation is necessary.

In order to accurately analyse the effects of different land use changes on runoff depth in the Songnen Plain, this study innovatively proposes a combination of hydrological scenario simulation and statistical attribution analysis (Fig. 1). Compared with previous studies, the proposed method achieves the effect of different land use changes on runoff depth without the influence of climate change. The study will begin by selecting a distributed hydrological model, such as the MIKE SHE/MIKE 11 model, known for its high accuracy in simulating the water cycle21. The runoff reduction method will then be employed to accurately measure and separate the impact of climate change and human activities on the change in runoff depth. Then, this work utilizes the spatio-temporal geographically weighted regression model (GTWR) to thoroughly analyze the impacts of various land use changes on runoff depth by differentiating the effects of climate change and land use change. The GTWR model is a powerful tool for examining the correlation between various land use changes and runoff depth in a small watershed from a micro-spatial perspective38. The study offers valuable information for the implementation of watershed development projects, regional planning initiatives, and the protection and management of water resources in places where data is lacking.

Fig. 1
figure 1

Location of the Songnen Plain and spatial distribution of land use.

Full size image

The study aims to achieve three specific objectives: (1) to precisely replicate the process of runoff depth changes in the Songnen Plain over the past 40 years; (2) to examine the influence of both climate change and human activities on the depth of runoff; and (3) to measure the effect of alterations in various land use types on the depth of runoff.

Materials and methods

Study area

The Songnen Plain is situated in the northeastern region of China, encompassing the southwestern area of Heilongjiang Province, the northwestern area of Jilin Province, and a small portion of the Inner Mongolia Autonomous Region. Geographically, it is positioned between 121°21′ and 128°18’E longitude and 43°36’-49°26’N latitude, covering an area of 18.28 × 104 km2 (Fig. 2a). The plain consists mostly of alluvial deposits from the Songhua and Nenjiang Rivers. The topography of the plain consists of a partially enclosed, asymmetrical basin characterized by elevated edges and a depressed center, which gradually slopes inward from the outer boundaries. The plain is situated in the East Asian monsoon climate zone, The Songnen Plain is situated in the East Asian monsoon climate zone, which is distinguished by a high average annual potential evaporation (1298.1 mm) and a low average annual precipitation (494.78 mm)23. The distribution of precipitation varies greatly in terms of space and time, and the dry environment restricts the availability of water in the catchment area. The average annual runoff at the Jiangqiao hydrological station over many years is 227.56 × 108 m3. During the flood season, which occurs between July and September, more than 60% of the runoff occurs24.

Fig. 2
figure 2

Spatial distribution of ___location and other information on the Songnen Plain.

Full size image

The basin suffers substantial erosion of soil and water as a result of the compact and fragile soil layer, along with the frequent intense rainfall during the flood season25. Efforts to mitigate these losses have been steadily rising since the late 1960s. By 2006, two substantial dams, capable of storing a combined total of 7.1 million m3 of water, had been built on the main river. Additionally, there were 12 notable smaller reservoirs and 885 check dams on tributaries. The entire area under control, including the dams and reservoirs, exceeded 61 km2 26. The basin consists of 329.67 km2 of terraces, 23.32 km2 of freshly constructed embankments, 1198.79 km2 of forested land, and 60.53 km2 of grassland27.

Distributed MIKE SHE/MIKE 11 model

MIKE SHE/MIKE 11 Model Principle and construction

The MIKE SHE/MIKE 11 model, derived from the SHE (Système Hydrologique Européen), is a sophisticated distributed hydrological model that accurately replicates the terrestrial aspects of the hydrological cycle. It encompasses essential processes like actual evapotranspiration, runoff, groundwater flow, and the interplay between different hydrological phenomena28. The model has multiple interconnected modules, with the key ones being Evapotranspiration (ET), Overland-Channel Flow (OC), Unsaturated Zone (UZ), Saturated Zone (SZ), Snow Melt (SM), and Aquifer-River Exchange (EX). Each submodule reflects the corresponding process of the hydrologic cycle. The model uses finite differences to solve the partial differential equations that depict the movement of water flow while simulating each distinct sub-module. The connection between the MIKE SHE model and the MIKE 11 model is established through the use of river linkages (MIKE SHE links)29. These links allow for the simultaneous simulation of the flow between surface water and groundwater, enabling the two models to be connected.

