North American Precipitation

From Climate Info
Revision as of 15:40, 15 January 2010 by Rood (Talk | contribs) (Created page with '__TOC__ ==Introduction== This is a project to investigate precipitation regimes in North America and Central America. This work is motivated by persistent problems in represen…')

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search


This is a project to investigate precipitation regimes in North America and Central America. This work is motivated by persistent problems in representing continental precipitation, and the critical need to represent continental precipitation and how it will change in a warming climate.

The original motivation for this work is as follows. Assume that precipitation in the U.S. can be categorized by regimes that are related to the mechanisms that form the precipitation. Examples are:

  • Wintertime western U.S. precipitation and its relation to synoptic-scale waves propagating from the Pacific and interacting with topography in the western half of the country.
  • Summertime precipitation in the eastern U.S., which relies on the Great Plains low-level jet as a moisture source. This moisture is organized in mesoscale clusters which have nighttime precipitation maxima.
  • Land-sea contrast, sea breeze, precipitation on the U.S. Southeast and Gulf Coasts.

There are other regimes as well.

Consider the summertime regime related to the low level jet. The water source for this jet is the Gulf of Mexico. The jet is trapped to the west by topography. The jet strength and location is correlated with the Bermuda High. It is reasonable to pose that for a climate model to provide robust information on precipitation in the eastern U.S., the low level jet needs to be represented. It is the major source of moisture for the interior of the continent.

Figure 1 is a motivating factor, which brings concreteness to the previous paragraph.
Figure 1: Wavelet analysis of moisture flux near San Antonio, TX, from GEOS-1 Re-analysis.
(from Schubert et al., 1998). This figure shows the time series of the moisture flux, and the frequency of the modes of the moisture flux. There is a distinct seasonal transition from spring to summer, which is characterized by the flux being in, strongly, synoptic-scale periods, to flux being, predominately, in diurnal scales. This links a climate parameter, seasonal continental moisture availability, to the representation of daily dynamical variability confined to the boundary layer.

From this perspective, climate change questions are, then, directly related to representation of topography and the quasi-stationary surface highs and lows. The representation of coherent boundary layer dynamics is important. Climate change questions are then focused on how the characteristics of these mechanisms change in the future. Topography does not change; hence, the behavior of the Bermuda High is brought into focus. There are questions of circulation strength, position, and variability. If the Bermuda High was more persistent in its location, this will lead to regions of persistent wetness and dryness in the U.S. and Canada.

Similar questions at the mechanistic level can be posed for all of the precipitation regimes.

The attention to mechanisms brings the focus of climate change to local rather than global processes. One place this impacts climate models is in the resolved dynamics. Previous research within this group has investigated the impact of the finite volume dynamical core and the spectral dynamical core in the Community Climate System Model. These two dynamical cores bracket the extremes of local and global representation of atmospheric fluid dynamics. The spectral method is well known to leave residual Gibbs oscillations in, for example, cloud fields near steep topography. There is "spectral rain" in the rain shadows of mountains. These phenomena persist in resolutions that are feasible for climate simulations.

The results from these initial investigations are here.

Bala et al.: Basic climate of CCSM: FV vs spectral (to appear J. Climate)

Bala et al.: Local Impacts in CCSM: FV vs spectral (submitted GRL)

Research Projects

North American / Mexican Monsoon

The North American / Mexican Monsoon is an important precipitation regime in the U.S. Southwest and Western Mexico. It is categorized by sharp spatial gradients. The ability to model the North American Monsoon (NAM) varies widely in models. One summary is given in the atlas of the North American monsoon Model Assessment Project. Figure 2 in this atlas List of Figures in Atlas shows significant differences in the spatial distributions form regional and global models. There are also differences between spectral and grid point models, and substantial over estimation of rain fall intensities.

The initial look at the model simulations from the finite volume (FV) AMIP simulations performed in 2005 at LLNL are encouraging.
Figure 2: 20 Year Average of Precipitation from "old" FV AMIP simulations
Both the geographical distribution and intensity are simulated with quality comparable to or better than the ones reported in atlas referenced above. North American monsoon Model Assessment Project Furthermore, as resolution is increased the simulation improves. Most of the models have enormous overestimates of the maximum rainfall rates; these simulations do not.

This initial performance motivates more research. The North American monsoon Model Assessment Project Atlas provides a list of diagnostics that should be viewed as standards. These papers by Douglas et al. and Higgins et al. provide analyses of the North American monsoon that define a standard.

Of special interest in the analysis to be pursued here is the moisture flux at and near the surface. From the preliminary graphics, there are major differences between the simulations at 1 degree and 1/2 degree resolution. There are also discernible differences between the two simulations at 1 degree resolution with spectrally smoothed and remapped topography. (see also Bala et al.: Local Impacts in CCSM: FV vs spectral (submitted GRL)) The differences are most notable in northern end of the Gulf of California and in the paths the moisture takes from the Gulf of Mexico. There is a long standing discussion of the observations of the source of moisture for the North American monsoon, and at least in the case of these simulations, the changes in moisture flux due to resolution should be identifiable. That is the first step.

