How well LandscapeDNDC reproduces the observations of the validation sites, condensed to one skill value per ecosystem and variable (0 = no skill, 100 = perfect) and followed from model version to model version.
Loading the scorecard…
Mean skill of all targets and its three parts for every recorded model run. Steps between runs are calculated on the targets both runs share, so added sites do not show up as a model change. Select a point to show that run.
Darker = better. Below the skill: whether the model is too high or too low, and the change against the compared run. Select a cell for the sites behind it.
The comparison uses the same site simulations and observations as the validation figures. A series is one treatment of one site, a location is the site; every location counts equally, however many treatments or measurement days it has.
A change between two runs can also come from new sites, corrected inputs or revised observations; the revisions of model inputs and measurements are listed with every run.