Establishing the demarcation of the research area border is an essential requirement for the creation of the model. The China Geological Survey delineated the boundary of the Songnen Plain using the research’s results, which precisely consider the surface watershed features and groundwater system of the plain30. Furthermore, it is imperative that the MIKE SHE/MIKE 11 model employ an identical projection coordinate system in order to guarantee seamless functionality of the model. This paper utilized ArcGIS 10.2 software to standardize the data to the WGS_1984 format and establish the projection coordinate system as WGS_1984_UTM_Zone_48N. Based on the accuracy requirements of the model run and the realistic situation, the reference data acquisition in this study ultimately uses a 500 × 500 m grid as the smallest cell size for the model.

Model input data preparation

The main input data for the MIKE SHE/MIKE 11 model include of climatic data, land use information, soil property data, groundwater measurements, river runoff measurements, and basic geographic data. This analysis relied extensively on the “China Surface Climate Standard Daily Value Dataset” acquired from the China Meteorological website for its meteorological station data. The majority of the data comprises daily mean values for precipitation, temperature, and crop evapotranspiration. The China Science Data Center offers land use data for model simulations with a resolution of 30 m for the years 1980, 1990, 2000, 2010, and 2020 (Fig. 2b). The data classifies regional land into categories such as cropland, forest land, grassland, water, urban and rural areas, industrial and mining areas, residential areas, and unused land. The National Earth System Science Data Centre (NESDC) provided the soil data under the name “Basic Attribute Dataset of China’s High Resolution National Soil Information Grid (2010–2018)”. Given the absence of measured soil property data, this study utilized the SPAW soil hydrological property software, developed by the USDA, to compute regional soil properties including soil saturated water content, field water holding capacity, wilting point water content, and saturated hydraulic conductivity28. The DEM data is obtained from NASA and NIMA’s SRTM elevation data, which has a spatial resolution of 30 m. The research findings from the China Geological Survey include data on the height of the submerged bottom plate, groundwater levels (Fig. 2c), and underlying hydrologic conditions. Refer to Table 1 for a detailed explanation of the specific statistics.

Table 1 Data required for the model and its characteristics.
Full size table

Model calibration and validation

The model is calibrated and validated using data on measured river discharge and groundwater levels. The study initially determined the most responsive model parameters through manual calibration. Subsequently, the class 4 groundwater calculation subdivision (Fig. 2d) were utilized as calibration areas, and the sensitive parameters for each zone were separately calibrated using automatic calibration based on the measured data. The chosen calibration parameters are the horizontal hydraulic conductivity, vertical hydraulic conductivity, soil infiltration coefficient, and saturation water content. These parameters are determined based on experience with simulation models and related literature31,32.

The model utilizes real hydrological and meteorological data, employing the sample split approach. The calibration period spans from 2005 to 2014, the validation period from 2015 to 2018, and the predicted values encompass the years 1980 to 2004 and 2019 to 2020. The primary metrics utilized for evaluating model performance consist of the correlation coefficient (R), root mean square error (RMSE), and Nash-Sutcliffe efficiency (NSE). The calibration targets for the observed and modeled data were set as NSE > 0.75, r > 0.85, and RMSE < 2 m, taking into account the scale and purpose of the model.

Runoff reduction method based on hydrological model

Mutation point identification

To evaluate the impacts of climate change and land use change on runoff depth, the study period is separated into two separate periods: a base period and an impact period. This division is accomplished by employing a mutation point test. The base period is defined as the period during which human activity has the least significant impact on hydrological processes. Conversely, the impact period is defined as the period during which human influence on hydrological processes is substantial. To undertake a thorough investigation of the influence of climate change and human activities on hydrological processes, one might compare the differences between the hydrological cycle processes during the base period and the impact period33. The Mann-Kendall method is a non-parametric test for the presence of mutations in sequences. Details of the method can be found in the reference34.

