Age of Forests and Plantations (Field Level @30m)

[kapildadheech]

Ticket Contents

Description

Estimating the age of forests and plantations provides key insights into forest dynamics and management. Knowing forest age helps differentiate indicators such as the timing and location of deforestation or degradation. Computation at 30m resolution allows field-level analysis, and vectorization enables integration with Area of Interest (AoI) and Micro Watershed (MWS) boundaries.

Goals

Goals

• Estimate the age of forests and plantations using Earth Observation data and available forest records.
• Generate raster outputs at 30m resolution representing forest and plantation age.
• Vectorize raster outputs into field-level or MWS-level polygons for analysis.
• Publish raster and vector outputs as Earth Engine assets with metadata.
• Enable visualization and temporal monitoring of forest age across AoI and MWS.
• Support integration with other forest health indicators such as deforestation or degradation events.

Expected Outcome

Expected Output

• Raster dataset (30m resolution) showing forest and plantation age (years) for AoI/MWS.
• Vectorized polygons with attributes:
  • Forest/plantation age
  • Area (ha)
  • Relevant metrics (mean age, max/min age per polygon)
• Published Earth Engine assets (raster + vector) with metadata (source datasets, processing date, resolution).
• Visualization in GEE color-coded by age (e.g., younger = light green, older = dark green).
• Validation report confirming coverage, accuracy, and alignment with forest records.

Acceptance Criteria

Acceptance Criteria

Data Acquisition

• Input datasets (forest inventory, plantation records, LULC time series) must be preprocessed and clipped to AoI/MWS boundaries.
• Resolution standardized to 30m.

Raster Computation

• Raster outputs must provide forest/plantation age per pixel.
• Entire AoI/MWS must be covered with no gaps.
• Age computation must allow detection of temporal changes related to deforestation or degradation.

Vectorization

• Raster outputs must be converted to vector polygons using reduceToVectors() in GEE.
• Each polygon must include:
  • Forest/plantation age
  • Area (ha)
  • Relevant metrics (mean/max/min age)
• Polygons must align with AoI/MWS boundaries.

Asset Publishing

• Raster and vector datasets must be published as Earth Engine assets.
• Metadata must include data sources, resolution, and processing date.

Quality & Validation

• Coverage check: 100% of forested areas within AoI/MWS included.
• Accuracy check: age estimates match reference forest/plantation records.
• Resolution check: raster outputs at 30m.
• Attribute check: all polygons include age, area, and metrics.
• Visualization in GEE shows correct distribution of younger vs older forests.

Implementation Details

Implementation Details

Data Sources

• Forest inventory datasets, plantation records, LULC time series.
• Sentinel-2, Landsat-8/9 imagery.
• AoI and MWS boundary polygons.

Processing

• Compute forest/plantation age using satellite time series analysis, tree cover change detection, or reference records.
• Generate raster outputs at 30m resolution for each year/period.
• Clip outputs to AoI/MWS boundaries.

Vectorization & Publishing

• Convert raster outputs into polygons using reduceToVectors().
• Include attributes: age, area, mean/max/min per polygon.
• Upload raster and vector layers as EE assets with metadata.

Visualization

• Color-code polygons in GEE (e.g., light green = young, dark green = old).
• Overlay with AoI/MWS boundaries for field-level insights.

Validation

• Compare outputs with reference forest and plantation records.
• Spot-check polygons for correct age classification.
• Generate validation report documenting coverage, accuracy, and attribute completeness.

Mockups/Wireframes

No response

Product Name

KYL

Organisation Name

C4GT

Domain

No response

Tech Skills Needed

Python

Organizational Mentor

@ankit-work7 @amanodt @kapildadheech

Angel Mentor

No response

Complexity

Low

Category

Backend

[harshala334]

Hi @kapildadheech @amanodt @ankit-work7 ,
I’d like to work on this issue. I have experience with Python, and geospatial data processing, can handle the complete pipeline, from preprocessing and raster computation to vectorization, and validation.
Could you please assign me this issue?

[aaditeshwar]

Thanks @harshala334 for the interest. Suggesting some points:

• First let us determine the methodology to use. See papers like https://www.nature.com/articles/s41597-022-01260-2 which spot the last period of disturbance using Landtrendr and BFAST algorithms, or https://www.sciencedirect.com/science/article/pii/S0034425720305381 which uses time-series of LULC and canopy height to identify, for a pixel, since when it has been forested or planted. This of course helps go back until when Landsat imagery is available, i.e. 1980s. To determine age for pixels that have been forested even before that, papers like https://essd.copernicus.org/articles/13/4881/2021/ estimate the current AGB, and AGB growth rates, to find since when the tree cover has been around.
• Will be easier to do this in stages: Let's do for age since 1980s for now, and later do the AGB based estimation for older forests.
• For a given AOI, build a pipeline then which uses tree cover LULC in the CoRE stack, and implements the Landtrendr/BFAST/canopy height methods, and produces raster outputs of the age of a tree cover pixel. The papers above may have published their code which can guide the implementation exercise.
• Produce sample outputs and validate them. Most of the papers above have produced datasets until the year the paper was published, and can provide a good validation.
• Finally integrate it into the CoRE stack by publishing it as an API. Further downstream steps would involve vectorization of the outputs. CoRE stack publishes most indicators at the micro-watershed level. Check out indicators for LULC, for example. Similarly, per-MWS indicators in this case could take the form of a distribution: extent of tree cover which is older than the 1980s, % which is older than 1990, % older than 2000, etc.
• The raster and vector datasets and then be consumed in apps and dashboards like Know Your Landscape.

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

[harshala334]

Hi @aaditeshwar ,
Thank you for the detailed guidance! I've reviewed the approach and understand the plan. Here's my understanding:

Proposed Approach:

1. Research Phase

• Study the three papers to understand LandTrendr, BFAST, and LULC time-series methods
• Check for published code repositories from the papers
• Prototype in GEE Code Editor with a small test AOI

2. Implementation

• Start with LULC time-series approach
• Implement LandTrendr
• Focus on forests since 1980s
• Leverage existing canopy cover and height modules

3. Validation

• Compare outputs with published datasets from the papers
• Calculate accuracy metrics (MAE, RMSE)

4. Integration

• Create computing/forest_age/ module
• Build REST API endpoints for computation and retrieving MWS-level stats
• Vectorize outputs to show age distribution (% older than 1980, 1990, 2000, etc.)

What's the expected timeline for this project? I can plan and divide the tasks accordingly.

[aaditeshwar]

Sounds great @harshala334 . Overall, I'd expect a month or two, and feel free to also connect directly with me at aseth@cse.iitd.ac.in to discuss any specifics.

[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.