Matt Forrest

Matt ForrestAug 19, 2026

This is how to use modern GIS, in 2026, and beyond. Or, how to make real, live data pipelines for spatial data. I am in the middle of writing a (very) long newsletter with the working title "The Modern GIS Roadmap". Why? I get this question all the time and there isn't one single resource that talks about how all this stuff works together. I am guilty of this too. I talk about technologies in isolation not as a group. And more importantly it's not about the parts, or how the parts work together. It's also the process and structure you build around it. The architecture and design. The result is the end product is way more impactful than the sum of it's parts. So how do (I think) you should build it. 1. You get data from many sources: servers, files, API, public data, etc. Outline all these and build 1 to 1 pipelines to move those to formats in step 2. 2. Pick your storage formats. I recommend GeoParquet, COG, and Zarr. These are the formats your data lake lives in, but not all outputs are in this format. 3. We have been using catalogs for a long time, Esri Server is a prime example. This is one level lower. Use Iceberg, STAC, DuckLake, or IceChunk are all examples (personal projects might skip this) 4. Now orchestrate. Design the steps. Use a multi-hop or medallion process. Pick any number of tools that work for you from chron jobs to managed services like Airflow/Astronomer, dbt, or others (lots of choices). Time and schedule. 5. Data contracts. For each layer (gold is important) what is your promise to the user (this data will tell you this, its updated every ___) 6. Now design your data layers. Each step. What happens. Be specific. 7. Think like a data engineer. Use the smallest amount of data and design for the end use case in mind (GIS data is different than web apps) 8. Spatial processing happens in the silver layer. Pick an engine that scales. (You can use different engines and different scales - scale from Python to full scale Wherobots or Apache Sedona). 9. Yes, you can use this (and should) for raster and climate data too. Again, pick the right engine at each stage. 10. Gold layer = data marts. Each mart = one use case. Output the right data in the right formats (PostGIS, GIS server, tiles, GeoParquet, etc). The final destinations can be for GIS, data science, analytics warehouses, SQL databases, analytics dashboards, web map tiles, even AI integrations. This for me is the final boss of geospatial. So much analysis we do is a snapshot in time. It's important but it limits what we can do with our data. Imagine we can deliver real tools with data updating every day. With our own insights and analytics. We are there. The pieces are there. It's happening at the data layer and every tool on top of it gets to benefit. I'll be sending out the full essay via my newsletter by the end of August. Make sure to sign up here: forrest.nyc

Matt ForrestAug 19, 2026

I just re-recorded one of my most successful videos from YouTube. There were a lot of updates from 2022, and this is what's inside. First off, I clearly didn't know a lot about making videos back then so how this video has close to 200k views is still strange to me. But I think it has to do with the content. I made this video because so many people message me about what they should learn and the roadmap. Now I can't make that specific to each person and everyone's experience will be specific to themselves, but these are the big things that helped me most and I think are good to learn for anyone in geospatial. I also wanted to address the addition of AI and how I think it should be used, not as a new tool, but something that can be a force multiplier of what you already know. And more importantly how to use it while learning the core skills. Here is a short list of what's inside. Also, a lot of technology has changed too! So those updates are here as well. Inside: 1. The single most important skill in 2026 is not a tool and not a language. Most GIS people are actively getting it wrong right now because of AI. 2. Why I still say learn SQL before Python, and why I doubled down after people pushed back on it the first time. 3. The two tools that dropped the barrier to entry for spatial SQL to zero. 4. My rule for how long to go without touching AI when you pick up a new tool. It is a specific number of hours and most people will not like it. 5. Three mistakes that slow people down. One of them did not exist when I made this video three years ago. 6. What GIS job postings looked like in 2022 next to what they look like now. 7. The three things your portfolio has to prove in 2026. A finished map proves none of them. 8. The one question that tells you whether you actually learned something or just prompted your way to an answer. If you want to check it out, I added the link to the comments below. Let me know what you think (and if you feel compelled to subscribe please do). Image is created with AI (trying to make my videos look as cool as Mark Rober's) 🌎 I'm Matt Forrest and I talk about modern GIS, earth observation, AI, and how geospatial is changing. 📬 Want more like this? Join 15k+ others learning from my daily newsletter → Link in my profile

Matt ForrestAug 27, 2026

Don’t let AI code for you without doing this. If you are using AI to code make sure you are able to explain back to yourself or someone else what the code is doing in plain language. Build that core skill first then add AI. Knowing coding like spatial SQL and Python is actually more valuable in today’s agentic coding world, not less.