Runoff reduction method

The base period and impact period are precisely categorized based on the precise ___location of the mutation point in runoff. The MIKE SHE/MIKE 11 model is used to simulate natural runoff by inputting climate data and land use data for the base period (Wpre). To simulate runoff under the influence of climate change, climate data for the impact period and land use data for the base period are entered into the model (Wsim). Similarly, to simulate runoff under the influence of land use change, climate data for the base period and land use data for the impact period are inputted(Wlucc). Lastly, to simulate runoff under the combined influence of climate change and land use change, climate data and land use data for the impact period are entered into the model (Wpost). The calculation involves determining the impact of human activities (ΔWH) and climate change on the change in runoff (ΔWC), as well as the combined effect of both factors (ηH, ηC)35. The specific formulae are:

$$\Delta {W_{all}}=\Delta {W_{S4}} - \Delta {W_{S1}}$$
(1)
$$\Delta {W_{climate}}=\Delta {W_{S3}} - \Delta {W_{S1}}$$
(2)
$$\Delta {W_{human}}=\Delta {W_{S2}} - \Delta {W_{S1}}$$
(3)
$${\eta _c}=\frac{{\Delta {W_{climate}}}}{{\Delta {W_{all}}}} \times 100\%$$
(4)
$${\eta _L}=\frac{{\Delta {W_{human}}}}{{\Delta {W_{all}}}} \times 100\%$$
(5)

Where Wpre is the change in natural runoff; Wpost is the change in runoff under the combined influence of climate change and human activities; Wsim is the change in runoff under the influence of climate change; and Wlucc is the change in runoff under the influence of human activities.

Spatial autocorrelation models

Global spatial autocorrelation

This study utilizes the spatial autocorrelation index, known as Global Moran’s I, to examine the presence of spatial agglomeration in the spatial data36. Global Moran’s I ranges from − 1 to 1, where a positive value suggests a tendency for clustering, with larger values indicating a stronger degree of clustering. Conversely, a negative value suggests a tendency for dispersion, with smaller values indicating a more pronounced tendency for dispersion. The specific formula for Global Moran’s I is as follows:

$$Moran \text{'}s\;I=\frac{{n\sum\limits_{{i=1}}^{n} {\sum\limits_{{j=1}}^{n} {{W_{ij}}\left( {{x_i} - \bar {x}} \right)\left( {{x_j} - \bar {x}} \right)} } }}{{\left( {\sum\limits_{{i=1}}^{n} {\sum\limits_{{j=1}}^{n} {{W_{ij}}} } } \right)\sum\limits_{{i=1}}^{n} {{{\left( {{x_i} - \bar {x}} \right)}^2}} }},\;\;\,(i \ne j)$$
(6)

Where: xi and xj are the changes in runoff depth at different times in the sub-basin; n is the number of spatial units in the sub-basin; Wij is the spatial weight matrix, using the neighbourhood criterion for measurement; and ‾x is the mean of x.

Bivariate spatial autocorrelation

This study used bivariate spatial autocorrelation analysis to examine the spatial association between various land use changes and their corresponding runoff depths37. The specific equations are:

$$Moran \text{'}s\;I_{{zy}}^{a}=\frac{{X_{y}^{a} - \overline {{{X_y}}} }}{{{\delta _y}}} \times \sum\limits_{{c=1}}^{n} {{W_{ac}}} \times \frac{{X_{z}^{c} - \overline {{{X_z}}} }}{{{\delta _z}}},\;\;\;(a \ne c)$$
(7)

Where \({{X_{y}}^{a}}\) is the value of attribute y in spatial cell a; \({{X_{z}}^{c}}\) is the value of attribute z in spatial cell c; \(\overline{x_{y}}\) and \(\overline{x_{z}}\) are the means corresponding to attributes y and z respectively; δy and δz are the variances corresponding to attributes y and z respectively; n is the number of spatial cells; Wac is represented as a spatial weight matrix using the neighbourhood criterion as the measure; attribute y represents the rate of change of different land uses; attribute z represents the amount of change in runoff depth.

