CalcModPerfMulti: Computes model performance statistics for WRF-Hydro flux...
In mccreigh/rwrfhydro: R tools for the WRF Hydro Model

Description Usage Arguments Details Value See Also Examples

CalcModPerfMulti calculates model performance statistics for flux output.

1
2
3

CalcModPerfMulti(flxDf.mod, flxDf.obs, flxCol.mod = "q_cms",
  flxCol.obs = "q_cms", stdate = NULL, enddate = NULL,
  writeOutPairDf = NULL, fileTag = NULL)

`flxDf.mod`	The flux output dataframe (required). Assumes only one forecast point per file, so if you have multiple forecast points in your output dataframe, use subset to isolate a single forecast point's data. Also assumes model output and observation both contain POSIXct fields (called "POSIXct").
`flxDf.obs`	The observed flux dataframe. Assumes only one observation point per file, so if you have multiple observation points in your dataframe, use subset to isolate a single point's data. Also assumes model output and observation both contain POSIXct fields (called "POSIXct").
`flxCol.mod`	The column name for the flux time series for the MODEL data (default="q_cms")
`flxCol.obs`	The column name for the flux time series for the OBSERVED data (default="q_cms")
`stdate`	Start date for statistics (DEFAULT=NULL, all records will be used). Date MUST be specified in POSIXct format with appropriate timezone (e.g., as.POSIXct("2013-05-01 00:00:00", format="%Y-%m-%d %H:%M:%S", tz="UTC"))
`enddate`	End date for statistics (DEFAULT=NULL, all records will be used). Date MUST be specified in POSIXct format with appropriate timezone (e.g., as.POSIXct("2013-05-01 00:00:00", format="%Y-%m-%d %H:%M:%S", tz="UTC"))
`writeOutPairDf`	character path/name for file to output the data frame constructed before calculating the statistics and various aggregations. NULL by default gives no output.
`fileTag`	A to be inserted in the file name, just before the file extension.

CalcModPerfMulti reads a model flux time series (i.e., created using ReadFrxstPts) and an observation time series (i.e., created using ReadUsgsGage) and calculates model performance statistics (Nash-Sutcliffe Efficiency, Rmse, etc.) at various time scales and for low and high fluxes. The tool will subset data to matching time periods (e.g., if the observed data is at 5-min increments and modelled data is at 1-hr increments, the tool will subset the observed data to select only observations on the matching hour break).

CalcModPerfMulti calculates the same statistics as CalcModPerf) but returns results as a single-row vs. multi-row dataframe. This is intended to streamline compiling statistics for multiple data runs or multiple sites.

Performance Statistics:
(mod = model output, obs = observations, n = sample size)

n: sample size
nse: Nash-Sutcliffe Efficiency

nse = 1 - ( sum((obs - mod)^2) / sum((obs - mean(obs))^2) )
nselog: log-transformed Nash-Sutcliffe Efficiency

nselog = 1 - ( sum((log(obs) - log(mod))^2) / sum((log(obs) - mean(log(obs)))^2) )
cor: correlation coefficient

cor = cor(mod, obs)
rmse: root mean squared error

rmse = sqrt( sum((mod - obs)^2) / n )
rmsenorm: normalized root mean squared error

rmsenorm = rmse / (max(obs) - min(obs))
bias: percent bias

bias = sum(mod - obs) / sum(obs) * 100
msd: mean signed deviation

msd = mean(mod - obs)
mae: mean absolute error

mae = mean(abs(mod - obs))
errcom: error in the center-of-mass of the flux, where center-of-mass is the hour/day when 50% of daily/monthly/water-year flux has occurred. Reported as number of hours for daily time scale and number of days for monthly and yearly time scales.
errmaxt: Error in the time of maximum flux. Reported as number of hours for daily time scale and number of days for monthly and yearly time scales).
errfdc: Error in the integrated flow duration curve between 0.05 and 0.95 exceedance thresholds (in native flow units).
errflash: Error in the Richards-Baker Flashiness Index

R-B Index = sum(abs(q_t - q_t-1)) / sum(q)

.

Time scales/Flux types:

t = native model/observation time step (e.g., hourly)
dy = daily time step
mo = monthly time step
yr = water-year time step
max10 = high flows; restricted to the portion of the time series where the observed flux is in the highest 10% (native time step)
min10 = low flows; restricted to the portion of the time series where the observed flux is in the lowest 10% (native time step)

A new dataframe containing the model performance statistics.

Other modelEvaluation: CalcModPerf, CalcNoahmpFluxes, CalcNoahmpWatBudg, CalcStatCategorical, CalcStatCont

## Take forecast point model output for Fourmile Creek (modStrh.mod1.fc) and a 
## corresponding USGS gage observation file (obsStrh.fc), both at an hourly time 
## step, and calculate model performance statistics. The model forecast point 
## data was imported using ReadFrxstPts and the gage observation data was
## imported using ReadUsgsGage.


## Not run: 
CalcModPerfMulti(modStrd.chrt.fc, obsStr5min.fc)

> Output:
t_n t_nse t_nselog t_cor t_rmse t_rmsenorm t_bias  t_msd  t_mae t_errfdc t_errflash
116  0.81     0.67  0.91  0.211       8.48    7.1 0.0313 0.1385     0.04     0.0132
dy_n dy_nse dy_nselog dy_cor dy_rmse dy_rmsenorm dy_bias dy_msd dy_mae dy_errcom dy_errmaxt dy_errfdc dy_errflash
 116   0.81      0.67   0.91   0.211        8.48     7.1 0.0313 0.1385         0          0      0.04      0.0092
mo_n mo_nse mo_nselog mo_cor mo_rmse mo_rmsenorm mo_bias mo_msd mo_mae mo_errcom mo_errmaxt yr_n
   5   0.92      0.82   0.96  0.1073       10.35     9.1 0.0324 0.0833      -1.2        0.2    1
yr_bias yr_msd yr_mae yr_errcom yr_errmaxt max10_n max10_nse max10_nselog max10_cor max10_rmse max10_rmsenorm max10_bias max10_msd max10_mae
    7.1 0.0313 0.0313         0          4      12      0.46         0.58      0.71      0.348          23.63       -2.8    -0.044     0.248
min10_n min10_nse min10_nselog min10_cor min10_rmse min10_rmsenorm min10_bias min10_msd min10_mae
     12   -385.99       -14.59      0.35     0.1678         538.78      112.4     0.056    0.0592

## End(Not run)