TL;DR
This study presents the first global map of over 3.4 million lakes and their dynamics over four decades, revealing a significant increase in lake area driven by reservoirs. Crucially, small lakes (<1 km²) are shown to dominate lake size variability and disproportionately contribute to increased carbon emissions, highlighting their overlooked role in global environmental change.
What Problem Does This Paper Address?
Accurately characterizing the world's lakes and understanding their long-term changes is vital for assessing impacts on water availability and carbon emissions. However, existing global lake datasets often rely on older imagery, lack seasonal and interannual variability, and struggle to accurately map small lakes (<1 km²), which are more dynamic and have significant ecological roles. This gap makes it difficult to quantify the true extent of lake changes and their consequences, particularly concerning carbon cycling.
Methods
The researchers developed a new global lake dataset called GLAKES by combining the Global Surface Water Occurrence (GSWO) dataset with a deep learning classification algorithm. This approach allowed them to identify lakes as small as 0.03 km², significantly improving the minimum mapping unit compared to previous datasets. They trained two specialized U-Net models (one for floodplains and one for normal areas) to distinguish lakes from rivers in satellite imagery. Post-processing steps were implemented to refine the lake boundaries, exclude ocean and river pixels, and differentiate between natural lakes and reservoirs using external datasets. Lake area changes were analyzed over three periods (1984-1999, 2000-2009, 2010-2019) using probability-weighted area calculations.
Datasets
- Global Surface Water Occurrence (GSWO) dataset: Used as the primary source for water presence information over nearly four decades (1984-2019) at 30m resolution.
- GSW monthly water history (MWH) data: Used for calculating water occurrence.
- Randolph Glacier Inventory 6.0 (RGI 6.0) and IMBIE Rignot dataset: Used to identify glacier-fed lakes.
- Permafrost distribution data (NASA): Used to identify permafrost-fed lakes.
- HydroBASINS dataset: Used to identify endorheic lakes.
- Georeferenced global Dam And Reservoir (GeoDAR) dataset: Used to distinguish reservoirs from natural lakes.
- Gridded Population of the World (GPW) dataset: Used to investigate the relationship between human population and lake area changes.
- Global River Widths from Landsat (GRWL) database and OpenStreetMap Water Layer (OSMWL) dataset: Used for masking land and excluding river pixels during label preparation.
- HydroLAKES polygon dataset: Used as a reference for lake labels.
The study introduces the GLAKES dataset as a new benchmark for global lake mapping, characterized by its global coverage, high spatial resolution, long-term dynamics, and improved detection of small lakes.
Key Insights and Ideas
- Reservoirs are the primary drivers of lake expansion: While climate change contributes, human-made reservoirs account for a substantial portion (56%) of the observed global lake area increase over the past four decades.
- Small lakes are surprisingly influential: Despite comprising only 15% of the global lake area, small lakes (<1 km²) dominate the variability in total lake size in half of the global inland lake regions. This suggests they are highly sensitive indicators of environmental change.
- Small lakes amplify carbon emissions: The increase in lake area, particularly due to small lakes, leads to higher lacustrine carbon emissions, highlighting their disproportionate impact on the global carbon cycle.
- Global lake area is increasing: Contrary to some localized observations, the study found a net increase in global lake area over the four-decade period.
Key Results
- Global lake area increased by +46,278 km² from 1984-1999 to 2010-2019.
- 56% of this expansion was attributed to reservoirs.
- Small lakes (<1 km²) accounted for 94.39% of the total number of lakes but only 15% of the total lake area.
- Small lakes dominated lake size variability in 49.9% of grid cells from 1980-1990s to 2000s, and 50.1% from 2000s to 2010s.
- Estimated global CO2 emissions from lakes are 226 Tg C yr⁻¹, and CH4 emissions are 1.4 Tg C yr⁻¹.
- Small lakes contribute disproportionately to emissions: 25% of CO2 and 37% of CH4 emissions are from small lakes, despite their smaller area share.
- Net increases in carbon emissions were observed: +4.81 Tg C yr⁻¹ for CO2 and +0.03 Tg C yr⁻¹ for CH4 from the 1980s-1990s to the 2010s.
Limitations and Future Work
The authors acknowledge several limitations:
- Lake definition: The study's definition of lakes includes some agricultural fields and parts of wetlands, which may affect precise interpretations.
- Temporal dynamics: While the study covers four decades, the influence of seasonal lake dynamics on area changes and carbon emissions is noted as a potential factor.
- Carbon emission upscaling: The accuracy of global carbon emission estimates relies on the representativeness of local flux measurements and the accuracy of the lake area data.
- CH4 emission pathways: The quantification of CH4 emissions through different pathways remains a challenge.
Future work could involve:
- Further refining the GLAKES dataset by excluding non-lake features like agricultural fields.
- More detailed analysis of seasonal lake dynamics and their impact on carbon emissions.
- Improving the accuracy of flux measurements for upscaling global carbon emissions.
- Investigating the specific mechanisms driving the high variability and emission contributions of small lakes.
Summary updated