wiki:InterpolEmiss

Interpolation over time of input data

The CMIP6 input that forces ground emissions of SO2, BC, POM, etc. is defined monthly for each cell grid of the globe. These monthly CMIP6 data shall be considered as corresponding to the 15th of each month. In order to avoid a "step" progression of the CMIP forcing, INCA operates a linear interpolation of the input data over time: this eventually yields a continuous, linear by pieces, evolution of the forcing timeseries.

Correction of CMIP6 aerosol emission inputs in order to recover monthly averages

Initially, a choice was possible between "no interpolation" (emi_interp_time flag set to 0 in inca.def), or else a simple point-to-point interpolation, linearly computed from each 15th to the next 15th.

A supplementary condition that was wished for was to consider the CMIP6 input as a monthly average, and to aim for a conservation of these monthly averages when linearly interpolating. If a simple linear interpolation of the CMIP6 input data is carried out (point-to-point between mid-month values), calculating monthly averages over [1st ; 30th] intervals does not yield the original prescribed CMIP6 value.

To solve this issue, we use a correction algorithm of the input data, adapted from [Sheng & Zwiers, 1998]. Through a matrix calculation, an adjustment of mid-month values is operated, in order to yield correct monthly averages, i.e.: the interpolated time series, once averaged over monthly intervals, shall give again the initial CMIP6 value.

The [Sheng & Zwiers, 1998] article details a rather simple algorithm to do so.

However, a point in particular involved a refinement of the method: for emissions of feeble amplitude, it is possible that the algorithm corrects the data (close to naught) downwards, and into a negative value.

To improve the method, and avoid possible inputs of emissions with negative values into the model, some data will be "blocked", following a method inspired by [Taylor & al., 2000].

The process goes as follows:

1/ any monthly value that would get corrected into a negative value by the S&Z algorithm shall point to a "blocked month": the initial CMIP6 value will then be kept, and affected to the whole month duration as a constant function of time;

2/ the time interpolation of the blocked months and their neighbours will differ from that of the other months: since the emission value is fixed and constant throughout a whole month, the linear interpolation will occur from the preceding 15th until the 1st of the blocked month, and from the 30th of the blocked month until the 15th of the following month.

The choice of months to be blocked, along with the computation of corrections on each month, are carried out in the script regrid_vN.sh (v6 on 12 June 2017).

This regrid script, starting from CMIP6 input data, produces a netcdf file sflx.nc containing corrected monthly emission values; the mksflx.F90 routine then deals with the time interpolation in itself. The output timeseries, once monthly averaged, shall yield the initial CMIP6 values.

Important note: if this interpolation mode is chosen (with correction of values in order to yield "correct" monthly averages, and hence blockage of some months), it is compulsory to set the emi_interp_time flag to 2 in inca.def. The boolean indication "blocked month" is not passed on to INCA with a flag, but is rather coded under the form of the sign of the data in sflx.nc: any negative value means "month blocked on that point", and any positive value means "classical interpolation". The mksflx.F90 routine was coded to take this sign information into consideration, apply the right interpolation scheme in the right points (and of course take the absolute value of any emission coded with a negative sign). It is thus important to keep the consistency between an sflx.nc file generated for a Sheng & Zwiers interpolation and the emi_interp_time flag: values 0 or 1 of this flag are not designed to deal with an sflx.nc file that would contain negative values.

Another improvement in the forcing of species was the altitude emission of part of the emissions: anthropogenic parts of emissions are kept as emitted from the ground, whereas the biomass burning parts are distributed over several levels above ground. This was implemented for SO2, POM, BC, NO2 and NH3 (under dvpt - June 2017).

References:

  • Sheng, J., & Zwiers, F. (1998). An improved scheme for time-dependent boundary conditions in atmospheric general circulation models. Climate Dynamics, 14(7), 609-613.
  • Taylor, K. E., Williamson, D. & Zwiers, F. (2000). The sea surface temperature and sea-ice concentration boundary conditions for AMIP-II simulations. PCMDI Report No. 60.
Last modified 7 years ago Last modified on 06/12/17 09:04:48