Attribution identification of terrestrial ecosystem evolution in the Yellow River Basin

2.1 Study area

The Yellow River basin (95°53′–119°05′E, 32°10′–41°50′N) is the second largest river in China, occupying approximately 795,000 km², which accounts for about 8.3% of China’s continent area (Figure 1). The watershed is located between northwest arid zones and southeast humid zones, most of the areas belong to arid and semi-arid continental monsoon climate. The average annual temperature is 6.74°C, with the characteristics of increasing from west to east and decreasing from north to south. The average annual precipitation is 487 mm, mainly concentrated in June to September and exhibited a decreasing trend from southeast to northwest. The Yellow River Basin covers 9 provinces (autonomous regions) of China, including more than 66 cities and 340 counties, with a total population of 116 million and a population density of 147 people/km². The terrain is high in the west and low in the east, with a highest elevation of 6,212 m and a minimum of −8 m. The basin is divided into three parts, the Plateau Glaciers in the upper reaches, the Loess Plateau in the middle reaches and the Yellow River Delta in the lower reaches.

Figure 1

Location, elevation, main cities, basin boundary and its different sections (upstream, midstream and downstream) of the Yellow River Basin.

2.2 Data

The data used in this article include the historical land use data, the historical meteorological data, digital elevation model (DEM) data, the socioeconomic data, the roads data and the statistics data of the Yellow River Basin (Table 1).

2.3 Methods

The research workflow includes data preparation, scenarios setting, land use simulation, index calculation and attribution identification (Figure 2). First, based on the historical meteorological data and land use data, the Mann–Kendall test and land use transfer matrix were used to detect the influence of climate change and human activities. Then, the scenarios of without any impacts (Scenarios 1), affected by climate change (Scenarios 2), by human activities (Scenarios 3) and by both impacts (Scenarios 4) were constructed. Second, the future land use simulation (FLUS) model was calibrated and verified, which was used to simulate the land use distribution in 2015 under the four scenarios. Third, the landscape patterns index and the ESV were calculated. Finally, according to the discrepancies of evaluation results between the scenarios, the impacts of climate change and human activities on terrestrial ecosystem were quantitatively evaluated.

1. Mutation detection

   The Mann–Kendall test is a nonparametric test method, which is widely used to detect the long-term change trend and mutation point of the precipitation and temperature [38,39]. The formulas were as follows:

   S k = ∑ i = 0 k r i ,

   r i = + 1 , x i > 0 , 0 , x i ≤ x j ,

   UF k = ( S k − E ( S k ) ) Var ( S k ) ,

   where k = 1, 2, …, n; E ( S k ) = k ( k + 1 ) / 4 , Var ( S k ) = k ( k + 1 ) ( 2 k + 5 ) / 72 . UF k was the statistical series calculated by the time series, and all the values would form a curve; then, the sequence was arranged in reverse order and repeated the above calculation process to get UB k . The significance level in this article was 0.05, and the critical value UF 0.05 = ± 1.96 . When the curve of UF and UB intersected at a certain point, which called the mutation point.

   According to the annual precipitation and temperature data of the Yellow River Basin from 1981 to 2015, we found that the average annual precipitation showed a decreasing trend with a decreasing rate of −2.1 mm/10a, but the change trend showed no significance (p > 0.05). While the average annual temperature showed a significant increasing trend (p < 0.05) with a rising rate of 0.4°C/10a. Since the temperature change was significant, and which would affect the frequency and intensity of precipitation [40], the temperature numerical data were used to detect the mutation point of climate change, and the mutation year appeared in 1994. Based on this, we divided the baseline period climate (1981–1994) and the variation period climate (1995–2015).

2. Scenarios setting

   We used the Arcgis 10.2 software to calculate the land use transfer matrix of the Yellow River Basin from 1995 to 2015. The changes in forestland, grassland, water bodies and other areas were mainly related to cultivated land and construction land. If the conversion of cultivated land and construction land to other land use types was prohibited, the changes of forestland, grassland, water area and unutilized land affected by human activities will be very small [41]. Therefore, the impacts of human activities were determined by whether the cultivated land and construction land changed in land use conversion. If the impacts of human activities were absence, the cultivated land and construction land were consistent with that in 1995. On the contrary, the conversion of cultivated land and construction land with other land use types was permitted.

   In the light of whether the land use types were affected by climate change and human activities, four scenarios were constructed, including without any impacts (Scenario 1), affected by climate change (Scenario 2), by human activities (Scenario 3) and by both impacts (Scenario 4). Where the discrepancies between Scenario 1 and Scenario 2 reflected the impacts of climate change, the discrepancies between Scenario 1 and Scenario 3 reflected the impacts of human activities and the discrepancies between Scenario 1 and Scenario 4 reflected the combined impacts of climate change and human activities.

