Leap Year Weather Records

[eurojosh]

I know this isn't a leap year, but I wondered if anyone knew how meteorological records reflect Feb. 29th on leap years. The existence of a day that only occurs once every four years has got to affect mean daily highs and lows, due to smaller sample set, as well as weather records. And I guess that Februaries that are 29 days long are, on average, a tiny amount warmer than those which are just 28.

What got me wondering about this is the little chart on the bottom of Accuwx point forecasts that show daily means, by month. The chart doesn't have a space for February 29th... so what data do they use when that date occurs?

Any insights?

[Kmlwx]

I know this isn't a leap year, but I wondered if anyone knew how meteorological records reflect Feb. 29th on leap years. The existence of a day that only occurs once every four years has got to affect mean daily highs and lows, due to smaller sample set, as well as weather records. And I guess that Februaries that are 29 days long are, on average, a tiny amount warmer than those which are just 28.

What got me wondering about this is the little chart on the bottom of Accuwx point forecasts that show daily means, by month. The chart doesn't have a space for February 29th... so what data do they use when that date occurs?

Any insights?

Good question! Hopefully one of the mets has an answer!

[RodneyS]

I know this isn't a leap year, but I wondered if anyone knew how meteorological records reflect Feb. 29th on leap years. The existence of a day that only occurs once every four years has got to affect mean daily highs and lows, due to smaller sample set, as well as weather records. And I guess that Februaries that are 29 days long are, on average, a tiny amount warmer than those which are just 28.

What got me wondering about this is the little chart on the bottom of Accuwx point forecasts that show daily means, by month. The chart doesn't have a space for February 29th... so what data do they use when that date occurs?

Any insights?

I believe that the answer to your question involves the difference between "average" and "normal" temperatures. I have Washington, DC daily temperature data back to 1872, and my calculations for 1872-2010 show that the average temperature on February 29 has been slightly lower than on February 28 -- 39.5 (degrees Fahrenheit) vs. 39.7. That has also been true of the 1971-2000 period that is currently used by the National Weather Service as the "normal" -- 42.8 vs. 43.5; as well as the more recent 1981-2010 period that will soon displace 1971-2000 as the new "normal" -- 40.7 vs. 41.3. However, the NWS 1971-2000 "normal" temperature for both days is 41.5 (50 degree high and 33 degree low).

Now, according to my calculations, during 1971-2010, the lowest daily average temperature in Washington, DC occurred on February 9 at 33.42, barely nosing out January 10 at 33.43. However, the normalization model that is used by the NWS shows temperatures in DC bottoming out in mid-January at 34.5 degrees (42 degree high and 27 degree low). By the time February 9 rolls around, the 1971-2000 normal temperature is 37 degrees (45 degree high and 29 degree low). I would guess that the reason for this disparity is that the normalization model used employs a moving average, along with an assumption that there are no long-term temperature anomalies. In other words, even though February 9 was the coldest day in DC during 1971-2000, that fact is considered by the NWS to have no predictive value. Indeed, for the more recent period 1981-2010 in DC, the coldest day was January 21, at 32.2 degrees, with February 9 considerably warmer at 36.5.

Accordingly, when you hear a TV weatherman refer to "today's average high and low temperatures", s/he is misspeaking because s/he is actually referring to the normal temperatures for that date. However, that still leaves unanswered the question of how the normalization model used by the NWS works -- particularly with respect to February 29. Perhaps a meteorologist who has been involved in the normalization process can address that question.