Relationship between land use and water quality in a tropical urban catchment of the Congo Basin: A case study of N'Djili River catchment

Rivers are abundant ecosystems and vital sources of life and livelihoods, serving multiple purposes to sustain both natural and human systems (Chan 2012). Water quality is a key factor that significantly influences socioeconomic, agricultural, public health, and industrial domains (Shi et al. 2017; Zhang et al. 2021). Assessing and monitoring water quality is hence very critical in protecting natural ecosystems whose degradation is driven by human activities (Mohammadpour et al. 2015; Ding et al. 2016; Wang et al. 2023).

Water quality assessment and understanding its impacts are important elements in the safeguarding processes of water resources (Mohammadpour et al. 2016). The main difficulty in preserving water quality arises from the presence of both point source and non-point source pollution (Fan et al. 2012). Point sources include the discharge from sewage treatment facilities, industrial wastewater, and non-point sources include runoff from urban and agricultural areas (Sliva & Dudley 2001). The unequal distribution of regional NPS pollution is exacerbated by intensive human activities and the intricate interplay between rainfall and landscape features, thereby amplifying the challenges associated with identifying these pollutants (Liu et al. 2016; Shi et al. 2017; Zhang et al. 2021). Land use is a major factor contributing to NPS pollution; numerous studies have employed land use types (LUTs) to model and evaluate NPS pollution (Prakash Basnyat et al. 2000; Xiao & Ji 2007; Maillard & Nádia 2008; Shen et al. 2013; Abdulkareem et al. 2018; Wang et al. 2021). Land use refers to the way in which land is utilized by humans for various purposes such as residential, commercial, industrial, agricultural, and recreational activities. It plays a significant role in water quality (Tu 2011, 2013).

The influences of LUT on water quality parameters can vary greatly depending on the scale at which they are examined (Shi et al. 2017). At a local scale, the type of land use adjacent to a water body can have a significant impact on water quality, whereas at a larger scale, the impacts of land use on water quality can become more complex and interrelated (Ding et al. 2016; Zhang et al. 2019). The buffer zone approach has often been applied at catchment, riparian, and site scales (Zhang et al. 2021; Wang et al. 2023).

Due to the multiple sources of water pollution and geographical characteristics combined with highly dynamic aquatic ecosystems in a river basin, the impact of LUTs on water quality parameters may vary from one area to another. In recent years, human activities have significantly degraded water resources ecosystems (Odume 2020; Brontowiyono et al. 2022).

The current study focused on the N'Djili River catchment system which is one of the Congo River tributaries (Tshimanga et al. 2021). In its lower course, N'Djili River crosses Kinshasa, the capital city of the Democratic Republic of the Congo. It is a very important source of water supply, satisfying almost 60% of the water demand of the city (Kimfuta 2013). However, there is a lot of pressure on the river due to the increasing human and livestock populations, watershed degradation from agricultural practices, informal settlements, and vegetation degradation. Despite the fact that there have been previous studies reporting water pollution in this catchment (Muteba 2002; Kimfuta 2013; Kakundika et al. 2022), none of these studies has found a concrete solution face to its water quality degradation. This study is the first in the study area to consider both satellite imagery and field observations to determine the relationship between land use and water quality. Hence, assessing the relationship between land use and water quality parameters becomes critical in trying to understand the impact of human activities on the river's ecosystem. This will further enhance the scientific capacity needed to develop effective strategies for its restoration and conservation.

Thus, the main objective of this study is to determine the relationship between land use and water quality variables in the N'Djili River catchment lower course. The physico-chemical and bacteriological quality of N'Djili river waters were first assessed. The water samples were collected both during the rainy and dry seasons. The study followed a source-to-sink approach, starting from an upstream point (Lemba Imbu – less populated) and moving downstream towards the sink (Masina – most populated).

The land use categories were determined at different sites based on buffer scaling using the Sentinel-2 land use/land cover data product (Karra et al. 2021). Furthermore, the study focused on field data collection to find out the different anthropogenic activities taking place in the surrounding areas. Finally, a redundancy analysis (RDA) was performed in view to determine the relationship between land use categories and water quality variables (Zhao et al. 2015; Ding et al. 2016; Batbayar et al. 2019; Zhang et al. 2019).

