Документ взят из кэша поисковой машины. Адрес оригинального документа : http://zebu.uoregon.edu/1999/es202/l12a.html
Дата изменения: Thu Feb 11 22:25:41 1999
Дата индексирования: Tue Oct 2 02:32:39 2012
Кодировка:
Поисковые слова: п п п п п п п п п п п п п п п п п п п п п п п п п п п п п

Data, data, data

ENVS 202 On-line Access

Human Impact on Climate does the data support that there is an effect?

Looking at more data to find significance or not:

Another Case: US Tornado Frequency:

This is the distribution of verified tornadoes in the American Midwest from the period of 1953--1993:

Is the apparent recent increase significant?

Excluding the last 4 years, the mean number of tornadoes is 747 with a dispersion of 162 and a mean error of 27
For the last 4 years, the mean number of tornadoes is 1172 with a disperion of 76 and a mean error of 38 (the mean error is larger because N is smaller)
Again the means are similar so we get
M1-M2 = 1172 -747 = 425
1.5E1 = 1.5*27 which is around 45
425/45 = 9 --> EXTREMELY SIGNIFICANT indicating a real effect has increased the tornado frequency --> observation or physical?

Another Case: Urbanization and the Weather

Does Urbanization produce a microclimate in the sense that urban areas are getting warmer.

Where the data is

Here are some results for Washington DC average Tempearture:

1960-1990: 14.78 +/- 8.5 N=360
1960-1970: 14.26 +/- 8.8 N=120
1970-1980: 15.25 +/- 8.3 N=120
1980-1990: 15.01 +/- 8.5 N=120

Note that the dispersions are all very similar in these periods. The square root of 120 (number of data samples) = 11.

Error in the mean for 1960-1970 = 8.8/11 = 0.80

Error in the mean for 1970-1980 = 8.3/11 = 0.75

(M1-M2)/1.5E1 = 15.25-14.26/(1.5*.8) = 1/1.2 = 0.83 --> no significant increase. A significant increase would be hard to detect in this data because the instrinsic dispersion is large.

Let's look at the case for Berkeley CA:

1960-1970: 13.67 +/- 3.1 N=120
1980-1990: 14.32 +/- 3.0 N=120

The error in the mean is a lot less because the disperion is less. The mean error is approximately 3/11 = .27

So for Berkeley we have: 14.32-13.67/1.5(.27) = .65/.40 = 1.65 --> still not significant but a lot more significant than the case for Washington DC.

This indicates that Berkeley is a better site for this study than Washington DC since the natural climate variations in DC are more severe and hence the signature of a small effect will be much more difficult to get out of the data.

Now, what do you make of this data?

Previous Lecture Next Lecture Course Page