OK wildfire

I'm devastated daily. When reading the news. Looking at the orange sky. Seeing the scenes of fires across BC, Canada and other regions breaks my heart. Our world is literally burning.

I took some time last night to sit and think about how the developments we've made in AI and data modeling could be used for environmental impact rather than making another bot to chat to (is a thing, won't waste time linking it).

These are thoughts and theories that I hope to develop out further in time. But for now, I believe that one or perhaps all of what I'm sharing can help us fight against the possibility of the world burning around us.

For those that aren't familiar, we have made tremendous leaps in Artificial Intelligence in computing. Allowing computers to model larger amounts of data, infer meaning from questions, allowing great companies like OpenAI build ChatGPT. These models are built by crunching a lot of data. Tooling is now available for developers and researchers to build out models faster with software and cloud hosted stacks. When I built systems like that in 2004, it took months. Today it's weeks. What's needed though is data.

I'm no expert, nor fire fighter, but for the sake of this article, I will consider that fire starts because the woods are dry, soils haven't retained irrigation as much as they would, so the trees become drier. Tech isn't going to put the fire out but could give us preventative tactics that we can have individuals work on during the off summer seasons.

Irrigation. We need to build better lightweight, lower powered data collectors for irrigation data. This would allow us to model how the water flows during our (very) rainy seasons here in BC. Technically we're a rain forest. This data can then help us understand the areas that are no longer able to retain the moisture as much. Could this then allow us to move and prep the land to ensure the water is retained for longer during those hot summer times. Perhaps designs of property, concrete, roads, etc are then put around the locations that are 'must have' healthy soil and tree root structure. Are we felling trees and areas that we shouldn't? Creating soil that is easier to dry out and become desert like.

Rainfall. With a better lightweight data collector, can we model out the rainfall and understand if a season didn't rain in high risk areas as much as others.

Imaging. Could we better predict hot spot, high risk areas. The starting points of the fires. With better thermal imaging or even just the images we might have already, could we find the hot spots, high risk areas. Not suggesting we put expensive thermal imaging satellites in space (we don't need more junk there). But model a couple of other things:

• How the wind moves through the trees, allowing us to understand a baseline on how much a tree should move based on healthy watering. A dry tree will obviously not move as much.
• The colour of the leaves and how fast they change. This obviously depends on the types of trees in the location being monitored. Maybe certain trees could be planted, knowing they give a better read on how the leaves would dry out.
• Modeling different filters on an image, haven't thought too much about this one. But similar to B&W photography where we used different colour lens filters to bring out different contrasts, could we do this with one image and gain more meaning from it. Could we see smoldering before a fire.
• Using the above could technically allow us to create a filter that assesses an image and extract context, probability on fire risk.

What could be done knowing this data would require collaboration with experts on fire prevention. I would imagine that we could move soils, tree root structures, plan out forest growth and irrigation flows.

But I'll leave this one here, as a whackier idea, could we place reservoirs that are in locations across our lovely parks, contained in a way where mosquitos can't be bred of course. Created from recycled plastics that we throw away daily.

To summarize, the investment would be in data collection, assessing it in different ways by merging different variables that we can do so with computing and lightweight powered data collectors. If you're researching or studying this area, I would very much like to chat.