Saturday, October 11, 2014

The Driest Day in U.S. History

** Update: Here is a posting about the wettest day(s) in U.S. history **

Has there ever been a day when no precipitation was observed in the Contiguous U.S.? It would be pretty remarkable if no precipitation fell over three million square miles. Let's take a look. Before we answer this question, we need to set some parameters as to what exactly a dry day is. So, here are the rules:


1) There is a fairly robust data set to use at the National Climate Data Center based on single calendar days. A very, very important caveat to this is the observation time for precipitation. Precipitation at Cooperative stations is usually observed between 7:00 a.m. and 8:00 a.m. This causes problems when precipitation falls between the observer's bedtime and the observation time the next morning. Which day does the precipitation get assigned to? Therefore, we will not use Cooperative stations in this analysis. Unlike Cooperative stations, WBAN stations always report data from midnight to midnight local standard time. In addition, they are almost always located at manned airport towers, so providing reliable data is especially important. Therefore, we will restrict the analysis to WBAN stations. Currently, there are about 1,200 WBAN stations across the U.S. (e.g., Chicago O'Hare Airport, Los Angeles Intl. Airport, etc.).

2) We are restricting the analysis to the Contiguous (Lower) 48 states. Stations in Alaska and Hawaii are often too sparse to capture significant patterns prior to 1950. In addition, there are locations in each state that receive precipitation nearly every day of the year. Therefore, Alaska and Hawaii are excluded from the analysis – although they are shown on the maps. (Note: None of the 139 stations in Alaska reported measurable precipitation on March 13, 1997 – the only time on record.).

3) A minimum number of stations must report observations on a day to be considered. If, for example, only 3 stations were used, many days per year would report no precipitation. So what should the minimum number be? We have arbitrarily decided that 100 stations is a minimum threshold for consideration. This ensures a wide enough network of stations to capture synoptic-scale weather event across the Contiguous U.S. The 100 station threshold means that every day since 1920 can be assessed (and some days back to 1915).

Let's run the numbers:

First, a word about data sources. My primary data source is the GHCN v. 3 database maintained by the National Climate Data Center. It is not a perfect database in that much station data are, unfortunately, not included. For example, there are approximately 40% more stations in the NOAA Regional Climate Center's xmACIS online application that report precipitation totals than in the GHCN database.

To the best of my knowledge, there is not a mechanism for programmatically accessing xmACIS so I am left with a systematic evaluation of GHCN data and then supplementing it with xmACIS data on a case-by-case basis.

After looking at all 18 million observations in the GHCN database and sorting them by date, we identified 10 calendar days when the combined total of all precipitation in the Contiguous U.S. was 0.10" or less. Table 1 shows those dates.

Table 1. List of lowest precipitation totals in the GHCN database and the number of stations reporting precipitation totals. Trace values are counted as 0.00" for mathematical purposes.

The "winner" is February 9, 1991. A combined 0.02" was reported for the entire Contiguous 48 states! Not only was it the winner, but the other stations that were in 2nd through 10th places all had fewer than 165 observing stations. This vastly increases the confidence that February 9, 1991, was indeed the driest day in modern U.S. history.

As stated earlier, the xmACIS application has a larger number of stations than the GHCN database that report precipitation so we manually looked at the xmACIS data on a state-by-state basis. A total of 714 WBAN stations in xmACIS reported precipitation on February 9, 1991. The grand total of all 714 stations in xmACIS was also 0.02". So there you have it.


It turns out that 2 stations each reported 0.01" of precipitation on February 9, 1991, in the Contiguous U.S. Those stations were Lemoore Reeves NAS, California, and Russell Municipal Airport, Kansas. An additional 41 stations reported a Trace of precipitation – but those are mathematically treated as 0.00" for summation purposes. Figure 1 shows the location of all stations that reported a precipitation total (including 0.00"). Alaska and Hawaii are shown on the map but are not included in the analysis.
Figure 1. WBAN stations reporting precipitation in xmACIS on February 9, 1991 (n=714).

As we stated in the Rules section, Cooperative stations were not included. Figure 2 demonstrates why. A small number of stations along the southeastern coast reported up to 0.50" of precipitation on February 9, 1991. However, a closer inspection of the Cooperative observer forms and satellite imagery shows that precipitation fell before midnight on the 8th in several places but was not observed and recorded until the next day

Figure 2. All stations reporting precipitation in xmACIS on February 9, 1991 (n=8,055).

The satellite image in Figure 3 is from the GOES 7 infrared instrument and was taken at midnight Eastern Standard Time (EST) on February 8/9, 1991. A complex of storms is seen leaving the North and South Carolina coasts. These storms produced up to 0.50" of rain just before midnight but were recorded in the following day's precipitation total. An image from the same satellite at 3 p.m. EST is shown in Figure 4.

In contrast to the southeastern coast, the precipitation dots in Figure 2 in the Pacific Northwest appear to be valid based on the perspective shown in Figure 4. Still, the decision to only use WBAN stations appears reasonable given the uncertainty of Cooperative precipitation totals due to the time of observation issue. It is safe to assume that the time of observation issue does not vary across the length of the historical record; so therefore, any one date is just as likely to have rainfall amounts assigned to an adjacent day, of just plain missed, proportionate to the amount actually captured by the WBAN stations.

Figure 3. GOES 7 image of the Contiguous U.S. on February 9, 1991 at 6:01 UTC.

Figure 4. GOES 7 image of the Contiguous U.S. on February 9, 1991 at 21:01 UTC.


I must admit to a slight disappointment to learn that there were no completely dry days to be found in the climate record. Oh well. That being said, is this methodology perfect? Not really. But it probably came to the correct conclusion. February 9th, 1991 was the driest day in modern history for the Contiguous U.S.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.