Time for a Change

NORTH51 Update - Mapping the unseen

Before we get into this article. I wanted to highlight NORTH51, the independent geospatial thought leadership event. N51 is about big ideas, and we have just published our program for this year’s event in April.

With sessions on Space, People, and Oceans, this year at North51, we are exploring ideas around mapping what we cannot or choose not to see. We welcome challenging, respectful discussions in the beautiful Canadian Rockies. If you want to take the temperature of the geospatial industry, come and join us.

Time for a change.

I’ve been discussing the sheer amount of data flowing out of Low Earth Orbit (LEO) for years, and that torrent of terabytes has only grown in magnitude. This firehose of data has prompted me to note that the biggest problems in Earth observation (EO), presently*, are in software, not sensors.

What I mean here is that we are amassing an enormous collection of surface reflectance readings organized into arrays that can be projected to model the surface of our planet. These we call images. From these arrays, we deduce either the status of a location or events that may have occurred.

While the status of a location may imply an absolute measurement. The notion of events implies a measurement of change. Given we are recording the surface of our planet at a furious rate, change has become the currency of the EO sector. This brings up three questions.

1. Why are we so poor at visualizing change?
2. Why is time treated so haphazardly?
3. Who is actually selling the product “change”?

Time must be a first-class citizen in geospatial.

Future geospatial platforms will be four-dimensional by nature. Already, we see reimagined platforms developed in geo-aligned sectors like drones, for instance, outside the more traditional GIS sector, consider the nature of both time and space equally. These sectors are unshackled from decades of planimetric technical and product debt. Free to approach the measurement of our landscapes with an untarnished eye.

But can we do better still? Returning to the two questions posed above, what are the main tools for interrogating change? Well, we have time sliders, and we have animated GIFs. This is a seemingly underwhelming selection of tools. But the central premise here is an assumption of the product. I have assumed that those who want to measure change also want to see that change geographically. This is one of the greatest assumptions in geospatial and GIS. This assumption I believe to be false. But a false assumption that drives a stake into the hearts of most geographers, EO scientists, and satellite builders. What if people don’t want to buy the visually appealing thing I’ve built? What if, instead, they just want to buy the informational product? Remember the Venn of geospatial people?

Gordon Logie, from Sparkgeo's data science team, demonstrates this conundrum in his work on drought in Southern Alberta (part 1, part 2).

He ably moves from video-based visualization (animated GIFs) to graphs. Ultimately, a financial analyst would use the graph for quantitative feedback. However, the initial video was necessary to build the algorithmic trust they needed to make that cognitive commitment.

An interesting and notable approach is evident from Hydrosat’s acquisition of Irriwatch. It uses a simple dashboard to indicate to a farmer which field needs to be watered. This is not a map—most farmers know where their fields are—just a dashboard fed by data from Hydrosat’s constellation. This is a simple and elegant EO-derived product.

Analysis Ready Time.

Analysis Ready Data (ARD) is a hot topic in EO. In essence, the dirty little secret is that different sensors are hard to compare. Unfortunately, this is true in almost every dimension, from geography to radiometry to time. Our industry must continue to build consensus on these dimensions. Thankfully, CEOS’s pragmatic approach to ARD is gaining some traction.

Timed to perfection.

We need that consensus because time in geospatial has every bit as much potential granularity as the XY and Z dimensions. This is as true of the applications as it is of the data collection.

Consider the basic EO product, “situational awareness.” As a snapshot of a location, this product is instantly out of date. When we consider the time an image might take to be delivered to the screen of an analyst or end-user, critical time might have passed. For military operations, this could mean lives; in the financial sector, this could mean profits; and in natural resources, this could mean pollutants. Time in all these applications is critical.

Then, we have the notions of change; change depends on context. Changes can happen slowly “over time” or quickly in “no time.” In both these cases, “time” is not defined, and whether it is slow or quick is dependent on the context of the subject in question. Urban development alone could be fast in Riyadh yet slow in Regina. Not to pick on our friends in Saskatchewan; I just like the alliteration, but things are built very quickly in Riyadh.

The central point here isn’t that “change is a cool product.” It’s that change is a multitude of different products, each with different characteristics determined by time. Additionally, the availability of suitable data is directly affected by our definitions of time. Finally, time is an unstructured dimension, one which we would be wise to consider as deeply as we consider geography. While we think about geographic alignment, do we consider temporal alignment with equal scrutiny?

To be clear, though, change is a cool product. In fact, I believe change is the EO product. The problem is that we so rarely sell it.

Change is time; time is change.

When we have a massive corpus of data, like EO data, a great way to synthesize it is to find useful dimensions along which to organize it. Time allows for that organization, but it also allows for the intermediate inferences to be made. Those inferences are the interpretations we ask our imagery analysts to make for us.

Typically, EO companies have avoided offering overtly change-oriented products. While some have ventured up the value chain, Planet being an excellent example with its planetary variables, few have offered an unabashed change-focused product. Even though, with high repeat imagery, change has always been the product’s manifest destiny. When something is being measured regularly, each absolute measurement becomes irrelevant against the richer understanding of the phenomena’s temporal behaviour.

Consider.

1. Time should always be a first-class citizen in geospatial analysis.
2. We need to be better at building 3 and 4-dimensional visualization products.
3. Measuring geography through time often means we lose the geographical representation, but that’s ok!
4. Change has always been EO’s best product, but we so rarely sell it.

Thanks for reading Strategic Geospatial! This post is public, so feel free to share it.

*independent of finding paying customers!