Next Article in Journal
Land Cover Trajectories and Their Impacts on Rainfall-Triggered Landslide Occurrence in a Cultivated Mountainous Region of Western Japan
Next Article in Special Issue
Water and Sediment Chemistry as Drivers of Macroinvertebrates and Fish Assemblages in Littoral Zones of Subtropical Reservoirs
Previous Article in Journal
Research on the Mechanism and Evolution Law of Delayed Water Inrush Caused by Fault Activation with Mining
Previous Article in Special Issue
Analyzing the Akinete Protein of the Harmful Freshwater Cyanobacterium, Dolichospermum circinale
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Water
Volume 15
Issue 24
10.3390/w15244210
water-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Effect of Land Use on Stream Water Quality and Biological Conditions in Multi-Scale Watersheds

by
Jong-Won Lee
,
Se-Rin Park
and
Sang-Woo Lee
*
Department of Forestry and Landscape Architecture, Konkuk University, Gwangjin-gu, Seoul 05029, Republic of Korea
*
Author to whom correspondence should be addressed.
Water 2023, 15(24), 4210; https://doi.org/10.3390/w15244210
Submission received: 30 October 2023 / Revised: 4 December 2023 / Accepted: 5 December 2023 / Published: 6 December 2023
(This article belongs to the Special Issue Ecological Monitoring and Assessment of Freshwater Ecosystems: New Trends and Future Challenges)
Browse Figures
Review Reports Versions Notes

Abstract

:
Understanding the relation between watershed land use and stream conditions is critical for watershed planning and management. This study investigated the effects of land use on stream water quality and biological conditions in sub-watersheds and micro-watersheds across the Han River watershed in South Korea. We developed random forest models for each water quality and biological indicator using the proportions of urban, agricultural, and forested areas. Our results indicate that water quality and biological indicators were significantly affected by forest area at both scales, and the sub-watershed models performed better than the micro-watershed models. Accumulated local effects were used to interpret the effect of each explanatory variable on the response variable. The plots for water quality and biological indicators with proportions of watershed land use demonstrated similar patterns at both scales, although the relation between land use and stream conditions was slightly more sensitive in micro-watersheds than in sub-watersheds. Urban and agricultural areas showed a lower proportion of water quality and biological condition variability in the micro-watersheds than in the sub-watersheds, while forests showed the opposite results. The findings of this study suggest that different spatial scales should be considered when developing effective watershed management strategies to maintain stream ecosystems.

1. Introduction

River and stream ecosystems are under pressure from various anthropogenic activities over multiple spatial scales [1]. In particular, anthropogenic land-use changes in watersheds are major factors hindering the integrity of river and stream ecosystems, including water quality and aquatic organism health. The negative effects and complex processes associated with urban and agricultural areas in watersheds cause changes to river and stream ecosystems at various spatial scales. To respond to these negative changes in stream ecosystems, it is necessary to understand the impact of various land uses in watersheds on stream ecosystems. Therefore, the impacts of land use in watersheds on water quality and aquatic organisms have been studied extensively.
In general, urban areas are the source of many anthropogenic pollutants that negatively affect streams [2,3]. Chemical pollutants, such as excess nutrients, heavy metals, and organic compounds, flow into stream ecosystems from urban areas due to increases in impermeable surfaces, such as pavements and rooftops [4,5]. Runoff with high concentrations of pollutants from industrial facilities or sewage treatment plants also acts as a stressor for aquatic organisms. The influx of these artificial pollutants destabilizes the physical and chemical runoff processes and disrupts the stream ecosystem [6,7,8]. For example, although nutrients can benefit aquatic organisms, heavy metals can have negative effects, and the combined effects of these stressors can disrupt ecosystem mechanisms [9]. Agricultural areas are known to promote eutrophication and harmful algal blooms in streams owing to excessive fertilizer runoff, sediment influx, and agricultural landscapes [10,11]. Conversely, forest areas and riparian vegetation in the watershed improve the stability of stream channels, regulate water temperature by preventing light from penetrating the canopy, and provide habitat and shelter for aquatic organisms. They also reduce nutrient runoff from the watershed, thereby mitigating erosion and eutrophication and improving biodiversity and ecosystem function by improving riparian vegetation ecosystem conditions and habitat quality [12,13].
The relationship between land use and river ecosystems is also related to the watershed size. Several studies have reported that the impact of land use on stream water quality and aquatic organisms varies across different spatial scales [14,15,16,17]. For example, the parameters on the riparian or reach scale may predict impacts on water quality and aquatic organisms better than those on the catchment scale. Dala-Corte et al. [18] and Pan et al. [19] reported that the land use adjacent to streams had a more important effect on chemical water quality and biotic communities compared to that observed from land use on the catchment scale across grasslands in southern Brazil and the Willamette Valley Ecoregion in Oregon, USA. Assessing land-use impacts on streams within these watersheds can allow us to more accurately estimate causal pathways in smaller watersheds, which are potentially vital for biodiversity and stream health [20]. However, studies have reported that watershed-scale land-use parameters better account for biotic diversity and water quality, including aquatic community patterns and species composition [7,17]. Tudesque et al. [21] explained that in the Adour–Garonne River in southwest France, the watershed scale is an important determinant of biological community structure. Ding et al. [14] and Zhang et al. [22] reported that the impact of land use on water quality in the Dongjiang River Basin and Three Gorges Reservoir area in China is better explained on the catchment scale than on the riparian, reach, or buffer scales. In addition, large-scale watersheds are more spatially dependent on land use due to the direct impact of the physical and chemical water quality and the cumulative anthropogenic influence on stream ecosystems related to water quality and aquatic organisms [15,21,23].
Recently, ecosystem research has used machine learning models to replace traditional statistical models that were previously built as description-driven limited models [24]. Machine learning, a branch of artificial intelligence, effectively overcomes the limitations of data-dependent bivariate and multivariate statistical methods, enabling prediction-driven models to estimate highly predictive models [8,25]. Unlike the general linear model, which traditionally predicts variables, tree-based random forest (RF) models make no assumptions about data distribution. They also integrate missing values with numeric or categorical predictors, tolerate general changes, such as the size of the data, remain resilient to predictor outliers [26], and perform well against complex and nonlinear interactions between variables [26]. The biggest advantage is that the prediction accuracy generally surpasses that of conventional methods. Another advantage of these RF models is the possibility of revealing complex nonlinear relations between land cover characteristics and stream water quality. Park et al. [8] found that these models accurately depict the complex nonlinear relation between landscape characteristics and stream water quality. Further benefits include the accuracy of machine learning models tested on new datasets and their application in predicting characteristics of aquatic ecosystems, such as the stream water quality. Moreover, such models can support future land-use planning scenarios to aid in policy decisions. Previous studies effectively employed the RF model to analyze the impact of land use on water quality and biological integrity. The relationships among stream water quality, aquatic organisms, and land use often exhibit disproportionate and nonlinear patterns, indicating abrupt points of change (thresholds). For example, significant changes in fish community composition and species abundance of 10.9% and 17.5%, respectively, were observed in a southeastern Queensland watershed [27]. These studies identified the point at which changes occur in the impact of land use, revealing an abrupt nonlinear ecological threshold [27,28,29]. Thresholds are needed to protect or restore water quality and aquatic organisms amid continuous land-use changes [21,29,30]. When ecological thresholds are surpassed, such as by pollutants flowing into streams, stream ecosystems can be damaged. Identifying these thresholds can help determine the impact of land use on streams and facilitate land-use planning [27].
This study aimed to quantify the relative importance and effects of the spatial scale of land use on the water quality and health of aquatic organisms in the Han River Basin. RF models were developed to evaluate the relations between land use and water quality and biological indicators. Most related studies have focused on forecasting changes and determining explanatory power according to spatial scale to identify the relation between land use and stream water quality and biological indicators [11,30]. In particular, the accumulated local effect (ALE) is used to compensate for the shortcomings of existing partial dependency plots (PDPs) and to display the results of the RF model more clearly. This can be a powerful tool for demonstrating the relation between highly correlated water quality and biological indicators. Our study advances the existing approach by analyzing differences in the impact of land use on water quality and aquatic organisms at multiple spatial scales and by comparing and analyzing the thresholds of land use according to different watershed scales. The results of this study can reduce uncertainty and provide insights for effective decision making in developing contaminated watershed management strategies and restoration plans. Different spatial scales should be considered when developing effective watershed management strategies to sustain stream ecosystems. This work is expected to help establish criteria for estimating land-use proportions to meet the health goals for stream ecosystems in watershed management.

