MCD upgrades and developments

Progress Report24 April2012

Ehouarn Millour, Luca Montabone, François Forget, Arnaud Colaitis,

Aymeric Spiga

• MCDv5 will provide a set of dust scenarios (topped with EUV upper atmosphere solar max, ave or min conditions) designed to “bracket reality”.

• Just as with MCDv4, there will be 4 classes of dust scenarios (with EUV low/ave/high conditions):• The MY24-30 standard year dust scenario, using available dust opacity measurements from instruments (TES/MGS, THEMIS/ODY, MERs, MCS/MRO)• A cold (low dust) scenario• A warm (high dust) scenario• A global planet encircling dust storm scenario

• Moreover the MY24-30 cases (including actual EUV input) will also be made available (as “add-ons”).

MCD v5 dust scenarios

Design of MCD v5 Building realistic dust scenarios for MY24-30

• We need to combine observations from TES (MY24-27), THEMIS (MY26-30), MERs (MY26-30) and MRO (MY28-30).

• We perform gridding of non-uniform observations on a regular grid (5°x5° lon-lat, 1 sol) using weighted binning

• We obtain more or less incomplete maps of total dust opacity, depending on the density of observations

• We interpolate the maps using kriging, and we finally obtain complete, daily maps of dust opacity on a regular grid.

See ESA contract WP12.2.1 report (35 pages) for details

Design of MCD v5 dust scenarios

• Combining the outputs of runs from all 5 “non-global dust storm” years (MY 24, 26, 27, 29, 30), we will generate the standard year climatology.

• In addition, the variability encompassed in all these simulations will be included in the MCDv5 improved large-scale perturbation scheme.

Design of MCD v5 dust scenarios

• The cold scenario: Very low amount of airborne dust. Dust opacity at a given season and location is taken as the minimum over the 7 martian years MY24-MY30 dust scenarios, moreover decreased by 30%.

Design of MCD v5 dust scenarios

• The warm scenario: Very high amount of airborne dust (but not a planet encircling dust storm event). Dust opacity at given season and location is taken as the maximum over the 7 martian years (excluding the global dust storm periods during MY25 and MY28), moreover increased by 30%.

Design of MCD v5 dust scenarios

Illustrative example: Dust opacity at the Meridiani Site

MCD v5 dust scenarios MCD v5 dust scenarios MCD v5 dust scenarios MCD v5 dust scenarios • The dust storm scenario: An extreme case of fixed high opacity (tau=5) combined with “darker dust” properties (ie: lower single scattering albedo ω) to cope with uncertainties on dust radiative properties; but designing an optimal case is not that obvious after all (see Luca Montabone’s talk).

Upper right: Nominal ωLower right: 90% of nominal ω in the blue, where uncertainties are significant.. Lower left: Temperature difference between darker and nominal dust: cooling near the surface, warming around 60km.

Large Scale Perturbation Model

• Which variables?: profiles (over all 50 vertical levels) on a reduced lonxlat grid (16x12, i.e. 22.5°x15°) for temperature, horizontal winds and surface pressure, are saved (1/day) during GCM runs.

• How is the information stored?: Using lonxlatxalt EOF (Empirical Orthogonal Functions) decomposition

AVE: average value NORM: Normalization const. PC: Principal Component

N: Number of EOFs retained (16x12x(3x50+1)=28992 elements for each EOFi and 669 elements for each PCi), which are ordered (by design) from “most” to “least” significant. But how many EOFs should be retained?

A technical issue: How many EOFs should be retained to yield a good enough variability?

Cross-checking with RMS from the MCD gives the answer:

The New Large Scale Perturbation Model

• Example: Values of RMS of surface pressure at VL2 site from the MCD, compared to measurements

A technical issue: How many EOFs should be retained to yield a good enough variability?

Cross-checking with RMS from the MCD gives the answer:

The New Large Scale Perturbation Model

• Example: Values of RMS of surface pressure at VL2 site when using the large scale perturbation model including more or less EOFs. N=200 is adequate.

Large scale variability model From the PCs are computed, using a 30 day running mean,

“smoothed” Principal Components PCsmooth.

Using just PCsmooth enables reconstructing the mean signal.

• Example: reconstruction of surface pressure near VL2 site.

General Idea: Build PCnew from PC and PCsmooth to evaluate the perturbation but at another instant:

Where ‘othertime’ is randomly determined and within a 30 sol window of ‘time’. This way we conserve the spatio-temporal coherence of transient phenomena.

The perturbed series is then always simply rebuilt as:

Moreover, EOFs and PCs from all the computed Mars Years scenarios can be used as seeds to inject realistic variability in the system and thus represent the full interannual variability

The Large Scale Perturbation Model

Comparison between RMS from MCDv4 and RMS computed from large scale perturbation model

Example: Northern Hemisphere Winter; density at 10km

• Using EOFS of higher spatial resolution (than 16x12 in lonxlat) would be an improvement; the “cost” of which needs to be further assessed…

Improving the MCD access software

• The new Martian thermals mass-flux parametrization gives access to quantities of scientific and engineering interest.

• MCDv5 will thus provide additional outputs such as:• The inversion altitude (ie: height) of the Martian

Planetary Boundary Layer (PBL),• The friction velocity u*,• The maximum vertical velocity reached in the

Martian PBL,• The maximum vertical eddy heat flux in the

thermals,• The (turbulent) vertical velocity variance

profile.

Improving the MCD access software

• In MCD v4, output (and thus computation!) of all extra-variables (surface temperature, Cp, air viscosity, ...) is triggered by the “extvarkey” flag (1: yes / 0: no).

• This input argument will be replaced by an “extvarkeys” array (100 elements, like the “extvar” array): Each element “extvarkeys(i)” of the “extvarkeys” array must be set to 1 to request that the corresponding extvar(i) be computed (and extvarkeys(i) set to 0 means that extvar(i) will not be computed).

• This should reduce the computational time of the “call_mcd” routine if only a few extvar(i) are needed.