Operational strategies to stabilize day-to-day variability in PlanetScope CI (agriculture)

Hi everyone,

We are using PlanetScope Surface Reflectance (PSB.SD, 8 bands) via Sentinel Hub for operational agricultural monitoring.

We calculate Chlorophyll Index Green (CI = NIR / Green − 1) and display daily maps to growers.

We are observing noticeable day-to-day variability in CI values that is not agronomically plausible (e.g., sudden drops followed by recovery the next day). This sometimes affects user confidence in the maps.

From Planet documentation, we understand that standard PSB.SD Surface Reflectance does not include BRDF normalization, so some inter-daily variability due to sun/view geometry is expected.

Our goal is not academic radiometric correction, but building a stable operational monitoring service.

For those using PlanetScope in agriculture:

1. How do you handle inter-daily variability in sensitive indices like CI or NDRE?

2. Do you apply geometry filtering (e.g., sun angle thresholds)?

3. Do you implement temporal smoothing or outlier detection in production workflows?

4. Has anyone tested ARPS-SR vs PSB.SD specifically for improving temporal stability in crop monitoring?

Any practical experiences or lessons learned would be greatly appreciated.

Thanks in advance!