Development of a new spatially complete and daily continuous lake surface water temperature dataset for Lake Vänern, Sweden

Abstract

Lakes are an important part of the world’s ecosystems. The ecological state of lakes is threatened by rising temperatures that affect the biological, physical and chemical cycles. Therefore, it is essential to monitor lake surface water temperature (LSWT) and its spa-tiotemporal variabilities. Currently monitoring LSWT employs three primary approach-es: in-situ measurements, satellite remote sensing, and reanalysis products/modelling. Each has its advantages and limitations. In-situ measurements offer accuracy at the point scale but suffer from inconsistencies and infrequent data collection. Satellite remote sensing provides relatively high spatial resolution but is affected by cloud cover and data gaps. Reanalysis products offer all-weather data but often at a coarse spatial resolu-tion, limiting their ability to capture fine spatial scale variations in LSWT. This study aims to develop a new spatially complete and daily continuous LSWT by fusing satellite LSWT product and reanalysis product for Lake Vänern, the largest lake in Sweden. The reanalysis product ERA5-Land providing hourly lake temperature at the spatial resolu-tion of 0.1° was used. Five existing satellite LSWT products were evaluated against in-situ measurements. The MODIS LSWT product was identified as the most suitable sat-ellite product to be fused with ERA5-Land data using the Enhanced Spatial and Tem-poral Adaptive Reflectance Model (ESTARFM). A bias correction was conducted to account for systematic bias resulting from the data fusion. The bias-corrected fused LSWT dataset was evaluated against in-situ measurements and showed higher accuracy than the MODIS and ERA5-Land data with a mean absolute error of 1.57 °C, root mean square error of 2.04 °C and R2 of 0.87. The spatial and temporal variations of the bias-corrected fused LSWT were in good agreement with the ERA5-Land and MODIS-derived LSWT, as well as with in-situ measurements. Finally, the bias-corrected fused LSWT product was used to investigate the spatial and temporal dynamics of Lake Vä-nern, revealing the development of a thermal bar and seasonal LSWT changes. This study demonstrated the good performance of the data fusion approach in generating a spatially complete and temporally continuous LSWT dataset. This approach is valuable for LSWT monitoring and further investigation of ecological changes in lakes associat-ed with shifting LSWT.

Abstract

Development of a new spatially complete and daily continuous lake surface water temperature dataset for Lake Vänern, Sweden

Keywords: Physical Geography, Ecosystem Analysis, Remote Sensing, Data Fusion, Reanalysis Product, Lake Surface Water Temperature, Hydrology

Lake Surface Water Temperature (LSWT) is an important indicator to monitor changes due to climate change on a regional scale. Changes in LSWT alter biological, chemical and physical processes, which affects the ecosystem of the lake. Currently, the methods to obtain LSWT are based on three main methods: in-situ measurements, satellite remote sensing and re-analysis products. However, all three methods come with their limitations, In-situ measurements have a sparse spatial and temporal network and are therefore not suitable to monitor spatial and temporal dynamics, satellite remote sensing is technically able to capture the spatial and temporal dynamics of LSWT, but cloud coverage results in data gaps. Re-analysis products generate hourly LSWT estimations under all weather conditions, although the coarse spatial resolution limits the method to capture small-scale spatial dynamics. Previous studies have used the advantage of satellite remote sensing and re-analysis products to overcome the current limitations of each product to generate a spatially complete and temporal dataset.

The study aims to develop a daily and complete LSWT dataset for Lake Vänern. Therefore, already existing LSWT products (GloboLakes, ARC-Lake, MODIS, TIRS, CGLOPS and ERA5-Land) for Lake Vänern were evaluated by comparing them to in-situ measurements. Further, a new LSWT dataset using ESTARFM was developed. The data fusion was done by combining the cloud-free days of a satellite product with the ERA5-Land data. To account for a systematic bias caused during the data fusion a bias correction was conducted. The generated product was then evaluated against in-situ measurements. Finally, the spatial and temporal dynamics of Lake Vänern were analyzed based on the generated dataset.

