Soil health maps
4 views
Skip to first unread message
Aaditeshwar Seth
Apr 2, 2026, 9:06:04 AM (3 days ago) Apr 2
to learning, ac...@googlegroups.com, CoRE stack: NRM, core-st...@googlegroups.com
Hi all,
Sharing this update from a project going on since last year. MTech students Aakash and Ratinder, with feedback from Nirzaree, worked on building soil health maps for N, P, K, and SoC. We used soil health cards data published by the government. This was point data for several hundred thousand lat/longs. A lot of cleaning had to be done, like to throw away impossible values, or points that didn't lie on cropping LULCs. Per-AEZ classifiers were then trained to predict N/P/K/SoC. We tried a range of features and finally what worked best were more of climatic features, soil type, and to some extent terrain features and vegetation features. This probably tells us that the models have learned some kind of a native soil health but given low R2 scores in several AEZs for several properties, there is much else unexplained stuff that we're not able to see from remote sensing alone. If this is the case then these maps can be relevant to relate indigenous cropping systems with the native soil type and climate of the area.
The report has links to the underlying datasets.
A note that global maps like from OpenLandmap do not capture the fine variation in SoC on cropping areas. They are better at distinguishing soil properties on cropping areas vs different types of forests, peatland in North Canada, etc.
We'll pull these into the CoRE stack soon too.
Adi
Co-founder, Gram Vaani; Co-founder, CoRE Stack
Reply all
Reply to author
Forward
0 new messages