In her report on the drop in Lake Mead’s water level (Physics Today, April 2008, page 16), Barbara Goss Levi concludes that when the impact of greenhouse gases and aerosols is included in global climate models, the drop in water level seems to be related to greenhouse gases and not to natural variability.

A NASA article about Lake Mead (see suggests that the current drop in water level is associated with a drought that began in that area of the Southwest in 1999. Furthermore, the NASA article points out that a more severe drought occurred in the same area over a five-year period in the 1950s.

I agree with Levi that local climate variations can contribute to repeated droughts in an area; that connection is substantiated by tree-ring data in the Southwest from back before human impact was even a factor. Since we cannot correlate those early droughts with increases in anthropogenic greenhouse gases, we must be careful with assertions that everything can be deduced from global climate-model simulations, especially when the simulations are downscaled to finer grids. While you can certainly make the block averages of properties equal in fine and course grids, you cannot reproduce the initial spatial correlations that would exist when starting with a fine-grid model based on actual data. Furthermore, from my experience in flow simulation models, I know that any forecasts derived from those models depend on grid cell size and orientation.

We also need to further understand the causes of repeated events, like droughts, in the climate of an area before we make the jump to a global climate model; the correlation lengths of local and global climate phenomena are not equivalent. We must be careful with our assertions of global climate change and be aware of the limitations in our models, especially in the absence of actual data.