tiger (version 0.2.2)

plots: Evaluation plots for temporal dynamics of model performance

Description

Create various plot to understand the temporal dynamics of model performance

Usage

box.plots(result, solution, show.measures = 1:num.measures,
             new.order = 1:solution, show.synthetic.peaks = FALSE,
             synthetic.peaks.col = c(2:8, 2:8), show.timestep = NA,
	     show.cell = NA,
                 ref = NULL, ref.new.order = new.order, ref.solutions =
                 solution, col.best.match = "black",
                 clusterPalette = rainbow(solution))
errors.in.time(xval, result, solution, rain.data = NULL, show.months
                 = FALSE, new.order = 1:solution, x.range =
                 1:length(xval), pmax = max(c(result$measured,
                 result$modelled), na.rm = TRUE), data.colors =
                 data.frame(measured = c("grey"), modelled =
                 c("black"), rain = c("black")), clusterPalette =
                 rainbow(solution), color.cut.off = 0, frac.max = 0.7,
                 frac.min = 0.4, grid.nx = 0, legend.pos = "topleft",
                 show.data = TRUE, show.errors = TRUE, show.data.model
                 = show.data, show.data.measured = show.data, ...)
peaks.in.clusters(result, solution,  new.order = 1:solution)
peaks.on.som(result, solution, clusterPalette=rainbow(solution), cell.size = 0.9, mfrow=c(2,ceiling(n.errors/2)), new.order=1:solution)
som.cluster(result, maxc = result$maxc, solutions = 2:maxc, new.order = NA)
som.measure.level(result, show.measures = 1:num.measures, mfrow = c(4, 3))
peaks.measures(result, show.measures = 1:num.measures,
                 synthetic.peaks.col = c(2:8, 2:8), mfrow = c(2, 3),
                 col.best.match = "black", do.out = rep(TRUE,
                 length(show.measures)), single.errors = FALSE,
                 show.legend = TRUE, show.main = TRUE, y.range = NULL)
scatterplot(measures, show.measures=1:num.measures)
p.validityIndex(result, validity.max)

Arguments

result

object returned from tiger

measures

data.frame from which to create a scatter plot. e.g. result$measures.uniform

solution

number of clusters to use for further evaluations (see also validityIndex)

show.measures

Vector with the indices of the performance measures to plot. Use result$names[show.measures] to see the names of the measures

new.order

New numbering to assign to clusters. See also change.order.clusters

show.synthetic.peaks

Show values of the synthetic peaks on top of the box plots.

synthetic.peaks.col

Colors to use for synthetic peaks.

do.out

vector of booleans indicating whether to exclude outliers when showing the plot

cell.size

fraction of the cell square to be filled with color

show.cell

the scores for a certain cell on the SOM can be ploted as blue line on the box plot (see examples)

x.range

Indizes of x-values to be plotted

y.range

Range for y axis

pmax

maximum discharge for definition of the plot range

frac.min

minimum of the y-range covered by color bars for cluster occurence

frac.max

maximum of the y-range covered by color bars for cluster occurence

clusterPalette

colors to use for the clusters

color.cut.off

Value of cluster occurence below which the color bar is set to transparent (for better readability)

legend.pos

Position of the legend

data.colors

Color definition for rainfall and runoff

show.timestep

timestep for which the values for the performance measures are to be plotted as black lines in the box plot

xval

Values to be plotted on the x-axis (e.g. POSIX-date)

show.months

Boolean indicating whether to add month ticks to x axis

mfrow

see par

ref

Reference solution to be ploted in grey on the box plot

ref.new.order

New numbering to assign to clusters for reference solution on the box plot

ref.solutions

Number of clusters for reference solution for which to plot the box plot

validity.max

Do not plot solutions with cluster numbers resulting above in a validty index above validity.max

col.best.match

Color to use for plotting the line indicating the position of the best match

rain.data

vector with rainfall data

show.data

boolean, indicating whether to show discharge data

show.data.measured

boolean, indicating whether to show measured discharge data

show.data.model

boolean, indicating whether to show modeled discharge data

show.errors

boolean, indicating whether to show error type bars

grid.nx

number of grid lines to be ploted (see grid)

maxc

Highest number of clusters to be plotted

solutions

Vector of solutions to be plotted

single.errors

Boolean, indicating weather different synthetic errors should be combined into a single plot or shown in multiple plots

show.legend

Boolean, indicating whether to show the legend

show.main

Boolean, indicating whether to show performance measure names as plot title

...

additional parameters passed to plot

Value

• used for the side effect of plotting results

Details

box.plots: for each performance measure, a box plot is created showing the values for each cluster

errors.in.time: occurence of the errors cluster along the time dimension

peaks.in.clusters: table of the position of the synthetic peak errors in the clusters.

som.cluster: Plot the location of the different clusters on the SOM for multiple numbers of clusters.

som.measure.level: Plot the distribution of the performance measures on the SOM.

peaks.measures: responce of the performance measures to the synthetic peak errors.

scatterplot: scatter plot of the performance measures

See package vignette for further details about which plot does what.

References

Reusser, D. E., Blume, T., Schaefli, B., and Zehe, E.: Analysing the temporal dynamics of model performance for hydrological models, Hydrol. Earth Syst. Sci. Discuss., 5, 3169-3211, 2008.

See Also

The package vignette

Examples

data(tiger.example)

new.order <- c(6,3,2,5,4,1)
correlated <- correlated(tiger.single, keep=c("CE","RMSE" ))

opar <- par(mfrow=c(3,5))
box.plots(tiger.single, solution=6, new.order=new.order, show.synthetic.peaks=TRUE)
box.plots(tiger.single, solution=6, new.order=new.order, show.cell=data.frame(x=1,y=1))
par(opar)
errors.in.time(xval=d.dates, result= tiger.single, solution=6, show.months=TRUE, new.order=new.order)
peaks.in.clusters(tiger.single, solution=6, new.order=new.order)
peaks.measures(tiger.single, show.measures=correlated$measures.uniform$to.keep)
scatterplot(tiger.single$measures.uniform, show.measures=correlated$measures.uniform$to.keep)