This study is the first step towards the development of the N'Djili River catchment water resource-integrated management approach. By applying the buffer scaling method, the study provides a comprehensive understanding of the relation between land use categories and water quality variables. But some questions need to be asked: what kind of relation can it be, which land use categories influence the water quality the most; and where exactly is the influence situated along the river. The study was able to find out the specific sources of pollution and their extent, thereby allowing for targeted interventions to reduce pollution and improve water quality in this catchment.

For the purposes of this study, a perimeter within the lower course of the N'Djili River catchment was delineated. The perimeter starts from the least populated area upstream (Site 1), extending to a highly populated area downstream (Site 10) (Figure 1). The sampling points were identified based on the potential sources of pollution along the river within the defined boundary. Ten study sites were selected. The sites were located using the Garmin brand global positioning system to obtain the geographical coordinates (Figure 1, Table 1, and see Supplementary Materials). The QGIS 3.10 software was used to create a map. Each site was carefully chosen to demonstrate the source-to-sink approach. Criteria such as accessibility, hydrographic distribution, and proximity to potential sources of pollution (e.g. sparsely populated areas, densely populated areas, presence of household waste, sandpits, farming area, and pig breeding) were considered when selecting the sites.

Table 1

Summary statistics for the water quality parameters on seasonal variations

Test Variables .	Summary statistics .
	Mean .		Median .		SD .		Skewness .		Kurtosis .		Coeff. Var .		Shapiro–Wilk test .		Kruskal–Wallis test .		Intl. standards .
	DS .	RS .	DS .	RS .	DS .	RS .	DS .	RS .	DS .	RS .	DS .	RS .	PDS .	PRS .	H .	P .	.
O2 (mgO2/l)	0.5	0.2	0.3	0.15	0.7	0.21	2	1.18	3.1	0.47	1.31	0.81	0.006	0.008	11.12	0.000	>4
COD (mgO2/l)	108	134,5	95	140	63.9	56.98	0.12	−0.42	−1.22	−0.45	0.59	0.42	0.000	0.087	9.15	0.002	≤30
TC (UFC/100 ml)	1,771	93,240	1,700	38,250	1,033.5	1,248.2	0.2	1.47	−1.6	0.72	0.58	1.33	0.168	0.000	14.29	0.000	—
FC (UFC/100 ml)	381.2	24,940	118.5	4,900	483.1	3,828.9	1.4	1.58	0.5	0.91	1.26	1.53	0.00	0.00	10.08	0.00	<101
pH	7.2	7.17	7.2	7.08	0.2	0.23	−0.06	0.55	−1.61	−1.29	0.04	0.03	0.44	0.15	1.12	0.28	pH > 6.5 pH < 9.5
EC (μs/cm)	28.9	27.79	19.9	21.4	25.05	19.84	2.05	2.47	4.09	6.73	0.86	0.71	0.00	0.00	0.28	0.59	—
T⁰ (⁰C)	26.3	28.42	26	27.7	1.4	2.14	1.7	1.99	3.9	4.79	0.05	0.07	0.02	0.00	6.82	0.00	25
Tb (NTU)	61.3	177.19	35.5	134.1	76	77.49	2.8	1	8.2	−1.13	1.24	0.43	0.00	0.00	10.57	0.00	<5
(mg /l)	2.38	2.95	0	0.15	4.9	5.46	1.8	2.55	1.9	7.02	2.07	1.85	0.00	0.00	0.28	0.59	<50
(mg /l)	0.3	0.15	0.1	0.06	0.5	0.21	3.09	1.62	9.7	1.48	1.77	1.35	0.00	0.00	1.46	0.22	<0.5

