sexta-feira, 23 de abril de 2010

Eurekagate

Yesterday's post about Eureka has led to some comments, regarding how important this issue is. Anthony Watts has made some very good points explaining why this is important. I would like to further expand on this.

Maximum temperatures are very important, as mean temperatures result from the average between the maximum and minimum temperatures of a day. That can be seen for the July 2009 values, where the mean temperature for July 14th, was 14.1ºC. This results from adding the minimum temperature (7.3ºC) to the maximum temperature (20.9ºC), and dividing the result by 2, thus giving 14.1ºC. Now, as can be seen from the hourly temperatures of the day, 14.1ºC was only surpassed at 19:00, with 14.4ºC. Now, the mean monthly temperatures seem to be derived from the daily means, giving the 8.0ºC mean temperature for July 2009. This value is the same as given in GHCN data (note: very large file!) and they confer with the National Climate Data and Information Archive, with only a minor difference in April, in all of 2009. Please notice the data in the GISS file:

4037191700062009 -340 -386 -386 -273 -98 40 80 56 -56 -234 -233 -296

Now, readers following this might wonder how much big the impact is when maximum temperatures have sky-rocketed, as we have shown yesterday. More in a little moment. First I had to be sure about how the mean temperatures were calculated. Reverse engineering the values for 2009, I found that the monthly mean is the mean of the daily means. This can be seen by calculating other months mean values, namely March 2009 and November 2009. This has an interesting impact, as the value rounding observed in each day's value, is incorporated in the monthly value. Please see the following table for the details on the July average:

DayMaxMinNCDIA
mean
Math
mean
July 110-0.154.95
July 27.82.154.95
July 37.61.94.84.75
July 49.34.46.96.85
July 56.22.64.44.4
July 69.52.96.26.2
July 78.62.75.75.65
July 810.12.56.36.3
July 910.43.97.27.15
July 1011.53.47.57.45
July 119.32.96.16.1
July 129.32.86.16.05
July 1319.82.811.311.3
July 1420.97.314.114.1
July 1515.74.510.110.1
July 1612.63.78.28.15
July 1714.438.78.7
July 1814.86.610.710.7
July 1913.46.710.110.05
July 2016.43.710.110.05
July 21152.68.88.8
July 2217.19.113.113.1
July 2313.86.810.310.3
July 24123.17.67.55
July 2514.47.310.910.85
July 2613.46.39.99.85
July 279.54.16.86.8
July 287.22.14.74.65
July 299.41.25.35.3
July 3013.14.48.88.75
July 3110.34.77.57.5
8.0067.981

Please notice that the math mean will always be lower than the NCDIA mean. How much depends on the specific values, but one would expect it to be 0.025ºC (15x0.05/30) on a 30 day month. This is in range with the difference observed for July 2009.

Now finally we can measure the impact of the strange temperatures in July 2009. If we admit that the July 13th maximum value of 19.6ºC is an absurd, we can use the second lowest value of that day, which is 13.7ºC. Some of you might argue that it is still a high value, as can be seen in the daily graph, but for simplicity, let's use it. And we'll use 14.4ºC as the maximum temperature for July 14, as this is the highest value for that day.

DayMaxMinNCDIA
mean
Math
mean
delta for
13 and 14
July 110-0.154.954.95
July 27.82.154.954.95
July 37.61.94.84.754.75
July 49.34.46.96.856.85
July 56.22.64.44.44.4
July 69.52.96.26.26.2
July 78.62.75.75.655.65
July 810.12.56.36.36.3
July 910.43.97.27.157.15
July 1011.53.47.57.457.45
July 119.32.96.16.16.1
July 129.32.86.16.056.05
July 1319.82.811.311.38.25
July 1420.97.314.114.110.85
July 1515.74.510.110.110.1
July 1612.63.78.28.158.15
July 1714.438.78.78.7
July 1814.86.610.710.710.7
July 1913.46.710.110.0510.05
July 2016.43.710.110.0510.05
July 21152.68.88.88.8
July 2217.19.113.113.113.1
July 2313.86.810.310.310.3
July 24123.17.67.557.55
July 2514.47.310.910.8510.85
July 2613.46.39.99.859.85
July 279.54.16.86.86.8
July 287.22.14.74.654.65
July 299.41.25.35.35.3
July 3013.14.48.88.758.75
July 3110.34.77.57.57.5
8.0067.9817.777

The value has come down from 8.0ºC to 7.8ºC. But one might argue that this max/min temperature simple calculations are really too simplistic. I moved on, gathering the hourly values for July 2009, in CSV format, and doing the necessary calculations. Another column emerges, and as can be seen below, another reality appears: mean monthly temperatures for July 2009 are at 7.6ºC!

