shahinmg / melt_lake_datacube

Greenland Ice Sheet melt lake presence time series using STACs

melt_lake_datacube

How to detect melt lake presence on ice sheets using STACs

The melt_lake_sample_tutorial.ipynb in the melt_lake_datacube shows how to sample melt lakes using Microsoft Planetary Computer's Sentinel 1 Radiometrically Terrain Corrected (RTC) catalog. melt_lake_detection_s1_rtc_ec2.py is the script I ran on my ec2 instance and saved the resulting xarray dataset as a csv surface_CW2019_lake_mean_da_asc_hv.csv in the output directory. The read_csv_from_flox.ipynb is the code I wrote to plot the time series plots. Those plots are also in the output directory. See example from a single lake below

Installation

Below are instructions for installation with a UNIX terminal. This is tested on my ubuntu 22.04 system.

To use this repo first clone the repository and change to the directory where the repo is located

git clone https://github.com/shahinmg/melt_lake_datacube.git

cd melt_lake_datacube

Within the melt_lake_datacube directory, install the packages in the environment.yml in a conda environment or create a new environment and install with conda env create --file environment.yml and activate the environment

Note: This can take a long time to create the environment

conda env create --file environment.yml

conda activate melt_lake_datacube

Note: I created this yml using conda env export > environment.yml. Sometimes I have issues with that when I create other environments from someone else's repo. If you have conflicts from this yml, let me know and/or make an issue or pull request.

Thoughts on improvements

I tried to make a GPU native version using cupy-xarray cupy_xarray_test.ipynb, which is not working as of now. We should also try to speed things up for when we scale up. Some thoughts might be to use the cohorts method in the group by. More on that is here.

We also need to verify some lakes with optical imagery to see what is actually happening in the visible spectrum as well.