Test Variables .	Summary statistics .
	Mean .		Median .		SD .		Skewness .		Kurtosis .		Coeff. Var .		Shapiro–Wilk test .		Kruskal–Wallis test .		Intl. standards .
	DS .	RS .	DS .	RS .	DS .	RS .	DS .	RS .	DS .	RS .	DS .	RS .	PDS .	PRS .	H .	P .	.
O2 (mgO2/l)	0.5	0.2	0.3	0.15	0.7	0.21	2	1.18	3.1	0.47	1.31	0.81	0.006	0.008	11.12	0.000	>4
COD (mgO2/l)	108	134,5	95	140	63.9	56.98	0.12	−0.42	−1.22	−0.45	0.59	0.42	0.000	0.087	9.15	0.002	≤30
TC (UFC/100 ml)	1,771	93,240	1,700	38,250	1,033.5	1,248.2	0.2	1.47	−1.6	0.72	0.58	1.33	0.168	0.000	14.29	0.000	—
FC (UFC/100 ml)	381.2	24,940	118.5	4,900	483.1	3,828.9	1.4	1.58	0.5	0.91	1.26	1.53	0.00	0.00	10.08	0.00	<101
pH	7.2	7.17	7.2	7.08	0.2	0.23	−0.06	0.55	−1.61	−1.29	0.04	0.03	0.44	0.15	1.12	0.28	pH > 6.5 pH < 9.5
EC (μs/cm)	28.9	27.79	19.9	21.4	25.05	19.84	2.05	2.47	4.09	6.73	0.86	0.71	0.00	0.00	0.28	0.59	—
T⁰ (⁰C)	26.3	28.42	26	27.7	1.4	2.14	1.7	1.99	3.9	4.79	0.05	0.07	0.02	0.00	6.82	0.00	25
Tb (NTU)	61.3	177.19	35.5	134.1	76	77.49	2.8	1	8.2	−1.13	1.24	0.43	0.00	0.00	10.57	0.00	<5
(mg /l)	2.38	2.95	0	0.15	4.9	5.46	1.8	2.55	1.9	7.02	2.07	1.85	0.00	0.00	0.28	0.59	<50
(mg /l)	0.3	0.15	0.1	0.06	0.5	0.21	3.09	1.62	9.7	1.48	1.77	1.35	0.00	0.00	1.46	0.22	<0.5

Bold values in table 1 are Shapiro-Wilk and Kruskal-Wallis test P values that are inferior to 0.05. Bold values = P<0.05.

View Large

Four sampling campaigns were carried out throughout the rainy and dry seasons for the assessment of seasonal water quality variability. A total of 80 water samples were collected from 10 sampling sites. The water samples were later taken in PET bottles for physico-chemical analyses and in labelled glass bottles previously sterilized for the collection of bacteriological samples. All samples were taken according to the procedures described in the water analysis manual by Rodier & Coll (2009). Five parameters were measured in situ, i.e. pH (using a pH meter HANNA) and temperature (T⁰), dissolved oxygen (O₂) with a WTW oximeter, and electrical conductivity (EC) with a conductivity meter HANNA and turbidity using a turbid meter HANNA. Tree physico-chemical (chemical oxygen demand (COD), nitrate (⁠⁠), and phosphate (⁠⁠) and two bacteriological parameters (total coliforms (TCs) and faecal coliforms (FC)) were measured in the laboratory (Table 2 Supplementary Materials).

Esri-Sentinel-2 land cover explorer 2022 imageries were used (Karra et al. 2021). Eight land use categories were found.

Three buffer scales were fixed around each site, at 100, 300, and 500 m. A total of 100 m was considered as the minimal distance at which river waters are directly impacted by its surrounding environment. We stopped at 500 m because at this distance all site's buffer scales start overlapping.

Water quality was evaluated based on the guide to assessing the status of inland surface waters (François et al. 2019). The Shapiro–Wilk data normality test was used to check if water quality parameters were in conformity with statistical normal distribution (Haidary et al. 2013). The Kruskal–Wallis test was used to find out if there are seasonal significant differences among the different water quality parameters (Woldeab et al. 2019). The XLSTAT life sciences 2023.2.0 software was used in these calculations (Lumivero 2023). A detrended correspondence analysis was performed to determine the type of model in which both land use indicators and water quality parameters were analysed (Zhao et al. 2015; Ding et al. 2016; Batbayar et al. 2019; Zhang et al. 2019, 2021). The RDA was carried out to achieve two objectives: firstly, to create ordination diagrams, known as biplots, to visualize the correlation between land use categories and water quality parameters and secondly, to determine the percentage of variation in water quality parameters that could be explained by several land use indicators. XLSTAT Life Sciences 2023.2.0 software was used to perform the RDA (Lumivero 2023).

