How is a vineyard like a street?

Towards an open data model for the wine industry

Last week, I had the pleasure of talking about mapping with a totally different group of people: the Australian wine industry.

At Platfarm’s first Collabriculture workshop, “Open Mapping and the Wine Industry”, I shared about linear referencing and building data standards for curb regulations, and how this work could be applied to vineyard management.

What do vineyards have in common with city streets? A lot more than you might expect. The image below shows the elements of a vineyard. If you take a step back, it looks very similar to a streetscape… but instead of streets lined with buildings, we have rows lined with vines, posts, and trellis wire.

Elements of a vineyard. Image courtesy of Platfarm

Because they are composed of rows, vineyards have the same fundamental need to map and locate assets or track activities in reference to a linear feature, in this case a row of grapevines.

For example, a grape grower may want to log the location of a broken irrigation drip line (“Row 2, left side, 30 metres along”). They may want to be able to take GPS telemetry from tractors and use it to determine how much fertilizer was applied to different areas (similar to aggregating scooter trips in a city). They may want to ensure that vendors are all using a common way to refer to the vineyard assets, so that data can be combined into a comprehensive model of inputs and outputs. All of this benefits from mapping the rows of vines, adding a linear referencing model, and building a common language for how to describe physical assets and agricultural activities (such as grape varietals, date of planting, and fertilizer application). It’s very similar to how we approached the need for a data standard for curb rules.

Drones, maps, and demos: Trying this out in real life

Thinking about applying methods to other contexts got us really excited. I wanted to do some hands-on experimentation to see how open source tools could be used for mapping vineyards. So I contacted my nearest vineyard for permission and set out to map the vines.

I wanted to ensure I had high-resolution imagery and I happened to have a drone kicking around the house, so I flew a couple of passes over the 16-acre fields of Faith, Hope and Charity Vineyards in Terrebonne, Oregon. This gave me aerial imagery with 7cm precision — far better than I’d actually need.

I probably could have made do with existing satellite imagery, but this was a fun excuse for a field trip. Right: drone imagery of the vineyard on top of existing satellite imagery of the general area (source: Maxar). Thanks to Faith, Hope and Charity Vineyards for letting me fly over your fields!

Drone imagery comes back as hundreds or thousands of individual images, which have to be combined into a single mosaic that’s “orthorectified” (georeferenced and adjusted for tilt, etc). I used OpenDroneMap to process the imagery, then uploaded the output to OpenAerialMap, which hosts and tiles the imagery so that local and web-based apps can access it. Now it’s ready to be mapped.

Tracing the vines into OpenStreetMap. The location of vines generally isn’t sensitive information since they can already been seen on satellite imagery. However, a vineyard may wish to keep other aspects of their vines or management practices private, rather than publish them openly.

I traced the vine rows into OpenStreetMap and tagged them as “tree rows” — there’s not an existing data schema for individual vineyard rows or grapevines.

OpenStreetMap is easy to learn and doesn’t require any specialized knowledge of GIS or other mapping programs; this is something that vineyards could start doing today if they are interested.

Next, I grabbed the OpenStreetMap data and fed it into existing SharedStreets tools in order to add linear referencing information to each vine row. This involved the SharedStreets Builder tool to create custom tiles, and the JS library to interact with them. The data came back as the same map geometries, but with reference IDs and other properties added on. While this may not sound thrilling, it means we can start to work with location data in new ways.

For example, let’s say I want to keep track of tractor activity to monitor fertilizer applications. Linear referencing allows me to take noisy GPS data points from the tractor activity and turn it into meaningful information for each section and row of vines. In the image below, dummy GPS data has been snapped to the vine rows and aggregated into counts for each 5m section of each row. This creates meaningful data that’s intuitive to understand.

Methods created for analyzing scooter activity can be repurposed to track vineyard management activity

Apps can also interact with linear-referenced data in real time to support vineyard management. For example, let’s say a vineyard worker comes across a broken irrigation line. Rather than tying a piece of flagging tape at the end of the row, or logging a GPS coordinate that may not be accurate enough to identify the correct row, the worker could use an app that’s built to log the vine ID, row number, and distance along the row where the irrigation line is located. This gives the essential information that a human would need in order to quickly locate the problem in the future, and it stores it in a way that a computer can interact with for routing or other purposes.

Interactive demo allows you to drop points or paths onto the vineyard and see how they’re matched, and what data comes back

To see a demo of how real-time matching works, check out this Observable notebook, which contains interactive demos for point and line matching, as well as the software code that powers the examples.

What’s next?

It’s exciting to see the Australian wine industry come together to discuss ways that vineyards, companies, and government entities can collaborate to advance the state of the practice and develop an open data model for vineyards.

While this may be a new venture for the wine industry, we hope that methods and tools developed for other contexts can spark ideas and even give folks a foundation to build from, rather than reinventing the wheel if it’s not necessary. We’re eager to watch where these conversations go, and we can’t wait to see what our friends Down Under develop!

Thanks to Oli Madgett and the Platfarm/Collabriculture teams for reaching out and inviting us to be involved in this conversation. You can check out the full session here, which also features content from Fiona Turner of Bitwise Agronomy and John Bryant of Mammoth Geospatial.

Want to map river networks and whitewater rapids? Railway junctions and stations? If you’ve got a use case for linear referencing tools, don’t hesitate to reach out — we’re happy to chat, even if it’s not necessarily about streets!