Add generation of land/water mask to gridded data processing

[DaniJonesOcean]

Background: Within environmental modeling, accurate delineations of land and water are crucial for various analyses, such as sensor placement, forecasting, and satellite gap-filling. We need a land/water mask for our project. This might be pretty straightforward, but it's worth laying out the full case below.

Issue: We need to be able to generate a gridded land/water mask programmatically, which can be integrated into the existing data preprocessing steps of GreatLakes-TempSensors. Currently, the determination of land vs. water regions relies on manual interpretation or external data sources that users need to integrate themselves.

Proposed Solution: Develop a method that can generate a land/water mask based on available environmental data. For instance, we might utilize non-observation points in our datasets or incorporate geospatial datasets that delineate these boundaries. The proposed solution should:

• Detect land/water boundaries and generate a corresponding binary mask as part of the gridded context set.
• Handle different data resolutions and modalities, ensuring consistency across diverse data sources used in DeepSensor.
• Operate efficiently to minimize the added computational load to data preprocessing. (Should be simple)

Discussion Points:

• What are the most reliable and efficient techniques for generating a land/water mask?
• Should we consider using existing libraries or datasets that provide this information or develop an in-house algorithm based on our data?
• How can we ensure the generated mask aligns with the data modalities and resolutions supported by DeepSensor?

Action Items:

• Research methods for generating land/water masks within gridded datasets.
• Assess potential external datasets for suitability.
• Design and implement a land/water mask generation feature in this repository
• Test the feature with various data sources to ensure accuracy and reliability.
• Update documentation to guide users on using the new land/water mask feature.

Potential Challenges: While integrating this feature, we need to consider the inherent data variability and missing values prevalent in environmental datasets. Ensuring the mask's accuracy across diverse datasets (e.g., in-situ measurements, satellite imagery) could be challenging.

[DaniJonesOcean]

@eredding02 Has produced an overall land/water mask

Now working on lake-specific land/water masks

[eredding02]

@DaniJonesOcean I have successfully created both lake-wide and lake specific masks for the Great Lakes. Using remotely sensed SST data, I set all values for lat/lon that a specific lake did not belong to to be Nan using xarray.where(). From there I set all Nan values to 0 and the rest to 1. Below are some of the resulting graphs.

[DaniJonesOcean]

@eredding02 Great! I'm glad to see these - nice work. Feel free to close this issue when you're ready