Table 1 shows the summary statistics of studied water quality parameters for both dry and rainy seasons, such as dissolved oxygen (O₂), chemical demand of oxygen (COD), TC, FC, pH, conductivity (EC), temperature (T⁰), turbidity (Tb), nitrate (⁠⁠), and phosphate (⁠⁠).

It also shows the p values of Shapiro–Wilk test (p < 0.05, data are not normally distributed) and Kruskal–Wallis test (p < 0.05, there is a significant seasonal difference between dry and rainy season data) for each parameter. Results were also compared to international standards, namely, the guide for assessing the status of inland surface waters (François et al. 2019).

Dry and rainy season data from the 10 sampling sites were used to make the boxplots. Each boxplot represents the mean, minimum, and maximum values of measured parameters during dry and rainy seasons.

Percentage of land use categories inside the three fixed buffer scales. QGIS 3.10 and Excel software were used to perform these spatial calculations. Built area is the major land use category at each scale. Table 2 shows the percentages of land use indicators at the whole study area scale but also at each buffer scale. For each buffer scale, the mean and standard deviation values were given using the 10 sampling sites' percentages values of land use indicators.

Urbanized areas accounted for 67.01%, rangelands 16.21%, trees 7.41%, crops 2.55%, and water 2.44%, and the total was about 95%. Considering the whole study area's spatial scale, some other smaller land uses occupied the remaining proportion (5%). These remaining land uses were excluded from our statistical analyses. As the spatial scale increased, built-up areas and range land proportions increased too, while the proportion of water as land use decreased. The crop land proportion decreased and then increased, while trees area increased and later decreased.

Percentages of land use indicators over the sampling sites per scale. QGIS 3.10 and Excel software were used to perform these spatial calculations. The proportion of urbanized area land increases from site 2 and decreases at site 10. Sites 8 and 9 are completely covered by urbanized area land. Crops land is present at sites 7 and 10 at 100 m buffer scale, with 46.98 and 41.66% coverage, respectively. Crops land also extends to site 6 at 300 m buffer scale and to site 3 at 500 m buffer scale (Figure 6).

RDA was used to analyse the relationship between land use categories and water quality parameters. Table 3 shows the results of the RDA test. Details of the constrained inertia carried by each RDA axis are provided in the table as their cumulative value percentage. It also shows the pseudo F and p values of the permutation test. The permutation test is a statistical method used to determine if there is a linear relationship between the response and explanatory variables. In this case, the p-value obtained from the test is significantly smaller than the predetermined risk threshold alpha 0.05. Therefore, we can confidently reject the null hypothesis, while still maintaining a very low risk of making an incorrect decision. Overall, 80% of the four water quality parameters for dry and rainy seasons were explained by the four land use indicators under investigation. The RDA analysis shows that land use indicators’ impact on water quality parameters has significant seasonal variation. The water quality parameters show that the variation recorded in all the axes during the rainy season is much higher (about 10%) than the percentage recorded during the dry season. The explained variation decreases in both the dry and rainy seasons (Table 3).

During the drying season, the first axis F1 showed a pollution concentration gradient (COD, FC, and TC decrease along the axis) and that it is positively correlated with urbanized area land use category and negatively correlated with trees and rangeland categories (Figure 7). At all buffer scales, dissolved oxygen (O₂) and urbanized area are negatively correlated, while O₂ is positively correlated with trees and rangeland use types. Site 1 is positively correlated with O₂, trees, and rangeland use, while sites 3, 5, 6, and 9 are positively correlated with urbanized area, COD, FC, and TC. Sites 4, 8, and 10 are strongly correlated with crop land, TC, and FC (Figure 7).

