Wanted to share this because it’s something lots of folks have discussed @DanT @kanedan29 @sudokita @dornawcox, hoping we can build some interest around this.
Hi all, Jon Maynard @jherrick @mstenta and myself met today to discuss more effective ways to benchmark soil data, in part for the Farmer Coffee Shop, but also more broadly from a data structure perspective. We also discussed how to include confidence intervals across the board for all data types.
We felt there was a pretty effective and immediately useful FarmOS integration possible as an MVP to show how to solve these problems. It feels achievable and integrated, so we’d like to give it a go!
- We chose SMAF as a framework in order to provide better contextualized soil health information. This will help people understand their soils relative to their potential, rather than relative to absolute values of all soils (which may be uninterpretable).
-
Create a data framework in order to provide different sources of input data (texture, taxonomy, precipitation, growing degree days based on standardized crop)
- Dynamic soil property - carbon, compaction, etc.
- Constant soil property - texture, soil taxonomy, temperature, precipitation
- Add uncertainty
- Update Modus’s data structure to jive with (2)… and propose to him on the changes we’re recommending. For example… can we add texture class, even though that’s not a lab test? Can we add ‘texture class from hand texturing’ and ‘texture class from soil map’? We also need to update Modus so that comparable IDs are noted, while creating space for confidence intervals or some confidence measure.
-
Get SMAF inputs from LandPKS (texture, soil taxonomy) and other sources (climate, etc.)?
- LandPKS can (via API) provide the highest likelihood value for texture/taxonomy.
- LandPKS recommends what you should do to clarify the texture (adds a question or two)
- As needed, one can then do in-field texture the soil and get a higher confidence taxonomy.
- Climate data can come from many other sources, and follow a similar pattern.
Next steps:
- (greg) Will make an initial mockup to more help communicate this (easier than words
- (jon) work on LandPKS API to be able to get higher-quality texture and class information from a given gps point.
- (greg) Once that’s done, we’ll follow up with Chelsea from Point Blue / Range C and Sarah McClure (Jon / Jeff can connect) for rangelands and landscape data commons. Can Point Blue’s confidence measurement structure be used here?
