[Discussion] Selecting the Weightings and Number of Zones when creating the potential management zones

[KipCrossing]

I'm going to keep pushing for clarity behind the decision-making process of selecting the weightings and number of zones when creating the potential management zones (via k-means). As I understand, it's a kind-of iterative process where the agronomist/farmer makes several attempts until the zones look (more or less) how they expect them to look, then checks if the data has been grouped properly based a visual histogram inspection. Several questions still linger:

• what is it we are attempting to manage?
• what do we do if we have no expectation on how the ones should turn out?
• should there be some pre-defined weightings based on what we are attempting to manage?
• should we take selecting the number of Zones away from the user (ie elbow method)?
• how damaging is it to the best spread of samples if the zones are wrong and we CLHC sample per zone?

It seems to me that this type of zone creation (and based on our 4/5 layers) is fairly new and therefore we can't rely on past user experience to make the right choices. 

IMO, these concerns should be answered by our ground-truthing working groups.

[Samuel22]

These are really good questions. When it comes to making zones, I'd recommend leaving the 'how' to the users of the tools (agtech integrators, agribusinesses). People will be using different datasets, so if we offer them up as a 'tool', users will best suit them to their business needs.

For example irrigation companies may use zones created with their own datasets such as EM38, or land management companies might use combinations of U-gamma and elevation in regions with granite to look for regolith changes. There are nearly an infinite number of ways to create zones depending on the use case, making it hard for us to define specific parameters around their use. There are broad recommendations in the Wiki for what to use for Farm vs Paddock scale zones, but these are not prescriptive.