During the rainy season, O₂, trees, and rangeland are well represented on axis F2. On axis F1, crops are positively correlated with FC and TC, while negatively correlated with urbanized areas and COD, which are also positively correlated. At all buffer scales, dissolved oxygen (O₂) and urbanized areas are negatively correlated, while O₂ is positively correlated with trees and rangeland use types (Figure 7). Site 1 is positively correlated with O₂, trees, and rangeland use, while sites 7 and 10 are positively correlated with crops land FC and TC. Sites 2, 3, 4, 5, 6, and 9 are positively correlated with urbanized areas and COD. The correlation between land use indicators and water quality variables is greater during the rainy season than in the dry season (Figure 7).

Four land use indicators were used, namely, urbanized area, crops, trees, and rangeland. Four water quality parameters were used, namely, dissolved oxygen (O₂), chemical demand for oxygen (COD), TC, and FC (TF). XLSTAT Life Sciences 2023.2.0 software was used to perform the RDA (Lumivero 2023).

The mean represents the average value, while the median identifies the middle value in an ordered sequence or positional average. Skewness indicates the symmetry of the data distribution, while kurtosis measures the degree of flatness around the mode of a frequency curve. The coefficient of variation provides a relative measure of variability within the sample (Parmar & Bhardwaj 2014). In addition, the Shapiro–Wilk and Kruskal–Wallis tests are utilized for assessing normality and seasonal differences, respectively (Table 1). During the dry season, higher dissolved oxygen levels may be attributed to reduced organic matter decomposition and increased oxygen solubility due to lower temperatures (Rajendiran et al. 2023). Conversely, the elevated COD levels (134.5 mgO₂/l) in the rainy season could be linked to intensified runoff carrying pollutants (Maulud et al. 2021), while the increased microbial load (TC and FC) may result from enhanced faecal inputs and nutrient transport during precipitation events (Zhang et al. 2020). The higher temperatures in the rainy season can accelerate biochemical reactions and microbial activity, influencing pollutant concentrations according to (Englert et al. 2015). Furthermore, heightened turbidity (177.19 NTU) during the rainy season may be associated with increased sedimentation and suspended solids carried by runoff (Nukapothula et al. 2022; Singh et al. 2024). The observed deviations from a normal distribution and significant differences (Table 1) indicated by the Kruskal–Wallis test across various environmental parameters, including dissolved oxygen (O₂), COD, TC, FC, temperature (T⁰), and turbidity (Tb), suggest a notable seasonal variation in water quality. Both dry and rainy seasons exhibit positive skewness for most parameters, indicating that the distribution of these parameters is skewed towards higher values (skewed to the right). However, pH and COD show negative skewness values, which suggest that their distributions are skewed towards lower values (skewed to the left). Kurtosis values greater than 3 indicate leptokurtic distributions, which means the distribution has heavy tails and is more peaked than a normal distribution. For the dry season, O₂, EC, T⁰, Tb, and exhibit leptokurtic distributions, indicating that they have heavier tails and are more peaked than a normal distribution. On the other hand, COD, TC, FC, pH, and have kurtosis values less than 3, suggesting they have lighter tails and are less peaked than a normal distribution (platykurtic curve).

The RDA results have shown that trees and rangeland had a positive impact on water quality parameters, and urbanized areas had a negative impact on them (Figure 7 and Tables 3 and 4 Supplementary Material). Urbanized areas have a positive impact on COD, FC, and TC. These results are similar to many of the previous studies. Dou et al. 2022 have recently worked on the spatial scale effects of landscape structure on water quality in the Yellow River. They found that the deterioration of river water quality is directly correlated with the expansion of urban areas and agricultural lands. Numerous studies have been conducted on the impact of land use change resulting from anthropogenic and socioeconomic activities on river water quality. All of these studies demonstrate that landscape categories associated with urban development and agricultural lands have a significant impact on changes in water quality. Furthermore, these studies confirm the negative influence of anthropogenic activities and land use change on river water quality (Wang et al. 2014; Schreiber et al. 2015; Xiao et al. 2016; Camara et al. 2019; Yadav et al. 2019; Tahiru et al. 2020; Brontowiyono et al. 2022; Chen et al. 2022).

