Landslide Susceptibility Map (Field Level @100m)

[kapildadheech]

Ticket Contents

Description

Landslide susceptibility mapping at field level (~100m resolution) helps identify areas prone to slope failure, enabling disaster risk management and land-use planning. Using GEE, susceptibility can be computed based on established methodologies and datasets, vectorized, and later refined for finer spatial scales.

Goals

Goals

• Implement landslide susceptibility methodology from existing https://www.sciencedirect.com/science/article/pii/S0341816223007440 using GEE.
• Generate raster outputs at ~100m resolution representing landslide susceptibility.
• Explore methods to compute susceptibility at finer spatial scales (<100m) for high-resolution mapping.
• Vectorize raster outputs into polygons for field-level analysis.
• Publish raster and vector outputs as Earth Engine assets with metadata.
• Enable visualization and spatial analysis of landslide-prone areas.

Expected Outcome

Expected Output

• Raster dataset (~100m resolution) showing landslide susceptibility.
• Vectorized polygons with attributes:
  • Susceptibility class (low, moderate, high)
  • Area (ha)
  • Relevant metrics (slope, curvature, land cover)
• Published Earth Engine assets (raster + vector) with metadata.
• GEE visualization highlighting landslide-prone zones.
• Validation report confirming coverage, accuracy, and classification.

Acceptance Criteria

Acceptance Criteria

Data Acquisition

• Input datasets (DEM, slope, curvature, LULC, rainfall, soil) preprocessed and clipped to AoI/MWS boundaries.
• Resolution standardized to ~100m.
• Derived topographic indices (slope, curvature, flow accumulation) computed.

Raster Computation

• Raster outputs computed using established landslide susceptibility methodology.
• Entire AoI/MWS covered without gaps.
• Classification thresholds documented (low, moderate, high susceptibility).

Vectorization

• Raster outputs converted to field-level polygons using reduceToVectors() in GEE.
• Each polygon includes:
  • Susceptibility class
  • Area (ha)
  • Relevant metrics
• Polygons aligned with AoI/MWS boundaries.

Asset Publishing

• Raster and vector datasets published as Earth Engine assets.
• Metadata includes source datasets, resolution, processing date, and methodology.

Quality & Validation

• Coverage check: all study areas included.
• Accuracy check: susceptibility classes validated against known landslide locations or historical records.
• Attribute check: all polygons include class, area, and metrics.
• GEE visualization confirms correct spatial distribution.

Implementation Details

Implementation Details

Data Sources

• DEM (e.g., SRTM 30m)
• LULC datasets
• Rainfall and soil data
• Historical landslide inventory
• AoI/MWS boundaries

Processing

• Compute topographic and hydrological indices (slope, curvature, flow accumulation).
• Apply weighted susceptibility model from literature.
• Generate raster outputs at 100m resolution.
• Explore methods for higher-resolution susceptibility computation (<100m).

Vectorization & Publishing

• Convert raster outputs to polygons using reduceToVectors().
• Include attributes: class, area, slope, curvature, land cover.
• Upload raster and vector layers as EE assets with metadata.

Visualization

• Color-coded raster and vector layers in GEE (low = green, moderate = yellow, high = red).
• Overlay with AoI/MWS boundaries for field-level inspection.

Validation

• Compare outputs with historical landslide events.
• Spot-check vector polygons for correct classification.
• Generate validation report documenting coverage, accuracy, and completeness.

Mockups/Wireframes

No response

Product Name

KYL

Organisation Name

C4GT

Domain

No response

Tech Skills Needed

Python

Organizational Mentor

@amanodt @ankit-work7 @kapildadheech

Angel Mentor

No response

Complexity

Medium

Category

Backend

[abhishek-8081]

Greetings Mentors,

I'm Abhishek Kumar, currently an LFX Mentee with CNCF working on the KubeEdge Ianvs project. I'm interested in contributing to this landslide susceptibility mapping task.

Through my LFX mentorship, I've been debugging complex compatibility issues, optimizing Python environments, and building CI/CD pipelines with GitHub Actions. I have a strong foundation in data processing and algorithm optimization, along with experience in environmental sustainability projects.

I'm confident I can implement the methodology from the referenced paper and deliver the expected raster and vector outputs with proper validation. Looking forward to contributing to this disaster risk management initiative!

