Lake surface water temperature from 1995 to present derived from satellite observations

The Lake Surface Water Temperature (LSWT) dataset from the Climate Data Store provides up to daily mid-morning LSWT values from 1995 to the present, derived from ATSR, MODIS, AVHRR and SLSTR satellite sensors. Bias adjustments ensure consistency across sensors, though gaps may exist. LSWT is a key Essential Climate Variable (ECV) influencing lake ecology and climate interactions. Developed mainly from the UK NERC GloboLakes and the ESA ARCLake projects, it supports research on long-term lake-climate dynamics.

Key strengths

• Spatial and temporal resolution: The dataset provides spatial resolution of 0.05° x 0.05° (approximately 5 km x 5 km) and temporal resolution (up to daily resolution with gaps), which is essential for capturing detailed variations in lake surface water temperature. This level of detail is beneficial for studying medium and large lakes to understand short-term and spatial variations of LSWT.
• Comprehensive coverage: The dataset offers global coverage, allowing for the analysis of lake temperatures across different climatic and geographical regions. This broad coverage supports comparative studies and global climate assessments. Currently, the dataset includes LSWT for ~2000 lakes distributed across the globe.
• Long-term data availability: With data available over several decades, this dataset is suitable for studying long-term trends and changes in lake surface temperatures. It supports historical analysis and future projections related to climate change impacts on freshwater ecosystems.
• Suitability for long-term/large regions climate studies: This dataset is well-suited to perform long-term variability studies and to compare lakes on different regions because the LSWT has been retrieved using the same algorithm. This algorithm employs observations from instruments which correspond to seven different satellites and periods of time (some overlapping). These instruments are: ATSR-2, MODIS, AATSR, AVHRR-A, AVHRR-B, SLSTR-A and  SLSTR-B. A per-lake bias correction has been applied to take into account the potential bias arising from the use of different instruments.
• Availability of per-pixel uncertainty and quality flags: This dataset provides for each pixel the uncertainty on the LSWT and a quality level which indicates the confidence that the assumptions made for the retrieval are fully met and therefore suitable for climate studies.
• Integration with other datasets: This dataset can be integrated with other climate and environmental datasets available in the Climate Data Store, enabling comprehensive and multifaceted environmental studies.

Key limitations

• Water detection confidence: Some water pixels may be missed and some non-water pixels may be misclassified, impacting the accuracy of water detection. The water detection pre-processing step is threshold based. A common set of thresholds has been applied for all lakes leading to potential inaccuracies due to varying lake reflectivities or cloud contamination.
• Quality flags: The LSWT quality levels range from 2 (suspect/marginal quality) to 5 (best quality). For most applications, it is recommend to use only quality levels 4 and 5. However, LSWT with quality levels = 2 and 3 are present in the product, and users can assess their usefulness for their own application.
• Emissivity bias: Assumes emissivity of fresh water for all lakes, causing potential bias in highly saline lakes.
• Surface vs. Sub-surface temperature: Retrieved LSWT is associated to skin temperature, with a likely cold bias (~0.2 K but it can vary among lakes) compared to sub-surface measurements.
• Spatial resolution limitations: Narrow lakes may contain few water-only pixels due to the 1 km resolution of sensors, affecting data quality.
• Temporal density variability: Due to frequent cloud coverage, observation frequency varies greatly between lakes; some lakes may have few or no high-quality LSWTs, especially in the tropical area.
• Historical data limitation: Lower temporal density of observations before the availability of MODIS (from year 2000) and later (in 2007) of global full-resolution AVHRR (MetOpA) data due to the narrower swath of earlier ATSR series instruments.

Example applications

• Climate change: Jennings and Carrea (2020) found that satellite-derived LSWT data, combined with modelling, reveals a warming trend in European lakes' surface water temperature, primarily driven by rising air temperatures. Using satellite data from 115 lakes (1995- 2018), they validated simulations showing that during the 2018 heatwave, mean and maximum LSWT were 1.5°C and 2.4°C higher, respectively, than the base-period average (1981 - 2010).
• Lake-atmosphere interactions: Amadori et al. (2024) used satellite-derived LSWT data to study how Lake Garda responds to atmospheric variability. By fitting a statistical model to temperature anomalies, they estimated the lake's mixed layer depth, finding that deeper areas respond more slowly. Their results show that satellite data not only track surface temperature but also reveal mixing processes, providing insights into lake-atmosphere interactions and spatial variations in thermal inertia.