Satellite data not available for some train/test samples

NASA Airathon
1

Hi there!

I opened this issue because there are data samples for which there is no satellite (MAIAC or MISR data available). For instance:
If you check the submission_format.csv, there are samples ranging from 2017-01-07T16:00:00Z. However, the most recent satellite file for MISR type is on 20171101T192120_misr_la_0.nc
And this is the case for many other samples:

Samples without MAIAC data for the training data:
11
Samples without MAIAC data for the testing data:
0
Samples without MISR data for the training data:
46
Samples without MISR data for the testing data:
4312

I just downloaded all the data available through the S3 entry point (a total of 5054 files for training data and 2679 for testing data)

What are we supposed to do with this problem?

Many thanks in advance!

Antonio Pérez Velasco

2

Hi @pereza - Thanks for sharing this exploration. Unfortunately, it’s true that there are many samples with missing satellite data. That’s just the reality of satellite orbits that this data is derived from. It’s part of what makes this challenge, well, a challenge! Here’s another thread where this is discussed: Strategy to impute missing data? - #2 by Carl_Malings

I would also suggest exploring how ancillary and historical data could help fill in the gaps. Good luck!