obs_eval_fluxnet2015.Rdata wrong data

[FrancescoGrossi-unimi]

obs_eval_fluxnet2015.Rdata object contains the daily mean instead of weekly and monthly mean

load(here("data/obs_eval_fluxnet2015.Rdata"))
obs_eval$ddf |> unnest(data) |> summarise(mean = mean(gpp, na.rm= TRUE))
obs_eval$xdf |> unnest(data) |> summarise(mean = mean(gpp, na.rm= TRUE))
obs_eval$mdf |> unnest(data) |> summarise(mean = mean(gpp, na.rm= TRUE))
obs_eval$adf |> unnest(data) |> summarise(mean = mean(gpp, na.rm= TRUE))

[stineb]

We'll have to be careful in eval_sofun(), then, making sure that we compare model outputs and observations, aggregated to the same level.
Does this mean that the aggregation of these variables is different in data/obs_eval_fluxnet2015.Rdata compared to the driver object available from FluxDataKit v3.4? If so, that might make the re-use of eval_sofun() for benchmarking with both datasets a slight bit more complicated. If so, you may introduce an argument with a default, for example:

eval_sofun(..., newdata = TRUE)

[FrancescoGrossi-unimi]

if we use the data from df_drivers_fluxnet2015.Rdata and compare it with obs_eval_fluxnet2015.Rdata we get this results

as we can see, the slope of xdaily and monthly is very low. If we multiply them respectegvely by 7 and 30 we get 0.91 and 1.02 which shuoldbe the true value.

If we also compare the daily mean with week and monlthy (chuck code above) we get similar results.

What I think is that in obs_eval_fluxnet2015.Rdata the weekly and montlhy mean are instead daily mean