Model Overview
Satellite Data: Sentinel-2 (Multispectral)
The Forest Canopy Height model estimates tree canopy height using optical satellite imagery. Forest canopy height is an important variable for ecological applications such as biomass estimation, biodiversity monitoring, forest restoration, and forest management.
The model also helps in estimating Diameter at Breast Height (DBH), making canopy height an important variable for carbon credit estimation.
Using a deep learning model, optical imagery is translated into canopy height estimates. The resulting spatial information can help forest managers and policymakers monitor forest growth, identify changes in forest health, assess biodiversity, and support the sustainable management and preservation of forest ecosystems.
Model Inputs
The model requires an Area of Interest and a Date of Interest for which canopy height needs to be estimated.
| Input Name | Description |
|---|---|
| Area of Interest | Area for which canopy height estimation is required. |
| Date of Interest | Month and year for which canopy height estimation is required. The selected date should fall within a leaf-on month. |
Important Guidelines
- The Area of Interest should be at least 2.650 km × 2.650 km.
- The Date of Interest should fall within a leaf-on month.
Model Output
The model generates two single-band raster outputs visualized over the selected AOI. These outputs provide estimated canopy height and the corresponding uncertainty at 10 m resolution.
The model also generates summary statistics showing bucket-wise mean and standard deviation values for the canopy height and uncertainty estimates.
Output Files
| S.No. | Output | Description | Download output as |
|---|---|---|---|
| 1 | CHE_ht (Forest Canopy Height) | Raster containing estimated forest canopy height values over the Area of Interest at 10 m resolution. | GeoTIFF, single-band raster |
| 2 | CHE_ht_unc (Forest Canopy Height Uncertainty) | Raster containing the uncertainty associated with the canopy height estimates at 10 m resolution. | GeoTIFF, single-band raster |
| 3 | CHE_ht_stats (Canopy Height Statistics) | Histogram showing bucket-wise mean and standard deviation of canopy height estimates. | JPEG |
| 4 | CHE_unc_stats (Canopy Height Uncertainty Statistics) | Histogram showing bucket-wise mean and standard deviation of canopy height uncertainty estimates. | JPEG |
Estimated Model Run Time
| AOI Size (Sq KM) | Estimated Model Run Time |
|---|---|
| 10 | 23 seconds |
| 20 | 25 seconds |
| 50 | 26 seconds |
| 100 | 39 seconds |
| 150 | 46 seconds |
| 200 | 51 seconds |
| 250 | 55 seconds |
| 350 | 66 seconds |
| 500 | 81 seconds |
| 1000 | 139 seconds |
| 1500 | 270 seconds |
| 2000 | 291 seconds |
| 2500 | 318 seconds |
| 3000 | 364 seconds |
| 3500 | 431 seconds |
| 4000 | 481 seconds |
| 4500 | 507 seconds |
| 5000 | 528 seconds |
Additional Details
| Parameter | Details |
|---|---|
| Minimum AOI Size | 2.650 km × 2.650 km |
| Maximum AOI Size | 5000 sq km |
| Geographies Supported | Provides reasonable estimates irrespective of geography, with high levels of uncertainty over highly undulating and mountainous terrains |
| Sensors Supported | Sentinel-2 |
| Type of Images Supported | Multispectral |
| Hyperspectral Images Supported | No |
| Resolution of Input Imagery | 10 meters |
| Model Type | Deep Learning model |
| Model Accuracy | Estimated within the range of 50 m |
| Model Height Limit | Saturates at 50 m height |
Limitations
The model has high uncertainties over highly undulating and mountainous terrains.
The model also saturates at average canopy heights of 50 m, meaning that it cannot reliably estimate canopy heights beyond this threshold.
Model Accuracy
The accuracy of the model is estimated within the range of 50 m. The model provides reasonable estimates irrespective of geography, while uncertainty increases over highly undulating and mountainous terrains.