Explore and visualize GLSEA and GLSEA3 surface temperature satellite data for the Great Lakes

[DaniJonesOcean]

Issue Description:

The GLSEA and GLSEA3 datasets contain detailed records of the Great Lakes' surface temperatures and ice from satellite observations. While GLSEA3 uses newer satellite data, GLSEA offers a longer historical record, starting from 1995. We plan to use these datasets to train DeepSensor, our neural process model. Your job is to dive into these datasets, get to know them well, and assess their potential for our project.

Tasks:

1. Dataset Access:

   • Familiarize yourself with and gain access to the GLSEA and GLSEA3 datasets located on the U-M Turbo Data Volume mounted on Great Lakes HPC at the following paths:
     • /nfs/turbo/seas-dannes/SST-sensor-placement-input/GLSEA_NETCDF
     • /nfs/turbo/seas-dannes/SST-sensor-placement-input/GLSEA3_NETCDF

2. Comprehend Dataset Structure and Metadata:

   • Visit the GLSEA project page for an in-depth understanding of the satellite data. References are available on that page with further information.

3. Data Exploration:

   • Conduct an exploratory data analysis on the datasets to discern temporal and spatial variations in surface temperature.
   • Examine how the GLSEA/GLSEA3 datasets deal with ice cover - does it mask out the near-surface temperature at locations that are covered with ice? How GLSEA/GLSEA3 handles this will have implications for our training process.
   • In combination with issue #15, create a spatial plot of surface temperature for one of the lakes (your choice).
   • Record any findings, including anomalies or interesting patterns, that could impact our understanding of the Great Lakes' surface environmental conditions. (This does not have to be in-depth. This is about getting familiar with the dataset and exploring it.)

4. Visualization and Interpretation:

   • Create visualizations that effectively communicate the dynamics of the Great Lakes' surface temperatures and ice cover using xarray and matplotlib.
   • Interpret these visualizations to extract patterns that could possibly be used to inform the DeepSensor training process.

Deliverable:

• A summary of your exploration, visualizations produced, and insights gained from the GLSEA and GLSEA3 datasets. (Comments and images attached to this issue would be fine.)
• Reflections on the process, including any obstacles encountered and how they were overcome, or if further assistance is needed.
• Thoughts on how this dataset could integrate with our other analyses, models, or influence the project direction.

Comparison with Great Lakes Coastwatch Node:

• If desired, you can compare some of your results with the GLSEA and ice concentration plots available on the NOAA Coastwatch Great Lakes Node, which is curated by GLERL

EDIT: Note that @eredding02 has produced some initial plots and attached them to issue https://github.com/CIGLR-ai-lab/GreatLakes-TempSensors/issues/12.

EDIT: Added a "plot the isolated lakes" point to the data exploration step.

[eredding02]

@DaniJonesOcean Along with the plots in issue 12, I have compiled my data exploration here. I looked at the severe winter of 2009, where I was able to see that ice cover corresponded to a value of 0.2C. We can use this information to create a binary mask for ice cover to feed to DeepSensor. We are curious how DeepSensor will handle the sudden variability of temperature, seen in the time series figures. I have also isolated each lake based on latitude and longitude(seen in slides), like cropping a picture. I have not yet plotted an individual lake in the same scale as the lakes as a whole.

[DaniJonesOcean]

@eredding02 Apologies for the delay - very nice work and great summary! I'm glad that we can select individual lakes now, and as you said, we're in a good position to create the binary ice masks.

Feel free to close this issue : )