Awesome! I'm a huge mapping and avalanche nerd, too. I didn't have much time to read through the sources of data you are using, but would it be possible/feasible to expand the covered area to places other than Norway?
-edit-
Sorry, I didn't see the part at the top about requesting new coverage areas!
Yes! The only thing that are needed to add another country is an elevation model, and there are global ones one can use if needed. The satellite data is already there for the whole globe. :)
I have not tried to run an instance with multiple elevation models though. It is possible the code need some modification to determine what area it should search for data in, but it shouldn't be too hard.
I'm actually planning to add Sweden rather soon (I'm from Sweden myself), but for that I will just merge the files from Norway's and Sweden's national geographical surveys into one file and use in the code.
This looks very cool. Speaking as someone who is not in to winter sports or mountaineering, what is this useful for? Identifying areas with high avalanche risk?
It's mainly useful for avalanche forecasters. Today, they mostly go out in the field, dig in the snow and look for old avalanches. While this methodology is really good, it doesn't really scale, and the risk of missing something important exists.
With this as a tool, I hope it will become easier to get an idea of the circumstances of a larger area, even where it is unfeasible to have personnel.
I also want to get data out of this to be able to make statistical conclusions regarding when and where avalanches are triggered, with regard to weather, time of year and so on.
They are used by people and infrastructure that are exposed to avalanche terrain. This usually involves alpine touring or snowmobiling, but in certain areas, e.g. Norway, a lot of roads and other infrastructure has been built right below dangerous slopes.
They are used to evaluate what risks and consequences that are involved in certain route choices. Should you choose to go up on a north or south facing slope today? Do we need to fire these gas cannons and trigger an avalanche today to avoid having an avalanche reach the rail tracks tomorrow?
Here is an epic photo from last year, shot in the northern part of my county, illustrating what happens when you don't take these factors into consideration.
I wonder if you can train a classifier to detect noise vs avalanche, just to help clean up the map from the blue noise. That would be a nice later addition.
My main objective is to create a proper automatic avalanche detector. For this however, I will need to get datasets of avalanches observed in the field.
I'm discussing with Swedish authorities to get access to their database, and the Norwegian is already open. But I feel those datasets are worse than I would like them to be. My dream dataset would have confirmed non-avalanches as well, to find out what a false positive looks like. One way to get this would be to fly a drone over the same area every day for a season, which I think is feasible.
With that said, if any one knows of some data I could use, open or not, please contact me!
The idea seems great, but I was not able to navigate the interface.
If I understood correctly you have access to past data from satellite. If yes, it would be fantastic resource to actually validate avalanches that happened. For that you would need to improve how to find the specific place as well - I could not figure it out.
e.g. I live in the Pyrenees and often go off-piste snowboarding and see avalanches, also sometimes I see some avalanches directly from my home as well. If you give me Pyrenees coverage and better UI I would love to play with it
What browser are you on? Because, IE and Edge does not work right now. This is simply because I do not have access to those browsers. I will probably just block them in the future.
If you use another browser, could you please be more specific about what did not work?
I'll see if I can add the Pyrenees in the coming weeks!
FWIW, there are a number of avy forecasting groups in the western US (e.g., utah, colorado, NWAC in Washington state,...), and they may be interested in this, you should reach out to them. Also, IIRC, University of Utah has an avi forecasting program.
Could you email me (using either the address in my profile or the one on the web page) more specific information about these groups, so that I can contact them?
17 comments
[ 3.6 ms ] story [ 49.1 ms ] threadI don't know what more to tell you, but I would love any feedback!
-edit- Sorry, I didn't see the part at the top about requesting new coverage areas!
I have not tried to run an instance with multiple elevation models though. It is possible the code need some modification to determine what area it should search for data in, but it shouldn't be too hard.
I'm actually planning to add Sweden rather soon (I'm from Sweden myself), but for that I will just merge the files from Norway's and Sweden's national geographical surveys into one file and use in the code.
With this as a tool, I hope it will become easier to get an idea of the circumstances of a larger area, even where it is unfeasible to have personnel.
I also want to get data out of this to be able to make statistical conclusions regarding when and where avalanches are triggered, with regard to weather, time of year and so on.
Some avalanche forecasting services:
https://avalanche.org/
https://www.avalanche.ca/map
http://www.avalanches.org/
http://www.varsom.no/
http://www.lavinprognoser.se/
They are used to evaluate what risks and consequences that are involved in certain route choices. Should you choose to go up on a north or south facing slope today? Do we need to fire these gas cannons and trigger an avalanche today to avoid having an avalanche reach the rail tracks tomorrow?
Here is an epic photo from last year, shot in the northern part of my county, illustrating what happens when you don't take these factors into consideration.
https://twitter.com/MacahanZeb/status/847504439220948993
My main objective is to create a proper automatic avalanche detector. For this however, I will need to get datasets of avalanches observed in the field.
I'm discussing with Swedish authorities to get access to their database, and the Norwegian is already open. But I feel those datasets are worse than I would like them to be. My dream dataset would have confirmed non-avalanches as well, to find out what a false positive looks like. One way to get this would be to fly a drone over the same area every day for a season, which I think is feasible.
With that said, if any one knows of some data I could use, open or not, please contact me!
As per today, it's still at zero for this winter, but there's also a historic list: https://www.slf.ch/en/avalanches/destructive-avalanches-and-...
If I understood correctly you have access to past data from satellite. If yes, it would be fantastic resource to actually validate avalanches that happened. For that you would need to improve how to find the specific place as well - I could not figure it out.
e.g. I live in the Pyrenees and often go off-piste snowboarding and see avalanches, also sometimes I see some avalanches directly from my home as well. If you give me Pyrenees coverage and better UI I would love to play with it
If you use another browser, could you please be more specific about what did not work?
I'll see if I can add the Pyrenees in the coming weeks!