Elsevier

Journal of Hydrology

Volume 599, August 2021, 126379
Journal of Hydrology

Research papers
Assessing the spatiotemporal variability of lake water quality using A novel multidimensional shape – Position similarity cloud model

https://doi.org/10.1016/j.jhydrol.2021.126379Get rights and content

Highlights

  • Effectively overcome the uncertainty of the lake water quality evaluation process.

  • Similarity of shape and position between sample clouds and level clouds are quantified.

  • A MSPSCM method is proposed to accurately reflect the actual situation of lake water quality.

Abstract

Cloud model theory provides a reliable method to effectively solve the problem of uncertainty associated with lake water quality assessments. To accurately match water quality parameters obtained from water samples and water quality standards, water quality parameters from water samples and water quality class levels were used as inputs to a reverse cloud generator algorithm to derive corresponding sample and level clouds. A multidimensional shape-position similarity cloud model (MSPSCM) was then developed to accurately evaluate lake water quality by considering shape and position similarities between the sample and level clouds. Using monthly water quality monitoring data from 2017 to 2019, spatiotemporal variability of water quality parameters of Nansi Lake in Shandong Province was analyzed, and the MSPSCM was used to further study the spatiotemporal variability of water pollution in Nansi Lake. Results showed that total nitrogen and total phosphorus were the main sources of pollution in Nansi Lake. Except for the severe pollution of the upper lake and its inflow waters (Class V water quality standard) in November 2017 and September 2019, Nansi Lake waters meet Class III water quality standard, and are suitable for drinking after being treated by a sewage treatment plant. Concentration of residential areas and industries around the upper lake is relatively high; large quantities of pollutants are discharged into the upper lake, resulting in considerably severe pollution of the upper lake (Class IV). Difference between water quality of Nansi Lake and that of its inflow water indicates that the purification ability of Nansi Lake should not be underestimated. In addition, compared with the existing cloud model used to evaluate lake water quality, the MSPSCM can more accurately reflect lake water quality, and provides a more flexible and effective method for lake water quality evaluation.

Introduction

Lakes play a very important role in freshwater storage in surface ecosystems (Wang et al., 2019a, Wang et al., 2019b, Wu et al., 2016). Rivers supply water to lakes and are the main transport channels of pollutants into lakes (Han et al., 2020, Wang et al., 2019a, Wang et al., 2019b). Lake pollution is becoming increasingly severe because of the long timescale of the water cycle, difficulty to protect water resources from pollution, high vulnerability of ecosystems, and strong dependence of human economic development on lake water resources (Egessa et al., 2020, Nong et al., 2020, Yang et al., 2016, Yao et al., 2020). Lake water quality assessment may be understood as a multi-criteria decision-making process with quantitative water quality parameters as input and qualitative assessment as output (Wang et al., 2018, Wu et al., 2017, Yao et al., 2019a). The process contains two types of uncertainties: (1) accuracy of water quality data and monitoring methods are constantly changing under the influence of hydrodynamic and biochemical factors, which introduces random uncertainty into water quality evaluation (Norris and Thoms, 1999); (2) water quality is affected by many water quality parameters, and the nonlinear and complex relationships among parameters introduce fuzzy uncertainty into the determination of the degree of water pollution (Yan et al., 2017). Development of efficient water quality evaluation methods to accurately identify spatiotemporal characteristics is key to comprehensive improvement of lake water and its environment.

Since the popularization of the water quality assessment index (Archibald, 1972), many researchers have developed assessment methods taking into account uncertainties involved in lake water quality assessment. These methods can be roughly divided into three categories: (1) numerical methods based on multivariate statistical analysis, such as Nemerow pollution index (NPI) (Chen et al., 2017), principal component analysis (PCA) (Platikanov et al., 2019), and analytic hierarchy process (AHP) (Singh et al., 2019); (2) methods based on fuzzy set theory and grey system theory, mainly including fuzzy mathematics evaluation method (FMEM) (Zhang et al., 2018a, Zhang et al., 2018b) and grey system evaluation method (GSEM) (Zhang et al., 2018a, Zhang et al., 2018b); (3) methods based on artificial intelligence, such as back propagation artificial neural network (BP-ANN) (Lu et al., 2016), support vector machine (SVM) (Ji et al., 2017) and long short-term memory network (LSTM) (Wang et al., 2019a, Wang et al., 2019b). While useful for water quality assessment, these methods lack comprehensive consideration of the different uncertainties involved in water quality assessment and their applicability is limited to specific conditions.