Spatio-temporal geographic weighted regression (GTWR) model

The GTWR model is derived from the GWR model, incorporating spatial and temporal factors. This model comprehensively incorporates the temporal aspects of the regression parameters of the independent variables and the spatial heterogeneity of the regression parameters across different spatial units, resulting in variation of the regression parameters based on geographical ___location38. The basic formula is:

$${y_i}={\beta _0}\left( {{u_i},\;{v_i},\;{t_i}} \right)+\sum\limits_{k} {{\beta _k}\left( {{u_i},\;{v_i},\;{t_i}} \right)} {X_{ik}}+{\varepsilon _i}$$
(8)

Where: (ui, vi) denotes the coordinates of the i sample point; ti denotes the observation time; yi denotes the value of the dependent variable at the i sample point; Xik denotes the value of the k independent variable at the i sample point; βk (ui, vi, ti) denotes the regression coefficient of the k independent variable at the i sample point; βo (ui, vi, ti) and εi are the regression constant and error terms for the i sample point, respectively.

Results

Simulation and analysis of runoff depth in the Songnen Plain over the last 40 years.

Model calibration and validation

The model was calibrated and validated using actual measurements of groundwater levels and river runoff data, yielding the calibration findings (Fig. 2).

The graphic demonstrates a strong correlation between the river runoff volumes predicted by the MIKE SHE/MIKE 11 model and the actual observed river runoff volumes. The patterns of increase and decrease in runoff are generally similar in both cases. Furthermore, the model slightly underpredicts flood peaks in the region, mainly because it does not accurately simulate the spring thaw process. This is a typical occurrence during the modeling process and has a minimal effect on the later impact study.

During the entire time (Table 1), the Nash efficiency coefficient and water balance error coefficient for the calibration period were 0.822 and 1.46, respectively. For the validation period, the corresponding coefficients were 0.759 and 2.462, demonstrating improved simulation results. The model also passed the 0.01 significance test based on the F-test results.

The simulation findings of the groundwater level (Fig. 3) demonstrate a strong correspondence between the simulated and observed groundwater levels for both the calibration and validation periods. In terms of the simulations’ general accuracy, the correlation coefficients between the observed and modeled groundwater data at the various monitoring stations were all higher than 0.8. Additionally, the Nash Efficiency Coefficients (NSE) were all higher than 0.6, with the majority of stations having NSE values above 0.8. Hence, the MIKE SHE/MIKE 11 model accurately replicates the water cycle phenomenon in the Songnen Plain. Evidently, the model is capable of acquiring precise data on changes in runoff depth.

Fig. 3
figure 3

Comparison chart of actual runoff at each station and simulated value.

Full size image

Spatial and temporal characteristics of runoff depth

The MIKE SHE/MIKE 11 model was employed to predict the variations in runoff depth in the Songnen Plain spanning from 1980 to 2020. Subsequently, the model was employed to compute the regional average values of runoff depth for various years. The results are shown in Fig. 4.

Fig. 4
figure 4

The comparison of actual groundwater level and simulated value at each site.

Full size image

The chart illustrates that the average depth of regional runoff across several years is 36.26 mm. The highest average annual runoff depth of 61.57 mm was observed in 2020, while the lowest average annual runoff depth of 16.1 mm occurred in 2012. Over the course of the study, the regional runoff depth exhibited a pattern of “rising-falling-rising,” with precipitation being the primary factor influencing the runoff depth.

Spatially (Fig. 5a), the average annual depth of runoff in the Songnen Plain varies from − 50.25 to 168.36 mm. The trend across the region indicates a steady decrease in runoff depth from the southwest to the northeast of the basin. Runoff depth exhibits high values primarily in proximity to low-lying and wetland regions, while low values are predominantly observed in places with flat terrain. The inter-annual variation of runoff depth (Fig. 5b) indicates that the overall runoff depth of the Songnen Plain is experiencing a little decrease, with an average rate of variation of -0.025 mm/a. The areas with high value are primarily situated in the southern and eastern regions of the Songnen Plain, whereas the areas with low value are found in the northwestern section of the plain and in the places where water conservancy facilities have been created.

Fig. 5
figure 5

Variation pattern of runoff depth in the Songnen Plain.

Full size image

Analysis of the contribution of climate change and human activities to runoff depth

Mutation point test analysis

The Mann-Kendall method was employed in this study to determine the point at which precipitation and runoff depth experienced a significant change in the study area. The findings are presented in Fig. 6.

Fig. 6
figure 6

Spatial distribution characteristics of mean annual runoff depth and its trend in the Songnen Plain.

