Impact of forest fragmentation on river water quality: an example from a typical subtropical hilly basin

Study area

The upper Ganjiang River basin is located between latitudes 24°29′N and 27°09′N, and longitudes 113°54′E and 116°38′E (Fig. 1). This region is characterized by a surround formed by the Wuyi, Yu, Zhuguang, Jiulian, and Dayu mountains, resulting in a terrain that exhibits a notable elevation gradient, with higher altitudes predominating in the southern section and lower elevations in the northern part. The landscape is predominantly hilly and mountainous, encompassing approximately 83% of the entire basin. The hydrological system within the basin is characterized by a convergence of water flows, predominantly facilitated by the two major tributaries, the Zhangjiang and Gongjiang Rivers. Upon merging, these tributaries contribute to the formation of the Ganjiang River, which serves as a critical source of both production and domestic water for the inhabitants of southern Jiangxi, particularly in Ganzhou City, Zhanggong District. The basin is situated within a subtropical monsoon climate zone, exhibiting high temperatures and significant rainfall during the summer months, contrasted by cooler temperatures and lower precipitation levels in winter. The mean annual temperature is recorded at 19.8 °C, while the average annual precipitation is approximately 1,318.9 mm.

Water quality data sources

Water quality data for the 15 monitoring sites (Figs. 1A and 1B) located in the upstream region of the Ganjiang River were sourced from the National Surface Water Quality Dissemination System (www.cnemc.cn/en/) for the periods of January 2021 and July 2021. The study sites are located at the mouths of rivers and are representative of the state of water quality in the sub-basins to which they belong. To ensure seasonal representativeness, data collection involved the selection of four distinct days in both January and July, with one day chosen per week, followed by averaging the values for each water quality indicator (monitoring station water quality data displayed in Table S3). The indicators assessed in this study encompass total phosphorus (TP, mg·L⁻¹), total nitrogen (TN, mg·L⁻¹), permanganate index (PI, mg·L⁻¹), chemical oxygen demand (COD, mg·L⁻¹), five-day biochemical oxygen demand (BOD₅, mg·L⁻¹), and ammoniacal nitrogen (NH₃-N, mg·L⁻¹). For comprehensive information regarding the detection methodologies employed, please refer to the document titled “Water and Wastewater Monitoring and Analysis Methods” (Committee, 2002). In addition, the water quality evaluation methodology used in this study is described in the Appendix.

Forest fragmentation data sources

Data pertaining to the basin and river systems were derived from the Hydrosheds products (https://www.hydrosheds.org/products), the 15 sub-basins were as shown in Fig. 2. The analysis of forest fragmentation utilized land use data with a resolution of 30 m for the year 2021 (https://zenodo.org/records/12779975). All sub-basins are dominated by forest land and cropland, and their forest cover has reached 69.34% to 92.58% (Table S4). Six landscape metrics was selected, including patch density (PD) and mean patch area (MPA), which represent the degree of landscape fragmentation and patch size, respectively; edge density (ED) and interspersion juxtaposition index (IJI), which reflect patch boundary characteristics and the relationships of spatial configuration; and the landscape division index (DIVISION) and effective mesh size (MESH), which measure overall landscape connectivity and fragmentation. These metrics allow for a more comprehensive understanding of the degree of forest landscape fragmentation (Lin et al., 2024; Ma et al., 2023; Pfeifer et al., 2017), which are described in detail as Table 1.

Statistical analysis

The Canadian Council of Ministers of the Environment Water Quality Index (CCMEWQI) was employed to assess the water quality in the upper reaches of the Ganjiang River, in accordance with the Class III surface water quality standards of China (where the safety thresholds are defined as PI < 6, COD < 20, BOD₅ < 4, NH₃-N < 1, TP < 0.2, and TN < 1, the exceeding these thresholds indicates that the surface water is unsuitable for drinking). Detailed calculation and evaluation methodologies are provided in Part 1 of the Appendix. To ensure the rigor of our analysis, the normal distribution on all datasets was tested and transformed the data that did not conform to the normal distribution. Furthermore, t-tests were employed to examine significant differences in water quality indicators across varying seasons. Redundancy analysis (RDA) as a methodological tool for elucidating the complex relationship between landscape metrics and water quality highlights its established efficacy within this domain of research (Clément et al., 2017; Wu et al., 2024; Xiao et al., 2016). Prior to the RDA analysis, the water quality indicators underwent Detrended Correspondence Analysis (DCA), revealing that the lengths of all four axes were less than 3, thereby validating the applicability of RDA to this dataset (Wu & Lu, 2021). RDA was analyzed with CANOCO 5.0 software (Microcomputer Dynamics, Inc., Metairie, LA, USA). The nonparametric change-point analysis (nCPA) was used to explore thresholds for landscape forest fragmentation metrics that lead to abrupt water quality changes. Statistically significant (p < 0.05) change points identified by the nCPA were determined by a permutation test (Bootstrap test method) and 95% confidence intervals for the change points were estimated to indicate that abrupt changes in nitrogen concentrations were associated with landscape fragmentation thresholds rather than random fluctuations (detailed calculations were in Part 2 of the Appendix). The above statistical analysis methods are all completed using R language, except RDA.

