VarEPS flux parameters

For applications requiring products starting from or spanning step T+240 users wishing to obtain products interpolated to regular lat/lon grids are advised to truncate the Leg 1 products to the N128 reduced Gaussian grid prior to interpolation to their choice of lat/lon grid. This will limit unnecessary impact from interpolation on the consistency of the fields and avoid noise appearing in flux fields when taking differences in time for timestep intervals starting from T+240 or spanning T+240, e.g. precipitation for the interval T+(252-240) or T+(246-234).

To indicate that Leg 1 products should be truncated to the N128 reduced Gaussian grid prior to interpolation to a lat/lon grid, the additional MARS keyword should be specified in the request:

RESOL = REDUCED GAUSSIAN 128

Computing differences of flux parameters

When computing differences for flux parameters between one timestep and an earlier timestep, be aware that interpolation to a lat/lon grid can amplify the GRIB packing difference and this may lead to spurious negative values in the final field. For further details, see:

http://www.ecmwf.int/publications/manuals/libraries/interpolation/precipitationFIS.html

In general, a more satisfactory result will be obtained by performing the subtraction on the reduced Gaussian grid and then interpolating the result to a lat/lon grid as a final step.

For time intervals spanning the truncation time step, the difference should be computed on the N128 reduced Gaussian grid of forecast Leg 2.

The following example shows one way this can be achieved.

Example MARS request to compute differences of flux parameters

For example, to evaluate the precipitation accumulated for the 12 hour period from T+234 to T+246 and interpolate the resulting field to a 1.0 degree lat/lon grid, a method based on the following MARS request is recommended:

# Retrieve the total precipitation field from MARS on
# the N128 reduced Gaussian grid for both forecast steps
retrieve,
  class    = od,
  stream   = enfo,
  expver   = 1,
  date     = -1,
  time     = 00:00:00,
  levtype  = sfc,
  param    = 228.128,
  type     = pf,
  number   = 1,
  step     = 234/246,
  grid     = 128,
  gaussian = reduced,
  target   = "precip.grib"
 
# Read the total precipitation for time step T+234
# into fieldset p228_234
read,
  source   = "precip.grib",
  param    = 228.128,
  step     = 234,
  fieldset = p228_234

# Read the total precipitation for time step T+246
# into fieldset p228_246
read,
  source   = "precip.grib",
  param    = 228.128,
  step     = 246,
  fieldset = p228_246

# Compute the precipitation accumulated in the 12 hour
# period from T+234 to T+246 on the N128 reduced Gaussian
# grid
compute,
  fieldset = tot_precip,
  formula  = "p228_246-p228_234",
  accuracy = 16

# Write the N128 field to the file "precip.N128"  
write,
  target   = "precip.N128",
  fieldset = tot_precip  

# Interpolate from the N128 reduced Gaussian grid to
# a 1.0 degree lat/lon grid and write the final field
# to the file "precip.1.0"
read,
  fieldset = tot_precip_ll,
  source   = "precip.N128",
  grid     = 1.0/1.0,
  target   = "precip.1.0"