Full size image

The diagram depicts the locations where changes in precipitation were identified using the M-K test from 1991 to 2012, as well as the areas where changes in runoff depth were observed from 1995 to 2010. This study examines the mutation points of precipitation and runoff depth in the Songnen Plain, taking into account the impact of human activities. The years 1995 and 2010 are selected as the mutation points to divide the runoff depth process into three eras. The time span between 1980 and 1995 is designated as the base period (T1), during which both human activities and climatic factors had minimal impact. Starting in 1996, the influence of human activities gradually intensified and began to interact with climatic factors in the runoff depth process, marking this period as the impact period. Since 2010, there has been an increased focus on water conservation projects, leading to the rapid building of water conservation facilities in the Songnen Plain and paddy field irrigation projects30. As a result, there has been a heightened impact on the water cycle caused by human activities. Hence, this study divided the impact period into two phases: impact period 1 (T2) from 1995 to 2010 and impact period 2 (T3) from 2011 to 2020.

Quantitative decomposition study of the impact of runoff depth

This study utilized the runoff reduction approach to analyze changes in runoff depth. Climate data and land use data were collected before and after the mutation point. These data were then incorporated into the calibrated MIKE SHE/MIKE 11 model to simulate the variations in runoff depth under different circumstances. The runoff depth contribution was computed using Eqs. (1) to (5), and the results are presented in Table 2.

Table 2 MIKE SHE/MIKE 11 model calibration period and validation period model performance.
Full size table

Table 2 shows that during impact period 1, the runoff depth decreased by 19.07 mm compared to the base period. Of this decrease, 3.18 mm (16.68% contribution rate) was caused by human activities, and 15.89 mm (83.31% contribution rate) was caused by climate change. During impact period 2, there was an overall increase in runoff of 7.49 mm compared to the base period. Of this increase, 5.24 mm (29.16% contribution rate) was caused by land use change, and 12.73 mm (70.84% contribution rate) was caused by climate change.

Spatially, In impact period 1, the areas with the highest levels of human activities affecting runoff depth are primarily found in the central region of the Songnen Plain (Fig. 7). This area experiences more severe land degradation, which has a significant impact on runoff, accounting for approximately 71.53% of the overall contribution. In impact period 2, the areas with the highest levels of human activities affecting runoff are mainly located in the central and western parts of the Songnen Plain. These areas are close to the construction of water conservancy facilities and river channels, contributing approximately 65.11% to the overall impact. To summarize, climatic change is the primary driver of variations in regional runoff depth. However, as human activities become more intense, their influence on runoff depth processes is progressively growing, with their contribution gradually rising from 20% to approximately 30%. The main factors contributing to changes in the regional hydrological cycle during impact period 1 are the reduction of wetland area, land degradation, urban expansion, and increased urban water demand. During impact period 2, the construction of water conservancy facilities and the substantial expansion of paddy fields, such as the Nilgir Water Conservancy Hub and Hadashan Water Conservancy Hub projects, have led to notable alterations in the vegetation cover of the region. Consequently, there have been more pronounced modifications in the depth of regional runoff.

Fig. 7
figure 7

Results of the mutation point testing for precipitation and runoff depth.

Full size image

Analysis of the spatial response of different land use changes to runoff depth

To investigate the impact of various changes in land use on the depth of runoff, the study period was divided into four phases: 1980–1990, 1991–2000, 2001–2010, and 2011–2020. The alteration in runoff depth caused by human activities (ΔWH) was determined for the years 1991–2000, 2001–2010, and 2011–2020 using the runoff reduction method. The study investigated the relationship between the rate of change in different land use types and the change in runoff depth caused by human activities, using the GTWR model and regional sub-watersheds as the computational units (ΔWH)39.

Spatial autocorrelation analysis

The global Moran’s I index for runoff depth changes at different time intervals within the Songnen Plain subwatershed was calculated using the GeoDa software, based on Eq. (7) (Fig. 8).

Fig. 8
figure 8

Contribution of climate change and human activities to runoff depth at different stages.

Full size image

From the image, it is evident that the global Moran’s I index of the Songnen Plain is consistently greater than 0 in various time periods. This suggests that the runoff depths at different times exhibit positive spatial autocorrelation and that the data display a certain level of clustering. Nevertheless, the global Moran’s I indices exhibit lower values, mostly as a result of the distinct geographical settings and placements of various watersheds.