The evaluation of the lake surface water temperature products showed that each product performed with high accuracy. It was found that ATSR2 and GloboLakes performed the best for Lake Vänern. Due to its technical advantages, it was decided to conduct the data fusion with MODIS. The dataset was able to estimate the lake surface water temperature without data gaps and with a high spatial resolution. The results indicate that the generated product outperformed the MODIS and had similar accuracy as the ERA5-Land data. Further, the analysis of the spatial and temporal dynamics of LSWT in Lake Vänern showed a development of a thermal bar and that the maximum temperatures were reached during July/August.

Advisor: Zheng Duan
Master's degree project 30 credits in Physical Geography and Ecosystem Science, 2023
Department of Physical Geography and Ecosystem Science, Lund University. Student thesis series INES nr 621

Popular Abstract

Development of a new spatially complete and daily continuous lake surface water temperature dataset for Lake Vänern, Sweden

Keywords: Physical Geography, Ecosystem Analysis, Remote Sensing, Data Fusion, Reanalysis Product, Lake Surface Water Temperature, Hydrology

Lake surface water temperature is an important indicator to monitor changes on a regional scale caused by climate change. Changes in lake surface water temperature alter biological, chemical and physical processes, which affects the ecosystem of the lake. Currently, the main methods to obtain lake surface water temperature are based on three approaches: in-situ measurements, satellite estimations and modeling. However, all three methods come with their limitations: the in-situ measurements only represent the temperature at a point rather than the whole lake. The satellite estimates are, in the best case, available for the whole lake but due to cloud contamination these estimates have missing data. The modeling products generate hourly lake surface water temperature estimations without data gaps, although the coarse spatial resolution limits this method to capture small-scale variations in lake surface water temperature. Previous studies have used the advantages of the satellite and modeling products to overcome the current limitations by combining the data to gain a high spatial resolution and complete dataset.

The study aims to develop a daily and complete lake surface water temperature dataset for Lake Vänern. Therefore, already existing lake surface water temperature products (GloboLakes, ARC-Lake, MODIS, TIRS, CGLOPS and ERA5-Land) were evaluated for Lake Vänern by comparing them to in-situ measurements. Further, a new lake surface water temperature dataset using a data fusion approach, the combination of two datasets, was developed. The data fusion was conducted using the cloud-free days of a satellite product and the estimations of the modeling data. To account for a systematic error caused from the data fusion a bias correction was conducted. The generated product was then evaluated against in-situ measurements. Finally, the spatial and temporal dynamics of Lake Vänern were analyzed based on the generated dataset.

The evaluation of the lake surface water temperature products showed that each product performed with high accuracy. It was found that ATSR2 and GloboLakes performed the best for Lake Vänern. Due to its technical advantages, it was decided to conduct the data fusion with MODIS. The results indicated that the generated product outperformed the MODIS and had similar accuracy as the ERA5-Land data. The dataset was able to estimate the lake surface water temperature without data gaps and with a high spatial resolution. Further, the analysis of the spatial and temporal dynamics of lake surface water temperature in Lake Vänern showed a development of a thermal bar and that the maximum temperatures were reached during July/August.

Advisor: Zheng Duan
Master's degree project 30 credits in Physical Geography and Ecosystem Science, 2023
Department of Physical Geography and Ecosystem Science, Lund University. Student thesis series INES nr 621

• author: Schultze, Anna
• supervisor: Zheng Duan
• organization: Dept of Physical Geography and Ecosystem Science
• course: NGEM01 20231
• year: 2023
• type: H2 - Master's Degree (Two Years)
• subject: Earth and Environmental Sciences
• keywords: Physical Geography, Ecosystem Analysis, Remote Sensing, Data Fusion, Reanalysis Product, Lake Surface Water Temperature, Hydrology
• publication/series: Student thesis series INES
• report number: 621
• language: English
• id: 9131238
• date added to LUP: 2023-06-29 08:48:53
• date last changed: 2023-06-29 08:48:53