Good day!
I would like to validate my Forest Fire Risk Map with hotspots from MODIS. I already have my FFRM (in .tif and .csv) and my MODIS data (in .shp and .csv). I already categorized my FFRM with z = 0 (False - for Very Low Risk to Moderate Risk) and 1 (True - High to Very High Risk). What should I do to generate an ROC curve with ArcGIS or QGIS? I can also use R.
Attached are images of the attribute tables for the FFRM and MODIS.
Thank you.
I don't know the exact tools you're using (ArcGIS or QGIS). I could only advise you to use Precision instead of False positive rate to create ROC, since it seems you have way more 1s than 0s.
You can use your original risk values (without classification into 0 and 1) versus the presence and absence (1 and 0) of real fire events. Then you can plot the ROC curve easily using software like SPSS or MedCalc.
Hi, you can use the ArcSDM tool under Arcgis, these links can be useful for you :
Application: https://www.youtube.com/watch?v=2XAhjJInDiI
Download tool: https://github.com/gtkfi/ArcSDM
Good luck
Good day. Thank you for answering.
Laura Dzelzkaleja I have more 0s (approximately 70% of the data) than 1s.
Mohammad Mehrabi Thank you for suggesting SPSS. I have seen the ROC function in the said software. However, I have no way to match specific pixel risk values with my MODIS data. Do you know of any method?
Goumrasa Abdeldjalil I have seen this method in past questions here in RG and I have tried it. Although I followed all of the steps stated in the video, I always get an error. I have tried it multiple times and still fail. Attached is a copy of the error message. Do you know how to solve this?
Thank you all for answering. :)
good greeting
Sorry, I fell into the same problem that you are in. Did you find a solution and what was the method you found in order to work with the program
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping