dude there is something very weird with the data, the test data is even weirder, I don't know if that's something to solve ourselve or shoud we just ignore this entirely. beside what do they mean with future data which we have no right to use.?

this is very confusing.

did you ever get to the bottom of this? I am finding it a bit counter intuitive if we are training models using the atmospheric data given and then feeding the test data to the model without any of the conditions to make the predictions from?

Aha, I get it now , I was confused with missing values in test data and how to predict them , it seems that time series forcasting is different from machine learning prediction and i have never play with time series data before . you can check this link: https://blogs.oracle.com/datascience/7-ways-time-series-forecasting-differs-from-machine-learning