To determine whether there is a spatial correlation between the variability of land use and the depth of discharge at different time periods, we utilized spatial bivariate autocorrelation methods (refer to Fig. 9) to investigate this relationship. The impact of a single alteration in land use on the depth of runoff was found to have a negative correlation with the pace of change in wetlands, paddy fields, and forests (Fig. 10). The respective Moran’s I indices of -0.359, -0.198, and − 0.188 indicate that the increase in wetlands, paddy fields, and woodlands partially hinders the generation of runoff depth. The rate of change of dryland, construction land, and unused land is positively correlated with the change in runoff depth. The respective Moran’s I indices of 0.142, 0.11, and 0.018 indicate that all three land types contribute to an increase in runoff depth.

Fig. 9
figure 9

Global autocorrelation of runoff change in the Songnen Plain 1980–2020. A indicates change in runoff depth from 1980 to 1990; B indicates change in runoff depth from 1990 to 2000; C indicates change in runoff depth from 2000 to 2010; D indicates change in runoff depth from 2010 to 2020.

Full size image
Fig. 10
figure 10

Bivariate autocorrelogram between interannual rates of change in different land uses and interannual changes in runoff depth. JLL indicates change in runoff depth; ST indicates rate of change in paddy land; HD indicates rate of change in dry land; LD indicates rate of change in forest land; CD indicates rate of change in wetland; JSDY indicates rate of change in construction land; WLYD indicates rate of change in vacant land.

Full size image

Contribution of different land use changes to changes in runoff depth

Constructs a Geographically Weighted Regression (GTWR) model by using the change in runoff depth (ΔWH) as the dependent variable and the various rates of land use change as the independent variables. Among them, the proportion of the water area was relatively modest and did not show significant variation. Therefore, the water area can be excluded as an independent variable. Based on the constructed GTWR model, the mean values of the regression coefficients (Table 3) and the spatial and temporal distribution maps of the regression coefficients (Fig. 10) for different land use types in different time periods were solved (Table 4).

Table 3 The impact of climate change and human activities on runoff depth.
Full size table
Table 4 Regression coefficients of different land use changes on runoff depth.
Full size table

Table 3 demonstrates that dryland, construction land, and unused land all have an impact on the depth of runoff. Among these, construction land has the greatest influence on runoff depth, with a regression coefficient of 6.21 × 10−1. This means that a 10% increase in the area of construction land in a small catchment leads to a 6.21 mm increase in runoff depth. Unused land and dryland have regression coefficients of 2.45 × 10−1 and 1.14 × 10−1 respectively, indicating their lesser contributions to runoff depth. The presence of paddy fields, forested land, and wetlands had a significant impact on reducing the depth of runoff. Among these, wetlands had the strongest inhibitory effect, with a regression coefficient of − 9.49 × 10−1. This means that a 10% increase in wetland area in the watershed resulted in a decrease in the depth of runoff by 9.49 mm. Paddy fields and forested land also had inhibitory effects, but to a lesser extent, with regression coefficients of − 6.46 × 10−1 and − 3.07 × 10−1, respectively.

In terms of spatial analysis (Fig. 11), the regression coefficients for drylands exhibit a positive trend, with values ranging from 0.2 to 0.4. This indicates that the expansion of dryland primarily contributes to the rise in the depth of runoff. The regression coefficients exhibit significant values mostly in regions adjacent to drylands, paddy fields, wetlands, and woodlands, primarily due to the higher water retention capacity of forests, wetlands, and paddy fields compared to drylands. The regression coefficients for paddy fields and wetlands exhibit distinct polarization, with low values ranging from − 0.9 to − 0.5 primarily in places furthest from the river and high values ranging from 0.2 to 0.4 primarily in areas adjacent to the river. The main reason for this is the greater amount of water present in wetlands and paddy fields, along with the larger volume of water that flows near the river channel, resulting in a greater depth of runoff. On the other hand, the ability of paddy fields and wetlands to hold water limits the depth of runoff from expanding in places that are far away from the river channel. The regression coefficients associated with woodland are predominantly negative, ranging largely between − 0.3 and − 0.1. This suggests that woodland has a substantial inhibitory impact on the increase in runoff depth. Both the regression coefficients for building land and unused land exhibit a positive value. The regression coefficients for building land often fall within the range of 0.5 to 0.8. Higher values are predominantly found in locations with greater urban land use, where the impervious layer is more extensive and hence more prone to form catchments. The regression coefficients for unused land predominantly fall within the range of 0.1 to 0.3. Higher values are primarily observed in the lower regions of the catchment. This is mainly due to the greater availability of water downstream and the comparatively weaker water retention capacity of unused land. Consequently, unused land is more prone to generating runoff downstream.

