South American Climate of the LGM:A Regional Modeling Study

Kerry H. CookDepartment of Earth and Atmospheric SciencesCornell University

Thanks to Edward Vizy and Nancy Saltzman

Goal: Explain the Climate Dynamics behind LGM Aridity Patterns

Mean Temperature of the Coldest Month Mean Annual Temperature

…. and temperature reconstructions as well

Regional climate modelingon a large domain - MM5 modified for climateapplications in the tropics

20 - 60 km horizontalresolution; 23 vertical levels;1 min time step

Year-long integrations withclimatology used forinitial and boundaryconditions: u, v,T, q, and surface conditions

Regional model domain and topography; shading interval is every 500 meters.

Validation of the present day simulation: Precipitation

Jan

July

Observations Climate Model

August

September

October

DJF 850 hPa geopotential heights (m) andwinds (m/s) from the (a) 1949 –2002 NCAR/NCEP reanalysis, (b) MM5 present day simulationinterpolated to NCEP’s 2.5 2.5 grid, andpresent day PMIP simulations from (c) ECHAM3(fixed SSTs), (d) UGAMP (fixed SSTs), (e) UKMO (calculated SSTs), and (f) GFDL (calculated SSTs).

The regional model representsthe present day South Americanclimatology much moreaccurately than the GCMsimulations that have been usedto study paleoclimate. Using GCMlateral boundary conditionsdegrades the present day simulationsignificantly.

This presents a problem: We can’tuse GCM lateral boundaryconditions for our LGM simulations. What to do?

Simulation of the LGM Climate:

SSTs

Vegetation (land use)

Atmospheric CO2

Orbital parameters

Initial conditions

Lateral boundary conditions

Surface Boundary conditions: Vegetation

Present day: USGS LGM: Crowley (2000)

We ran simulationswith each of theseLGM SST anomalies,and compared theresults with the land-based proxy data.

2 of the SST distribu-tions essentially shutdown the monsoon.

“Line” and CLIMAPproduced similarresults.

We chose “line”.

CLIMAP 1981 “Core”, Schäfer-Neth and Paul (2003)

“Line”, Paul and Schäfer-Neth (2003) GCM, Shin et al.(2003)

Back to the problem of the lateral boundary conditions:

Using GCM conditions on the lateral boundaries seriously degrades the simulation of South American climate. (Also found by Seth and Rojas 2004, and more generally by Pielke et al 2005).

In setting the lateral boundary conditions for the LGM simulations, we arenot too concerned about eliminating remote effects on South American climatemore concerned about the consistency between the LGM SSTs and the atmosphere on the boundaries and over the “nudging” region.

Modify the present day lateral boundary conditions so they are dynamically consistent with the LGM surface boundary conditions (SSTs).

• Ran model with present day lateral boundary conditions• Used interior points to develop a method to modify the points on the boundary:

- adjust low-level temperature, retain lapse rate- test for geostrophy (reasonable except within a few degrees of

equator) - adjust surface wind based on interior points, and use the thermal wind relation to propagate the difference vertically - assume constant relative (not absolute) humidity

The resulting differences in the lateral boundary conditions were small everywhere except over the tropical Atlantic. But the modeled LGM solution with present day lateral boundary conditions captured this as well.

P – E Differences Precipitation Differences

LGM minus Present Day Simulations

Solid: Present day Dashed: LGM

Monthly Mean Precipitation (mm/day)

Region 1 Region 2

Region 3 Region 4

Simulated Winds and Specific Humidity at 850 hPa: October

Present day LGM

Vertical Profile of Moist Static Energy at 5ºS and 60ºW

March, present day “dry” –“wet”

MSE increasing with height => stability; low-level decreases inMSE stabilize the atmosphere against convection

Solid: MSEDash : sensibleDot/dash: latentDot: geopot

MSE = cpT+Lq+gz

Moist static energy

Sensible heat content

Latent heat content

Differences in MSE (solid),sensible (dashed) and latentheat (dotted) terms

P-E Difference

A region without increasedaridity during the LGM

Annual precipitation differences from the present day simulation

LGM SST forcing aloneLGM vegetation forcing alone (a deforestation experiment)

Precipitation in Region 5

Present dayFull LGM

Present day LGM veg alone LGM SST alone

A close up view of Region 5 in May

Surface elevation and 910 hPa winds

Precipitation differencesin May

… agrees with Roni’s result thathow deforestation occurs isrelevant to the sign of the precipitationresponse

surface temperature differences 870 hPa height and wind anoms

Difference between LGM-vegetation-only and precipitation simulations May

Vizy, E.K., and K.H. Cook, 2005: Evaluation of LGM SST reconstructionsthrough their influence on South Americanclimate. In press at J. Geophysical Research – Atmospheres.