Information and evaluation of water quality in the upper Ganjiang River basin

The basic information, spatial distribution characteristics and water quality evaluation of the six water quality indicators in the upper Ganjiang River basin were shown in Table 2, Figs. 3 and 4, respectively. It was found that there were significant differences in BOD₅ and TN concentrations between summer and winter, while no differences in PI, COD, NH₃-N and TP (Table 2). The spatial distribution of the water quality indicator showed that the concentrations of PI, COD, BOD and TP at all stations were below the limits of China’s Class III surface water quality standards in both winter and summer (green and blue points, see Appendix for detailed description). Concentrations of NH₃-N at the S2 monitoring site exceeded the limits of China’s Class III surface water quality standards during the winter. For TN, the concentrations at S1 and S11 in summer also exceeded the limits of China’s Class III surface water quality standards, and the exceedance was more serious in winter, with some concentrations exceeded the limits of China’s Class III surface water quality standards at some sites (S1, S2, S6, S9, and S11) (Fig. 3).

The evaluation of the Canadian Water Quality Index (CCWEWQI) found that the water quality at most of the monitoring sites in the region was generally at good and excellent levels (Figs. 4A and 4B), while the water quality at the S4 sample site (located in the urban area of Ganzhou City) was at fair level only in the winter (Fig. 4D).

Forest landscape fragmentation metrics

The basic information of forest landscape fragmentation metrics in the sub-basins represented by 15 monitoring sites in the upper Ganjiang River basin was presented in Table 3 and Fig. 5. The results indicated that the mean values of PD, ED, MPA, IJI, DIVISION, and MESH for the forested landscapes in the 15 sub-basins were 1.27/100 ha, 31.67 m/ha, 85.17 ha, 21.90%, 0.46, and 241,077.01 ha. The maximum value of PD of the forest occurred at S15 (2.27/100 ha) and minimum value at S4 (0.40/100 ha). The maximum ED value was found at S13 (49.08 m/ha) and minimum value at S4 (15.51 m/ha). The maximum and minimum MPA values were found at S4 (233.06 ha) and S15 (30.49 ha), respectively. The minimum value of IJI was found at S8 (7.57%) and the maximum value was found at S11 (28.82%). The minimum value of DIVISION was found at S4 (0.15) and the maximum value was found at S12 (0.82). Most of the monitoring sites had MESH values below the mean, with a median of only 102,012.14 ha, and only five sites had higher MESH values (S1, S2, S3, S5, S9).

Relationship between forest landscape fragmentation metrics and water quality

RDA was used to explore the effects of forest fragmentation metrics on overall water quality indicators (Table 4 and Fig. 6). Six forest fragmentation metrics accounted for over 40% of the variance in overall water quality, demonstrating a more pronounced explanatory capacity during the winter season (55.47%) compared to summer (41.21%) (Table 4). This difference highlights the critical role of forest fragmentation in influencing water quality in different seasons.

In summer, MPA and MESH of forest were negatively correlated with most of the river water quality indexes, PD, ED and DIVISION were positively correlated with most of the river water quality indicators, while IJI was only positively correlated with TP, PI and BOD5, and negatively correlated with TN, NH3 and COD. MESH contributed the most (20.9%), followed by DIVISION (20.4%) and IJI the least (6.9%). In winter, IJI, ED, PD and DIVISION were positively correlated with most water quality indicators, while similar to summer, MESH and MPA were negatively correlated with most water quality indicators. IJI had the highest contribution (44.9%) and the lowest contributing fragmentation metrics was MPA (3.9%) (Fig. 7).