Fig. 11
figure 11

Distribution of GTWR model local regression coefficients between land use change rate and change in runoff depth.

Full size image

Discussion

Impacts of climate change and human activities on runoff depth

Quantifying the effects of climate change and human activities on changes in runoff depth is important for the scientific management and use of water resources40. At present, a large number of studies have been carried out at home and abroad on the patterns of runoff changes in watersheds under changing environments and the attribution of runoff changes, such as studies on the influence of changing environments on hydrological processes in watersheds41, the quantification and separation of the contribution of the main factors of runoff changes in watersheds42, and the possible impacts and uncertainties of climate change on future runoff in watersheds43. Compared to existing studies, this study uses a coupled surface-groundwater model to simulate the process of water cycle change and uses river runoff data and groundwater level change data to calibrate and validate the model, which improves the accuracy and reliability of the study and can better explain the intensity of the water cycle response to climate change and human activities. This study found that the impact of human activities on runoff depth changes was more significant in areas where wetland degradation was evident in the central and western parts of the region and in areas close to rice fields and cities. This is mainly due to the continuous reclamation of wetlands into dry fields as a result of human activities, which has a lesser effect on impeding runoff and thus causing changes in runoff, and the expansion of paddy fields and urban areas, which causes changes in runoff. In addition, this study found that the effects of climate change and human activities on runoff behave differently during the dry and wet periods of the basin. During the dry period (impact period 1), when precipitation is relatively low, human activities have a greater impact on runoff changes in the central and western parts of the Songnen Plain, which is a semi-arid region. During the high water period (impact period 2), precipitation is abundant and climate change is the main factor leading to increased runoff, but in areas of the basin where land use changes are evident, such as near paddy fields, water facilities, and cities, the basis for land use change impacts dominates (Fig. 7).

Analysis of the mechanisms driving land use change on runoff depth

Disentangling the effects of human activities on runoff depth can help to understand the mechanisms driving changes in the water cycle. Human activities such as the expansion of urban areas, the diversion of water for irrigation, and the drastic reduction of wetlands are changes that can alter regional water cycle processes and thus affect runoff depth in catchments44. In this study, based on quantifying the effects of regional climate change and human activities on runoff depth, the effects of different land use types on runoff depth changes were decomposed using a GTWR model. The study found that a 10% increase in the rate of change of construction land, dry land, and unused land in the watershed resulted in an increase in runoff depth of 6.21 mm, 2.45 mm, and 1.14 mm, respectively, while a 10% increase in the rate of change of wetland, paddy, and forest land resulted in a decrease in runoff depth of 9.49 mm, 6.46 mm, and 3.07 mm, respectively.

Since 1995, human activities have had a far greater impact on the environment than changes in nature itself45. Before 2010 (the impact period 1), the average precipitation in the Songnen Plain was 79.16 mm below the multi-year average, leading to an annual reduction in regional runoff depth (Fig. 5). Moreover, the implementation of water-stopping projects in the upper sections of numerous rivers has curtailed river runoff, resulting in a reduction of flow, particularly in the Tao’er River and the Hualin River, which have seen persistent cessation of flow46. Subsequent to 2010, during the impact period 2, the implementation of extensive water conservancy projects progressively enhanced the water transfer capacity of the Songnen Plain. The Hadashan Reservoir and Nilgi Reservoir progressively redirected water to the southwestern region of the plains, enhancing the capacity for water resource utilization in the Songnen Plain and resulting in an increase in runoff depth from the plains47. Besides, the expansion of the basin’s irrigation zones and impervious surfaces is occurring, with the enlargement of irrigation areas resulting in heightened irrigation, while the growth of impervious surfaces contributes to increased surface runoff and diminished infiltration, subsequently elevating the depth of runoff in the region. Overall, as human activity has increased, its impact on runoff depth has gradually increased. This study analyzed the effects of various land use patterns on runoff depth using a spatio-temporal geographically weighted regression model (GTWR), with findings aligning with prior research10,11. This study examines the correlation between land use change and runoff depth within a small watershed and will subsequently conduct a comprehensive analysis of runoff depth variations across different land use patterns to identify the optimal land use strategy for enhancing regional soil and water resource conservation.