Cook, K.H., and E. K. Vizy, 2005: South American climate during the LastGlacial Maximum: Delayed onset of the South American monsoon.Submitted to J. Geophysical Research – Atmospheres.

Current projects:

(1) Dynamical interactions between the high Andes and the restof South America. What is the paleo-record in the high Andestelling us about the Amazon and subtropical South America?

(2) Work with a PVM to explore consistency between prescribedvegetation forcing and the modeled climate, and translate climate intopotential vegetation.

That was an overview of recent results:

Landuse categories:6 urban/cropland7 grassland,8 shrubland10 savanna,13 evergreen broadleaf forest,14 evergreen needleleaf forest16 water,19 barren or sparsely vegetated20 tundra24 snow/ice

Present day simulation

LGM Simulation

60 km outer grid resolutionwith 20-km resolution nested

Topography

Precipitation is not Validating Well in theHigh Andes Nested Domain in the

Present Day Simulation

We are not capturing thewetting signal in thehigh Andes in the LGMsimulation

20-km resolution is not fineenough for this simulationdesign ….

Another project underway …. using a PVM to translate theclimate produced by the regional climate model into vegetationcategories

(1) To provide a more direct comparison with some of the geologicalproxy data

(2) To better understand the implications of the climate differencessimulated by the model

(3) To “free” ourselves from uncertainties in specifying the vegetationdistribution in simulating LGM climate (iteration)

Potential Vegetation Model(Oyama and Nobre 2004)

Left: initial vegetationRight: vegetation after one iteration

Top: “Core” LGM SSTsBottom: “Line” LGM SSTs

Top: One iteration fromCrowley initial vegetation,LGM SSTs

Bottom: One iteration fromPresent day initial vegetation,LGM SSTs

Different initial conditions onvegetation might cause differ-ent solutions, as in Oyamaand Nobre 2003. Esp. notethe eastern Amazon.

.

Even if we use present day SSTs and this initial vegetation, the eastern Amazon does not become forested.

Cutoffs for rainforest:

TC 15.5C h 0.8 s 0.81

s = seasonality indexh = wetness index

Conclusions

A somewhat different modeling approach to studying paleoclimate,using regional models and our knowledge of how the present dayclimate works to evaluate the quality of the paleoclimate simulation.(We are working on LGM South America and the AHP.)

The approach has plusses and minus – as do other approaches tounderstanding paleoclimate such as GCM modeling and theanalysis of various proxies.

Compared with GCMs:+ Better simulation of South American climate, finer resolution,

able to resolve interactions across space scales (relatively large domain with a relatively fine resolution).

- global teleconnections are not considered

Compared with proxy data:+ constrained by physics (Navier-Stokes eqns), produces fields

that are internally consistent- model dependent results

We need all of these approaches.

We find that –• There is large-scale drying in the Amazon basin during the LGM,delivered in the form of annual precipitation reductions on the orderof 30%.• In the Southern Hemisphere, this drying is due primarily to ashortening of the rainy season, and a lengthening of the dry season.• The shortening of the dry season is caused by a delay in the onsetof the monsoon. The monsoon starts later because the tropicalAtlantic is cool, so the buildup of moisture/MSE is delayed.

Cooling of more than about 2K in the tropical Atlantic shutsdowns the monsoon completely in this model.

In the Northern Hemisphere, in which summer precipitation is moreITCZ-like in its circulation, drying is also due to the fact that thelow-level convergence is dryer – again associated with cool tropicalAtlantic.

We find that the first-order forcing is from the SSTs, but there areinteresting and important regional responses related to vegetationforcing. For example, the region of increased precipitation along theEquator when the large-scale circulation anomaly interacts with aRegional orographic feature.