Abrupt change

River water quality was found to be primarily influenced by TN concentrations in previous water quality assessments (Fig. 3). Therefore, nCPA analyzed the thresholds at which forest landscape fragmentation metrics lead to abrupt changes in TN concentrations (Fig. 8). It was found that the values of PD, ED, IJI, DIVISION and MESH were consistent in winter and summer when the probability of abrupt change in TN concentration reached 100%, that is to say they were higher than 1.83/100 ha, 39.4 m/ha, 28.1%, 0.77 and 7.89e+0.5 ha, respectively. However, there was a difference in the thresholds at which MPA led to abrupt changes in TN concentration in winter and summer, with winter (>143.8) the threshold was higher than in summer (>128.9).

Water quality assessment analysis in the upper Ganjiang River basin

The quality of river water is closely related to the health of the people in the basin and the sustainable development of production and life (Ibáñez & Peñuelas, 2019; Ma et al., 2020; Pfeifer et al., 2017). The CCWEWQI evaluation method was utilized to analyze the overall status of water quality at various monitoring sites (Fig. 4). The results found that the river water quality in the upper Ganjiang River Basin was generally at a good level (most of the monitoring sites had good and excellent water quality), which is similar to the results of previous studies (Chen et al., 2021; Peng, Mingdong & Shaonan, 2023; Song et al., 2024). However, the water quality at monitoring site S6 was assessed as fair in winter, which may be due to the fact that this monitoring site is located in the suburban area of Ganzhou City. On the one hand, there are frequent agricultural activities in the suburban area, and the excessive use of chemical fertilizers and pesticides can lead to the entry of these chemicals into the water body through surface runoff or groundwater infiltration, resulting in eutrophication and other pollution problems; on the other hand, with the expansion of the city and the increase of population in the suburban area, the discharge of domestic wastewater also increases. If sewage treatment facilities are inadequate or have insufficient capacity, the direct discharge of domestic sewage into water bodies can lead to water quality pollution. For individual water quality indicators, the concentrations of PI, COD and BOD₅ in the river of the upper Gangjiang River basin were all within safe limits (not exceeding the Class III water quality standard) (Fig. 3). The concentrations of these three indicators can directly or indirectly represent the concentration of organic matter in the river, and the lower their concentrations, the lower the content of organic matter in the river. Generally speaking, the sources of organic matter in rivers include three main aspects, which are domestic sources (e.g., detergents), agricultural sources (e.g., fertilizers, pesticides, agricultural fertilizers), and industrial sources (e.g., paper, tanneries, petrochemicals) (Aguilar-Torrejón et al., 2023; Singh, Singh & Singh, 2021). The low river organic matter in this watershed may be due to the low organic matter content of the red soil, which does not allow for the input of excess organic matter into the river, in addition to the small area of cultivated land and the low input of organic matter from agricultural sources. Nitrogen and phosphorus are important indicators of eutrophication in water bodies (Akinnawo, 2023). The concentrations of TP were low at the monitoring points in this study, again less than the Class III water quality standard. However, high nitrogen concentrations are common for the region, especially in winter, for example, all five monitoring points (S1, S2, S6, S9, and S11) exceeded the Class V water quality standard, of which three monitoring points (S1, S2, and S6) were mainly located within the urban area of Ganzhou City, and thus focused on the TN (production and domestic sources) inflow around the monitoring points, in addition to receiving the TN inflow from the upper Ganjiang River basin. This suggested that TN was a key factor affecting river water quality in this study area, and for this reason, we specifically focused on TN as a research object to analyze the thresholds for abrupt changes in TN concentrations due to forest landscape fragmentation.

Impacts of forest landscape fragmentation metrics on water quality

Generally, forests play a highly positive role in purifying river water quality. Both trees and shrubs can significantly contribute to the interception of runoff pollutants, and they can also absorb contaminants, including nutrients, through their root systems (Ellison et al., 2017; Qiu et al., 2023). However, with the increasing intensity of human activities, forests are gradually becoming fragmented (Fischer et al., 2021; Liu et al., 2019). Research indicated that the fragmentation of forests in most temperate and subtropical regions—primarily in northern Eurasia and South China—has declined between 2000 and 2020 (Ma et al., 2023). The fragmentation of forest landscapes (e.g., ED, PD) has been widely recognized as a contributing factor to the deterioration of river water quality (Clément et al., 2017; Qiu et al., 2023). Therefore, this study delves into the impact of forest landscape fragmentation on the water quality of the upper Ganjiang River basin, which is characterized by high forest cover (>76%), during both winter and summer seasons.

