Introduction
The Western hemisphere exhibits a well-defined intertropical convergence zone (ITCZ) of deep convective rainfall and cloudiness, especially in northern summer. However, this structure, and its associated climate variability, is sub­stantially distorted in the American longitudes by land-atmosphere interactions. These interactions occur across the many scales of landscape structure, from the sharp ranges of the Andes to the breadth of Amazonia. However, they share the primary pulse of land-atmosphere interactions everywhere, especially at low latitudes: the diurnal cycle.

Project Goals and Methods:
To investigate the physical processes responsible for landscape-influenced rainfall structure in the tropical Americas, and the techniques necessary to represent them properly in climate models, we turned to a nested-grid regional (mesoscale) land-atmosphere model, the Penn State/ NCAR MM5. Our domain structure cen­ters on western Colombia, the rainiest locality on Earth (e.g. Poveda and Mesa 2000), and includes the whole of northwest South America at 72km, two-way nested with smaller subdo­mains resolved at 24km and 8km (Fig. 1). Finally, a 2km grid is used with one-way nesting to resolve the landscape structure and convection in just the region of Pacific Colombia.

Figure 2 shows satellite observations of cloud clusters (contiguous regions of cold cloud tops in infrared imagery) for the region of our 24km resolution domain. The size of the plotted ovals indicates the size of an observed cloud cluster, while color indicates the hour at which each cloud cluster was observed (orange-red = afternoon, blue-green = late night / early morning). The con­vective clouds over land tend to exhibit an afternoon peak, with some interesting exceptions, while the offshore clusters tend to occur in the early morning. September shows the most widespread convection over both land and ocean in Fig. 2, so our further inquiries center on August-September, specifically of 1998, a period fairly typical of longer-term climatology, but with especially diverse synoptic activity.

Modeled vs. observed Amazon diurnal cycle:
At the largest scale, the simulated diurnal cycle over central America is in broad agreement with observations (Fig. 3), with afternoon convection over land crossing the coast to become morning rains offshore, especially in the concavity of western Colombia and the Gulf of Panama. An especially striking aspect of Fig. 3 is the 3 westward-moving north-south bands of rainfall, which come through clearly in this composite diurnal cycle. These near the eastern boundary of the model and take 3 days to cross south America. They have a rather artificial look to them, but there may be at least a hint of a similar phenomenon in observations (Fig. 4).

One obvious difference between Figs. 3 and 4 is that the average intensity of South American convection has a larger diurnal cycle in the satellite data than in the model rainfall. This suggests that the model convection over the Amazon may be too much under the dynamic control of traveling wind disturbances, and not sufficiently modulated by night-day differences in surface heating and surface fluxes. Whether this reflects some weakness of the land surface model, or simply the atmosphere model, is not yet clear. 

Finest-scale results in Pacific Colombia:
Figure 5 shows results from the 2km ‘cloud-resolving’ domain over western Colombia and the adjacent ocean. Results have been averaged over a range of latitudes over which the coastline and are approximately north-south and uniform (top panel). The solar heating pulse over land (second panel) drives a sea breeze, which penetrates into the Atrato valley, between the low coastal range and the Andes (third panel), where it triggers late-afternoon rainfall (bottom panel). In the evening, this rainfall propagates back toward the coast, along with the retreat of the sea breeze, as observed in nature. A secondary offshore rain event of ~300 km scale erupts rather suddenly after midnight. This appears not to be a simple case of the land-initiated convec­tion merely intensifying after crossing the coast. Rather, a distinctive mesoscale dynamics of the offshore region appears to be at work. This phenomenon is being investigated further.

Ongoing and future experiments:
We are diagnosing these model simulations at the various resolutions of the nested grids, and running perturbation experiments, in order to clarify the physical processes and model strengths and weaknesses. For example, in one experiment we halved the soil moisture, reflecting an uncertainty in soil moisture initialization. The result was decreased rainfall over Brazil, but increased rainfall over Colombia. Apparently in Colombia, drier land drives a stronger diurnal circulation, drawing in more moisture.

Results are being written in a coordinated series of 3 manuscripts, on the satellite climatology, control run and validation, and physical process experiments, to be submitted in 2001.

Contacts for Principal Investigators:

Brian E. Mapes
bem@cdc.noaa.gov

Return to CLIVAR PACS Home Page