2. Materials and Methods

2.1. Study Area

The Han River, located at 126°–129° E and 36°–38° N, originates from Taebaek Mountain and flows into the Yellow Sea. It is the second largest river in Korea, spanning a length of 494 km, and has a basin area of approximately 35,770 km2. Spring (March–May) and autumn (October–November) are generally sunny and dry owing to the influence of mobile cyclones. Summer (June–September) accounts for two thirds of the total annual precipitation, with high temperatures due to the influence of the North Pacific high pressure. Winter (December–February) is cold and dry owing to the influence of temperate cyclones. Overall, the Han River basin comprises highlands over 1000 m above sea level on the east and lowlands on the west [31], with an average annual precipitation of 1208.3 mm. The main type of land cover comprises forests, occupying approximately 68% of the land, while urban and arable land account for approximately 17% and 7%, respectively. In this study, the Han River basin was divided into large- and small-scale watersheds (Figure 1). The large-scale watershed unit (3rd–4th-order streams) was determined by the Ministry of Environment for water management, encompassing an area ranging from 38.96 km2 to 447.87 km2. The small-scale watershed unit (1st–2nd order streams) was defined by local governments based on stream management goals and ranges from 0.11 km2 to 12.97 km2.

2.2. Data Source and Preprocessing

2.2.1. Water Quality and Biological Indicators

In Korea, the National Aquatic Ecology Monitoring Program (NAEMP) has developed evaluation standards and sampling protocols for monitoring rivers and streams. This program evaluates the overall ecological health status of streams, including the chemical water quality, biological condition, habitat quality, riparian vegetation, and land use around the stream [32]. For stream evaluation, the NAEMP investigates the ecological health and water quality of the five major streams and tributaries in Korea twice a year (spring: April–May; autumn: October–November) using aquatic organisms. The ecological health of streams was evaluated using the Trophic Diatom Index (TDI), Benthic Macroinvertebrate Index (BMI), and Fish Assessment Index (FAI) based on biomonitoring at 3035 sites nationwide. In addition, the Riparian Vegetation Index and Habitat and Riparian Index were included to evaluate the riparian vegetation and habitat. Table 1 shows the equations used to estimate the biological indicators used in the NAEMP. The NAEMP also measures various water quality parameters at the biomonitoring sites. Biochemical oxygen demand (BOD5), ammonia/ammonium (NH3-N), nitrate (NO3-N), total nitrogen (TN), total phosphorus (TP), phosphate (PO4-P), and chlorophyll-a (Chl-a) were measured. Water quality was analyzed using an official test method conducted by the Korean Ministry of Environment. In this study, the biological indicators TDI, BMI, and FAI were used.
In this study, biological indicators and water quality were measured in the same manner across both watershed scales. The data used were obtained from basic environmental survey projects conducted by the Han River Watershed Environmental Management Office between 2018 and 2022 to diagnose the health status of streams. Table 1 shows the equations used to calculate the values for biological indicators.

2.2.2. Land Use/Land Cover

The land-use proportion of the stream watershed was determined using a land cover map provided by the Environmental Geographic Information Service. This land cover map had a spatial resolution of 5 m, and a 1:25,000 topographic map was used for the analysis in ArcGIS 10.6.1. The same land cover maps were applied to both the large- and small-scale watersheds. These maps were used to estimate the effects of land use on water quality and biological indicators by extracting urban, agricultural, and forest areas, which are the major land uses in the watershed, from the seven classified categories. Urban areas included residential, industrial, commercial, cultural, sports, leisure, transportation, and public facilities. Agricultural areas included cultivation facilities, orchards, and other croplands, while forested areas included broadleaf trees, conifers, and mixed forests. The impact of land use was assumed to remain constant throughout the study period.

2.3. Statistical Analysis

2.3.1. Independent Two-Sample T-Test and Pearson’s Correlation

In this study, three analytical methods were used to estimate the impact on stream ecosystems on the watershed scale. Initially, an independent two-sample t-test was performed to verify the statistical differences between large- and small-scale watersheds. Actual differences in the mean values for water quality (BOD, TN, and TP), biological indicators (TDI, BMI, and FAI), and proportion of land use (urban, agricultural, and forest areas) across scales were identified. Furthermore, a correlation analysis was performed before applying the RF model to determine the impact of land use on water quality and biological indicators in both watersheds. In particular, PDPs, which are often used in RF model analysis, were affected by correlations between factors. If the correlation between variables is strong, then a bias may occur in the interpretation of the PDP [33]. Therefore, a correlation analysis was performed before visualizing the relations between the variables based on the RF model. Correlations were calculated for each water quality variable, biological indicator, and land-use proportion based on Pearson correlation analysis. The correlation coefficient ranges from −1 to +1. A value of 0 indicates that there is no statistical association between the variables, a value closer to +1 indicates a strong positive association, and a value closer to −1 indicates a strong negative association [34]. The parameters used in this study exhibited a normal distribution, except for the urban proportion. However, since this study investigates threshold values of water quality and biological indicators according to the actual urban proportion, this proportion was used without conversion.

