Visualise systematic model errors against covariates

Visualise systematic model errors against covariates.

# S3 method for hydromad
event.xyplot(
  x,
  events,
  formula = ~log2(e(Q, mean) + 0.01) + log2(e(lag(Q, -2), first) + 0.01) + log2(e(U,
    max) + 0.01) + e(E, mean),
  extract = residuals,
  with.U = TRUE,
  ...,
  panel = panel.superpose,
  panel.groups = panel.groups.funs,
  abline = list(h = 0),
  pch = ".",
  ylab = "residual flow sums in event windows (mm)",
  data = NULL
)

# S3 method for hydromad.runlist
event.xyplot(
  x,
  events,
  formula = ~log2(e(Q, mean) + 0.01) + log2(e(lag(Q, -2), first) + 0.01) + log2(e(U,
    max) + 0.01) + e(E, mean),
  extract = residuals,
  with.U = TRUE,
  ...,
  panel = panel.superpose,
  panel.groups = panel.groups.funs,
  abline = list(h = 0),
  pch = ".",
  ylab = "residual flow sums in event windows (mm)",
  data = NULL
)

Arguments

x

a hydromad or hydromad.runlist object.

events

event sequence produced by eventseq, or a vector defining continguous groups on the specified variables.

formula

formula defining the covariates to plot, as passed to the formula method of event.xyplot. It may refer to any of the variables in the model data frame (observed(x, select = TRUE)); additionally it may refer to the value U, for the effective rainfall series derived from the model. Finally julian may be referred to (from data time index).

extract

a function to apply to x to extract the response variable; by default this is residuals but could be e.g. fitted or function(x) residuals(x, boxcox = TRUE).

with.U

to include modelled effective rainfall U as a covariate.

...

further arguments passed to event.xyplot and on to xyplot and the panel function.

panel, panel.groups, abline, pch, ylab

passed to xyplot.

data

ignored.

Value

this function returns a trellis object which can be plotted.

See also

event.xyplot, xyplot, eventapply

Author

Felix Andrews felix@nfrac.org

Examples

data(Cotter)
x <- Cotter[1:1000, ]
mod <- hydromad(x,
  sma = "scalar",
  routing = "armax", rfit = list("sriv", order = c(2, 1))
)
ev <- eventseq(x$P, thresh = 3, inthresh = 1, indur = 5)
event.xyplot(mod, events = ev)
#> Loading required namespace: mgcv

event.xyplot(mod,
  events = ev,
  extract = function(x) residuals(x, boxcox = TRUE)
)


foo <- event.xyplot(mod,
  events = ev,
  ~ sqrt(e(P, max)) + sqrt(e(rollmean(lag(P, -1), 20, align = "left"), first))
)
dimnames(foo)[[1]] <- c("sqrt. peak rain (mm/day)", "mean 20-day ante. rain")
foo