Uncertainty and prospects

Although this study provides an in-depth analysis of the contribution of runoff depth change to climate change and subsurface change in the Songnen Plain and quantifies the impact of different land use changes on runoff change, there are still some uncertainties. Firstly, there are uncertainties in the method used to analyze the attribution of runoff depth change. In this study, the analysis of the subsurface drivers of runoff depth change assumes that precipitation, runoff, and subsurface changes are independent of each other, but in fact, the three are interrelated. Changes in the subsurface cause a response in atmospheric motion, resulting in changes in precipitation, and changes in precipitation cause changes in the runoff process, and these effects are not considered in this study. Secondly, the data processing process introduces uncertainty. Due to the small number of meteorological stations in the catchment, the representation is somewhat poor. To reduce this uncertainty, national meteorological data were used to obtain the meteorological element characteristics of the catchment by spatial interpolation, but uncertainty still exists.

In summary, this study quantifies the sensitivity and intensity of runoff depth responses to climate change and human activities and also decomposes the impact of subsurface changes on runoff processes under human activities, revealing to some extent the driving mechanisms of runoff depth changes. The constructed coupled surface-subsurface water model is suitable for assessing the response of climate change and human activities in the catchment to changes in runoff depth, and the model has some applicability to the Songnen Plain. Human activities affect hydrology mainly through land use, soil conservation, rainwater storage, and other methods that alter the underlying surface of the catchment, resulting in changes in the runoff depth regime. In general, in areas of low human activity, natural climate change is the main factor influencing changes in hydrological processes. Precipitation increases runoff depth, while temperature has a dissipative effect on water resources. Compared to the existing literature, this study is innovative in its analytical approach to quantitatively identifying the impacts of climate change and human activities on the hydrological cycle. However, the mechanisms of human activities at runoff depth are more complex, and the quantitative decomposition of different anthropogenic factors at runoff depth needs further investigation.

Conclusion

This study is based on the coupled surface-groundwater model MIKE SHE/MIKE 11 to simulate the process of water cycle changes in the Songnen Plain over the past 40 years; the runoff reduction method is used to quantify the impact of climate change and human activities on runoff depth ; and the spatio-temporal geographically weighted regression model is used to quantitatively study the impact of different land use changes on runoff depth. The main results are as follows:

The simulated runoff depth results for the Songnen Plain reached R2 and NSE above 0.75 in both the baseline and validation periods, and the model also passed the 0.01 significance test according to the F-test results. Overall, the MIKE SHE/MIKE 11 model has high simulation accuracy and can simulate the runoff depth of the Songnen Plain well.

Over the last 40 years, the average runoff depth in the Songnen Plain has been 36.26 mm, showing a trend of “increase, decrease, increase”. Spatially, the high-runoff areas are mainly located near low-lying areas and wetlands, while the low-lying areas are mainly located near flat areas.

The contribution of climate change to runoff depth is more significant than that of human activities. In impact period 1, runoff decreased by 19.07 mm, of which the reduction due to human activities was 3.18 mm, a contribution of 16.68%, and the reduction due to climate change was 15.89 mm, a contribution of 83.31%; in impact period 2, runoff increased by a total of 7.49 mm compared to the base period, of which the reduction due to human activities was 5.24 mm, a contribution of 29.16%, and the increase due to climate change was 12.73 mm, a contribution of 70.84%.

The effect of different land use types on runoff depth is in the following order: construction land > dryland > unused land > wetland > paddy land > woodland. A 10% increase in the rate of change of construction land, dryland, and unused land in the catchment will lead to an increase in runoff depth of 6.21 mm, 2.45 mm, and 1.14 mm, respectively; a 10% increase in the rate of change of wetland, paddy, and woodland will lead to a decrease in runoff depth of 9.49 mm, 6.46 mm, and 3.07 mm, respectively.