Enhanced Water Quality Monitoring and Estimation Using a Multi-Modal Approach

Abstract

Remote sensing through satellites and internet of things (IoT) technology are two widespread techniques to assess inland water quality. However, both these techniques have their limitations. IoT provides point data, which is insufficient to represent entire water body, especially if the water body has complex terrain and hydrology. Through remote sensing, we can sample data of a large area, but data acquisition is constrained by satellite. Revisit time and quality of estimates can be affected by image resolution. Moreover, non-optical properties that might affect water quality cannot be sensed through satellites. To complement this, GIS data from labs can be useful for providing higher resolution and accurate data and can be used as ground truth. Thus, in this chapter, the authors aim to integrate both these data collection techniques followed by estimation and prediction through machine learning models. The accumulated datasets are used to train machine learning (ML) models deployed at a server. The selected ML model is an artificial neural network with train accuracy of 97% and test accuracy of 95%.

Unmanned surface vehicle for intelligent water quality assessment to promote sustainable human health

“…A recent study (Zubair et al 2022) integrated three data modalities (GIS, satellite imagery, and IoT nodes) to predict the water quality holistically using time series analysis. Khan et al (2022) worked along the same lines and singled out a multimodal approach to classify the water quality, acquiring data from IoT nodes and satellite imagery. The artificial neural network (ANN) outperformed SVM and random forest with an accuracy of 97% in classifying water quality.…”

“…The latest research has mostly focused on the use of the IoT and satellite imagery for water quality monitoring along with laboratory methods as conventional laboratory methods are expensive, time-consuming, and non-real-time. Both IoT nodes and manual sampling provide point data, which are insufficient to represent the entire water body, and satellite imagery has its own drawbacks other than its revisit time, as it can only measure optically active parameters (Sagan et al 2020) and is affected by climatic effects (Khan et al 2022). The proposed methodology addresses the limitations of previous methods and techniques and provides a holistic method to monitor water quality.…”

“…Recently, IoT sensors have been widely utilized to assess water quality in real time, saving a great deal of time when compared to manual laboratory techniques. However, IoT nodes provide point data, which is insufficient to represent the water quality of a large water body (Khan et al 2022). In addition, difficult terrain makes certain data sites inaccessible, and the presence of certain bacteria in water may cause health risks to those collecting samples (Koparan et al 2018) or installing sensors.…”

“…With a coarser spatial resolution, monitoring of small and medium inland lakes becomes difficult (Murray et al 2022). Moreover, satellite imagery is also affected by several environmental and climatic effects (Khan et al 2022) such as cloud cover (Murray et al 2022).…”