EDA Mosaics🔗
AI Ready Mosaic Creation Parameters🔗
EDA's AI-Ready Mosaics (ARMs) are produced using proprietary algorithms to create cloud free, temporally coherent images ready for machine learning algorithms or mapping activities. The customization of sources, location, time, and method allow for a mosaic suited for specific analysis purposes.
ARM mosaics are produced with 6 bands from the four possible input sources:
- Sentinel-2A
- Sentinel-2B
- Landsat-8
- Landsat-9
All input data is level 2A meaning atmospherically corrected, surface reflectance products.
Single Source (Sentinel-2A/B) Mosaics🔗
| ARM Band Name | Sentinel-2 A/B Band | Approx. Center Wavelength (um) |
|---|---|---|
| coastal | Band 1 - Coastal | 0.443 |
| blue | Band 2 - Blue | 0.490 |
| green | Band 3 - Green | 0.560 |
| red | Band 4 - Red | 0.665 |
| rededge1 | Band 5 - Vegetation Red Edge | 0.705 |
| rededge2 | Band 6 - Vegetation Red Edge | 0.740 |
| rededge3 | Band 7 - Vegetation Red Edge | 0.783 |
| nir | Band 8 - NIR | 0.842 |
| nir08 | Band 8A - NIR | 0.865 |
| swir16 | Band 11 - SWIR | 1.610 |
| swir22 | Band 12 - SWIR | 2.190 |
Dual Source (Sentinel-2A/B + Landsat8/9) Mosaics🔗
| ARM Band Name | Sentinel-2 A/B Band | Landsat 8/9 Band | Approx. Center Wavelength (um) |
|---|---|---|---|
| coastal | Band 1 - Coastal | Band 1 - Coastal | 0.443 |
| blue | Band 2 - Blue | Band 2 - Blue | 0.490 |
| green | Band 3 - Green | Band 3 - Green | 0.560 |
| red | Band 4 - Red | Band 4 - Red | 0.665 |
| nir08 | Band 8A - NIR | Band 5 - NIR | 0.865 |
| swir16 | Band 11 - SWIR | Band 6 - SWIR 1 | 1.610 |
| swir22 | Band 12 - SWIR | Band 7 - SWIR 2 | 2.190 |
Area of Interest (AOI)🔗
Defines the geographic extent used to build the mosaic and is a key component for mosaic generation price.
Time of Interest (TOI)🔗
The time of interest will dictate the amount of available data the mosaic system can draw from. Limiting the TOI can deliver poor results if there is no cloud-free data within the specific AOI and TOI.
Mosaic Settings🔗
Resolution Selection🔗
Preview Mosaic🔗
The preview mosaic is used to ensure the combination of data source, AOI, and TOI is going to be viable for a given mosaic region. It will allow you to see the expected output of a Full Resolution mosaic in a fraction of the time. While the Preview doesn't apply all the improvements of Full Resolution (geometric, radiometric correction, aerosol-optical thickness weighting, or deep-learning cloud masks), it can still be used to get a sense of the expected cloud cover and visual consistency.
Once the parameters are to your liking, a full resolution mosaic can be produced with the same settings.
Full Resolution🔗
This is the setting to use for a full mosaic product. Note this will have a price associated with it and may still produce clouds if the settings for input data are too narrow. It is advised you preview any mosaic first to ensure usability.
Pixel Selection🔗
Best Measurement🔗
ARM's 'Best Measurement' extends work by White et. al (2014) and includes pixel-by-pixel weighting for several factors such as: sensor platform, scene content (clear, cloud, cloud shadow, water, snow), spatial distancing from measurement contamination, and aerosol optical thickness. This algorithm can produce highly consistent results where the goal is to choose the most representative sample for a given time period.
Peak Normalized Difference Vegetation Index (Coming Soon)🔗
ARM's 'Peak NDVI' seeks to maximize the vegetative signals from the mosaic process, targeting conditions with the most vigorous vegetation signal.
Peak Burn Severity (Coming Soon)🔗
ARM's 'Peak Burn Severity' seeks to maximize the response from burned pixels in order to map fire extent and degree of burn over large vegetated regions. To achieve peak burn severity, the Normalized Burn Ratio (NBR) is used to inform the pixel selection process. This mode can be sensitive to date selection and should be informed by knowledge of the local fire conditions and timing.
Percentile🔗
ARM's Percentile selection uses a purely statistical approach to identify common pixels from the stack of images. This approach can be very effective with larger volumes of data. This is a classical approach to pixel selection, but generally has poorer results compared to Best Measurement.
Source🔗
Select if your mosaic will be made using only Sentinel-2A/B or the combination of Sentinel-2A/B and Landsat-8/9.