3. The FLUS model simulation and validation

   In this article, the FLUS model was used to simulate and predict the land use distribution (Figure 3), which was an integrated model for multi-type land use scenario simulation by coupling anthropogenic and nature effects [42,43]. First, the system dynamics model was used to predicted the land use demand under various climatic and socioeconomical driving factors. Then, the artificial neural network was used to calculate the adaptability probability, which aimed to obtain the sinuous relationships between land use types and various human and nonhuman driving factors. In addition, the model adopted an adaptive inertia competition mechanism based on roulette wheel selection method, which dealt with the uncertainty and complexity of land use conversion effectively, and made the FLUS model with a high simulation accuracy [44,45,46].

   Based on the land use data in 1995, we selected natural factors (such as DEM, slope and aspect), social and economic factors (such as population density and GDP), meteorological factors (such as temperature and precipitation) and other driving factors to calculate the adaptability probability of the land use types (Figure 3) and then simulated land use distribution in 2015. The actual data of 2015 were used to verify the simulation results. The Kappa coefficient was 0.85, which indicated that the FLUS model could better simulate the spatial distribution of the land use in the Yellow River Basin. In the next work, we simulated and predicted the land use distribution under the four scenarios by adjusting the input driving force and parameters of the model.

4. Landscape patterns

   According to the landscape patterns features and the practical principle in the study area, the watershed ecological landscape patterns indices were analyzed from the type level and landscape level. Where the type level includes the number of patches (NP), the edge density (ED), the landscape shape index (LSI) and the mean patch area (AREA_MN), the landscape level comprises the ED, the LSI, the Contagion (CONTAG) and the Shannon’s diversity index (SHDI). The value of each landscape indices was calculated by Fragstats 4.2 software, and the ecological significance and calculation method of each index were detailed in the references [47,48].

5. ESV

   The equivalent coefficient method was the most representative method to quantitatively evaluate the ecosystem services functions. In this article, the ecosystem services types and the equivalent coefficient were determined based on the research results of Costanza et al. and Xie et al. [49,50]. Combined with the value calculation method of the equivalent coefficients proposed by Xie et al. [51], we obtained the ESV per unit area of different land use types in the Yellow River Basin (Table 2).

Figure 2

The research workflow of the attribution identification of terrestrial ecosystem evolution.

Figure 3

The framework of FLUS model and the process of land use simulation.

3.1 Land use simulation of different scenarios

The land use distribution and the proportions of different land use categories in the Yellow River Basin by 2015 under the four scenarios are shown in Figure 4. Compared with scenario 1 (Figure 4a), the forestland area in scenario 2 (Figure 4b) was decreased by 0.1%, the grassland area was increased, the water bodies and unutilized land were changed slightly, which indicated that climate change led to a decrement in the forestland and an increment in the grassland. The area of water bodies and construction land in scenario 3 (Figure 4c) and scenario 4 (Figure 4d) were consistent, with an increase of 1.3 and 20.6%, while the other land use types showed diverse trends. In scenario 3, the area of cultivated land and grassland were decreased by 2.0 and 0.3%, and the area of forestland was increased by 1.2%. In scenario 4, the area of cultivated land and grassland were decreased by 1.8 and 0.3% and the area of forestland was increased by 1.1%. These differentially showed that under the dual effects of climate change and human activities, the area of cultivated land was decreased sharply, while the forestland and construction land were increased drastically. Diverse climatic conditions led to different variations in forestland, cultivated land, grassland and unutilized land.

Figure 4

Spatial distribution and the proportions of land use types in the Yellow River Basin by 2015 under different scenarios. Where (a–d) represent the land use distribution under the scenario 1, scenario 2, scenario 3 and scenario 4, respectively.

3.2 Impacts of climate change and human activities on landscape patterns

1. The impacts on the type level

   In terms of type level, the landscape pattern indices of each land use types under different scenarios are shown in Figure 5. Compared with scenario 1, the ED and LSI were changed little in scenario 2, the NP of forestland and unutilized land were increased by 0.09 and 0.77%, the AREA_MN were decreased by 0.19 and 0.76%, the NP of grassland was decreased by 0.67% and the AREA_MN was increased by 0.70%. The results showed that the fragmentation of vegetation landscape was increased due to climate change. Scenario 3 and Scenario 4 showed a similar trend, the NP and AREA_MN of cultivated land as well as construction land were changed the most, the LSI of cultivated land, water bodies and construction land were increased significantly and the ED of grassland was decreased the most. These alterations showed that the dominant species of grassland landscape was decreased, the shape of cultivated land, water bodies and construction land became more irregular and the boundary length was decreased due to human activities.

