[Open Wells - Water Monitoring for Drinking Water Security]

[craigdsouza]

Change Detection - Open Wells in India

Context

Hello all! I work as an independent researcher with various non-profits on water related themes. There is a lot of interest in groundwater management in India since 80% of the country depends on groundwater and it's reserves are fast depleting. Many non-profits and government as well are attempting sporadic measurements of groundwater in open wells across the country. I am keen on understanding whether Planet imagery can complement this effort.

To share further details, most open wells are large enough (6m diameter or more) that I presume they will show up on an NDWI layer of Planetscope imagery, I don't know for sure though, I am trying to get access to imagery for non-profit use cases. I would like to do small proof of concept using this hack. Maybe an NDWI trendline across a year of imagery would work?

I sent this message to Jonathan Evens (Planet) and he suggested the following:
Hi Craig, thanks for reaching out. This is a very interesting and important use case. I agree, I believe the right approach would be to first use NDWI to see if you can see the open wells and how distinguishable they are. If your AOI is small, you could even purchase a small Analytic SR or Normalized basemap online and perform an NDWI operation in Explore. A more sophisticated customer model development approach could then serve as a second step.

below is what I'm hoping for

What would a hack team for this project work towards in 2 days?
Attempt to do a POC to determine when during the year an open well runs dry.

How would this project use Planet's data & platform?
Use PlanetScope Imagery Analytic SR/ Basemaps, to pull a time series of NDWI for a number of open wells. I have a kml of well locations I can share. Link will be added here shortly.

What obstacles or blockers do you think this project might run into?
I have very limited experience with Planet's APIs. However I am very familiar with GEE. Accessing data might be my biggest challenge.

why should people hack on THIS project in particular?
If we succeed with a POC here a number of institutes would love to pursue this as a larger project. This data would also be really beneficial to the country's National Drinking Water Programme.

If you're interested in hacking on this project, add a 👍 reaction to this post.

[oh-data-sci]

this interests me, but how would wells look in satellite imagery? how can such pixels be identified from other, dark pixels? would we need a labelled data set, and are such available to train a detection model?

[ASUsparky]

Hello Craig, wondering if you need a teammate for this idea?

[craigdsouza]

this interests me, but how would wells look in satellite imagery? how can such pixels be identified from other, dark pixels? would we need a labelled data set, and are such available to train a detection model?

hey @oh-data-sci I'm guessing that since individual Planet pixels are 3 m and wells are larger than 6 m most often, the NDWI index should show a consistently high value in October each year (that's the time of year when the monsoon ends and the wells are usually full)

[craigdsouza]

Hello Craig, wondering if you need a teammate for this idea?

Hey @ASUsparky most definitely, I'm still new to Planet imagery, just thought working on an idea would be the best way to learn.

[craigdsouza]

to anyone who is keen on discussing this I'm thinking about scheduling a zoom at 11am (PDT) . Will post a link here

[ASUsparky]

Hello Craig, wondering if you need a teammate for this idea?

Hey @ASUsparky most definitely, I'm still new to Planet imagery, just thought working on an idea would be the best way to learn.

I am new to it as well. Let's explore together :)

[craigdsouza]

here's a link to dropbox for my AOI, Pabal village in central India, and also a kml with approximately 250 open wells . https://www.dropbox.com/sh/ekhbwwdyx70fhs8/AAA_ZF1XXkV7vPTvgv5dX8Fma?dl=0

[oh-data-sci]

hey @oh-data-sci I'm guessing that since individual Planet pixels are 3 m and wells are larger than 6 m most often, the NDWI index should show a consistently high value in October each year (that's the time of year when the monsoon ends and the wells are usually full)

ok, and then you'd just need to find pixels that are consistently dark in october and not in the dry season. and that way you could tell it apart from say dark rocks, roads, painted roofs, etc?

[oh-data-sci]

this interests me, but how would wells look in satellite imagery? how can such pixels be identified from other, dark pixels? would we need a labelled data set, and are such available to train a detection model?

hey @oh-data-sci I'm guessing that since individual Planet pixels are 3 m and wells are larger than 6 m most often, the NDWI index should show a consistently high value in October each year (that's the time of year when the monsoon ends and the wells are usually full)

but that would still not easily separate out wells from other dark pixels, say roads of areas of e.g. wet, dark rocks or painted house roofs?

[craigdsouza]

hey @oh-data-sci I'm guessing that since individual Planet pixels are 3 m and wells are larger than 6 m most often, the NDWI index should show a consistently high value in October each year (that's the time of year when the monsoon ends and the wells are usually full)

ok, and then you'd just need to find pixels that are consistently dark in october and not in the dry season. and that way you could tell it apart from say dark rocks, roads, painted roofs, etc?

hmm.. i'm skeptical if just looking for single dark pixels would work, but I'm thinking consistently high NDWI values in October would be a better indicator, and low values in May

also hey, I have kml of 250 open wells georeferenced, shared via dropbox above, it's not a really large training set, but it's something to start with.

[oh-data-sci]

nice. i downloaded and am reading these kml files now.

[craigdsouza]

for anyone interested , i am doing a zoom call to discuss this at 11 am PDT

Craig Dsouza is inviting you to a scheduled Zoom meeting.

Topic: [Open Wells - Water Monitoring for Drinking Water Security]
Time: Oct 15, 2020 11:00 AM Pacific Time (US and Canada)

Join Zoom Meeting
https://zoom.us/j/4047470413?pwd=Um5EeWw5dks0dWh2bDI1T044Y3RxUT09

Meeting ID: 404 747 0413
Passcode: 6aUiui

[skyprince999]

This is interesting! Couldn't join your team. But good luck.
Would love to connect with you on LinkedIn-

https://www.linkedin.com/in/aakash-gupta-5ky/

[craigdsouza]

This is interesting! Couldn't join your team. But good luck.
Would love to connect with you on LinkedIn-

https://www.linkedin.com/in/aakash-gupta-5ky/

thanks @skyprince999 , will connect with you next week!

[skyprince999]

@craigdsouza you were searching for some resources to get started with GEE. Did you find it? Were you able to pull the data & analyze it on GEE server?