Aaron Zuspan
@aazuspan.dev
🛰️🌲🔥 Researcher at Oregon State University / Earth observation and fire / open source geospatial / Python / Earth Engine
aazuspan.dev
aazuspan.dev
bsky.app/profile/did:...
Nothing fancy at all. Turns out Earth's surface color is pretty predictable.
Nothing fancy at all. Turns out Earth's surface color is pretty predictable.
The model is a stupidly simple pixel-based random forest trained with a grid of sampled points (below) using downscaled bioclimatic predictors derived from temp and precip (via WorldClim) and stretched MODIS RGB.
I also tried predicting in HSV and LAB space, but RGB looked the best 🤷♂️
I also tried predicting in HSV and LAB space, but RGB looked the best 🤷♂️
December 10, 2024 at 2:36 AM
bsky.app/profile/did:...
Nothing fancy at all. Turns out Earth's surface color is pretty predictable.
Nothing fancy at all. Turns out Earth's surface color is pretty predictable.
The model is a stupidly simple pixel-based random forest trained with a grid of sampled points (below) using downscaled bioclimatic predictors derived from temp and precip (via WorldClim) and stretched MODIS RGB.
I also tried predicting in HSV and LAB space, but RGB looked the best 🤷♂️
I also tried predicting in HSV and LAB space, but RGB looked the best 🤷♂️
December 8, 2024 at 8:15 PM
The model is a stupidly simple pixel-based random forest trained with a grid of sampled points (below) using downscaled bioclimatic predictors derived from temp and precip (via WorldClim) and stretched MODIS RGB.
I also tried predicting in HSV and LAB space, but RGB looked the best 🤷♂️
I also tried predicting in HSV and LAB space, but RGB looked the best 🤷♂️
For comparison, here's the actual and predicted modern imagery from MODIS. A climate-only model definitely misses some topographic and soil detail, but seems to capture land cover surprisingly well.
December 8, 2024 at 8:15 PM
For comparison, here's the actual and predicted modern imagery from MODIS. A climate-only model definitely misses some topographic and soil detail, but seems to capture land cover surprisingly well.
Great work! Having an app to explore the data is really useful.
November 13, 2024 at 6:27 PM
Great work! Having an app to explore the data is really useful.
If you're curious about implementation, I wrote a blog post about processing this with #EarthEngine and #BigQuery.
www.aazuspan.dev/blog/summari...
www.aazuspan.dev/blog/summari...
Summarizing 51 Years of Landsat Data
I analyzed 11 million Landsat scenes to test out the new Earth Engine - BigQuery interface.
www.aazuspan.dev
November 13, 2024 at 1:28 AM
If you're curious about implementation, I wrote a blog post about processing this with #EarthEngine and #BigQuery.
www.aazuspan.dev/blog/summari...
www.aazuspan.dev/blog/summari...
Good point - much easier math at the equator! So in Swahili time, the clock is effectively counting how long the sun has been up/down? That's fascinating!
November 11, 2024 at 8:00 PM
Good point - much easier math at the equator! So in Swahili time, the clock is effectively counting how long the sun has been up/down? That's fascinating!
Fun fact, I did zero input validation 👍
November 10, 2024 at 6:56 AM
Fun fact, I did zero input validation 👍
There's also an interactive version where you can calculate travel distance for any date/location
www.aazuspan.dev/daylight-hou...
www.aazuspan.dev/daylight-hou...
Daylight Hours
Calculate the daylight hours at a given latitude and date, and see how far you need to travel to maintain today's day length.
www.aazuspan.dev
November 10, 2024 at 6:56 AM
There's also an interactive version where you can calculate travel distance for any date/location
www.aazuspan.dev/daylight-hou...
www.aazuspan.dev/daylight-hou...