Submissions for GOAT / RFC Hackathon, Dec. 1 - 2

Hi GOATees new and old!

This post is the place for applicants to the biddy GOAT Hackathon in Southbridge MA, Dec 1 - 2!

What’s that? check out the details!

Post the idea to bring to the Hackathon

Reply to this thread with the idea you’d like to bring to the hackathon. If you need a reminder of the mission and some examples, re-read the hackathon invite here.

In your post, include:

  • the concept
  • any work you’ve already done towards achieving it
  • who’s on your team (or if you are looking for people to join, or if it’s just you that’s ok)
  • what you think the final product/proof of concept/pitch might be.

RFC folks and other knowledgeable community members will be around to give advice, suggestions and help to get you ready.

Or - post something you wish someone would bring

So if you are a BFA member, or involved on the doing ag (not building ag tech) side of things, and there is a ag tech related topic of project you’d love to see, please post it here. These ideas will help stimulate participants and get them barking up the right tree.

Ok - post some ideas and let’s get the discussion going!

Ok, I’m going to cheat and submit one myself :slight_smile:

So there is a ton of evidence that compaction is a major predictor of yield, and compaction can be of particular concern in no-till type fields. It’s a meaningful piece of data to relate to a variety of factors and worth testing correlations to nutrition (probably as a suite of items, not just compaction by itself).

Problem is, compaction is measured using a penetrometer which looks like this… it’s fairly easy to use - stick into ground, read pressure as you go deeper. But, it’s super hard to interpret - how much pressure indicates a compaction zone? How consistently do you press? And it’s hard to measure depth, and you usually have to do this in several points int he field, so it becomes a pain.

As such, they aren’t as widely used in ag as one would expect given how important we know compaction is.

So… I would like to remove the interpretation layer, and clarify the method, so that people can quickly and consistently take compaction measurements using a penetrometer. There are in fact some companies doing this - and it looks very cool . Unfortunately, it’s expensive ($1500 euro) and not open source.

There is an extremely cheap digital penetrometer on Alibaba - - which has an RS232 out port, and continously measures depth (wow, China!). I’d like to hook the RS232 up to the phone via USB OTG and push the data directly into the Our Sci app, so we can handle the interpretation on the phone side of things.

I think this would be doable in two days, so long as I can get the part here from China.

This is interesting. These devices seem to work simply by measuring physical resistance in psi. As this video points out, there are several variables that can affect the force required to insert the probe, both within a field and from day to day. Might there be a more reliable unit for measuring compaction?

I don’t know of any… most people just collect several data points in several locations to deal with the variation… it’s not perfect, but it does strongly relate to outcomes like yield so we know it’s meaningful.

Got the Chinese one - $270 and digital! - will test in the next few days to see if we can get useful USB data out of it.

Compaction is absolutely meaningful. I’m just surprised that physical resistance is a good means of measuring compaction, given all the other variables that contribute to soil hardness.

Cool! Excited to see that penetrometer @gbathree!

FWIW Cornell’s soil lab can optionally include penetrometer readings in its Comprehensive Assessment of Soil Health. They ask for up to 20 readings (10 at 0-6" depth, 10 at 6-18") and provide some good instructions and general info here:

And you can borrow one from them too:

I’m looking forward to this! Here are a few things I would love to work on, if there’s time:

In general I’d love to just come work on stuff with other folks! See what excites us in the moment!

So cool! I want to create a survey so people can run that assessment on Our Sci, I’d also kind of like to put all the soil assessments side by side to understand their utility in different applications. I feel like there are a million, and it’s a little confusing. It’d also help us identify if we want to replicate one for our application, or what pieces we want to take.

Aaaaaaaalso… @mstenta do you know someone from Cornell that I could talk to about pushing their dataset to ? I would love to see what AI experts could do in terms of using some of the simpler data types + metadata to predict the more complex data types and/or outcomes. If someone from Cornell is coming, I would love to work on them with it. We could get someone form crowdAI involved too (even if over skype or something).

Let me know.

@mstenta do you know someone from Cornell that I could talk to about pushing their dataset to ?

Yea: me. :slight_smile:

I built their system for managing soil test applications, results, PDF report generation, etc. I can connect you to Aaron at the lab, if you want to talk to him about it. Data privacy is the big consideration, though. It would probably need some kind of “opt-in” for the folks submitting samples.

K - pulling in Mohanty then to see what he thinks. Also suggesting that crowdAI folks get involved, perhaps if they were at the hackathon we could make some good intros and get something cool done with the data we have. Agreed on the opt-in, though I’d suggest opt-out with caveat that all personal data is purged by default if shared. Sounds like an ideal dataset to play with.

@gbathree I would also love to sit down with you and @mdc to get a crash course in Open Data Kit, if we have time…

1 Like

So Mike, Dan T., Manuel and I talked yesterday and we are going to create a vision and roadmap to handle farm data collection and research broadly.

We have several active use cases (Sieg Snapp’s lab and their work in Malawi, the Bionutrient Food Association and in-field sample data collection, and Ankita’s work at USDA ARS), which have significant overlap. In short, they logically need FarmOS’s data structure and longitudinal perspective, but also require Our Sci’s wider range of data inputs (point sensor data, non-standard question types, etc.) and/or visualization options. These cases are about long-term data collection, engaging farmers in their own data (instance), but also creating cross-farm research with appropriate data handling (who can access that cross-farm data and when).

We hope by roadmapping where we can go, we can clearly identify, prioritize, and (hopefully) more easily fund the features that make cross-farm research and data collection the norm. We invite others who have a similar vision, or who can add features to the pile, to take part!! Just let us know!