2.3.2. RF Regression Algorithm

The RF regression algorithm was used to predict and evaluate the relation between predictor variables (land-use proportion) and response variables (water quality and biological indicators) using non-parametric ensemble machine learning. The goal of RF regression is to create a real tree-like ensemble from which multiple regression trees are built to produce a regression [35,36,37].
The RF model derives the final prediction result using the average of the prediction results from each tree model through multiple decision trees. In the model training process, which is based on bagging (bootstrap aggregation), several decision tree models are gathered to form a forest, and the model is trained [35,36]. Bagging applies bootstrapping to extract the training dataset by allowing duplication in the entire dataset so that each tree model uses a different training dataset. It is a prediction method that summarizes the results of each model by making the size as large as the number of original data [37]. A prerequisite for improving the ensemble model’s performance is to secure diversity through bagging and assuring randomness by dividing it into random subspaces. It considers only these variables by randomly selecting fewer variables than the number of original variables. Random predictor selection reduces the correlation between trees and variance and, in particular, depends only on the number of predictors selected by the model user [38,39]. The RF model considers the mean and variance of the out-of-bag error difference between the out-of-bag error of the original dataset and a random mixture of the values of a particular variable. This identifies the importance and choice of variables to be used in the model. The variable importance in RF determines the extent to which each variable contributes to the accuracy and node impurity improvement. In this study, when the accuracy of the tree constructed via changing the order of specific variables was reconstructed, it was assessed as the mean squared error, which is the average of the differences in the reduced accuracy. To optimize the RF model, the numbers of regression trees and input variables per node were adjusted [39,40]. The number of regression trees was 200, and 1/3 of the total number of variables were optimized via setting them as random sampling variables. In this study, the data were analyzed via division into training and test data—70% and 30%, respectively. The model was trained on the training data and then the root-mean-square error and the mean absolute error were used to assess the accuracy of the model performance.
RFs do not yield single trees that can be graphed; however, the results can be expressed as PDPs or ALEs [29]. PDPs are commonly used to visualize the effects of predictors in black-box supervised learning models. This parameter shows the marginal effect on the prediction result of the machine learning model, and the relationship between the target and a feature can be determined via plotting this effect. In addition, it can be used to graphically characterize the relation between the probability of existence of one predictor after averaging the effects of other predictors in the model [41]. Therefore, the PDP provides a graphical depiction of the marginal effect of a variable on the regression [42]. Moreover, PDPs show the average effect of the features on the predictions and can be clearly interpreted if there is no correlation. They are also easy to implement in a plot and can analyze the relationships between features and predictors [33,43]. Thus, many studies using RF models have employed PDPs [8,44,45]. However, the most vulnerable PDP assumes that there is no correlation between a feature and other features. If there is a correlation among the features, then new data will be generated in the distribution regions where the actual probability is low, and the result will differ from the real values. Therefore, the ALEs can be selected as an unbiased choice for the PDP when the features are correlated. In addition, the ALEs visualize the main effects of individual predictors in the black-box supervised learning model and explain how each characteristic value averages and affects the prediction of the model [46]. Moreover, they operate conditionally instead of constraining the distribution, thereby solving the problem of independence. Because a correlation was observed among water quality, organisms, and land use in this study (Table 2), the ALE was used [43,46]. The RF regression model and ALEs were constructed using the RF ALE Plot packages in R Studio [46,47].

3. Results

3.1. Descriptive Statistics and Independent Sample T-Test Analysis

Table 2 and Table 3 describe the water quality variables, biological indicators, and land-use proportions used in this study. The average BOD, TN, and TP concentrations were lower in the large watersheds compared to those in the small watersheds. However, the maximum BOD and TN concentrations were higher in the small watersheds compared to those in the large watersheds. In both large and small watersheds, urban areas had a relatively low average proportion compared to that in farmland and forests, with forests showing a relatively high proportion. As for biological indicators, the TDI, BMI, and FAI values were better in large watersheds compared to those in small watersheds. According to the Ministry of Environment’s aquatic ecosystem health evaluation, the average TDI, BMI, and FAI had grades of C, B, and B for large watersheds, respectively, and C, B, and C for small watersheds, respectively.
Regarding aquatic ecosystem health, grades A and B suggest the presence of relatively abundant trophic diatoms sensitive to nutrients, many benthic macroinvertebrate species vulnerable to organic pollution, and a higher abundance of fish species sensitive to environmental changes. Grades D and E indicate the opposite, with the Ministry of Environment classifying streams with these grades as impaired. Grade C signifies a normal stream that is neither healthy nor damaged.
The independent sample t-test results for the differences in water quality, biological indicators, and land-use proportion according to the watershed scale are shown in Table 4. Water quality (BOD, TN, TP), biological indicators (TDI, BMI, FAI), and land use (agricultural, forest) presented statistically significant differences according to the watershed scale.

3.2. Correlation Analysis

Water quality and biological indicators were correlated with land-use properties in multiple watersheds (Figure 2). In the large- and small-scale watersheds, the relations between all variables were found to be significant. Correlations between land-use proportion, water quality, and biological indicators in the large-scale watersheds were greater than those in the small-scale watersheds. In urban and agricultural areas, a positive correlation was observed with the water quality parameters BOD, TN, and TP, while a negative correlation was observed with the biological indicators TDI, BMI, and FAI. In particular, in the large- and small-scale watersheds, forest areas showed a stronger correlation between water quality and biological indicators than that for other land uses/land covers.

3.3. RF Models for Water Quality and Biological Indicators in Large- and Small-Scale Watersheds

RF models were created for each watershed scale for the water quality and biological indicators, and the performance of each model was compared (Table 5). Overall, better root-mean-square error and mean absolute error values were observed for the large watersheds, and better TN values were only observed in small watersheds.

3.4. Analysis of the ALE Plots

The ALE plot shows the effect of the land-use proportion on the water quality prediction probability (Figure 3). In large-scale watersheds, the water quality concentration increased sharply when the urban proportion was approximately 10% or higher. In agricultural areas, the BOD and TN concentrations were likely to increase when the agricultural proportion was approximately 5–10% or higher. However, the TP concentration was likely to increase when the agricultural proportion was approximately 25% or higher. In small-scale watersheds, the water quality concentration increased when the urban proportion was in the range of 5–10% or higher, and the agricultural proportion was in the range of 20–30% or higher (Figure 4). This means that the threshold proportion for water quality may vary depending on the watershed scale. The likelihood of water quality deterioration gradually decreased as the proportion of forests increased on both scales.
Figure 5 and Figure 6 illustrate the impact of the land-use proportion on the predicted probabilities of TDI, BMI, and FAI. In large-scale watersheds, all indicator values decreased when the urban proportion was approximately 5% or higher. In agricultural areas, the TDI decreased by more than 25% and the BMI and FAI populations decreased by approximately 5%. In small-scale watersheds, the TDI was likely to decrease when the urban proportion was greater than 3%, and BMI and FAI were more likely to increase when the proportion was greater than 10%. In agricultural areas, the TDI increased nonlinearly up to approximately 30% and then decreased. The BMI decreased when the agricultural proportion was over 10%, and the FAI increased and decreased nonlinearly until the agricultural proportion reached approximately 30% and then decreased. In large-scale watersheds, the forest proportion gradually increased over the 25–35% range, and in small-scale watersheds, the forest proportion increased over 40–50%.

