GFDLDataPortal LAS scenario - Hudson Bay #1

[karlmsmith]

Reported by @kevin-obrien on 13 Aug 2007 23:19 UTC From Keith's email

--------
On page 272 of G&G it is written:

   "We obtained the mean monthly temperature and total
   monthly precipitation data for the 1961-1990, the
   2040-2069, and the 2070-2099 time-slices, roughly
   corresponding to the 1 x CO2, 2 x CO2, and 3 x CO2
   climates, respectively."

...and on the previous page they noted...

   "From 1900 to 1990, the models were run using historical
   increases in greenhouse gases and sulphate aerosols,
   while the simulations beyond 1990 were obtained using
   the IS92a transient forcing scenario of the Intergovernmental
   Panel on Climate Change (IPCC)."

Regarding the 1961 to 1990 time periods, for CM2.0 we offer external users not one, but 3 climate of the 20th century experiments:
   # GFDL CM2.0 20C3M-0 Run 1 climate of the 20th Century
   # GFDL CM2.0 20C3M-0 Run 2 climate of the 20th Century
   # GFDL CM2.0 20C3M-0 Run 3 climate of the 20th Century
And for CM2.1 we offer 5:
   # GFDL CM2.1 20C3M-1 Run 1 climate of the 20th Century
   # GFDL CM2.1 20C3M-1 Run 2 climate of the 20th Century
   # GFDL CM2.1 20C3M-1 Run 3 climate of the 20th Century
   # GFDL CM2.1 20C3M-1 Run 4 climate of the 20th Century
   # GFDL CM2.1 20C3M-1 Run 5 climate of the 20th Century
The user may eventually want to look at all of these.

Regarding the two 21st century periods (2040-2069, and the 2070-2099) the climate modeling community no longer uses the IS92a scenario, so we will substitute the SRES A2 scenario. For CM2.0 that is
   # GFDL CM2.0 TRESA2-0 Run 1 SRES A2
and for CM2.1 it is
   # GFDL CM2.1 SRESA2-1 Run 1 SRES A2

Next the user must determine the latitude x longitude domain to use for the Hudson Bay analyses.  Look at Fig. 2 of G&G for an idea as to the region of interest.  At the bottom of page 273 of G&G it is written...
   "In the CGCM1 model, we defined the Hudson Bay region
   by the 40 grid points located between 53.81?N to 68.65?N
   and 71.25?W to 97.5?W"
...so the user will need to select a similar range for the GFDL CM2
analyses. As noted by G&G...
   "The co-ordinates of the Hudson Bay region slightly differ from model
to model..."
The extent of these differences is documented in their Table 1.

Having determined the experiments and lat x long region of interest, the user is ready to proceed to make some maps (to see if things "look right") and to do some simple calculations to provide a small set of numbers to compare with G&G Table V (top of page 283).
   TABLE V: "Projected changes in temperature (?C) in the
   Hudson Bay region for the 2040-2069 and 2070-2099 time-
   slices from six coupled ocean-atmosphere GCMs"
The rightmost column is "Annual average", so the user looks to do that first.  The variable of interest is "Surface Air Temperature (tas; air_temperature)" under "atmos  > annual".

Let's focus on CM2.1 first.

*1* The user will want to compute "tas" time averages over Hudson Bay for both 2040-2069 and 2070-2099 of the CM2.1 SRES A2 experiment and map them as XY maps to see that they make sense.

*2* The user will want compute regional spatial averages of the time averages from 1, resulting in single numbers that are presumed to be the 30-year average surface air temperature for the Hudson Bay region.

*3* The user will want to compute "tas" time averages over Hudson Bay
for 1960-1099 of the CM2.1 H2 experiment and make an XY map.

*4* The user will want compute a spatial average of the 1960-1999 time
averages from 4.

*5* To match the rightmost column of Table V, the user will need to subtract the 1960-1990 time average from each of the two 21st century periods.  This can be done by hand, since there are only a few numbers, given the spatial and temporal averaging done in 2 and 4.

*6* The user also wants to see a map of the differences, not just the
single regional averages.  Is the temperature change uniform, or is
there a dominant spatial pattern?  Does the north warm faster than the south? Does the air over the land warm faster than the air over the bay itself? To do this the user will need to subtract the mapped XY values from step 3 from the maps produced in step 1.

*7* The user will want to repeat steps 1 through 6 using the CM2.0 SRES A2 experiment and the CM2.0 h1 experiment instead of CM2.1.

*8* The above steps all focused on the annual average.  However, the other 4 columns of G&G Table V are labeled...
Winter Spring Summer Autumn
...implying that a series of 3-month averages need to be computed over the three different 30 year periods.  The user fails to see how this can be done in LAS.   ;-(

PS - I don't think G&G explicitly state which months comprise the different seasons, so we will assume they used the most common Northern Hemisphere atmospheric definitions of
  Winter = Dec Jan Feb (DJF)
  Spring = Mar Apr May (MAM)
  Summer = Jun Jul Aug (JJA)
  Autumn = Sep Oct Nov (SON)
Since winter spans two calendar years, we'd assume that December of 1959 is used in computing the 1960 DJF average. So, the monthly time series from Dec 1959 through Nov 1999 would be used in computing the seasonal averages for the late 20th century. 

Migrated-From: http://dunkel.pmel.noaa.gov/trac/las/ticket/170

[karlmsmith]

Modified by @kevin-obrien on 13 Aug 2007 23:19 UTC

[karlmsmith]

Comment by @kevin-obrien on 14 Aug 2007 20:47 UTC

Running through these tests:

step 1 - no problem

step 2 - I believe can't be done because of [http://porter.pmel.noaa.gov/trac/las/ticket/172 the error in averaging in XY an already t-averaged variable]

step 3 - as in step 1, no problem

step 4 - same as step 2

step 5 - can't complete because step 2,4 cannot be done

step 6 - can't be done because[http://porter.pmel.noaa.gov/trac/las/ticket/124 user-defined variables cannot be compared ]

step 7 - didn't see any point in trying this on all the other experiments

step 8 - need to add averaging selection tools to the UI constrain page

[karlmsmith]

Modified by @kevin-obrien on 7 Feb 2008 21:36 UTC

[karlmsmith]

Modified by @kevin-obrien on 8 Feb 2008 18:09 UTC

[karlmsmith]

Modified by @kevin-obrien on 15 Feb 2008 22:23 UTC