Spend less data scientists’ time cleaning data: High-quality cloud masks for Sentinel2, Landsat, and others available today 

Earth Observation has a long history of serving agriculture, from governments to many types of agribusinesses. Recently, new advances have equipped providers with a greater ability to drive even more value. 

In producing high-quality geospatial analytics, one of the major challenges for providers is clouds, shadows, and haze on satellite imagery.  

Crop monitoring by satellite relies on visible and near-visible wavelengths of light. Clouds mask the color and brightness of different vegetation regions, creating an obstruction of reflections in cloud-covered regions. Based on cloudy images, the NDVI map will have distorted values. The vegetation on the fields will look stressed and non-homogenous on the NDVI map. 

Standard NDVI map showing a red anomaly in the field which could be interpreted as a problematic area (June 22). 

Cloud-free picture. (June 27). 

That’s why the application of an appropriate cloud masking process is key for understanding field variability and avoiding false interpretations.  

To deliver reliable data to its customers, EarthDaily Agro has its own proprietary Cloud Masking process calibrated for agricultural areas with the best-fitted parameters for this type of monitoring. EarthDaily Agro Cloud Masking model has been calibrated and validated in agricultural areas all over the world. 

The accuracy of the EarhDaily Agro cloud mask is higher than the cloud mask of other providers, allowing for a reduction in under-detection and keeping a high quality on cloud over-detection to not miss clear areas. 

Recently we’ve deployed our new Auto Clear Mask (ACM) in Python. This new version improves the performance: twice as fast for Sentinel2 images and 4 times faster for Landsat 8 & 9 compared to the last version! 

In terms of accuracy, this new version is better for cloud detection and reduces the problem of over-detection of cloud objects. 

We invite you to consult in GitHub the study contains a Python notebook comparing cloud masks from different providers: 

  • ACM: cloud mask generated by EarthDailyAgro 
  • SCL : Sentinel-2 L2A Scene Classification 
  • WQR: Reference mask based on human-corrected ACM mask (World Quality Reference) 

EarthDaily Agro Cloud Mask guarantees the quality of our analytics, but they will also be available to our customers who have their own image processing chain as a Data as a Service.  

With these high-quality cloud masks, you will be sure to have the right analytics to make a good decision! 

If you want to know more about EarthDaily Agro Auto Clear Mask or EarthDaily analytics, contact our team