4. Discussion

4.1. Land-Use Thresholds for Water Quality and Biological Indicators

The stream water quality and biological indicators responded nonlinearly as the proportion of land use increased, indicating critical values. Identifying the critical points at which water quality and biological indicators respond to changes in the proportion of land use in a watershed is important for land-use planning. Therefore, many studies have identified associated ecological thresholds [48,49].
This study visualized the nonlinear relationship of land-use proportions with water quality and biological indicators using ALE main effect plots (Figure 3, Figure 4, Figure 5 and Figure 6). The ALE main effect plot showed that urban and agricultural areas have a negative effect on water quality and biological indicators on both watershed scales, while forest areas have a positive effect on water quality and biological indicators. In addition, the plot showed that the critical point of the impact of land-use proportion on water quality and biological indicators was different for different watershed scales.
This study confirmed that the range of urban proportions over which BOD, TN, and TP concentrations began to increase varied depending on the watershed scale. In large watersheds, the water quality concentration increased in 35–45% of urban areas, and in small watersheds, the water quality concentration increased in 5–10% of urban areas [1]. This is similar to previous studies in which water quality concentrations increased when the urban fraction was 10–50%. Additionally, in this study, the BOD, TN, and TP concentrations were highest when the urban proportion was 35–45% and 5–10% in the large and small watersheds, respectively. Tromboni and Dodds [50] showed that in Brazil, maximum nutrient concentrations are reached at 10–46% urban area, and water quality concentrations may further increase as urbanization accelerates. In the case of biological indicators in urban areas, a decrease in biological indicator values was observed when the urban area proportion was over 5% and 3% in large and small watersheds, respectively, with a rapid decrease observed for large watersheds. This result is similar to that of previous studies. For example, studies have reported a sharp decrease in benthic macroinvertebrate total taxa richness, Ephemeroptera, Plecoptera, and Trichoptera taxa richness, and the Shannon–Wiener diversity index in approximately 3–15% of the urban areas of relatively small watersheds (1–70 km2) [51]. Additionally, in Korea’s riparian area, the fish community composition is low when the urban proportion is 2–19% [27]. This suggests that to preserve the biological status and minimize the impact on the development of large and small watershed urban areas, the proportion of development sites should be calculated according to the watershed scale.
In agricultural areas of large-scale watersheds, the values of water quality and biological indicators decreased at approximately 5–25% of the percentage of agricultural land use. Generally, the anthropogenic input of pollutants from agricultural areas is a major cause of increased nitrogen and phosphorus concentrations in streams [52]. For example, non-point pollutants entering streams from agricultural areas include fertilizers, sediments, nitrogen-fixing crops, and animal waste [53,54]. However, the threshold was found to be low compared to that in other studies. For example, Roth et al. [17] observed no change in the fish community in Wisconsin, USA, until the cropland proportion of the watershed reached 50%; Grimstead et al. [55] found that the abundance of benthic macroinvertebrate communities decreases when croplands exceed approximately 70% of the catchment area; and D’Amario et al. [56] found that the concentrations of nitrogen and phosphorus also increased with an increase in the agricultural area of the watershed and found that the excess concentration occurred at approximately 34–43%. An increase in agricultural areas leads to an increase in stream nutrients, which also affects algae and benthic macroinvertebrates [56]. However, previous studies [17,55] were conducted in intensive agricultural areas in large-scale watersheds. Intensive agricultural areas generally have negative impacts on water quality and aquatic organisms. However, riparian vegetation in these regions can effectively buffer the negative impacts of agricultural activities. Although this study did not consider the impact of riparian vegetation on stream water quality and aquatic organisms, it shows that large-scale watersheds have different impacts on streams than small-scale watersheds, even with a small percentage of disturbed land use/land cover within the watersheds. These characteristics may be related to the land-use/land-cover composition and structure of the watersheds, including riparian vegetation [57]. High landscape diversity in agricultural areas and forests in riparian areas can improve water quality by reducing nutrient and organic material losses [58,59,60,61]. It has been reported that agricultural areas in small watersheds have a negative impact on stream water quality by increasing land-use diversity and fragmenting the natural environment [17]. Land-use proportions alone cannot capture the diverse and complex responses of watersheds to spatial patterns and scales [62]. Contaminant sources in agricultural areas can be negative for aquatic organisms, but the threshold value can change due to the influence of several other factors, such as nutrients flowing into streams; habitats, including sediment delivery systems; and government policies [63].
In the case of forests, water quality and biological indicators increased as the forest proportion increased. These results are consistent with those of previous studies showing that forested areas play an important role in maintaining stream biological integrity and water quality [64]. Brogna et al. [65] found that watershed forests were highly positively correlated with stream and biological water quality. Forests have various benefits for streams, including acting as filters to remove pollutants carried by surface runoff and reducing nutrient concentrations. In this study, the ALEs showed a decrease in water quality concentration in approximately 20–25% of the large- and small-scale watersheds, and an increase in biological indicators in 40–60% of the watersheds. Clément et al. [29] found that having at least 50% forested areas could promote good water quality in Canadian rivers. A previous study suggested an environmental conservation goal of approximately 60% forest cover [8]. In conclusion, securing a forest cover proportion of 25–60% is the minimum requirement for maintaining stream water quality and biological integrity.

4.2. Watershed Scale

The association between land use and water quality and biological indicators assessed using the RF model showed that land use had a significant impact on water quality and biological indicators. In the RF model, large-scale watersheds showed better predictions than small-scale watersheds. This is consistent with other studies showing that the cumulative effect of land use throughout the watershed influences stream water quality and biological conditions [1,66]. In general, catchment scale is reported to have an indirect effect on aquatic organisms by integrating all environmental factors that appear in small watersheds, such as the riparian and reach scales [67,68]. These results may be due to the contribution of all land use in the watershed to the stream ecosystem due to geographical factors, such as the slope and elevation of the watershed, gravity, and various patterns of land use [66]. However, studies have also reported a stronger relation between water quality and biological indicators on smaller scales [69]. Larger watersheds are difficult to estimate because the paths of pollutants moving into streams are more complex and diverse than those in smaller watersheds [14,15,70]. Such differences in the scale of the study results may be due to the effects of water quality parameters and spatial resolution differences in the research design [3,7,71,72]. It is clear that the structure and function of stream ecosystems, including stream water quality and aquatic organisms, depend on the spatial scale [73,74]. This is an important criterion in watershed management and restoration. Current watershed management aims to synthesize the characteristics of watersheds and manage individual streams. However, goals or management strategies based on the watershed may ignore the watershed characteristics of individual streams. This implies that a watershed plan must be set differently for each scale. Land use at various spatial scales affects stream ecosystems through a series of processes. Therefore, understanding the relationship between land use, stream water quality, and aquatic organisms within a watershed is an important part of watershed planning, management, and restoration [24,63]. Understanding the impact of land use on the water quality and biological properties of streams and effectively managing watersheds require insight into their spatial scales. To understand the impact of land use on stream water quality and aquatic organisms and to manage watersheds effectively, it is necessary to consider the spatial scale.

5. Conclusions

This study found that the relation between land use and water quality and biological indicators in the Han River Basin is nonlinear and that critical points appear differently depending on the scale. The results showed similar patterns of relationships between land use and water quality and biological indicators in both the large and small watersheds. As shown in the correlation analysis results, urban areas and agricultural areas had a negative impact on water quality and biological indicators, while forests had a positive impact. The ALE plot results showed that water quality and biological indicators fluctuated in urban areas at lower land-use ratios compared to those of agricultural areas. Both water quality and biological indicators showed more sensitivity to the land-use ratio in small watersheds than in large watersheds. In forested areas, the water quality and biological indicators fluctuated within a certain land-use ratio range, regardless of the watershed scale. In particular, in large watersheds, when the forest proportion was over approximately 25%, the biological indicators increased rapidly. Water quality showed a decreasing trend as the forest rate increased, and when the forest rate was over 25%, it decreased sharply. Therefore, we conclude that achieving at least 25% forest area is necessary to maintain water quality and biological health. The results of this study suggest that to develop effective watershed management strategies to maintain stream ecosystems, various spatial scales must be considered, and a minimum forest area of 25% or higher must be maintained.

Author Contributions

J.-W.L. designed the research, performed the data analysis, and wrote the manuscript, interpreted the results of the analysis. S.-R.P. edited the manuscript; S.-W.L. and S.-R.P. reviewed and finalized the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the Korea Environment Industry and Technology Institute through The Decision Support System Development Project for Environmental Impact Assessment, funded by the Korea Ministry of Environment (No. 2020002990009). This paper and the research were supported by the Korea Forest Service (Korea Forestry Promotion Institute) (FTIS 2021331A00-2223-AA01).