Best regards,
Abhishek Kumar

[vibhorjoshi]

hello i am vibhor joshi i am recently working on similar project About
Accurately extracting building footprints at large scale is challenging due to heterogeneous imagery, limited labels, and compute constraints. We present a GPU-accelerated hybrid GeoAI framework that transforms large rasters into normalized 3-channel patches with paired binary masks and processes them through a seven-step pipeline: data preparation.

link-https://github.com/vibhorjoshi/Fusing-Brains-and-Boundaries

i have experience in working with data i think i can contribute to your project

[aaditeshwar]

Thanks @abhishek-8081 , @vibhorjoshi for your interest. Will be great if you can self-organize to work together - it will help. Some comments to get started:

• Study the landslide susceptibility paper to understand the raster classes (per pixel) and vector properties (per AOI) that can be obtained from the dataset.
• Similarly study the existing CoRE stack datasets to understand how raster and vector outputs are presented. Take a look at LULC as an example for which which pan-India rasters are vectorized on micro-watershed (MWS) units, one tehsil at a time.
• In the CoRE stack, we are currently making pan-India outputs available as GEE assets, which are clipped using GEE libraries at the tehsil level and processed to obtain vector outputs at the MWS level for each tehsil.
• Once you have understood the above, the rest should be straightforward. The landslide susceptibility map is already available for India. You can write pipelines to clip it to tehsils, vectorize it, and ingest them for use in the Know Your Landscape dashboard and other applications.
• At a later stage, you can study what underlying datasets are dynamic in nature and would merit recomputation of the entire map on a regular basis. A fresh pipeline can be built at that point to not just use the published pan-India map and clip it, etc. but to compute it afresh for all of India say once a year. For this part, it will be good if you can try to identify a faculty mentor.

@kapildadheech , @amanodt , @ankit-work7 , @nirzaree-cfpt , FYI.

[vibhorjoshi]

ok

[vibhorjoshi]

I have one doubt. Can you provide me with a contribution guideline on how to contribute to import further changes made .

[vibhorjoshi]

@aaditeshwar i have created a PR request; please review the changes and tell me if any further changes required and please assign it to me

[aaditeshwar]

Hi all, Do please join the discord channel and googlegroup linked from this announcement of the CoRE stack innovation challenge, and do participate in the challenge too: https://core-stack.org/core-stack-innovation-challenge-1st-edition/. Our first community call on Friday 3-4pm will be very useful to get started.

[kapildadheech]

@amanodt Can you please review pr request and take this forward

[nirzaree-cfpt]

@vibhorjoshi : Hi Vibhor, we are unable to access the output files. Could you share outputs run through your pipeline, so that we can review them first?

[nirzaree-cfpt]

@vibhorjoshi : Also Vibhor, hope you know that we've been having weekly dev calls on Fridays at 3pm. Will be good to join these calls and discuss there too. Details here: https://core-stack.org/core-stack-innovation-challenge-1st-edition/

Join the Googlegroup and Discord channel.

[aaditeshwar]

Hi @vibhorjoshi and @abhishek-8081, We've been having weekly calls to discuss various ongoing contributions. Can you join the next call and take us through your methodology and implementation so far. Please do go over the recordings https://core-stack.org/core-stack-innovation-challenge-1st-edition/ and join the Google group https://groups.google.com/u/2/g/core-stack-dev and discord channel https://discord.gg/FpMArraM2X. We'll meet next Friday at 3pm. Please let us know if you can join.

[amit-spatial]

Hi all,

We'll have our dev call at 3pm on Friday as usual. There has also been a lot of interest on some issues we had created on github to add new datasets and pipelines to the CoRE stack. So we wanted to put out some guidelines to make this experience enriching and productive for all of us.

Many of the issues we had created and a wishlist of datasets we had published some time back are very domain specific and need to be technically correct before they can be included in the stack. For context, any of the layers currently on the CoRE stack that required some kind of geospatial algorithms or machine learning went through several months of testing, sometimes years, to examine the outputs carefully, understand when and where the models are performing well or not, how to handle missing data (satellite data can have cloud cover and needs specific methods to address this), and so on.

So if you are keen to pick up such issues then please share your intermediate progress with us during the dev calls. Help us understand how you are approaching the issue, show us the results, discuss problems, and it'll become much easier then to absorb your contributions. If you can inform us in advance about anything specific you'd like to discuss then we can also arrange for somebody who understands that area really well to also come on the call. Amongst so many research scholars and students who have contributed to the CoRE stack, there is a lot of expertise on LULC, computations on DEM layers, tree cover analysis, rainfall, drought, indexes like NDVI, census data, commodity pricing data, agriculture marketing, etc.

We are working hard to put up a developer website asap with all required documentation in one place, but here are two MUST read links before you begin to take up any substantial tasks.

Contribute datasets, pipelines and tools

Core-Stack Innovation Challenge

See you in our developer calls!

Zoom Link
Meeting ID: 856 5729 0380
Passcode: 317250