Loading...

Mapping the spatiotemporal variability of salinity in the hypersaline Lake Urmia using Sentinel-2 and Landsat-8 imagery

Bayati, M ; Sharif University of Technology | 2021

860 Viewed
  1. Type of Document: Article
  2. DOI: 10.1016/j.jhydrol.2021.126032
  3. Publisher: Elsevier B.V , 2021
  4. Abstract:
  5. The spatiotemporal dynamic of salinity concentration (SC) in saline lakes is strongly dependent on the rate of water flow into the lake, water circulation, wind speed, evaporation rate, and the phenomenon of salt precipitation and dissolution. Although in-situ observations most reliably quantify water quality metrics, the spatiotemporal distribution of such data are typically limited and cannot be readily extrapolated for either long-term projections or extensive areas. Alternatively, remotely sensed imagery has facilitated less expensive and a stronger ability to estimate water quality over a wide range of spatiotemporal resolutions. This study introduces an adaptive learning model that leverages in-situ measurements, satellite imagery, and machine learning algorithms to estimate the spatiotemporal changes of water surface SC in saline lakes. We demonstrate the applicability of the framework by using high-resolution Sentinel-2 and Landsat-8 satellite data in the Lake Urmia (LU), where 130 points were sampled in April and July 2019 for the model calibration and validation purposes. The results showed that Artificial Neural Network (ANN) yields the most accurate relationship between SC and water surface reflectance (R2 = 0.94; NRMSE = 6.9%; MAE = 12.3 ppt) as compared to Adaptive Network-based Fuzzy Inference System (ANFIS), and Multiple Linear Regression (MLR) models. A sharp distinction in SC between the LU's northern and southern parts was observed during high flows due to the limitation on north–south water mixing imposed by the lake's middle causeway. Nevertheless, the impact of causeway on water circulation at the other seasons is minimum. A hysteretic relationship was also found between the lake water volume and the lake-averaged SC due to the distinct role of bed-salt dissolution and high evaporation rate at different seasons. A tight agreement between the SC variations in the southern part and the whole lake proved that the southern part dominantly controls SC dynamics in the whole lake. © 2021 Elsevier B.V
  6. Keywords:
  7. Causeways ; Dissolution ; Evaporation ; Flow of water ; Fuzzy inference ; Fuzzy neural networks ; Lakes ; Learning algorithms ; Linear regression ; Machine learning ; Precipitation (chemical) ; Satellite imagery ; Water quality ; Wind ; Adaptive network based fuzzy inference system ; Model calibration and validation ; Multiple linear regression models ; Spatio-temporal dynamics ; Spatio-temporal resolution ; Spatiotemporal distributions ; Spatiotemporal variability ; Water surface reflectance ; Saline water ; Artificial neural network ; Concentration (composition) ; In situ measurement ; Lacustrine environment ; Landsat ; Mapping method ; Precipitation (chemistry) ; Salinity ; Satellite data ; Sentinel ; Spatiotemporal analysis ; Iran ; Lake Urmia
  8. Source: Journal of Hydrology ; Volume 595 , 2021 ; 00221694 (ISSN)
  9. URL: https://ieeexplore.ieee.org/document/9497112