DayMaxMinNCDIA
mean
Math
mean
delta for
13 and 14
hourly
mean
July 110-0.154.954.953.895
July 27.82.154.954.954.305
July 37.61.94.84.754.753.691
July 49.34.46.96.856.856.564
July 56.22.64.44.44.44.255
July 69.52.96.26.26.25.659
July 78.62.75.75.655.654.773
July 810.12.56.36.36.34.636
July 910.43.97.27.157.157.241
July 1011.53.47.57.457.457.564
July 119.32.96.16.16.15.027
July 129.32.86.16.056.055.573
July 1319.82.811.311.38.258.527
July 1420.97.314.114.110.8511.464
July 1515.74.510.110.110.19.65
July 1612.63.78.28.158.159.927
July 1714.438.78.78.77.95
July 1814.86.610.710.710.711.164
July 1913.46.710.110.0510.0510.823
July 2016.43.710.110.0510.0512.845
July 21152.68.88.88.87.809
July 2217.19.113.113.113.113.973
July 2313.86.810.310.310.310.15
July 24123.17.67.557.557.995
July 2514.47.310.910.8510.8510.809
July 2613.46.39.99.859.859.286
July 279.54.16.86.86.87.118
July 287.22.14.74.654.653.991
July 299.41.25.35.35.34.686
July 3013.14.48.88.758.757.614
July 3110.34.77.57.57.56.482
8.0067.9817.7777.595

For some strange reason, data for this station doesn't include the first two hours of the day. Seems like temperatures should be colder in that period, despite the fact that there is always a sun up there in July. Couldn't help testing this one too. So, with no surprise, temperatures went down again:

DayMaxMinNCDIA
mean
Math
mean
delta for
13 and 14
hourly
mean
24 hour
mean
July 110-0.154.954.953.8953.804
July 27.82.154.954.954.3054.217
July 37.61.94.84.754.753.6913.633
July 49.34.46.96.856.856.5646.471
July 56.22.64.44.44.44.2554.346
July 69.52.96.26.26.25.6595.492
July 78.62.75.75.655.654.7734.742
July 810.12.56.36.36.34.6364.504
July 910.43.97.27.157.157.2417.108
July 1011.53.47.57.457.457.5647.592
July 119.32.96.16.16.15.0274.954
July 129.32.86.16.056.055.5735.717
July 1319.82.811.311.38.258.5278.208
July 1420.97.314.114.110.8511.46411.004
July 1515.74.510.110.110.19.659.717
July 1612.63.78.28.158.159.92710.138
July 1714.438.78.78.77.957.713
July 1814.86.610.710.710.711.16410.921
July 1913.46.710.110.0510.0510.82310.742
July 2016.43.710.110.0510.0512.84512.638
July 21152.68.88.88.87.8097.846
July 2217.19.113.113.113.113.97313.95
July 2313.86.810.310.310.310.1510.246
July 24123.17.67.557.557.9958.083
July 2514.47.310.910.8510.8510.80910.729
July 2613.46.39.99.859.859.2869.608
July 279.54.16.86.86.87.1187.104
July 287.22.14.74.654.653.9914.1
July 299.41.25.35.35.34.6864.563
July 3013.14.48.88.758.757.6147.429
July 3110.34.77.57.57.56.4826.421
8.0067.9817.7777.5957.540

So, instead of 8.0ºC, we are down to 7.5ºC! Well, I'll stop for now. Before it gets too cold!

Note: The photo at the beginning of the post was given by Tim Clark, in the Watts Up With That post, and is available here.
Note2: Reference to GISS was replaced with the correct reference to GHCN, above.