Estimation of satellite-derived lake water surface temperatures in the western Mediterranean: Integrating multi-source, multi-resolution imagery and a long-term field dataset using a time series approach

Salvatore G P Virdis; Nooch Soodcharoen; Antonella Lugliè; Bachisio M Padedda

doi:10.1016/j.scitotenv.2019.135567

Estimation of satellite-derived lake water surface temperatures in the western Mediterranean: Integrating multi-source, multi-resolution imagery and a long-term field dataset using a time series approach

Sci Total Environ. 2020 Mar 10:707:135567. doi: 10.1016/j.scitotenv.2019.135567. Epub 2019 Nov 18.

Authors

Salvatore G P Virdis¹, Nooch Soodcharoen², Antonella Lugliè³, Bachisio M Padedda³

Affiliations

¹ Department of Information & Communication Technologies, School of Engineering and Technology (SET), AIT Asian Institute of Technology, P.O. Box 4, Klong Luang, Pathum Thani 12120, Thailand. Electronic address: virdis@ait.ac.th.
² Department of Information & Communication Technologies, School of Engineering and Technology (SET), AIT Asian Institute of Technology, P.O. Box 4, Klong Luang, Pathum Thani 12120, Thailand.
³ Department of Architecture, Design and Urban Planning (DADU), University of Sassari, Via Piandanna 4, 07100 Sassari, Italy.

PMID: 31780156
DOI: 10.1016/j.scitotenv.2019.135567

Abstract

Lake surface water temperature (LSWT) is a key parameter to help study the environmental and ecological impacts of climate change. In this work, we measured the LSWT of 1 natural and 23 artificial lakes located on the island of Sardinia in the western Mediterranean, which is a region where changes in climate are projected to have significant impacts. By integrating multi-source and multi-resolution datasets of the Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat and long-term in situ temperature observations, we detected, measured, and analysed the LSWT trends during the period of 2000-2018 across all the investigated lakes. Methodologically, we demonstrated that a simplified approached based on Planck's equation for Landsat thermal infrared (TIR) data could be a valid alternative to radiative transfer equation retrieval methods for the retrieval of LSWT without loss of accuracy. Moreover, we demonstrated that rescaled and independently validated MOD112A-derived LSWT showed good accuracy, efficiently filled the spatial and temporal gaps in long-term in situ LSWT, and could be used for long-term LSWT trend detection and measurement. All 24 lakes showed an annual warming trend of +0.010 °C/y, warming winter trend of +0.013 °C/y, and cooling summer trend of -0.038 °C/y during the period of 2000-2018. This study demonstrated that the measured trend rates could be explained by and were strongly correlated with the climatology of Italy for the 2000-2018 period. Finally, we demonstrated the key role and the importance of the availability of long-term in situ temperature datasets. The approach used in this study is up-scalable to other medium to low-resolution TIR sensors as well as to other long-term monitoring sites, such as LTER-Italy, LTER-Europe, or ILTER sites.

Keywords: Climate change; In situ long-term dataset; LSWT; Mann-Kendall test; Mediterranean Sea; Nash-Sutcliffe efficiency; Sen's slope; Time series.