Repository navigation
Forecast series for the fingerprint plots #171
Description
Split out from #129. We tested four ways to show NEE after 2021 at six sites: a Theil-Sen trend, a least-squares trend, ARIMA/ETS and CMIP6 model output. None predicted 2017 to 2021 better than the plain average of the earlier years. FLUXCOM NEE has no significant trend over 2001 to 2021, so there is nothing to extend.
Plan: the app will not compute a forecast from past data. If we add a forward view, it will read CMIP6 land-carbon output (gpp, ra and rh from at least three models) and show the change from today to 2041 to 2060 as a separate layer, off by default and labelled "scenario, not a forecast".
Activity
lazarusA commentedon Nov 3, 2026
sorry, being very busy lately... this will need some research and thinking (for the forecast part) before making some decisions.
The one thing that could be easier to do at the moment is the original issue, trends across years from the fingerprints, namely, means, median, std. Once you have for the example the mean across them, you could do the difference to get anomalies, namely, which year was an extreme?
Keep the app still focus on fingerprints, it needs to that well, before moving into more advanced features.