This is an R package containing sepcifically design metrics to assess and benchmark Land Surface and Vegetation Models against observations, but can also be applied to any simulation vs observation comparison.
There are 8 metrics used for different types of comparison
- Normalised Mean Error and Normalised Mean Squred Error (NME, NMSE) for most comparisons (inc gridded, temporal and site based comparisons)
- Manhattan Metric and Square Chord Distance (MM, SCD) for "item" comparisons (i.e, fractional cover of different land surfaces)
- A Discrete version of MM and SCD (DMM, DSCD) for item comparison for discete, one value items per cell/site, and items are described by affinity to traits.
- Mean Phase Difference (MPD) for comparisons of osillaiting processes (i.e, seasonal or diurnal variations
Scores of metrics are easy to interpret for model everulation and inter-model comparisons. Null models help aid score evalutation futher.
To insatll the package, first make sure you have R/Rstudio running and you have installed and loaded devtools:
install.packages('devtools')
library(devtools)
Then run the following command to install and load Benchmark Metrics:
install_github('douglask3/benchmarkmetrics/benchmarkMetrics')
library(benchmarkMetrics)
To see help files and examples of how to use:
?benchmarkMetrics
Info on the main functions can be found using:
?NME
?MM
?DMM
?MPD
click here for documentation.