LAGOS-US LANDSAT: Remotely sensed water quality estimates for U.S. lakes over 4 ha from 1984 to 2020
Abstract Broad-scale, long-term water quality (WQ) studies are critical for understanding increasing pressures on inland waters but remain rare due to cost. The LAGOS-US LANDSAT dataset provides 37-year remote sensing-derived WQ estimates for thousands of U.S. lakes ≥ 4 ha (1984–2020). WQ estimates...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-07-01
|
| Series: | Scientific Data |
| Online Access: | https://doi.org/10.1038/s41597-025-05600-w |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Broad-scale, long-term water quality (WQ) studies are critical for understanding increasing pressures on inland waters but remain rare due to cost. The LAGOS-US LANDSAT dataset provides 37-year remote sensing-derived WQ estimates for thousands of U.S. lakes ≥ 4 ha (1984–2020). WQ estimates use machine-learning models with Landsat imagery and ground-truthed Water Quality Portal data (LAGOS-US LIMNO). The dataset includes: (a) 45.9 million whole-lake water reflectance (six bands and 15 band ratios); (b) 740,627 matchups from 13,756 lakes with in situ data for six WQ variables: chlorophyll, Secchi depth, true color, dissolved organic carbon, total suspended solids, or turbidity; and (c) predictions for each WQ variable across lake-time combinations with quality imagery. Two random forest models were fit for each variable: Holdout-data (75/25 spatially representative train-test split) and Full-data (trained on all data). Variance explained for the Full-data predictions ranged from 20.7% for TSS to 63.7% for Secchi depth. Imagery underwent cloud and pixel quality control, and workflow components were validated guiding future research. |
|---|---|
| ISSN: | 2052-4463 |