Data Availability Statement

The data supporting the findings of this investigation can be obtained from [Water Environment Information System]. Readers should contact the corresponding author for details.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Allan, J.D. Landscapes and riverscapes: The influence of land use on stream ecosystems. Annu. Rev. Ecol. Evol. Syst. 2004, 35, 257–284. [Google Scholar] [CrossRef]
  2. Diem, J.E.; Hill, T.C.; Milligan, R.A. Diverse multi-decadal changes in streamflow within a rapidly urbanizing region. J. Hydrol. 2018, 556, 61–71. [Google Scholar] [CrossRef]
  3. Schiff, R.; Benoit, G. Effects of impervious Cover at multiple spatial scales on coastal watershed streams 1. J. Am. Water Resour. Assoc. 2007, 43, 712–730. [Google Scholar] [CrossRef]
  4. Wicke, D.; Matzinger, A.; Sonnenberg, H.; Caradot, N.; Schubert, R.-L.; Dick, R.; Heinzmann, B.; Dünnbier, U.; von Seggern, D.; Rouault, P. Micropollutants in urban stormwater runoff of different land uses. Water 2021, 13, 1312. [Google Scholar] [CrossRef]
  5. Hamid, A.; Bhat, S.U.; Jehangir, A. Local determinants influencing stream water Quality. Appl. Water Sci. 2020, 10, 24. [Google Scholar] [CrossRef]
  6. Teurlincx, S.; Kuiper, J.J.; Hoevenaar, E.C.; Lurling, M.; Brederveld, R.J.; Veraart, A.J.; Janssen, A.B.; Mooij, W.M.; de Senerpont Domis, L.N. Towards restoring urban waters: Understanding the main pressures. Curr. Opin. Environ. Sustain. 2019, 36, 49–58. [Google Scholar] [CrossRef]
  7. Shi, P.; Zhang, Y.; Li, Z.; Li, P.; Xu, G. Influence of land use and land cover patterns on seasonal water Quality at multi-spatial scales. CATENA 2017, 151, 182–190. [Google Scholar] [CrossRef]
  8. Park, S.R.; Kim, S.; Lee, S.W. Evaluating the relationships between riparian land cover characteristics and biological integrity of streams using random forest algorithms. Int. J. Environ. Res. Public Health 2021, 18, 3182. [Google Scholar] [CrossRef]
  9. Johnson, R.C.; Jin, H.; Carreiro, M.M.; Jack, J.D. Macroinvertebrate community structure, secondary production and trophic-level dynamics in urban streams affected by non-point-source pollution. Freshw. Biol. 2013, 58, 843–857. [Google Scholar] [CrossRef]
  10. Liu, B.; Chen, S.; Liu, H.; Guan, Y. Changes in the ratio of benthic to planktonic diatoms to eutrophication status of muskegon lake through time: Implications for a valuable indicator on water Quality. Ecol. Indic. 2020, 114, 106284. [Google Scholar] [CrossRef]
  11. Smucker, N.J.; Vis, M.L. Using diatoms to assess human impacts on streams benefits from multiple-habitat sampling. Hydrobiologia 2010, 654, 93–109. [Google Scholar] [CrossRef]
  12. Sweeney, B.W.; Newbold, J.D. Streamside forest buffer width needed to protect stream water Quality, habitat, and organisms: A literature review. J. Am. Water Resour. Assoc. 2014, 50, 560–584. [Google Scholar] [CrossRef]
  13. Turunen, J.; Elbrecht, V.; Steinke, D.; Aroviita, J. Riparian forests can mitigate warming and ecological degradation of agricultural headwater streams. Freshw. Biol. 2021, 66, 785–798. [Google Scholar] [CrossRef]
  14. Ding, J.; Jiang, Y.; Liu, Q.; Hou, Z.; Liao, J.; Fu, L.; Peng, Q. Influences of the land use pattern on water Quality in low-order streams of the Dongjiang River Basin, China: A multi-scale analysis. Sci. Total Environ. 2016, 551–552, 205–216. [Google Scholar] [CrossRef]
  15. Buck, O.; Niyogi, D.K.; Townsend, C.R. Scale-dependence of land use effects on water Quality of streams in agricultural catchments. Environ. Pollut. 2004, 130, 287–299. [Google Scholar] [CrossRef]
  16. Lammert, M.; Allan, J.D. Assessing biotic integrity of streams: Effects of scale in measuring the influence of land use/Cover and habitat structure on fish and macroinvertebrates. Environ. Manag. 1999, 23, 257–270. [Google Scholar] [CrossRef]
  17. Roth, N.E.; Allan, J.D.; Erickson, D.L. Landscape influences on stream biotic integrity assessed at multiple spatial scales. Lands. Ecol. 1996, 11, 141–156. [Google Scholar] [CrossRef]
  18. Dala-Corte, R.B.; Giam, X.; Olden, J.D.; Becker, F.G.; Guimarães, T.d.F.; Melo, A.S. Revealing the pathways by which agricultural land-use affects stream fish communities in South Brazilian grasslands. Freshw. Biol. 2016, 61, 1921–1934. [Google Scholar] [CrossRef]
  19. Pan, Y.; Herlihy, A.; Kaufmann, P.; Wigington, J.; Van Sickle, J.; Moser, T. Linkages among land-use, water Quality, physical habitat conditions and lotic diatom assemblages: A multi-spatial scale assessment. Hydrobiologia 2004, 515, 59–73. [Google Scholar] [CrossRef]
  20. Oeding, S.; Taffs, K.H.; Cox, B.; Reichelt-Brushett, A.; Sullivan, C. The influence of land use in a highly modified catchment: Investigating the importance of scale in riverine health assessment. J. Environ. Manag. 2018, 206, 1007–1019. [Google Scholar] [CrossRef]
  21. Tudesque, L.; Tisseuil, C.; Lek, S. Scale-dependent effects of land Cover on Water physico-chemistry and diatom-based metrics in a major river system, the Adour-Garonne Basin (south Western France). Sci. Total Environ. 2014, 466–467, 47–55. [Google Scholar] [CrossRef]
  22. Zhang, J.; Li, S.; Jiang, C. Effects of Land Use on water Quality in a River Basin (Daning) of the Three Gorges Reservoir Area, China: Watershed versus riparian Zone. Ecol. Indic. 2020, 113, 106226. [Google Scholar] [CrossRef]
  23. Young-Kyu, S. Comparison of water Quality between forested and agricultural subcatchments in Daegwallyong Area. Korean Geogr. Soc. 2004, 39, 544–561. [Google Scholar]
  24. Lee, J.W.; Lee, S.W.; An, K.J.; Hwang, S.J.; Kim, N.Y. An estimated structural equation model to assess the effects of land use on water Quality and benthic macroinvertebrates in streams of the Nam-Han River system, South Korea. Int. J. Environ. Res. Public Health 2020, 17, 2116. [Google Scholar] [CrossRef]
  25. Wang, F.; Wang, Y.; Zhang, K.; Hu, M.; Weng, Q.; Zhang, H. Spatial heterogeneity modeling of water quality based on random forest regression and model interpretation. Environ. Res. 2021, 202, 111660. [Google Scholar] [CrossRef]
  26. Hastie, T.; Tibshirani, R.; Friedman, J.H.; Friedman, J.H. The Elements of Statistical Learning: Data Mining, Inference, and Prediction; Springer: Berlin/Heidelberg, Germany, 2009; p. 2. [Google Scholar]