The cloud model proposed by Li et al. (2009) can convert qualitative descriptions into quantitative values, effectively mitigate impacts of uncertainties on accuracy of evaluation results, and overcome the limits of the traditional water quality assessment methods described above (Wang et al., 2016a). Using a forward and a reverse cloud generator, the model converts qualitative concepts to quantities, generating the quantitative cloud characteristics of Expectation (Ex), Entropy (En), and Hyper entropy (He) (Wang et al., 2014). In the model, clouds are formed from cloud droplets, which are distributed in the domain space within the maximum and minimum boundaries (Fig. A1). The forward cloud generator derives the degree of water pollution from qualitative information of water quality status, while the reverse cloud generator calculates the eigenvalues of Ex, En and He from values of water quality parameters obtained from samples (Yao et al., 2019a). Expectation is the expected spatial distribution of cloud droplets in the region, and is a quantitative characterization of water quality. Entropy represents the degree of dispersion of cloud droplets and the range of cloud droplets in the domain space, and is an effective measure of the uncertainty associated with the qualitative concept of water quality. Hyper entropy is a measure of entropy uncertainty, which reflects the degree of condensation between cloud droplets (Ren et al., 2017).

The normal cloud model is the most basic version of the cloud model with normal (Gaussian) membership function. It has been widely used in lake water quality assessments (Wang et al., 2016a, Wang et al., 2016b, Yang and Wang, 2020, Yao et al., 2019a). However, uncertainties from monitoring methods or environmental parameters are not taken into account when uncertainties are only reflected by cloud characteristics (Wang et al., 2016a, Wang et al., 2016b). The multidimensional similarity cloud model proposed by Yao et al. (2019a) effectively solves this problem. It uses values of the cloud characteristics of each evaluation index to form a vector, and calculates the cosine of the angle between the vectors to determine water quality class level. However, when Ex is far greater than En and He, En and He are ignored, which may lead to large deviations in evaluation results. In fact, similarity measurements of connotation and extension of the concept of qualitative description are included in cloud model similarity measurements (Yang and Wang, 2020). In lake water quality evaluations based on normal cloud models, connotation refers to interval range of water quality classes, that is, cloud position; extension refers to uncertainty associated with water quality class, that is, cloud shape. Similarity or sameness between clouds is determined by similarity between cloud shapes and positions. However, the similarity cloud model proposed by Yao et al. (2019a) lacks detailed consideration of similarity measurements of cloud shapes. Thus, in this study, we used the random weighting method (Yao et al., 2019b) to randomize water quality monitoring data to reduce the impact of uncertainties on accuracy of water quality evaluations. From this, we developed a multidimensional shape-position similarity cloud model (MSPSCM) that considers similarity measurements of cloud positions and shapes to capture water quality quantitatively and accurately.

The objectives of this study are: (1) to develop the MSPSCM to comprehensively consider similarity measurements of cloud positions and shapes to capture lake water quality status more accurately; (2) to apply the model to explore the spatiotemporal variability of water quality of a lake and its inflow, and accurately identify the main factors underlying lake water quality deterioration and main pollution areas. Generally, this study aims to propose a model that can accurately quantify lake water quality, and thus provide more reliable technical support for effective monitoring and control of lake water quality.

Section snippets

Study area

Nansi Lake is in the northern part of Huaihe Basin (116°34′–117°21′ E, 34°37′–35°20′ N), and extends along the northwest–southeast axis (see Fig. 1). It is 126 km long in the north–south direction and 5–25 km wide in the east–west direction. Lake waters cover an area of 1266 km2; average water depth is about 1.46 m; annual average air temperature is 14.2 °C; average annual precipitation is about 700 mm; precipitation between June and September accounts for about 70% of total annual

Analysis of water quality parameters

To accurately obtain the main factors affecting the water quality of the study area, annual average concentrations and variability of water quality parameters were derived by using the collected data from 2017 to 2019 (see Table 2). In addition, differences between annual average concentrations of water quality parameters at 18 monitoring sections were tested using one-way analysis of variance (one-way ANOVA) (Varol, 2019) (see Fig. 4).

Statistical variables of DO concentration in Table 2 show

Conclusion

To improve the match between water quality parameters obtained from samples and water quality standards in multidimensional cloud model studies of lake water quality assessments, a reverse cloud generator was used to process sample data and water quality class level interval to generate corresponding sample clouds and level clouds, and a MSPSCM was established from the shape and position similarities between sample and level clouds. Spatiotemporal variability of five selected water quality

Declaration of Competing Interest

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.

Acknowledgements

This study was supported by the National Natural Science Foundation of China (Grant No. 51879006).

References (37)

Cited by (19)

  • Detecting interdisciplinary semantic drift for knowledge organization based on normal cloud model

    2023, Journal of King Saud University - Computer and Information Sciences
View all citing articles on Scopus
View full text