2. The impacts on the landscape level

   At the landscape level, the landscape pattern indices of different scenarios are shown in Figure 6. Compared with scenario 1, landscape pattern indices of scenario 2 was changed slightly. Scenario 3 and Scenario 4 were displayed homologous change characteristics, in which the LSI was increased by 0.46 and 0.44%, the CONTAG was decreased by 1.71 and 1.70% and the ED and SHDI were changed slightly. The results indicated that the entirety landscape patterns in the Yellow River Basin were less affected by climate change, while the ecological function of the Yellow River Basin was impaired under the action of human activities, and the landscape patterns tended to be complicated, fragmented and dispersed.

Figure 5

The change rate of landscape patterns indices of type-level under different scenarios in the Yellow River Basin. Where FL, WL, GL, WB, CL and UL represent the farmland, woodland, grassland, water bodies, construction land and unutilized land, respectively. ΔS2–S1, ΔS3–S1 and ΔS4–S1 represent the discrepancies of scenario 1 with scenario 2, scenario 3 and scenario 4, respectively. The NP, ED, LSI and AREA_MN represent the number of patches, the edge density, the landscape shape index and the mean patch area, respectively.

Figure 6

The change rate of landscape patterns indices in landscape level under different scenarios in the Yellow River Basin, where ΔS2–S1, ΔS3–S1 and ΔS4–S1 represent the discrepancies of Scenario 1 with Scenario 2, Scenario 3 and Scenario 4, respectively. The ED, LSI, CONTAG and SHDI represent the the number of patches, the landscape shape index, Contagion and Shannon’s diversity index respectively.

3.3 Impacts of climate change and human activities on ESV

1. Changes of the ESV under different scenarios

   The ESV per unit area as well as the proportions of different ecosystems under four scenarios are shown in Figure 7. The total ESV of scenario 1 (Figure 7a) was expected to be 237.87 billion dollars by 2015. Compared with scenario 1, the ESV of scenario 2 (Figure 7b) was decreased by 0.03 billion dollars, the proportion of ESV in the grassland was increased and forestland was decreased. Scenario 3 (Figure 7c) and Scenario 4 (Figure 7d) were increased by 0.56 and 0.50 billion dollars, respectively. In Scenario 3, the cultivated land and grassland were reduced by 0.43 and 0.28 billion dollars, while the forestland and water bodies were increased by 0.72 and 0.55 billion dollars. In Scenario 4, the cultivated land and grassland were decreased by 0.40 and 0.31 billion dollars, and the forestland and water bodies were increased by 0.67 and 0.55 billion dollars. These discrepancies indicated that the ESV of forestland decreased due to the impacts of climate change. Under the combined influences of climate change and human activities, the ESV of forestland and water area were increased, while the cultivated land and grassland were decreased.

2. Changes in the values of different ecosystem services

   Figure 8 shows the changes in the values of different ecosystem services in the Yellow River Basin under four scenarios. Compared with Scenario 1, the regulating services such as erosion prevention were reduced in Scenario 2 and the other ecosystem services functions were changed small. The provisioning services in Scenario 3 and Scenario 4 were weakened, which reduced by about 0.3%. While the regulation services were improved significantly, especially the function of the climate regulation and regulation of water flows.

Figure 7

Spatial distribution and the proportions of ESV in the Yellow River Basin by 2015 under different scenarios. Where (a), (b), (c) and (d) represent the ESV under the Scenario 1, Scenario 2, Scenario 3 and Scenario 4, respectively.

Figure 8

Changes in the values of different ecosystem services in the Yellow River Basin under different scenarios, where ΔS2–S1, ΔS3–S1 and ΔS4–S1 represent the discrepancies of Scenario 1 with Scenario 2, Scenario 3 and Scenario 4, respectively.

4.1 Effects of climate change on landscape patterns and ecosystem services function

Since 1995, climate change was notable in the Yellow River Basin. The average temperature was increased by 0.4°C/10a, which was significantly higher than the growth rate in China (0.25°C/10a) [52]. The average precipitation was decreased by −2.1 mm/10a but presented the characteristic of large spatial-temporal difference. Concurrently, the frequency of extreme events was increased [4,40], such as the agricultural and ecological droughts. In accordance with the simulation results of scenario 1 and scenario 2, under the impacts of climate change, the forestland area was decreased by 0.1% and the grassland area was increased in the Yellow River Basin from 1995 to 2015, but these changes showed great spatial heterogeneity. In the upper reaches of the Yellow River, the forestland area was decreased by 0.2%, the grassland and unutilized land area were increased. While in the middle and lower reaches, the impacts of climate change were weakened, the forestland was reduced by 0.04%.