This study is the first in N'Djili River catchment to consider both satellite imagery and field observations to determine the relationship between land use and water quality. As most sites were positively correlated to the urbanized area, site 1 as the control site was negatively correlated to this area (Figure 7). Most of these sites were densely populated and recorded high COD, FC, and TC levels. In 2022, Chen et al. evaluated the impact of urbanization on the water quality of four typical urban rivers in Tanzania and found that there is a strong correlation between large cities and the high pollution level of water bodies, suggesting that as the population rapidly increases, the pollution levels also tend to rise. Urbanized areas are the major source of water pollution in the N'Djili River catchment. Sewage, waste, and garbage are dumped into the river in their raw state without prior treatment (Joseph et al. 2023).

Vegetation plays a crucial role in reducing water pollution (Mohammadpour et al. 2014), and areas with trees or rangeland cover tend to have better water quality due to reduced runoff and soil erosion (Xiao et al. 2016). Site 1 was positively correlated to trees and rangeland, and this site recorded a low level of pollutants. Crop lands are positively correlated with FC and TC. Sites 7 and 10 are positively correlated with crop lands. While conducting the field survey at these sites, it was observed that activities such as pig farming, substantial agricultural operations utilizing chemical fertilizers, as well as the use of poultry and cattle droppings as manure were taking place. The contamination of rivers with microbes is significantly influenced by the runoff from the catchment area. The most significant levels of microbial contamination were found in downstream sites (most populated) and areas where intensive agricultural practices were being carried. These results are similar to those of Schreiber et al. (2015), who have worked on the impact of land use on microbial surface water pollution.

The RDA results show that water quality parameters are better explained during the rainy season than during the dry season (Table 3). Indeed, large and small businesses, gas stations, garages, slaughter houses, and storm water runoff create additional pollution risks, and these results are similar to those of Xiao et al. 2016 who have made a multi-scale analysis of the relationship between landscape patterns and urban river water quality in different seasons and found that some water quality parameters such as COD are influenced by seasonality. From site 1 (source) towards site 10 (Sink), urbanized areas increasingly negatively impact water quality (Figures 6 and 7).

The rapid growth of Kinshasa city has led to more paved surfaces, which are impermeable and hence prevent rainwater from infiltrating into the ground; this leads to an increase in the runoff amount. This runoff carries pollutants along the way from buildings such as sewage and waste into the river, resulting in more pollutants, bacteria, heavy metals, and other contaminant levels in the N'Djili river waters, posing risks to aquatic life and human health (Kakundika et al. 2022).

The scale plays a crucial role in understanding the relationship between land use indicators and water quality. As the scale increases, the impact of various land use categories on water quality decreases (see Table 3, the percentage of explained variation in all axes decreases from 100 to 500 m scale for both seasons). Specifically, the influence of land use indicators on water quality is more pronounced at a 100 m scale compared to a 500 m scale. This suggests that the immediate surrounding environment has a more direct impact on river waters. In light of this, when developing strategies for water quality management in this region, it may not be necessary to consider a larger spatial scale, as the finer scale assessments appear to be more relevant and informative.

At all scales, urbanized areas were more closely related to water quality because the study area is situated in a highly urbanized area. At the higher scale of 500 m buffer, after urbanized areas, crop land had relatively negative effects on water quality; this is due to the fact that at this scale, crop land extended from sites 7 and 10 to other sites, including sites 6 and 3 (Figure 6).

This study demonstrates innovation by strategically employing the site buffer scale approach to overcome substantial data gaps in hydrological information across the Congo Basin, particularly in the context of limited historical water quality data. The method effectively captures spatial heterogeneity, considering variations in land use patterns and their impact on water quality at multiple distances from the site. By creating buffers at various spatial scales, the research enhances accuracy and provides a nuanced understanding of the correlation between land use categories and water quality parameters. This approach not only addresses data gaps but also aids in identifying critical zones influencing water quality, supporting informed decision-making for land use management and water quality protection.

