**1. Introduction**

The quality of water supplied by potable water treatment plants is crucial for ensuring consumers' health. Turbidity is one of the most critical parameters to determine this quality, as it indicates the presence of colloidal, mineral, and organic substances in the water [1].

The presence of turbidity in water stimulates the proliferation of bacteria and provides coverage and food for pathogens in the water. Moreover, it hinders proper water disinfection and risks consumer health [2]. For this reason, entities supplying water for human consumption in Colombia must ensure compliance with the water quality standards established in Resolution 2115 of 2007, which establishes that turbidity should not exceed 2 Nephelometric Turbidity Units (NTU) [3].

Currently, there are various pieces of equipment and instruments, such as turbidity tubes, Secchi disks, and spectrophotometers, which are used to measure the turbidity of water. However, according to studies carried out by [4,5], and the project "Vulnerability

**Citation:** Fernandez Alvarez, V.; Granada Salazar, D.; Figueroa, C.; Corrales, J.C.; Casanova, J.F. Estimation of Water Turbidity in Drinking Water Treatment Plants Using Machine Learning Based on Water and Meteorological Data. *Environ. Sci. Proc.* **2023**, *25*, 89. https://doi.org/10.3390/ ECWS-7-14326

Academic Editor: Lampros Vasiliades

Published: 3 April 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

and Risk in Drinking Water Systems of Cauca–Aquarisc", it has been demonstrated that some municipal water treatment plants in Colombia face difficulties in determining the physicochemical conditions of the water. These limitations prevent treatment plants from measuring raw water parameters, including turbidity. It is essential to know the turbidity, as this parameter is crucial for establishing the criteria required at each stage of the purification process and carrying out a planned and precise water purification process [6].

An effective strategy to address this limitation is establishing a model for estimating source water turbidity to support operators in decision-making. Several studies support this assertion, including those conducted in [7–10], where artificial neural networks were developed to predict or estimate turbidity concentrations and other water quality parameters. In addition, studies such as [11,12] performed regression analyses considering satellite images and data acquired by citizen scientists for turbidity prediction. Moreover, more research related to the topic was found, but it was noted that research of this type was scarce in tropical zones.

The lack of exploration in these areas creates a gap in the analysis of the behavior of meteorological variables in the type of climate that occurs and the impact it has on water quality parameters, such as turbidity. Colombia is one of the countries belonging to the tropical zone, so it has a relatively varied climatic classification due to the lack of seasonality. Therefore, creating a local experience based on what has been evaluated by countries with different climatic zones than Colombia would contribute to research in analyzing the behavior of meteorological and hydrological variables about the turbidity parameter.

Therefore, this study evaluates the performance of models that have not been implemented in tropical zones and compares them to determine their applicability in environments with hydrological and meteorological characteristics similar to the local environment, to expand experience in this type of climatic zone for the estimation of turbidity and its relationship with various parameters.

### **2. Materials and Methods**