  27. Chen, K.; Olden, J.D. Threshold responses of riverine fish communities to land use conversion across regions of the world. Glob. Chang. Biol. 2020, 26, 4952–4965. [Google Scholar] [CrossRef]
  28. Munsch, S.H.; Andrews, K.S.; Crozier, L.G.; Fonner, R.; Gosselin, J.L.; Greene, C.M.; Harvey, C.J.; Lundin, J.I.; Pess, G.R.; Samhouri, J.F.; et al. Potential for ecological nonlinearities and thresholds to inform pacific salmon management. Ecosphere 2020, 11, e03302. [Google Scholar] [CrossRef]
  29. Clément, F.; Ruiz, J.; Rodríguez, M.A.; Blais, D.; Campeau, S. Landscape diversity and forest edge density regulate stream water Quality in agricultural catchments. Ecol. Indic. 2017, 72, 627–639. [Google Scholar] [CrossRef]
  30. Foudi, S.; Spadaro, J.V.; Chiabai, A.; Polanco-Martínez, J.M.; Neumann, M.B. The climatic dependencies of urban ecosystem services from green roofs: Threshold effects and non-linearity. Ecosyst. Serv. 2017, 24, 223–233. [Google Scholar] [CrossRef]
  31. Chang, H. Spatial analysis of water quality trends in the Han River Basin, South Korea. Water Res. 2008, 42, 3285–3304. [Google Scholar] [CrossRef]
  32. Lee, S.-W.; Hwang, S.-J.; Lee, J.-K.; Jung, D.-I.; Park, Y.-J.; Kim, J.-T. Overview and application of the National Aquatic Ecological Monitoring Program (NAEMP) in Korea. Ann. Limnol. Int. J. Lim. 2011, 47, S3–S14. [Google Scholar] [CrossRef]
  33. Greenwell, B.M.; Boehmke, B.C.; McCarthy, A.J. A simple and effective model-based variable importance measure. arXiv 2018, arXiv:1805.04755. [Google Scholar]
  34. Baak, M.; Koopman, R.; Snoek, H.; Klous, S. A new correlation coefficient between categorical, ordinal and interval variables with Pearson characteristics. Comput. Stat. Data Anal. 2020, 152, 107043. [Google Scholar] [CrossRef]
  35. Breiman, L. Random forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef]
  36. Ratolojanahary, R.; Houé Ngouna, R.H.; Medjaher, K.; Junca-Bourié, J.; Dauriac, F.; Sebilo, M. Model selection to improve multiple imputation for handling high rate missingness in a water quality dataset. Expert Syst. Appl. 2019, 131, 299–307. [Google Scholar] [CrossRef]
  37. Grömping, U. Variable Importance Assessment in Regression: Linear Regression versus Random Forest. Am. Stat. 2009, 63, 308–319. [Google Scholar] [CrossRef]
  38. Prasad, A.M.; Iverson, L.R.; Liaw, A. Newer classification and regression tree techniques: Bagging and random forests for ecological prediction. Ecosystems 2006, 9, 181–199. [Google Scholar] [CrossRef]
  39. Probst, P.; Wright, M.N.; Boulesteix, A. Hyperparameters and tuning strategies for random Forest. WIREs Data Min. Knowl. 2019, 9, e1301. [Google Scholar] [CrossRef]
  40. Wang, L.; Zhou, X.; Zhu, X.; Dong, Z.; Guo, W. Estimation of biomass in wheat using random forest regression algorithm and remote sensing data. Crop J. 2016, 4, 212–219. [Google Scholar] [CrossRef]
  41. Cutler, D.R.; Edwards, T.C., Jr.; Beard, K.H.; Cutler, A.; Hess, K.T.; Gibson, J.; Lawler, J.J. Random forests for classification in ecology. Ecology 2007, 88, 2783–2792. [Google Scholar] [CrossRef]
  42. RColorBrewer, S.; Liaw, M.A. Package ‘Randomforest.’; University of California: Berkeley, CA, USA, 2018. [Google Scholar]
  43. Molnar, C. Interpretable Machine Learning; Lulu.com: Morrisville, NC, USA, 2020; ISBN 0244768528. [Google Scholar]
  44. Yu, Q.; Ji, W.; Prihodko, L.; Ross, C.W.; Anchang, J.Y.; Hanan, N.P. Study becomes insight: Ecological learning from machine learning. Methods Ecol. Evol. 2021, 12, 2117–2128. [Google Scholar] [CrossRef]
  45. Stritih, A.; Senf, C.; Seidl, R.; Grêt-Regamey, A.; Bebi, P. The impact of land-use legacies and recent management on natural disturbance susceptibility in mountain forests. For. Ecol. Manag. 2021, 484, 118950. [Google Scholar] [CrossRef]
  46. Apley, D.W.; Zhu, J. Visualizing the effects of predictor variables in black box supervised learning models. J. R. Stat. Soc. B 2020, 82, 1059–1086. [Google Scholar] [CrossRef]
  47. Liaw, A.; Wiener, M. Classification and regression by RandomForest. R News 2002, 2, 18–22. [Google Scholar]
  48. Barnosky, A.D.; Hadly, E.A.; Bascompte, J.; Berlow, E.L.; Brown, J.H.; Fortelius, M.; Getz, W.M.; Harte, J.; Hastings, A.; Marquet, P.A.; et al. Approaching a state shift in earth’s biosphere. Nature 2012, 486, 52–58. [Google Scholar] [CrossRef]
  49. Groffman, P.M.; Baron, J.S.; Blett, T.; Gold, A.J.; Goodman, I.; Gunderson, L.H.; Levinson, B.M.; Palmer, M.A.; Paerl, H.W.; Peterson, G.D.; et al. Ecological thresholds: The key to successful environmental management or an important concept with no practical application? Ecosystems 2006, 9, 1–13. [Google Scholar] [CrossRef]
  50. Tromboni, F.; Dodds, W.K. Relationships between land use and stream nutrient concentrations in a highly urbanized tropical region of brazil: Thresholds and riparian zones. Environ. Manage. 2017, 60, 30–40. [Google Scholar] [CrossRef]
  51. Wang, B.; Liu, D.; Liu, S.; Zhang, Y.; Lu, D.; Wang, L. Impacts of urbanization on stream habitats and macroinvertebrate communities in the tributaries of Qiangtang River, China. Hydrobiologia 2012, 680, 39–51. [Google Scholar] [CrossRef]
  52. Ni, X.; Parajuli, P.B.; Ouyang, Y.; Dash, P.; Siegert, C. Assessing land use change impact on stream discharge and stream water Quality in an agricultural watershed. CATENA 2021, 198, 105055. [Google Scholar] [CrossRef]
  53. Savci, S. An agricultural pollutant: Chemical fertilizer. Int. J. Environ. Sci. Dev. 2012, 3, 73–80. [Google Scholar] [CrossRef]
  54. Mangadze, T.; Wasserman, R.J.; Froneman, P.W.; Dalu, T. Macroinvertebrate functional feeding group alterations in response to habitat degradation of headwater austral streams. Sci. Total Environ. 2019, 695, 133910. [Google Scholar] [CrossRef]
  55. Grimstead, J.P.; Krynak, E.M.; Yates, A.G.; Land Cover, S.-S. Thresholds for conservation of stream invertebrate communities in agricultural landscapes. Landsc. Ecol. 2018, 33, 2239–2252. [Google Scholar] [CrossRef]
  56. D’Amario, S.C.; Rearick, D.C.; Fasching, C.; Kembel, S.W.; Porter-Goff, E.; Spooner, D.E.; Williams, C.J.; Wilson, H.F.; Xenopoulos, M.A. The prevalence of nonlinearity and detection of ecological breakpoints across a land use gradient in streams. Sci. Rep. 2019, 9, 3878. [Google Scholar] [CrossRef]
  57. Marzin, A.; Verdonschot, P.F.M.; Pont, D. The relative influence of catchment, riparian corridor, and reach-scale anthropogenic pressures on fish and macroinvertebrate assemblages in French Rivers. Hydrobiologia 2013, 704, 375–388. [Google Scholar] [CrossRef]
  58. Uuemaa, E.; Roosaare, J.; Mander, Ü. Scale dependence of landscape metrics and their indicatory value for nutrient and organic matter losses from catchments. Ecol. Indic. 2005, 5, 350–369. [Google Scholar] [CrossRef]
  59. Jones, K.B.; Neale, A.C.; Nash, M.S.; Van Remortel, R.D.; Wickham, J.D.; Riitters, K.H.; O’Neill, R.V. Predicting nutrient and sediment loadings to streams from landscape metrics: A multiple watershed study from the United States mid-Atlantic region. Landsc. Ecol. 2001, 16, 301–312. [Google Scholar] [CrossRef]
  60. Gergel, S.E.; Turner, M.G.; Miller, J.R.; Melack, J.M.; Stanley, E.H. Landscape indicators of human impacts to riverine systems. Aquat. Sci. 2002, 64, 118–128. [Google Scholar] [CrossRef]
  61. Fitzpatrick, F.A.; Scudder, B.C.; Lenz, B.N.; Sullivan, D.J. Effects of multi-scale environmental characteristics on agricultural stream biota in Eastern Wisconsin 1. J. Am. Water Resour. Assoc. 2001, 37, 1489–1507. [Google Scholar] [CrossRef]
  62. Hawkins, C.P.; Norris, R.H.; Gerritsen, J.; Hughes, R.M.; Jackson, S.K.; Johnson, R.K.; Stevenson, R.J. Evaluation of the use of landscape classifications for the prediction of freshwater biota: Synthesis and recommendations. J. N. Am. Benthol. Soc. 2000, 19, 541–556. [Google Scholar] [CrossRef]
  63. Meador, M.R.; Goldstein, R.M. Assessing water Quality at large geographic scales: Relations among land use, Water physicochemistry, riparian condition, and fish community structure. Environ. Manag. 2003, 31, 504–517. [Google Scholar] [CrossRef]
  64. De Mello, K.; Valente, R.A.; Randhir, T.O.; dos Santos, A.C.A.; Vettorazzi, C.A. Effects of Land Use and Land Cover on water Quality of Low-Order Streams in Southeastern Brazil: Watershed versus riparian Zone. CATENA 2018, 167, 130–138. [Google Scholar] [CrossRef]
  65. Brogna, D.; Dufrêne, M.; Michez, A.; Latli, A.; Jacobs, S.; Vincke, C.; Dendoncker, N. Forest cover correlates with good biological water Quality. Insights from a regional study (Wallonia, Belgium). J. Environ. Manag. 2018, 211, 9–21. [Google Scholar] [CrossRef]
  66. Hunsaker, C.T.; Levine, D.A. Hierarchical approaches to the study of water Quality in rivers. BioScience 1995, 45, 193–203. [Google Scholar] [CrossRef]
  67. Villeneuve, B.; Piffady, J.; Valette, L.; Souchon, Y.; Usseglio-Polatera, P. Direct and indirect effects of multiple stressors on stream invertebrates across watershed, reach and site scales: A structural equation modelling better informing on hydromorphological impacts. Sci. Total Environ. 2018, 612, 660–671. [Google Scholar] [CrossRef]
  68. Forio, M.A.E.; Burdon, F.J.; De Troyer, N.; Lock, K.; Witing, F.; Baert, L.; De Saeyer, N.; Rîșnoveanu, G.; Popescu, C.; Kupilas, B.; et al. A bayesian belief network learning tool integrates multi-scale effects of riparian buffers on stream invertebrates. Sci. Total Environ. 2022, 810, 152146. [Google Scholar] [CrossRef]
  69. Zhou, T.; Wu, J.; Peng, S. Assessing the effects of landscape pattern on river water Quality at multiple scales: A case study of the Dongjiang River watershed, China. Ecol. Indic. 2012, 23, 166–175. [Google Scholar] [CrossRef]
  70. Nash, M.S.; Heggem, D.T.; Ebert, D.; Wade, T.G.; Hall, R.K. Multi-scale landscape factors influencing stream water Quality in the State of Oregon. Environ. Monit. Assess. 2009, 156, 343–360. [Google Scholar] [CrossRef]
  71. Pratt, B.; Chang, H. Effects of land Cover, topography, and built structure on seasonal water Quality at multiple spatial scales. J. Hazard. Mater. 2012, 209–210, 48–58. [Google Scholar] [CrossRef]
  72. Allan, J.D.; Johnson, L. Catchment-scale analysis of aquatic ecosystems. Freshw. Biol. 1997, 37, 107–111. [Google Scholar] [CrossRef]
  73. Meyer, J.L.; Strayer, D.L.; Wallace, J.B.; Eggert, S.L.; Helfman, G.S.; Leonard, N.E. The contribution of headwater streams to biodiversity in river networks 1. J. Am. Water Resour. Assoc. 2007, 43, 86–103. [Google Scholar] [CrossRef]
  74. Wipfli, M.S.; Richardson, J.S.; Naiman, R.J. Ecological linkages between headwaters and downstream ecosystems: Transport of organic matter, invertebrates, and wood down headwater channels 1. J. Am. Water Resour. Assoc. 2007, 43, 72–85. [Google Scholar] [CrossRef]
Water 15 04210 g001
Figure 1. Monitoring sites and land use/land cover in the Han River basin. (a) Large-scale watersheds showing 3rd–4th-order stream basins and (b) small-scale watersheds showing 1st–2nd-order stream watersheds.
Figure 1. Monitoring sites and land use/land cover in the Han River basin. (a) Large-scale watersheds showing 3rd–4th-order stream basins and (b) small-scale watersheds showing 1st–2nd-order stream watersheds.
Water 15 04210 g001
Water 15 04210 g002
Figure 2. Relationships among biological indicators, water quality parameters, and land use in the large-scale watersheds (n = 177) (left figure) and small-scale watersheds (n = 157) (right figure) expressed as Pearson correlation coefficients (** p < 0.01, * p < 0.05).
Figure 2. Relationships among biological indicators, water quality parameters, and land use in the large-scale watersheds (n = 177) (left figure) and small-scale watersheds (n = 157) (right figure) expressed as Pearson correlation coefficients (** p < 0.01, * p < 0.05).
Water 15 04210 g002
Water 15 04210 g003
Figure 3. Accumulated local effect plots for random forest regressions of water quality parameters and land-use variables in large-scale watersheds. ALE, accumulated local effect; BOD, biochemical oxygen demand; TN, total nitrogen; TP, total phosphorous.
Figure 3. Accumulated local effect plots for random forest regressions of water quality parameters and land-use variables in large-scale watersheds. ALE, accumulated local effect; BOD, biochemical oxygen demand; TN, total nitrogen; TP, total phosphorous.
Water 15 04210 g003
Water 15 04210 g004
Figure 4. Accumulated local effect plots for random forest regressions of water quality parameters and land-use variables in the small-scale watersheds. ALE, accumulated local effect; BOD, biochemical oxygen demand; TN, total nitrogen; TP, total phosphorous.
