Dataset Information

Machine learning approach towards explaining water quality dynamics in an urbanised river.

ABSTRACT: Human activities alter river water quality and quantity, with consequences for the ecosystems of urbanised rivers. Quantifying the role of human-induced drivers in controlling spatio-temporal patterns in water quality is critical to develop successful strategies for improving the ecological health of urban rivers. Here, we analyse high-frequency electrical conductivity and temperature data collected from the River Chess in South-East England during a Citizen Science project. Utilizing machine learning, we find that boosted trees outperform GAM and accurately describe water quality dynamics with less than 1% error. SHapley Additive exPlanations reveal the importance of and the (inter)dependencies between the individual variables, such as river level and Wastewater Treatment Works (WWTW) outflow. WWTW outflows give rise to diurnal variations in electrical conductivity, which are detectable throughout the year, and to an increase in average water temperature of 1 [Formula: see text] in a 2 km reach downstream of the wastewater treatment works during low flows. Overall, we showcase how high-frequency water quality measurements initiated by a Citizen Science project, together with machine learning techniques, can help untangle key drivers of water quality dynamics in an urbanised chalk stream.

SUBMITTER: Schafer B

PROVIDER: S-EPMC9295889 | biostudies-literature | 2022 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Machine learning approach towards explaining water quality dynamics in an urbanised river.

Schäfer Benjamin B Beck Christian C Rhys Hefin H Soteriou Helena H Jennings Paul P Beechey Allen A Heppell Catherine M CM

Scientific reports 20220719 1

Human activities alter river water quality and quantity, with consequences for the ecosystems of urbanised rivers. Quantifying the role of human-induced drivers in controlling spatio-temporal patterns in water quality is critical to develop successful strategies for improving the ecological health of urban rivers. Here, we analyse high-frequency electrical conductivity and temperature data collected from the River Chess in South-East England during a Citizen Science project. Utilizing machine le ...[more]

PMID: 35854053

Similar Datasets

Project description:BackgroundUrbanisation has facilitated changes in the hydrological cycle, sediment dynamics and habitat loss and it has had a tremendous impact on river ecosystems. Moreover, the invasion of non-native species reduces the number of native species with the progress of urbanisation, thereby resulting in the homogenisation of fish fauna and significant decrease in diversity. However, the effects of urbanisation on fish fauna vary depending on the region, degree of urbanisation and number of years since the construction of the city. Hence, it is necessary to collect information on how fish fauna changes due to urbanisation in different regions. The target site of the present study is the Tsurumi River, of which approximately 85% of the basin is urbanised and many of the river channels have been affected by straightening and concrete channelling. Monitoring of biota has been conducted mainly in downstream areas; however, data on fish fauna in upstream areas and tributaries of the river, which have been substantially affected by urbanisation, are lacking.New informationSurveys were conducted at 71 sites in the Tsurumi River Basin during summers and winters, focusing on river channels that have been straightened or converted to concrete channels due to urbanisation. As a result of this investigation, 10 families, 28 species and 9,335 individuals were collected. Some of the fish collected included rare species, such as Tanakialanceolata, Lefuaechigonia and Pseudobagrustokiensis, indicating that, even in rivers that have been severely impacted by human activities, there are still rare indigenous species living there, albeit in limited numbers. In addition, Misgurnus sp. (clade B2), Micropterussalmoides, Lepomismacrochirus and Poeciliareticulata were confirmed to be non-native species. Misgurnus sp. (clade B2) was widespread; however, populations of carnivorous non-native alien species were small. The data are all accessible from the document "database_fish_urban_tsurumi" (https://ipt.pensoft.net/manage/resource?r=database_fish_urban_tsurumi).

Project description:In this study, different versions of feedforward neural network (FFNN), Gaussian process regression (GPR), and decision tree (DT) models were developed to estimate daily river water temperature using air temperature (Ta ), flow discharge (Q), and the day of year (DOY) as predictors. The proposed models were assessed using observed data from eight river stations, and modelling results were compared with the air2stream model. Model performances were evaluated using four indicators in this study: the coefficient of correlation (R), the Willmott index of agreement (d), the root mean squared error (RMSE), and the mean absolute error (MAE). Results indicated that the three machine learning models had similar performance when only Ta was used as the predictor. When the day of year was included as model input, the performances of the three machine learning models dramatically improved. Including flow discharge instead of day of year, as an additional predictor, provided a lower gain in model accuracy, thereby showing the relatively minor role of flow discharge in river water temperature prediction. However, an increase in the relative importance of flow discharge was noticed for stations with high altitude catchments (Rhône, Dischmabach and Cedar) which are influenced by cold water releases from hydropower or snow melting, suggesting the dependence of the role of flow discharge on the hydrological characteristics of such rivers. The air2stream model outperformed the three machine learning models for most of the studied rivers except for the cases where including flow discharge as a predictor provided the highest benefits. The DT model outperformed the FFNN and GPR models in the calibration phase, however in the validation phase, its performance slightly decreased. In general, the FFNN model performed slightly better than GPR model. In summary, the overall modelling results showed that the three machine learning models performed well for river water temperature modelling.

Dataset Information

Machine learning approach towards explaining water quality dynamics in an urbanised river.

Publications

Machine learning approach towards explaining water quality dynamics in an urbanised river.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets