digitalearthpacific/dep-seagrass

AI Summary: The task is to correct the creation of the probability Cloud Optimized GeoTIFF (COG) for seagrass habitat mapping. Specifically, the nodata value (currently 255) needs to be explicitly set within the COG metadata to ensure proper handling of areas without data.

AI Summary: Create two geoJSON files, each containing approximately 100 surf points for two specified locations. These files will supplement an existing surf mask by adding surf points in areas currently not covered. Specific locations are assigned to different individuals (@Maivunijale, @Sosi19301, @elenoab99) with a deadline of Thursday at 12pm.

AI Summary: The task involves scaling a seagrass habitat mapping model for wider deployment. This includes creating two Python functions (print_tasks.py and run_tasks.py), a Docker image, a YAML workflow, a STAC collection, a Datacube product, and a Datacube configuration. The primary focus is on creating a seagrass model with two output layers (presence/absence and probability) for 4-6 study sites. Prior experience with Docker, YAML, STAC, and Datacube is beneficial.

AI Summary: Clean, validate, and verify geotagged points for the Paunganisu area, mirroring the process used for Bootless Bay's seagrass habitat mapping. This involves data processing and quality control to ensure accurate representation of seagrass extents.

AI Summary: Debug a Python function (`do_prediction`) within a seagrass habitat mapping workflow. The function, used to generate predictions with a mask, is producing incorrect results. The issue likely lies in data handling (reordering or transposing arrays) within code cells 22-25 of a Jupyter Notebook. The task involves analyzing the code, identifying the error in data manipulation, and correcting it to produce the expected output.

AI Summary: Debug a Dask-parallelized xarray operation (`xr.apply_ufunc`) within a Jupyter Notebook used for seagrass habitat mapping. The operation calculates GLCM features from image patches and is stuck on a specific cell, potentially due to computational cost despite using parallelization. The task involves investigating the cause of the slowdown (e.g., insufficient resources, code inefficiency, data issues), potentially optimizing the code, and ensuring the operation completes successfully.

AI Summary: Review and improve a Jupyter notebook used for seagrass habitat mapping. The notebook processes geospatial data, incorporates texture data, and uses machine learning for prediction. The task involves reviewing existing code, integrating masking functions from other contributors, streamlining workflows, and ensuring accuracy before scaling to other datasets.

AI Summary: Document the most useful GLCM window sizes used in seagrass habitat mapping, including supporting images and justifications, for project record-keeping. This is in response to a discussion with the product development team and wider community on enhancing seagrass products.