Better method to QC (or gap fill) precipitation data

[wwieder]

Do we need a better method to QC precipitation data?

A number of sites seem to have individual years with unexpectedly high or low precipitation values (e.g. NOGP 2018-2020). In some cases these low values are preventing the model from spinning up. If they are correct, then we need to look at this on the CLM side, but how can we check for reasonableness of the input data from NEON?

I've created a summary table of current precipitation provided in the V1 data and visually flagged years that seem unexpectedly high / low.

How do we identify 'bad' data from NEON and what do we replace it with?

• Nearby station data?
• Reanalysis product?

Regardless, I suggest we distribute precipitation evenly over the period for which we have data (e.g. hourly or daily to the 30 minute CLM timestep.

[wwieder]

Adrianna also made these plots comparing monthly total to PRISM data.

[wwieder]

Finally, Andrew Richardson made this plot showing data that Alison and Adrianna put together from CPER with NEON (blue), ClimateNA extracted from PRISM (red), and HQ data (Green).