The Georgia Drought
A Sign of Climate Change?
Georgia made national news in 2007. The main stream media cited an unprecedented drought. The State of Georgia imposed Drought Level 4 ("extreme drought", the highest level) water use restrictions. And indeed, those driving through northeast Georgia could see reservoirs and lakes with an eerily low water level.
What is going on? Are we witnessing yet another sign of man-made global warming? Are we heading towards a new desert in the Southeast USA?
Without doubt, 2007 was a year with low rainfall. But wouldn't the reservoirs store enough water to bridge one, even two years with low rainfall? Let us look at a few figures. Georgia has a surface area of approximately 59,425 square miles (153,910 square kilometers) and an average annual rainfall of 52 inches (1321 mm). In theory, therefore, a total of 203.4x109 m3 of rain hit the surface of Georgia over one average year (109 = one billion). With a population of close to 10 million and an average per capita and per day consumption of 0.636 m3 (168 gallons, excluding agriculture) (ref), total water consumption by the population is 2.32x109 m3 per year. This is about 1.14 percent of the total rainfall. It is probably safe to say that Georgia has enough water.
What caused the water scarcity, then? I am avoiding the word "drought" because even those thirty-something inches of rainfall in 2007 are plenty of humidity. Here is a diagram that shows the annual rainfall for the last century (ref). The years 2006 and 2007 are highlighted in red, and the horizontal dashed red line highlights the 2007 rainfall level. The thick blue line represents the long-term trend, found by linear regression 1.
Yes, 2007 was a year with relatively low rainfall. But it is not unprecedented. The year with the lowest rainfall since 1900 is 1954 with 31 inches. This is 20% less than 2007. The years 1904 and 1932 had also lower rainfall than 2007. And the years 1921, 1925, 1927, 1955, 1999, and 2006 all were within 1 inch of the 2007 rainfall level. In fact, a two-year sequence of low rainfall - very similar to 2006-2007 - occurred in 1954-1955.
A trend? Statistically, average rainfall has increased by 2.5 mils (2.5 thousands of an inch or 1/16 mm) per year. This is an increase of rainfall of less than the thickness of an average human hair per year. Moreover, this increase is not statistically significant (P=0.899). From a statistical standpoint, therefore, rainfall in Georgia has not changed over the last 100 years.
Can we blame the 2007 "drought" on climate change? There is no evidence that any hypothesized climate change is related to the water scarcity in Georgia. Because there is no evidence that the amount of rainfall has changed.
Let us focus on a different development, that is, the population growth in Georgia. Not only is Atlanta the fastest growing metropolis in the world (ref), but Georgia population has literally exploded over the last 40 years as the following diagram shows (ref):
A nonlinear regression curve is fitted into the population data, an exponential growth function. Exponential growth is a natural unrestrained growth function, typically seen with bacteria. This regression tells us that the population of Georgia doubles every 35 years (R2=0.994). It is easy to see that water demand follows proportionally, i.e., doubles every 35 years, too.
The State of Georgia has done tremendous efforts to increase water supply by building new reservoirs, particularly in the last decade. The graph below shows the approximate development of new reservoir volume in million cubic meters per year (data kindly provided by GA DNR):
Since 1960, a total of 215 million cubic meters of reservoir volume was newly built in Georgia. In the same time, however, water demand caused by population growth increased by approximately 1400 million cubic meters, more than six times the added reservoir volume. Building enough reservoirs is challenging. Not only places this a high burden on the state budget, but many sites where reservoirs can be reasonably built are already used.
So what can be done? Clearly, first of all, water consumption can be reduced. Less washing cars, less lawn watering. Restrictions that are mandated by Drought Level 3 or 4 anyway.
Before we move along this path, let us first take a look at how it should not be done. An example is the University of Georgia in Athens and its Ad-Hoc Task Force on Water Resources. In the report of the ad-hoc task force, several suggestions for saving water can be found. Let us look at some of them:
Click on the image for a full-size version
So the poor guy who is trying to wash off the hydrochloric acid spill is supposed to consider water conservation while burning from the acid? These suggestions show not only a lack of appreciation for the situation, but also a blatant lack of common sense.
On the other hand, reducing common water-consuming tasks such as lawn irrigation and car washing does reduce water consumption. It is also said that the older parts of the Atlanta water distribution system have some leaks with a 20% savings potential. But even if we assume - generously - that water consumption can be reduced by one third, population growth will eat up all these savings within one to two decades. The problem of water scarcity has no easy solution.
Maybe a market approach can be tried? An approach where water gets more expensive when availability is low? A monthly "water socket" could still be provided at fixed prices, but the elevated prices for excess consumption could nicely encourage homeowners to collect roof runoff for irrigation, or install low-flush toilets. In any case, a market-based approach to stabilize water prices and water consumption needs to be carefully crafted. Allowing oligopolist or state suppliers to dictate prices increases the temptation to simply increase revenue through elevated prices. A key component of the market approach is competition that allows the consumer to choose the best water supplier. Furthermore, any reform of the water market should be done under the premise of lowest possible taxes since a tax increase would simply stifle the economy. Of course, a cynic could claim that with lower growth, less consumption and reduced migration into Georgia, this would certainly alleviate the water problem. But a weak economy is bad for all people, especially the poor. Therefore, increased taxes are bad. But why not allow private enterprises to build reservoirs and sell the water?
Unfortunately, the fact remains that any water savings will eventually be eaten up by population growth. But it gives Georgia a few more years to install new reservoirs.
There will be a solution, because Georgia has enough water.
And there is no relationship to any alleged climate change.
We can see that 2008 was a year with almost average rainfall (less than 5% below average). And 2009 counts to the three years with highest rainfall on record. In fact, only 1929 and 1964 had higher rainfall than 2009. Whereas 2006 and 2007 had a cumulative annual rainfall 21% below the average, the combined average 2006-2009 was only 3% below the long-term average. Statistically, these four years do not differ from the long-term average (P=0.68). Obviously, there is no unusual weather in Georgia.
Why the scare, then?
1 Linear regression is a method to obtain a trend. When a linear regression model is applied, the question is asked: "How much did the observed variable change per time interval?". In the example of rainfall, we asked: By what amount - in average - did the rainfall change per year? The underlying assumption is that the average change is the same every year. Is this a simplifying assumption? Yes, it is. However, it is a good starting point. If we discover a significant linear trend, we may continue the analysis and ask, is there any nonlinear component, such as an acceleration in recent years? If we fail to discover a linear trend, on the other hand, there is no need to further investigate more complex models, because "no change" remains "no change", no matter what model we use. To this end, linear regression does not only tell us the trend (i.e., the amount of rainfall change per year), but also (1) how good this model fits the data, and (2) the probability that the observed trend is just a fluke of the data. These values are known as the R2 and P-values, respectively.