Research indicated that six forest landscape fragmentation metrics can explain 41.21% of the variability in river water quality during the summer; however, the explanatory power increases by 14.26% for winter water quality changes (Table 4). This difference is primarily attributed to local climate and topography. The upper Ganjiang River basin is characterized by a subtropical monsoon climate, with high temperatures, abundant rainfall, and frequent storms during the summer (He & Liu, 2016). In a fragmented forest landscape, the interception capacity of the forest is significantly reduced, and the steep hilly terrain may exacerbate the entry of surface pollutants into the river. Conversely, during the winter, precipitation is lower, and the surface runoff’s erosion capacity is far less than in summer (Liu, Zhang & Zhang, 2018; Sun et al., 2024).

In this study, MESH was negatively correlated with most river water quality indicators, and its contribution is 20.9% in summer and 16.6% in winter (Figs. 6 and 7). The MESH reflected the size of the area of unblocked patch types in the landscape. The larger the MESH value, the more connected the patch, the lower the degree of fragmentation, and the better the ecological connectivity (Fischer et al., 2021; Lin et al., 2024). It is shown that the increase of forest MESH value is beneficial to river water quality in the upper reaches of Ganjiang River basin, which is rarely found in related studies. The difference in the MESH value between winter and summer may be due to the size of the canopy. In summer, the forest is rich in foliage, and the canopy of trees is higher than that of winter (Primicia et al., 2013). IJI was only positively correlated with TP, PI and BOD₅ in summer water quality indicators (the contribution rate was 6.9%), while it was positively correlated with most river water quality indicators in winter (the contribution rate was as high as 44.9%) (Figs. 6 and 7), indicating that IJI had a higher impact on winter water quality. Generally speaking, the higher the IJI value of the forest, the higher the degree of interleaving between the land use types in space, and the forest is disturbed by other land use, and the boundary is more complex (Wu & Lu, 2021). Different from this study, Wu & Lu (2021) showed that IJI of riparian buffer forest was negatively correlated with water quality indicators in dry season, which may be related to the difference of water quality parameters and research scale. In the two seasons, PD, ED and DIVISION were positively correlated with most of the water quality indicators, indicating that the higher their value, the greater the adverse impact on river water quality (Figs. 6 and 7). This result was also reflected in previous studies (Clément et al., 2017; Shehab et al., 2021; Zhang et al., 2022a). In general, forest fragmentation has a profound impact on the water quality of rivers in the region, and it is extremely important to strengthen the management of forests to avoid excessive disturbance.

Abrupt change analysis and management recommendations

Abrupt change analysis was conducted to investigate the key thresholds of six forest landscape fragmentation metrics that led to abrupt changes in TN concentrations of the upper Ganjiang River basin. It was found that the thresholds of five forest landscape fragmentation metrics, namely PD, ED, IJI, DIVISION and MESH, leading to abrupt changes in TN concentration in both seasons (winter and summer) were consistent, which indicated that there was no obvious seasonality in the thresholds of these variables affecting TN abrupt changes (Fig. 8). In this study, MESH and DIVISION of forests were the top contributors to the summer water quality change with 20.9% and 20.4%, respectively, but their impacts on water quality were completely opposite, and an increase in the value of MESH favored an increase in the river water quality, while an increase in the value of DIVISION led to a deterioration in the river water quality. Therefore, based on the thresholds for the mutation analysis, it is recommended that in the Upper Ganga River basin forest should exceed 7.89e+0.5 ha for MESH and on the contrary, the value of DIVISION should be controlled within 0.77. In winter, the IJI contribution of forests amounted to 44.9%, which had a much greater impact than the other five forest fragmentation metrics, and since it was positively correlated with most of the water quality indicators, it is necessary to recommend that the IJI value of forests in the upper Ganjiang River basin must be less than 28.1%. Given that the thresholds for MESH, DIVISION and IJI have no seasonal effect, the thresholds are applicable in both winter and summer. Although nonparametric change point analysis (nCPA) identifies fixed thresholds for landscape fragmentation metrics, it was recognized that these thresholds may oversimplify the reality of changes in water quality. Changes in water quality are often influenced by a combination of factors, including climatic conditions, land use types, and human activities. Therefore, fixed thresholds should be viewed as an initial reference rather than an absolute standard of judgment.