Handling Missing Values in Historical Data


I have noticed that for the historical observations, there are many null values for precipitation and snow.

I believe I have read somewhere that a null value is genuinely missing data and not 0. However, i am unsure as i cannot find the page i read this on and many null values seem like they occur at points where the value is unlikely to be more than 0. For example, solar radiation is periodically null for night hours as far as i can observe.

For snow however, there are instances of snow depth being populated with values while snow remains null for the entire date range.

For precipitation, it becomes harder to determine whether the nulls are due to 0 values or whether there is a high volume of missing data, in which case there is a data quality problem.

Question 1: Are all null values instances of missing data and NOT recordings of 0? Does this rule vary across variables such as solar radiation, wind gust, precipitation?

Question 2: Is there a standard recommended methodology to infill missing observations for variables such as precipitation? I have attempted histfcst, but nulls still pull through.


Example request: 




Please sign in to leave a comment.