Here is a collection of figures of moisture flux at 850 mb. Old figures from LLNL analysis The nomenclature for the figures: c resolution is 1 degree, d resolution 1/2 degree, and s is spectrally smoothed topography and r is remap topography. There are three things that strike me as interesting in these figures.

  1. The precipitation in NW Mexico, SW Arizona. The 1 degree run has surface moisture flux in the northern part of the Gulf of California, but no precipitation. The 1/2 degree run has moisture flux and precipitation in northwest Mexico, suggesting the possibility of the existence of local dynamics that brings the moisture inland, then rain.
  2. The moisture flux on the eastern side of Mexico changes substantially with the increase from 1 degree to 1/2 degree resolution. What is striking is the intrusion of moisture into the interior in preferential regions which are assumed to be strongly related to topography. Also notable is the large difference in flux on the east and west side of Mexico.
  3. The changes between the 1 degree and 1/2 degree resolution in the sea-breeze rain on the eastern Mexican coast and along the U.S. Gulf and Southeast also call for attention.

Here is a directory of figures from more new runs of fvCAM and some comparisons with NARR data. These runs are made with the current research version of CAM and includes substantially different physics than the "old runs" linked above. The physics may be altered before the next official release of CAM and the results, interpretation, and reporting that we do needs to be as a trusted collaborator in concert with collaborators at NCAR, esp. Phil Rasch.

The results from these runs are largely consistent with the "old runs." We have not looked at the diurnal variation of precipitation, which is one place that we might expect differences. Here are a couple of figures that highlight some things to think about.

In Figure 3 is the precipitation from the NARR re-analysis and three resolutions of CAM. These resolutions are, approximately, 2 degree, 1 degree, 0.5 degree; the FV grid spacing of CAM was altered to support computational optimization.
Figure 3: 20 Year Average of August Precipitation and Wind from "new" FV AMIP simulations (NARR is only 10 years.)
Consistent with the "old" runs the 0.5 degree simulation has precipitation in Northwest Mexico and in Arizona. Of special note is the development of a southerly component of the wind in the northern portion of the Gulf of California. There is, with this, consistency with the re-analysis; there is fascinating flow into the Southwest and across the California mountains.

On the eastern side of Mexico and in Texas, the winds are once again fascinating and their change with resolution is promising. The magnitude of the wind is, visually, biased high. In eastern New Mexico, there is a more westerly component of the winds in the simulation. (I think this version of the model may be running the new Bretherton PBL?) The precipitation on the eastern side of the mountains is not as well represented. A glaring weakness is the eastern coastal precipitation in southern Mexico. Map of topography shows that this precipitation is located in what appears to be a coastal plain between the mountains and the sea. Hence, it does not seem to be topographical. (Also note from Bala, the possibility of persistent downward motion in the model. This vertical velocity field, omega, needs some thought!)

In order to get some grasp on the moisture transport the dew point is shown in Figure 4.
Figure 4: 20 Year Average of Dewpoint and Wind from "new" FV AMIP simulations (NARR is only 10 years.)
Again, in northwestern Mexico and Arizona, the figure is intriguing. The turn of the wind to have a strong southerly component leads to transport of water into the continent. This appears to be qualitatively correct. The intrusion of moisture into the continent is not deep enough. The improvement of the continental moisture in the U.S. Southwest with resolution is substantive.

On the eastern side of the Rockies and the Sierras and in the Mississippi River the simulation of the dew point is more interesting in its shortcomings. In the spirit of there being a dry line, representative of the contrast between the Gulf moisture and the western Great Plains dryness, the behavior of the model in New Mexico and eastern Colorado as resolution increases is enticing. The underestimation of dew point in eastern Oklahoma and the Mississippi Valley is curious. The winds are strong from the south, and would be expected to be bringing more than adequate moisture to the continent. (Look at other variables? Here is temperature; it's hot in Texas and Oklahoma.)The low dew points are related to high continental temperatures, which are high biased in both the old and the new physics run.

Here is dewpoint from only one year from a degree run. The plotting of the wind vectors has failed due to software (we think!). Compared with the NARR, there are a number of interesting features developing. (Note 50 and 25 km bracket the NARR.) The wetness of the Gulf of California is simulated, suggesting that at this resolution, this moisture source is represented. Looking at the Baja Peninsula the topography is starting to worm down the continent. It is very interesting what is happening internal to eastern Mexico. Localized biases are developing, in stark contrast to well represented moisture over the oceans.These biases are increasing with resolution. (Obviously, they require more analysis.)