[GSoC Project Proposal]: Extend CrocoLake's available datasets

[enrico-mi]

Project Description

CrocoLake is a datalake gathering several physical and biogeochemical ocean observations, with the goal of providing an efficient format and a unified interface for data assimilation and ocean modeling activities. CrocoLakeTools contains the python modules to convert existing datasets to CrocoLake's structure (i.e. parquet format with common schema), and a unified interface to load them in the same dataframe.

This project consists in taking an existing dataset that is not yet included in CrocoLake and developing or adapting existing modules to convert it to CrocoLake's format. The list of possible datasets to include has already been put together by the mentor. The dataset to convert will be chosen together with the mentee, based on their experience and familiarity with the format and size of the original database. The project can then be tailored to the mentee's interests and skills: from the conversion of a single-file csv dataset, to the multi-processing conversion of a dataset containing multiple netCDF files.

Expected Outcomes

• CrocoLake's success depends on the amount of observations that it can serve in a uniform manner, hence the importance of this project.
• Adding more modules to convert datasets increases the examples on how to build one, reducing the efforts of future users that are interested in adding their own datasets.
• CrocoLakeTools is at its infant stage, and this project would also provide feedback on its accessibility to new users.

Skills Required

Python (pandas, xarray), git

Additional Background/Issues

CrocoLakeTools current version is here.

Mentor(s)

Enrico Milanese (enrico.milanese@whoi.edu, @enrico-mi)

Expected Project Size

175 hours

Project Difficulty

Intermediate

[MathewBiddle]

@enrico-mi can you provide a link to the CrocoLakeTools GitHub repository?

[enrico-mi]

Hi @MathewBiddle -- I'm working on publishing it and sharing the link by the end of the day

[enrico-mi]

@MathewBiddle I have added a link to the repo

[RATED-R-SUNDRAM]

HI @MathewBiddle , @enrico-mi I am interested in working on this project , My background includes familiarity with python (over 4 years including packages such as pandas, numpy, xarray, sklearn, etc.) , Git/Github , Cloud technologies(Azure) , and other data lifecycle management skills ( building data pipelines, warehousing etc. ),

I would love to get a headstart and guidance on where to get started on this project.

[RATED-R-SUNDRAM]

Hi @MathewBiddle . @enrico-mi , I've brushed up my understandings of xarray and NETCDF files along with certain conventions for fair data, If you can provide some headstart and guidance on where to get started on this project , I can start contributing to this issue

[enrico-mi]

Hi @RATED-R-SUNDRAM , thank you for your interest in the project. Our application with Google is still pending so we cannot get started yet -- we should have news in a couple of weeks and we'll be able to get started then in case. Thank you for your patience.

[RATED-R-SUNDRAM]

Sure @enrico-mi , looking forward to it

[Mirandazhu02]

Hi @enrico-mi, @david11133,

My name is Miranda Zhu. I'd like to contribute to the CrocoLake project. As a junior studying data science at UC Berkeley, I have a strong foundation in Python and experiences working with databases. After reviewing the repository, I have some questions about how to best contribute to the project:

1. Looking at both the ARGO and GLODAP converters, I notice they handle different data sources and formats – ARGO processes already-converted parquet data with quality filtering, while GLODAP converts directly from CSV. For the CPR dataset with its rich biological parameters, which approach would you recommend as more suitable, and what specific data preprocessing challenges should I anticipate given the biological nature of the data?

2. The CPR dataset has 422 parameters, which is significantly more than the current datasets in CrocoLake. How would you suggest I approach the parameter mapping strategy to maintain consistency with TRITON standards while accommodating this large number of biological variables?

3. As a new contributor to CrocoLakeTools, what would be the most effective way to begin my work? Would you recommend first setting up the development environment with the latest code from the 'from-private' branch, exploring one of the existing converters in detail, or perhaps starting with documentation improvements while gaining familiarity with the codebase?

[david11133]

Hi @Mirandazhu02 ,

Just to clarify, I'm not a mentor, but I'm a contributor, just like you! Currently, I'm working on the CPR converter and am happy to share what I know.

To get started, I highly recommend setting up the development environment with the latest develop branch and exploring the existing converters, like GLODAP. This will give you a solid understanding of the workflow. Contributing to documentation or tackling smaller fixes is also a great way to familiarize yourself with the codebase.

If you're interested in adding a new conversion module, I suggest using existing modules, like converterGLODAP.py or converterSprayGliders.py, as references. Following their structure will make the process much easier.

Once Enrico is available, he'll be happy to provide example datasets to help you get started.

Best of luck!

[enrico-mi]

Hi @enrico-mi, @david11133,

My name is Miranda Zhu. I'd like to contribute to the CrocoLake project. As a junior studying data science at UC Berkeley, I have a strong foundation in Python and experiences working with databases. After reviewing the repository, I have some questions about how to best contribute to the project:

Hi @Mirandazhu02 , great to hear of your interest!

1. Looking at both the ARGO and GLODAP converters, I notice they handle different data sources and formats – ARGO processes already-converted parquet data with quality filtering, while GLODAP converts directly from CSV. For the CPR dataset with its rich biological parameters, which approach would you recommend as more suitable, and what specific data preprocessing challenges should I anticipate given the biological nature of the data?

CPR is also .csv so the GLODAP converter should be a good reference to start.

3. The CPR dataset has 422 parameters, which is significantly more than the current datasets in CrocoLake. How would you suggest I approach the parameter mapping strategy to maintain consistency with TRITON standards while accommodating this large number of biological variables?

At the moment we're keeping all the parameters, and the mapping is done on the coordinates (latitude, longitude, time). We'll trim the parameters later, as I am waiting on colleagues' feedback on that point.

5. As a new contributor to CrocoLakeTools, what would be the most effective way to begin my work? Would you recommend first setting up the development environment with the latest code from the 'from-private' branch, exploring one of the existing converters in detail, or perhaps starting with documentation improvements while gaining familiarity with the codebase?

I have added a CONTRIBUTE.md file since you wrote this message, so you can look at that now for more info.
Please note that @david11133 is already working on the CPR dataset at the moment, so I suggest you to look at other datasets (e.g. Saildrones or GDP) and that you reach out over email to discuss what you feel comfortable tackling.