Figure 4. Accumulated local effect plots for random forest regressions of water quality parameters and land-use variables in the small-scale watersheds. ALE, accumulated local effect; BOD, biochemical oxygen demand; TN, total nitrogen; TP, total phosphorous.
Water 15 04210 g004
Water 15 04210 g005
Figure 5. Accumulated local effect plots for random forest regressions for biological indicators and land-use variables in the large-scale watersheds. ALE, accumulated local effect; FAI, Fish Assessment Index; TN, total nitrogen; TP, total phosphorous.
Figure 5. Accumulated local effect plots for random forest regressions for biological indicators and land-use variables in the large-scale watersheds. ALE, accumulated local effect; FAI, Fish Assessment Index; TN, total nitrogen; TP, total phosphorous.
Water 15 04210 g005
Water 15 04210 g006
Figure 6. Accumulated local effect plots for random forest regressions for biological indicators and land-use variables in the small-scale watersheds. ALE, accumulated local effect; FAI, Fish Assessment Index; TN, total nitrogen; TP, total phosphorous.
Figure 6. Accumulated local effect plots for random forest regressions for biological indicators and land-use variables in the small-scale watersheds. ALE, accumulated local effect; FAI, Fish Assessment Index; TN, total nitrogen; TP, total phosphorous.
Water 15 04210 g006
Table 1. Equations for biological indicators used in the National Aquatic Ecology Monitoring Program.
Table 1. Equations for biological indicators used in the National Aquatic Ecology Monitoring Program.
Biological IndicatorsEquations
TDI
(Trophic Diatom Index)		 T D I = 100 − { ( W M S × 25 ) − 25 }
WMS: weighted mean sensitivity
W M S = ∑ A j · S j · V j / ∑ A j · V j
A j : proportion (relative abundance) of species in sample, %
S j : pollution sensitivity of species, 1 ≤ S ≤ 5
V j : indicator value of species, 1 ≤ V ≤ 3
BMI
(Benthic Macroinvertebrate Index)		 B M I = 4 − ∑ i = 1 n s i   ·   h i   ·   g i ∑ i = 1 n h i   ·   g i ×   25
i : number assigned to the species
n : number of species
s i : unit saprobic value of the species i
h i : frequency of the species i
g i : indicator weight value of the species i
FAI
(Fish Assessment Index)		 F A I = s u m   o f   8   m e t r i c s
Metric 1 (M1): number of Korean native species
Metric 2 (M2): number of rifle benthic species
Metric 3 (M3): number of sensitive species
Metric 4 (M4): percentage of tolerant species
Metric 5 (M5): percentage of omnivores
Metric 6 (M6): percentage of insectivores
Metric 7 (M7): amount of native species
Metric 8 (M8): percentage of fish abnormalities
Table 2. Descriptive statistics of the water quality parameters, biological indicators, and land-use proportion of small watersheds in the Han River basin.
Table 2. Descriptive statistics of the water quality parameters, biological indicators, and land-use proportion of small watersheds in the Han River basin.
VariablesMin.Max.MeanS.D.
TDI (0–100)13.887.259.917.2
BMI (0–100)23.794.065.317.9
FAI (0–100)6.310056.317.8
BOD (mg/L)0.67.82.21.1
TN (mg/L)0.9710.283.371.42
TP (mg/L)0.0060.3900.0850.074
Urban (%)0.068.610.411.0
Agricultural (%)0.286.128.518.0
Forest (%)0.096.754.322.0
Notes: n = 157. S.D., standard deviation; Min., minimum; Max., maximum; BOD, biochemical oxygen demand; TN, total nitrogen; TP, total phosphorus; TDI, Trophic Diatom Index; BMI, Benthic Macroinvertebrate Index; FAI, Fish Assessment Index.
Table 3. Descriptive statistics of water quality parameters, biological indicators, and land-use proportion for large watersheds in the Han River basin.
Table 3. Descriptive statistics of water quality parameters, biological indicators, and land-use proportion for large watersheds in the Han River basin.
VariablesMin.Max.MeanS.D.
TDI (0–100)11.498.765.118.0
BMI (0–100)27.094.672.216.8
FAI (0–100)3.210066.821.1
BOD (mg/L)0.55.91.60.8
TN (mg/L)1.048.162.881.20
TP (mg/L)0.0070.3680.0510.049
Urban (%)0.077.87.714.5
Agricultural (%)0.010017.013.8
Forest (%)0.010069.321.6
Notes: n = 177. S.D., standard deviation; Min., minimum; Max., maximum; BOD, biochemical oxygen demand; TN, total nitrogen; TP, total phosphorus; TDI, Trophic Diatom Index; BMI, Benthic Macroin-vertebrate Index; FAI, Fish Assessment Index.
Table 4. T-test for water quality, biological indicators, and land-use proportion between large and small watersheds.
Table 4. T-test for water quality, biological indicators, and land-use proportion between large and small watersheds.
VariablesLevenet-Valuep-Value
FSig.
TDI (0–100)0.0010.9812.6690.008
BMI (0–100)2.1060.1483.6260.000
FAI (0–100)9.4410.0024.9310.000
BOD (mg/L)14.0430.000−5.3270.000
TN (mg/L)1.100.29−3.410.00
TP (mg/L)24.0570.000−4.8190.000
Urban (%)1.2370.267−1.8740.062
Agricultural (%)20.9910.000−6.4550.000
Forest (%)0.1260.7236.2570.000
Notes: BOD, biochemical oxygen demand; TN, total nitrogen; TP, total phosphorus; TDI, Trophic Diatom Index; BMI, Benthic Macroinvertebrate Index; FAI, Fish Assessment Index.
Table 5. Water quality and biological indicator prediction performance of random forest models.
Table 5. Water quality and biological indicator prediction performance of random forest models.
ScaleEvaluateBODTNTPTDIBMIFAI
Large-scaleRMSE0.400.800.02412.719.5211.26
MAE0.300.590.0179.907.007.91
Small-scaleRMSE0.690.630.04415.4512.2916.06
MAE0.550.530.03212.319.7512.81
Notes: RMSE, root-mean-square Error; MAE, mean absolute error, BOD, biochemical oxygen demand; TN, total nitrogen; TP, total phosphorous; TDI, Trophic Diatom Index; BMI, Benthic Macroinvertebrate Index; FAI, Fish Assessment Index.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Lee, J.-W.; Park, S.-R.; Lee, S.-W. Effect of Land Use on Stream Water Quality and Biological Conditions in Multi-Scale Watersheds. Water 2023, 15, 4210. https://doi.org/10.3390/w15244210

AMA Style

Lee J-W, Park S-R, Lee S-W. Effect of Land Use on Stream Water Quality and Biological Conditions in Multi-Scale Watersheds. Water. 2023; 15(24):4210. https://doi.org/10.3390/w15244210

Chicago/Turabian Style

Lee, Jong-Won, Se-Rin Park, and Sang-Woo Lee. 2023. "Effect of Land Use on Stream Water Quality and Biological Conditions in Multi-Scale Watersheds" Water 15, no. 24: 4210. https://doi.org/10.3390/w15244210

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop