Meta Resampling Question

I pose a meta resampling question: what is the optimal resampling spatial resolution for S2A data for ingestion within the biophysical processor for LAI output? Since most bands (B3 through B8a, B11, B12) used for vegetation applications have a spatial resolution of 20m, I would say this would be an optimal resampling resolution; however one could argue that all bands at some point are used to get to the end product of LAI, I presume, so resampling at 60m would be the optimal choice for this argument. The goal of obtaining the highest spatial resolution of 10m given that 1/3 bands are used for vegetation applications could be an argument as well.

I would lean towards resampling at 20m, given that 2/3 of the vegetation bands are acquired at this resolution, it strikes a good balance between resolution and processing time and would be easier to defend if questioned based on the aforementioned statements. I would like to hear others thoughts on the matter.

It is hard to give a good one-size fits all answer as different vegetation has different spatial and time scales. I work with oceans which are the extreme case of instability. As soon as you do time-averages you end up with much smoother (in space) images except in areas where you only have one image without cloud. The most convincing approach is to run your analysis using several resolutions. The ask: How would my conclusions differ if I only used one resolution?

In practice, you may need more than one resolution: higher resolution may be helpful to understand edge effects in a qualitative way while lower resolution is useful when looking at areal means.

1 Like