Jordan Lee, Collab Ag
A companion piece to Bindi Isbister’s “Georeferenced Boundaries for Farm Automation” — ACTFA On Track, June 2026
Bindi Isbister laid it out clearly in the last issue of the ACTFA On Track newsletter – boundaries shift, brands disagree, and the moment you add a contractor or swap software, things stop lining up. If you haven’t read Bindi’s article, start there – provide link to Bindi’s article Jun 26.
What I want to add is the layer underneath it. The brand inconsistencies Bindi described have a root cause most operators never see, and it’s getting harder to ignore as more technology stacks up on farm.
That cause is GPS datums.
What’s a datum and why does it matter?
A datum is the reference grid your guidance computer uses to place itself on Earth. Different systems use different grids, and if they’re not on the same one, they won’t agree on where your CTF lines are, even when both screens are reading “RTK accurate.”
Australia has two national datums currently in use. GDA94 was fixed to the Australian tectonic plate’s position in 1994. GDA2020 updated that to 2020. The plate moves roughly 7 cm northeast every year, so over 26 years that’s 1.5 to 1.8 m of offset for the same physical point on the ground.
On a CTF farm, that’s not a small miss.
Which brand is on which datum?
John Deere dealer networks and base stations commonly broadcast in WGS84, the global GPS datum, sitting within 0.1 to 0.2 m of GDA2020. CNH and Trimble have historically referenced GDA94.
Most modern machines across most brands are capable of connecting to an NTRIP correction service. NTRIP is simply how a machine receives RTK-level correction data over the internet rather than from a physical base station you own on-farm.
For John Deere users running a mixed fleet, this opens up the opportunity to connect via Agra GPS CRG, the first non-Deere receiver compatible with John Deere GPS systems, allowing John Deere machines to connect to NTRIP services alongside the rest of the fleet.
None of these systems are wrong. Each is accurate in its own reference frame. The problem is when they’re expected to follow the same lines or act on the same prescription maps, and nobody has checked whether they’re all speaking the same positional language.
It’s not just a mixed-brand problem, it can happen within the same brand
John Deere’s own StarFire RTK and SFRTK use different georeferenced positions, meaning they don’t automatically line up with each other. A mixed fleet of older and newer JD machines running different correction signals faces the same side-by-side paddock alignment problem as a mixed-brand fleet.
There is a software conversion available, and most JD dealers can do it. The problem is they don’t typically offer it proactively when you buy a new machine or when older machines are still in the fleet. You have to know to ask.
This is also a timely issue. Some dealer-operated RTK networks are being wound back, leaving growers who have built their CTF system around that network signal facing an unplanned upgrade, new receivers, SFRTK subscriptions, or their own base station, potentially across a large number of machines. For operations running several machines, that’s a significant and unexpected capital conversation.
But for many CNH machines, there’s a second problem before datum even comes into it
A lot of CNH machines are still running Trimble RangePoint RTX as their correction service. RangePoint RTX is not RTK.
RTK delivers around 2 cm of repeatable accuracy. RangePoint RTX delivers approximately 50 cm. That’s not a configuration issue, it’s a fundamentally different class of positioning technology. A machine on RangePoint RTX cannot hold a CTF traffic lane at RTK precision, full stop, regardless of datum.
So for those operators, the problems are stacked. First, the correction service isn’t delivering the accuracy CTF requires. Second, if they upgrade to RTK, the datum alignment question is still waiting for them. Two separate issues and conflating them makes both harder to solve.
RangePoint RTX does what it was built to do. The issue is using it in a context where the accuracy class doesn’t fit the machinery fleet system in which it is going to operate.
Add a drone and it gets interesting
Drones used for mapping and prescription work receive their positioning corrections in one of two ways. Some connect to an NTRIP service, referencing GDA2020 through the state CORS network. Others use their own portable base station, similar to a surveying base station, set up in or near the paddock. If that base station isn’t georeferenced to a known datum point, it creates its own local coordinate system that may not align with anything else on farm.
It is worth noting that Trimble receivers are used across surveying, construction, and multiple ag brands, not just CNH. A drone or portable base station running Trimble technology isn’t automatically on GDA94, it depends on how it was configured and for what purpose.
A drone maps the paddock, identifies a weed patch, and generates a prescription spray map referenced to GDA2020. That map goes to the boom sprayer, which is running CTF lines off a GDA94-referenced base.
The sprayer’s guidance is excellent. The prescription map is accurate. But they are 1.5 m apart in the paddock. The sprayer targets the wrong zones, and even with a buffer applied around the weed patch, a 1.5 m offset can put the application outside the target zone entirely. The weeds stay put. Neither system logs an error. Neither knows it’s out of step with the other.
The same issue applies to autonomous vehicles and autonomous retrofit kits, each bringing their own GNSS receiver and correction source into an already mixed environment.
Whose job is it to ask these questions at point of sale?
This is the part that stays with me.
When a grower buys new machinery, is the sales consultant asking what correction service is running on the other machines? Whether they’re on CTF? What datum the base station uses? How this new machine needs to integrate with what’s already on the farm?
In most cases, ‘no’, and it’s not entirely the salesperson’s fault. They’re trained to sell equipment, not audit digital farm systems. But the point of purchase is often the only moment when these questions could be addressed proactively, before the equipment arrives, before the season runs, before the drift has had three years to quietly compound.
Unpicking it is hard work after the fact. The same gap exists on the agronomy side – a prescription map and the machine executing it may never have a conversation about what datum either is working in.
Who owns this conversation is unresolved. What I do know is that the cost of not having it upfront keeps going up.
Three questions worth asking now
- What correction service is every machine on the farm running, and what accuracy class does it deliver?
- What datum is each correction source referenced to, and has anyone checked whether they all agree?
- If you’re bringing drones or new machinery onto the farm, has anyone checked how they align with what is already there?
The boundary problem Bindi described is real. The datum and accuracy class layers underneath it are what make it harder, and more costly, than it first appears.