Predicting Optical Water Quality Indicators from Remote Sensing Using Machine Learning Algorithms in Tropical Highlands of Ethiopia

Abstract

Water quality degradation of freshwater bodies is a concern worldwide, particularly in Africa, where data are scarce and standard water quality monitoring is expensive. This study explored the use of remote sensing imagery and machine learning (ML) algorithms as an alternative to standard field measuring for monitoring water quality in large and remote areas constrained by logistics and finance. Six machine learning (ML) algorithms integrated with Landsat 8 imagery were evaluated for their accuracy in predicting three optically active water quality indicators observed monthly in the period from August 2016 to April 2022: turbidity (TUR), total dissolved solids (TDS) and Chlorophyll a (Chl-a). The six ML algorithms studied were the artificial neural network (ANN), support vector machine regression (SVM), random forest regression (RF), XGBoost regression (XGB), AdaBoost regression (AB), and gradient boosting regression (GB) algorithms. XGB performed best at predicting Chl-a, with an R2 of 0.78, Nash–Sutcliffe efficiency (NSE) of 0.78, mean absolute relative error (MARE) of 0.082 and root mean squared error (RMSE) of 9.79 µg/L. RF performed best at predicting TDS (with an R2 of 0.79, NSE of 0.80, MARE of 0.082, and RMSE of 12.30 mg/L) and TUR (with an R2 of 0.80, NSE of 0.81, and MARE of 0.072 and RMSE of 7.82 NTU). The main challenges were data size, sampling frequency, and sampling resolution. To overcome the data limitation, we used a K-fold cross validation technique that could obtain the most out of the limited data to build a robust model. Furthermore, we also employed stratified sampling techniques to improve the ML modeling for turbidity. Thus, this study shows the possibility of monitoring water quality in large freshwater bodies with limited observed data using remote sensing integrated with ML algorithms, potentially enhancing decision making.

A Machine Learning Approach for the Estimation of Total Dissolved Solids Concentration in Lake Mead Using Electrical Conductivity and Temperature

“…Best-performing models have the sample points lying around the bisector line, indicating best estimates [64]. From the plots, it is shown that all models perform similarly, with R 2 ranging from 77 to 82%, with standalone models such as LR and ANN having the highest values of 82%, which is contrary to many studies reporting ensemble methods such as the RF and XGBoost to be best-performing models compared to standalone methods such as ANN, SVM, and LR [57,111,114], although studies such as [115,116] also found standalone ANN models to be better performers of WQPs than ensemble methods such as the RF. The findings from this study, therefore, mean that considering just one or few metrics in assessing the model performance may not necessarily give a better insight into the actual performance and that ensemble ML does not necessarily perform better than the standalone ML in every scenario.…”

“…In comparison, researchers in [111] found that the RF was the best-performing model for estimating TDS, with R 2 of 0.79, RMSE of 12.30 mg/L, MARE of 0.082, and NSE of 0.80, although they used remote sensing images as their features, and also conducted their research in a different water body (Lake Tana) located in Ethiopia (no external validation was carried out). Our research produced R 2 of 0.80, RMSE of 35.25 mg/L, MARE of 0.0342, and NSE of 0.80 on the external validation for the RF.…”

“…It is seen from the results that the GBM ensemble method generally performs better than all the models based on the metrics used, with relatively lower ranks across the datasets (average rankings of 3.0, 1.7, and 4.6, respectively, for the training, testing, and external validation datasets and an overall average of 3.1) due to its underlying operations of using complex relationships to iteratively build trees and learn from the errors of preceding weaker trees [109,111]. These results, therefore, demonstrate the reliability and efficiency of the GBM in estimating TDS from EC and temperature.…”

Implementation of remote sensing algorithms to estimate TOC, Chl-a and TDS in a tropical water body; Sanalona reservoir, Sinaloa, Mexico

“…In Sanalona study achieved an accurate model for TDS (R 2 = 0.9375) when using only with 3 bands (B2, B3, B5). By using remote sensing algorithms, Leggesse et al, (2023) obtained R 2 values of 0.78, 0.79 for Chla and TDS (respectively) when predicting the water quality of Lake Tana, Ethiopia using the OLI/LANDSAT-8 satellite images.…”

Measurement of Total Dissolved Solids and Total Suspended Solids in Water Systems: A Review of the Issues, Conventional, and Remote Sensing Techniques

“…Increased concentrations of TDS and TSS in water bodies limit them from serving their purpose for drinking, power generation, industrial cooling, supporting biodiversity, ecosystem services, recreation, transportation routes, waste disposal, agriculture production, irrigation, energy production, regional planning, and fish farming [7][8][9][10][11][12][13][14][15][16][17]. Impairment of water bodies by parameters such as TDS and TSS is caused by climate change, development, and urbanization associated with surface imperviousness resulting from increased population, and contamination caused by rapid and uncontrolled environmental changes including drought, wastewater discharges, nutrient pollution, sediments, and changes in land use and land cover which results in negative impacts such as the proliferation of harmful blue-green algae, accelerated eutrophication, and extreme turbidity among others which have negative implications on the sustainability of the limited water resources [3,6,[18][19][20][21][22][23][24][25][26][27][28][29].…”