The present study is the first work to provide a relationship between land use and water quality, using high spatial resolution (10 m) and a buffer zone approach in a River catchment of the Congo Basin. This study is fundamental for the establishment of integrated water resources management strategies in the study area. However, a limitation arises as the study does not fully consider the temporal fluctuations introduced by rainfall patterns, overlooking potential influences on runoff, sediment transport, pollutant wash-off, and some externals events (Wang et al. 2013; Luo et al. 2020; Zhang et al. 2020, 2022; Sent et al. 2021). Hydrological uncertainty also, driven by factors like precipitation variations, land use shifts, and climate fluctuations, significantly impacts the length and variability of water quality samples, posing challenges in environmental monitoring. This uncertainty leads to unpredictable changes in water flow patterns and contaminant transport, complicating sample collection due to erratic temporal and spatial pollutant distributions. As elucidated by Dimitriadis et al. (2021), this behaviour underscores the need for extensive and high-resolution sampling to accurately quantify variability on daily, seasonal, and annual scales. According to them, longer sampling durations may be needed to encompass diverse hydrological conditions, resulting in increased variability in water quality parameters. Consequently, accurately characterizing water quality becomes inherently challenging, demanding adaptable monitoring strategies to address the impacts of hydrological uncertainty and maintain the reliability of environmental assessments. Future research should incorporate temporal rainfall patterns perspective and streamflow to comprehensively assess the impact of extreme weather events on water quality dynamics in river catchments. Also, as the redundancy analyses are based on linear correlations, which creates some difficulties in small or medium data lengths, such as statistical bias and standardization error, it is recommended for future studies to look at the so-called cross-climacogram (Koutsoyiannis 2019), which is a tool proven to have smaller bias and error compared to the linear-(auto) correlation (Dimitriadis & Koutsoyiannis 2015).

This study was about the spatio-temporal impacts of land use categories on river water quality parameters in an urban catchment. The study found that water quality parameters were affected by land use to a certain degree. In particular, the high influence of urbanized areas was significant. In fact, multiple factors affected water quality and these included seasonality, scaling, and land use indicators. Land use categories' effects on water quality parameters were more significant during the rainy season than the dry season. The total variation explained by land use indicators on water quality parameters was higher at the 100 m buffer scale than at the 500 m scale. The study indicated that urbanized areas had negative effects on water quality parameters, while trees and rangelands had a positive impact. Scale plays a crucial role in understanding the relationship between land use indicators and water quality. As the scale increases, the impact of various land use categories on water quality decreases (see Table 3, the percentage of explained variation in all axes decreases from 100 to 500 m scale for both seasons). Specifically, the influence of land use indicators on water quality is more pronounced at a 100 m scale compared to a 500 m scale. This suggests that the immediate surrounding environment has a more direct impact on river waters. In light of this, when developing strategies for water quality management in this region, it may not be necessary to consider a larger spatial scale, as the finer scale assessments appear to be more relevant and informative. These results provide key insights into the multiple-scale impact of major land use categories on the N'Djili River catchment's lower course water quality, thus allowing for targeted interventions to reduce pollution and improve water quality in this part of the Congo River basin. It takes into consideration the effective impact of land use on water quality, thereby providing a valuable and concrete reference for the development of an integrated management strategies plan for this river's water resources. A systematic analysis of the sources and nature of pollution and their interactions with N'Djili River waters should be conducted to determine the extent of its water pollution so as to develop a water resources integrated management strategy for minimizing pollution in this area.

This research was sponsored by the African Water Resources Mobility Network (AWaRMN) (Grant No. 2019–1973/004 – 001), which has received funding from the Intra-African Academic Mobility Scheme of the European Union. The authors are grateful to God for his grace and mercies throughout this work. The authors are thankful to the Congo Basin Water Resources Research Center – CRREBaC. The authors thank for their help all the persons who have provided data or information to carry out this work, precisely the National Central Laboratory of REGIDESO in Kinshasa.

Zouera Sani: writing – original draft, methodology, analysis; Pr. Haddy Mbuyi, Pr. Raphael Tshimanga, Pr. Twaha Basamba, and Pr. Nelson Odume: Supervision – writing – review and editing.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.