Accuracy of remote sensing-derived SST values compared with in-situ measurements in Eastern Mediterranean (Cyprus)

Abstract

Sea surface temperature (SST) is an essential climate variable that plays a significant role in regulating climate and its variability; therefore, assessing the quality of satellite-derived information is important. Remote sensing can provide sea surface data in high spatial and temporal resolution and with large spatial coverage. This allows new discoveries and applications in oceanography and marine ecology and investigation of sensitive regions as the Mediterranean Sea. However, a validation of satellite data with in-situ seawater temperature (SWT) measurements is needed to capture environmental variability at small spatial scales and in near-shore environments, particularly in regions where SWT records are absent. The aim of this work is to compare remote-sensing SST values with in-situ SWT recorded with data loggers around the island of Cyprus (Eastern Mediterranean). More precisely, we aimed to evaluate the ability of Global 1-km Sea Surface Temperature (G1SST) dataset to detect overall variability and intra-seasonal variability of SWT. In-situ SWT data were collected from -4m and -25m depths at four locations over different time-periods (between 2013 and 2016). A daily, global blended Level 4 SST data set of ultra-high resolution (1km) derived from the Global High-Resolution SST Pilot Project (GHRSST-PP) was used for the validation. The satellite image database provided by NASA Physical Oceanography Distributed Active Archive Center (PODAAC) web servers was in network common data form (NetCDF) format. A comparison between daily mean SST and daily mean SWT for all sites and seasons pooled together yield a very high correlation and biases. The stronger correlations with almost a perfect data fit obtained from nearshore shallow sample locations, while the weaker correlations derived from deep-water areas. Moreover, satellite-derived data presents a tendency to over-estimate SWT variability in all seasons while strong correlations (r > 0.80) are presented in all sample locations on cold seasons. In the other hand, during the hot seasons, weaker correlations are presented mostly on deep- water locations. In overall statistics analysis and taking into account the high correlation coefficients, G1SST data proved to be a reliable proxy of SWT and mostly for studies requiring temperature reconstruction in areas where in-situ SWT observations are not available or a time series is required to identify seasonality in the record.