CMIP5 Model Analysis

Editor's Notes

• #3 Things that play a role: -ENSO (largest known forcing of interannual variability) -strength, structure, and variability of SALLJ Large-scale thermally direct circulationw ith a continental rising branch and an oceanic sinking branch, land-atmosphere interactions associated with elevated terrain and land surface conditions, surface low pressure and an upper level anticyclone, intense low-level inflow of moisture to the continent, and associated seasonal changes in precipitation Note the seasonal shift in convection across the South America Continent Prominent feature of the SAMS is the presence of a northwest-southeast oriented band of clouds and precipitation that originate in the Amazon and run toward the subtropical Atlantic
• #4 Drawbacks associated with GPCP (5 by 5 degree resolution) -satellite (1979-2005) -rain gauge station density is another, one gridded area could be dependent upon a few things, recording techniques, possible elimination of extreme values
• #5 Did this time span so we can fully see how the models pick up on the onset and demise of SAMS -want to know spatial pattern of the onset/demise, rather than just peak months. Much of interannual variability is due to discrepencies in the date of the onset/demise, thus if you don’t get this right, cant expect to get interannual or mean conditions correct Prominent feature of the SAMS is the presence of a northwest-southeast oriented band of clouds and precipitation that originate in the Amazon and run toward the subtropical Atlantic HadGEM does the best job of replicating the orientation of the band of convection while also matching appropriate intensity. Prevents build up of the max except along the Peruvian coast which all models have the same bias to do so. Ability to resits a preffered max MIROC5 and CCMS4 overestimate precipitation at preferred locations CCSM4 pocket in northeastern Brazil MIROC5 illustrates a max along central to southeastern Brazil With the demise, HadGEM2 illustrates too strong of a demise, not observations see a slow track retreat of this band, not a complete shutoff of max convection. With this model in the months of March and April, a band of convection cannot be noted across the area. The CCSM4 does not really show any change to it’s pocket of precip in the northeast portion of Brazil, shows strong persistence in this max, possible interaction between ITCZ and land, near the mouth of the Amazon MIROC5 actually provides a fairly well nice representation of the slow retreat of the system.
• #6 As from before, the HadGEM appears to do the best job of spatially representing precipitation across the greater amazon basin. All models appear to want to pick up on some form of a double max located near columbia, over amplication of this secondary eastward max. The CCSM4 does not pick up on the southward extend of the monsoon season and the MIROC5 has too southward of an extension (think though, if ccsm4 does not extend much southward, should we expect large varaiblity values that low) Had’s propensity to moderate things will hurt it later
• #7 A model is not based upon its ability to repoduce climatology, rather its measured based on its ability to reporoduce the variability within atmospheric states Absolute Average deviation is a statistical measure of disperision, finding the average anomalous precipitation -it has been noted areas of high AAD, have high interannaul varaib -does not tell us whether it is over or underestimating, just the ability of the model to measure varaiiblity -only problem with this is the selection of central tendency (Urugauy precipiation is associated with El Nino, also norhter SA and southern Brazil, northeast Argentina) -from this can note possible abilities of the model to replicate ENSO forcings -can note from this that in some areas, the models are heavily overestimating varaibility from the mean, or from year to year -once again the HadGEM appears to do the best job in keeping variability to a minimum over land, not the case of the oceans though where it produces double all sorts of ITCZ -both the MIROC and CCSM4 produce regions of large varaibility in association iwht their little pockets of maximum precip -not much variability int hat secondary maximum near Columbia -no models picked up on the zonal band of variability across the Amazon/Brazil
• #8 Another advantage to this topic is that we know that all the models tried to make a better visuallization of the ENSO, but was this sufficient, compare models to the next and see
• #9 Greatest discrepency is located near the Amazon Delta -show a rise in precipitation associated with the ‘summer month when observations is whoign a steady drop. Off on initial conditions and -to much dependence upon the summer/seasonal cycle, problem with dynamics in location to the equator, resolution at the equator NWSA, too much of reaction to the NH summer seaso -look at the other plots, the HadGEM appears to stay with GPCP most consistently of the bunch -consistently through the plots, it appears as though MIROC5 and CCSM4 are very sensitive to the hemispheric winter/summer whether it be dry or wet patterns.
• #10 Backing away from the equator, the CCSM4 and MIROC5 seem to have a better handle on the seasonal months Southern portions of the continetn, seemed ot do a much better job of replication, possibly due to distance away from the equator
• #11 40 to 70 20 to 0 which I took to be representative of the monsoon season Note: Wave Pattern
• #12 Data can show good correlations, however corresponding IMFS may show little in terms of synchronization. The reason isthat the underlying physical processes that dictate the large-scale interactions between atmosphere and ocean differ on various timescales. Extract physically meaningfual signals from the data While the HadGEM appeared to get the best spatial correlation, you can see here that the oscillations/forcings are off -both the HAD and MIROC show an inability to represent the annual cycles within the first IMF early on (is this solely because of our time scale?) Higher IMFs see lower frequency events -and when it comes to these lower frequency modes, the models are able to produce anytype of a cycle if not evne of the wrong sign. The CCSM4 is the only model that is able to reproduce these oscillations consistenly at both high and low end frequencies
• #13 Cannot qualittatively say that one model is better than the other, rather that certain are better in terms of one thing or the other All models have their strenghts/weaknesses