Gap-fill NDVI
Sensors Supported: Sentinel-2
The Gap-Filled NDVI Model is designed to fill in missing NDVI values caused by cloud cover, ensuring continuous assessment of farm health. The model leverages Sentinel-1 data to predict missing NDVI values. The model calculates the average NDVI with each farm boundary and fills in gaps caused by cloud cover.
Input
The below inputs are required to be provided by the user. The gaps in the given date range caused due to cloud cover will be gap filled.
| Name | Description |
|---|---|
| Farm Boundary | Boundaries Extracted from the Farm Boundary Model |
Start date | The start date in “YYYY-MM-DD” format |
End date | The end date in “YYYY-MM-DD” format |
Output:
Model generate two outputs i.e. Classified map and its statistics. These files will be exported in .geojson and tabular format respectively.
| timestamps | num of timestamps | mean ndvi | gap filled ndvi | number of timestamps with gaps | number of timestamps gaps filled | percent of gaps filled |
|---|---|---|---|---|---|---|
| 2021-01-03, 2021-01-04, 2021-01-08, 2021-01-11 | 4 | nan, 0.4873, nan, nan | 0.4697, 0.4873, 0.3934, 0.3561 | 3 | 3 | 100 |
| 2021-01-03, 2021-01-04, 2021-01-08, 2021-01-11 | 4 | nan, 0.47, nan, nan | 0.6204, 0.47, 0.4619, 0.6516 | 3 | 3 | 100 |
USE CASE Example
Input
Farm Boundary Model output is used here as an input.
Along with Farm Boundary output, Time frame or Date range also need to be provided as an additional input since with this the mean NDVI values will be estimated.
| Name | Description |
|---|---|
Start date | The start date in “YYYY-MM-DD” format |
End date | The end date in “YYYY-MM-DD” format |
Output:
Model generate two outputs i.e. Classified map and its statistics. These files can be downloaded as .geojson and tabular format respectively.
On the platform we can see the NDVI graphs for the selected farm.
Additional Details
Minimum size of Area of Interest required (in sq km): 10
Maximum size of Area of Interest supported (in sq km): 5000
Geographies supported: All geographies
Sensors Supported: Sentinel-2
Model Accuracy: 75% to 90%, depending on the geography
Important Note:
- The crop type, the soil and moisture do influence the Sentinel-1 backscatter and thereby the estimated gap filled NDVI values.
- The number of days of cloud free data does also influence the outputs as this data be used in building the model.