A vital resource for life and the advancement of human
civilization is water. The current water supplies are facing
serious problems due to the world's expanding population and
rising demand for clean water. Large-scale assessment of
Surface Water Quality Parameters (SWQPs) and routine
monitoring of water bodies are essential; on the other hand,
remote sensing offers broad geographical and temporal
coverage. This research models SWQPs by using ground truth
water quality data, satellite data, and Artificial Neural
Networks (ANNs). The surface reflectance value of the water
area has initially been produced using atmospheric correction
methods, such as Quick Atmospheric Correction (QUAC),
Dark Object Subtraction (DOS), Fast Line of Sight
Atmospheric Analysis of Hypercubes (FLAASH), and
Atmospheric Correction (ATCOR). The coefficient of
determination (R2) and Root Mean Squared Error (RMSE)
values for each atmospherically adjusted technique and the
Landsat8 reference data were then calculated using a Python
tool. Second, this research builds a Backpropagation Neural
Network (BPNN) model based on Landsat8 for modeling Total
Organic Carbon (TOC) and Total Dissolved Solids (TDS) using
the most accurate atmospherically adjusted technique,
FLAASH, with ANN. Thirdly, different epochs (800–10,000)
were used to train the created BPNN models in order to evaluate
their effect on the training procedure. The models will continue
to be accurate and cost-effective if the epochs are set to 1200,
which will save processing time and computational expenses
for subsequent projects.