There were several reasons for these complex spatial–temporal changes. In the upstream high-altitude regions, the vegetation ecosystems were extremely sensitive and fragile in respond to climate variation. The rising temperature prolonged the growing season of vegetation and strengthened the photosynthesis, which improved the vegetation coverage, and led to the enhancement of ecosystem services function and landscape patterns [53,54]. While in the other arid and semi-arid areas, climate warming induced an increase in evaporation, coupled with the decreased precipitation, all of which aggravated the shortage of soil moisture and resulted in the decrease of vegetation coverage, then impaired the ecosystem services [24,55,56]. In addition, extreme drought events were occurred frequently. Studies have proved that prolonged drought may cause trees to die due to the lack of moisture for a long time and reduce the survival rate of trees [25,57]. According to Bulletin of Flood and Drought Disaster in China [58], there were five large-scale droughts in the Yellow River Basin during 1995–2015. Especially the North China Plain drought in 2009, with the features of long duration and wide range, which had serious impacts on the vegetation ecosystems. In conclusion, the vegetation ecosystem was damaged due to climate change in the Yellow River Basin, the fragmentation of vegetation landscape was intensified, which resulted in the reduction of ESV, and the ecosystem regulation services became weakened, such as the erosion prevention.

4.2 Effects of human activities on landscape patterns and ecosystem services function

During the study period, the population density in the Yellow River Basin increased from 121.11 people/km² (1995) to 149.62 people/km² (2015), and the GDP per unit area increased from 467.7 yuan/km² (1995) to 715.0 yuan/km² (2015). In line with the simulation results of scenario 1 and scenario 3, under the impacts of human activities, the area of forestland, water bodies and construction land were increased by 1.2, 1.3 and 20.6% while the cultivated land and grassland were decreased by 2.0 and 0.3%. However, these variations exhibited different characteristics in the upper, middle and lower reaches of the Yellow River Basin. In Qinghai, Gansu and other mid-upstream cities, the economic development was relatively backward, and with a low population concentration, while the soil erosion and ecological degradation were serious. The local government had implemented large-scale projects of the Grain for Green Project to protect the ecological environment, which resulted in the forestland area increased rapidly, the arable land and grassland were decreased [30,59]. According to statistics, since the implementation of the Grain for Green Project in 1999, the cumulative afforestation area reached 18.9 million hm². The ecological services functions such as the soil and water conservation in the mid-upstream area were significantly improved, but resulted in the singularization of ecological function and weakened the landscape diversity [32].

In Henan, Shandong and other mid-downstream cities, many large irrigation districts have been built due to the merits of population concentration, flat terrain and convenient irrigation, such as the Yellow River irrigation district, which resulted in a large number of grassland and water bodies converted into arable land [31]. Simultaneously, to cooperate with the population growth and economic development, a large number of cultivated land and ecological land were occupied by the construction land [27]. In addition, some reservoirs have been built for the purpose of irrigation and flood prevention, such as the Xiaolangdi Hydro Project (total storage capacity 12.65 billion m³), as well as some rubber dams have been built for restoring the aquatic ecosystem, such as the water surface engineering [60]; all of these measures have increased the area of water bodies area, which improved the ecosystem regulation of water flows function significantly. The expansion of arable land and construction land in mid-downstream impaired the ecosystem function, but these losses were offset by the increased water bodies and forestland. In general, the total ESV were increased by 0.56 billion dollars affected by human activities, in which the provisioning services function was weakened, and the regulating services and supporting services were improved, but the ecosystem services function tends to be unitary. In terms of landscape patterns, the ecological landscape patterns became more complex and irregular, landscape diversity became more weakened and the anti-interference ability decreased.

4.3 Relationship between landscape patterns and ESV

Studies have shown that there was a close relationship between landscape patterns and ESV. Landscape patterns changed the material circulation and the energy flow in landscape by affecting the types, area and spatial distribution of various ecosystems and then affected the supply and maintenance of ecosystem services function [61,62]. However, the emphases of the two analysis method were different. The landscape patterns index was an important quantitative indicator reflecting the landscape structure composition, spatial configuration and landscape heterogeneity, such as the landscape fragmentation degree and landscape anti-interference ability, which highly concentrated the landscape patterns information and described the spatially discontinuous data [10,47]. While the ESV was a monetary method to quantify the value of ecosystem services function and evaluate the production and living services provided by the ecosystem, including provisioning services, regulating services, supporting services and cultural services [48,49]. Therefore, our study selected the landscape patterns and ESV as the evaluation indicator, which would have a more comprehensive understanding of the impacts of climate change and human activities on the ecosystem of the Yellow River Basin.