Statistical Remote Sensing for Prediction of Inland Water Quality Parameters for Shatt Al-Arab River in Iraq

Abstract

Abstract In this study, the empirical approach besides the methods multilinear regression was followed to generate the predictive mathematical models to estimate the water quality parameters values in the surface water of the Shatt Al-arab River which is located southern part of Iraq. And to show both benefits and the viability of using Landsat 8 optical spectral images to estimate some of the water quality parameters concentration. The daily water quality data archive for the of the total dissolved solids (TDS), conductivity (E.C), Nitrate (NO3), and potential hydrogen ions (pH) of the water in four seasons (winter, spring, summer, and autumn) distributed along three years (2013, 2014, and 2015), was collected from four ground stations along of the Shatt Al-arab River. The objective of the study was to generate seasonal empirical mathematical models for the (open time) that can be used every year, without the need for calibration every time. Optical data were corrected to remove radiometric and atmospheric error sources effects prior to the developing the models. Multiple regression analysis between measured water quality parameters of the ground stations and the reflectance of the pixels corresponding to the sampling stations was used to generate these models. Determination coefficients (R2) of the proposed mathematical models were between 0.83-0.99. The percentage error between predicted and measured values for these models were between 0.03% -12%. The results of this work indicate the novelty of the approach used to generated these mathematical models for the open time for any year but in the each season. These models are reliable to estimate the spatial and temporal variation of TDS, E.C, NO3, and pH. So models generated from Landsat 8 can be used as a tool to facilitate the environmental, economic, and social management of the surface waters bodies like a Shatt Al-arab River.