Water quality assessment using Landsat 8 and Sentinel-2 : a case study of the Umdloti estuary, KwaZulu-Natal, South Africa.
| dc.contributor.advisor | Gebreslasie, Michael Teweldemedhin. | |
| dc.contributor.author | Ellero, Matthew. | |
| dc.date.accessioned | 2019-07-23T13:00:47Z | |
| dc.date.available | 2019-07-23T13:00:47Z | |
| dc.date.created | 2018 | |
| dc.date.issued | 2018 | |
| dc.description | Master of Science in Environmental Sciences. University of KwaZulu-Natal, Durban, 2018. | en_US |
| dc.description.abstract | Estuaries are amongst the most productive and important ecosystems on the planet but are vulnerable to physico-chemical alterations within their waters. Water quality and health monitoring therefore requires the timely retrieval of physico-chemical concentrations. Whilst accurate, water quality retrieval performed through field sampling is often expensive, time consuming and unsuitable across larger spatial scales. remote sensing offers a possible solution to these problems. This study therefore assessed the ability of Landsat 8 and Sentinel-2 remotely sensed imagery in estimating physico-chemical health within the Umdloti Estuary, South Africa. Sampling of the Umdloti Estuary was conducted over winter and spring where the in situ conditions of temperature, turbidity, secchi disk depth (SDD), salinity, electrical conductivity (EC), total dissolved solids (TDS), chlorophyll-a (chl-a), dissolved oxygen (DO) and pH were retrieved through field and lab testing. Remote sensing algorithms were thereafter used to estimate the values of these parameters. Results from the comparison of the two approaches showed that temperature and turbidity were able to be accurately retrieved with best respective coefficients of determination (R2) of 0.96 and 0.97 and root mean squared error (RMSE) of 2.648 °C and 2.944 NTU. Chl-a, TDS and EC had respective inaccurate R2 values of 0.031, 0.576 and 0.037 but accurate RMSE of 0.902 μg/l, 638.159 mg/l and 1.801 μS/cm. These parameters were poorly modelled but accurate absolute concentrations could be estimated. Salinity, DO, pH and SDD estimation was poor. These parameters had a respective R2 values of 0.45, 0.53, 0.23 and 0.0007 as well as RMSE of 5.84, 1.91 ppm, 2.15 and 1.43 m. The optical inactivity of these physico-chemical parameters and unique complexity of estuarine waters were likely culprits in failed estimation. Positively, algorithms modified by this study showed greatly increased accuracy and future promise in the estimation of every parameter except secchi disk depth and pH. The estuary was determined to be in poor health due to a lack of improvements in physico-chemical conditions since it was last classified as poor. In future, studies should continue to refine algorithms for use in Umdloti Estuary and its health should be safeguarded. | en_US |
| dc.identifier.uri | https://researchspace.ukzn.ac.za/handle/10413/16390 | |
| dc.language.iso | en | en_US |
| dc.subject.other | Water quality. | en_US |
| dc.subject.other | Physiochemical properties. | en_US |
| dc.subject.other | Landsat 8. | en_US |
| dc.subject.other | Sentinel 2. | en_US |
| dc.title | Water quality assessment using Landsat 8 and Sentinel-2 : a case study of the Umdloti estuary, KwaZulu-Natal, South Africa. | en_US |
| dc.type | Thesis | en_US |