search

bgctw / REddyProc

Processing data from micrometeorological Eddy-Covariance systems
57 stars 31 forks source link

Provide vignette on aggregating uncertainty #23

Closed bgctw closed 5 years ago

bgctw commented 5 years ago

There are continuous support request on how to aggregate the uncertainty to daily and annual aggregates. Would be good to provide an example with the package.

lsigut commented 5 years ago

I have already done that as a part of my Summary workflow that I plan to share very soon. We could link it here once I put it on GitHub. Any comments welcome of course.

lsigut commented 5 years ago

The complete example including uncertainty aggregation is available here: https://github.com/lsigut/EC_workflow The relevant part is inside siteyear_Summary_2018-09-17.R.

bgctw commented 5 years ago

Dear Ladislav,

As I see you aggregate uncertainty with neglecting correlations, i.e. assuming independent measurements. To my experience, this will in many cases underestimate uncertainty in annual aggregates quite a lot.

Yours Thomas

lsigut commented 5 years ago

Thank you for the feedback! I was actually surprised that indeed the resulting uncertainties were quite low. I used rather simplistic approach, aggregating "in quadrature" mostly based on the Section Uncertainty Quantification in book Aubinet et al., 2012. Do you know about some reference for uncertainty aggregation including correlations?

bgctw commented 5 years ago

I created a vignette in #25. The markdown version is also in the repo and can be directly displayed. @lsigut What is your opinion?

bgctw commented 5 years ago

Markdown version of the vignette now in linked from master branch

lsigut commented 5 years ago

I created a vignette in #25. The markdown version is also in the repo and can be directly displayed. @lsigut What is your opinion?

This looks great Thomas. Thank you for explanations and nice demonstration. I will use it for my work and will implement it in the Summary EC workflow. I omitted correlations but also 1.96 stdDev.