PS 3-37 - lakeCoSTR: a tool to access and synthesize Landsat lake surface water temperature

Monday, August 15, 2022

5:00 PM – 6:30 PM EDT

Located in: ESA Exhibit Hall

Presenting Author(s)

Bethel G. Steele

Senior Research Specialist
Cary Institute of Ecosystem Studies
Millbrook, New York, United States

Coauthor(s)

CH

Christina Herrick

University of New Hampshire
Durham, New Hampshire, United States
JB

Jennifer A. Brentrup

Cary Institute of Ecosystem Studies and Dartmouth College
Millbrook, New York, United States
KC

Kathryn L. Cottingham

Dartmouth College
Hanover, New Hampshire, United States
MD

Mark J. Ducey

University of New Hampshire
Durham, New Hampshire, United States
DL

David A. Lutz

Dartmouth College
Hanover, New Hampshire, United States
MP

Michael W. Palace

University of New Hampshire
Durham, New Hampshire, United States
MT

Michael Thompson

University of New Hampshire
Durham, New Hampshire, United States
Jessica V. Trout-Haney

Postdoc
Dartmouth College
HANOVER, New Hampshire, United States
KW

Kathleen C. Weathers

Cary Institute of Ecosystem Studies
Millbrook, New York, United States

Background/Question/Methods

For researchers not familiar with remote sensing data, gathering and processing these data are difficult tasks. Acquisition and synthesis of remote sensing data involves technical skill in coding (often in multiple programming languages), appropriate application of atmospheric correction, and often requires the integration of spatially heterogeneous remote sensing data with in situ data obtained at specific geographic locations. To increase accessibility of remotely-sensed surface temperature data for lakes, we developed lake Colab Surface Temperature Retrieval (lakeCoSTR). lakeCoSTR is a user-friendly, cloud-based script that gives ecologists who lack specialized training the ability to access the Landsat Collection 2 temperature product for lakes. The tool can also create a ground-truthing dataset with paired Landsat-in situ data when in situ data are available. In an example application of the tool, we retrieved lake surface water temperature (LSWT) data for Lake Sunapee, NH, USA, a lake with a long monitoring history, and used the Landsat-in situ comparison dataset from lakeCoSTR to assess the remotely sensed estimates of LSWT. As a further comparison, we processed all available Collection 1 Landsat thermal band data and used a single-channel algorithm to estimate LSWT, thereby identifying improvements of Landsat Collection 2 over Landsat Collection 1 data.

Results/Conclusions

Application of lakeCoSTR to Lake Sunapee increased accessibility and usability of Landsat-derived LSWT data. We found that focusing on Landsat scenes with narrow temperature distributions (kurtosis < 2) was an effective filter to remove possible cloud contamination. Kurtosis-filtered Collection 2 data at Lake Sunapee acquired using lakeCoSTR required no calibration to in situ data during July through September, when bias was less than 0.50℃ and mean absolute error (MAE) was less than 0.75℃ for each Landsat mission per month. In contrast, LSWT data derived from the Collection 1 single-channel algorithm always required calibration with in situ data prior to use. The MAE for the complete kurtosis-filtered Collection 2 dataset (May through November) was 0.60℃ overall, compared to 0.94℃ for the Collection 1 data. We documented temporal trends in Lake Sunapee surface water temperature within and among seasons. For example, in July and August the surface temperature, as measured in situ and with uncalibrated Collection 2 LSWT data, has risen at a rate of 0.08 and 0.07°C year^-1, respectively, since the early 1980s. As climate change continues to alter LSWT, lakeCoSTR can increase accessibility and usability of 40 years of Landsat-derived temperature data to answer pressing ecological questions.