Meteorology Final Project PDF

Preview

Summary

Utilizes data collected outside of class and then summarizes the process and materials learned in class through a capstone project....

Description

Channel 5 High 90 80 70 60 50 40 30 20 10 0 20

f(x) = 0.84 x + 7.42 R² = 0.45

30

40

50

High Linear60(High)

70

80

90

NWS Low 70 60 50 40 f(x) = 0.16 x + 24.59 R² = 0.19

30 20 10 0

Low

0

10

20

30

40

50Linear (Low) 60

70

80

90

100

AccuWeather High 90 80

AccuWeather Low 70

70 60 50

60 f(x) = 0.11 x + 46.78 R² = 0.06

40 30

50 40 30

20 10 0

0

10

20

30

40

High Linear60 (High) 70 50

f(x) = 0.16 x + 24.59 R² = 0.19

20 80

10 0

Low

0

10

20

30

40

(Low) 70 50Linear 60

80

90

100

Channel 2High 90 80 70 60 f(x) = 0.11 x + 46.78 R² = 0.06

50 40 30 20 10 0

High

0

10

20

30

40

(High) 70 50Linear 60

80

90

100

Channel 4 High 90 80 70 60 50 40 30 20 10 0

Channel 4 Low 70 60 f(x) = 0.07 x + 48 R² = 0.02

50 40 30 High

0

10

20

30

40

Linear (High) 50 60 70

80

90

100

f(x) = 0.13 x + 25.44 R² = 0.12

20 10 0

0

10

20

30

40

50

60

Linear ()

70

80

90

100

1. The station with the most accurate forecasts is the National Weather Service. The National Weather Service has a mean error of +1 degree for the high and -2 degrees for the low. This puts their mean error as one of the lowest of all the channels. The National Weather Service is the most accurate because it has a very accurate read on high temperatures and one of the strongest low temperature forecasts. The slope of the high temperature is 1.0417 making it close to 1 which would make it a perfect forecast. The low temperature slope is 0.8674 making it close to the 1 slope. The standard deviations for high and low are 4 and 7 which are the lowest values of all the channels indicating a small spread in the data points. 2. There is a difference in high and low forecast accuracy across all stations. The National Weather Service has a highly accurate high temperature, but the low temperature falls off. The data becomes much more spread with the low temperature and the reliability falls. 3. The statistics are not biased by busted forecasts. The amount of data collected lends itself well to outliers. The common theme is that the data points that might be considered outliers just increase the spread of the data. An example would be from AccuWeather where a high was forecast at 60 but the actual high turned out to be 80. The outlier gets balanced out by other inaccurate forecasts and spreads the data. 4. The low forecast temperatures for channel 4 from 2/3 to 3/15 held consistently large inaccurate data. This trend did not occur across all stations. There are brief few day periods of heightened accuracy throughout all the channels, but the trends do not occur across all channels. Accuracy and inaccuracy appear to be fairly random. One interesting accuracy occurrence comes from 1/28 where all channels are accurate between 1 and 2 degrees, but channel 5 is off by 21 degrees. The next day the temperature dropped 20 degrees. Channel 5 appeared to predict the drop early. 5. The stations do not use persistence forecasts for extended periods of time. Channel two uses a persistence forecast on occasion like on 1/19 and 1/20 and AccuWeather uses persistence on 3/21 and 3/22. Because no station uses the persistence method for a long period of time it is difficult to determine whether the station is using a persistence forecast or if it is just a coincidence that the persistence shows up. 6. Missing data has a significant impact on the data quality. By removing a portions of the data it can paint an unfinished or biased picture. With missing data points, the outliers start to have a significant impact on the spread of the data and in turn messes with the forecast predictability. When comparing forecast accuracy between stations this becomes an issue because there will be forecasts that are busted by all the stations and if one station does not have that bust it skews their data in a positive manner. It can be filled with a persistence forecast. The persistence forecast allows missing points to be filled and does not distort the data too much. I filled the

missing points for the National Weather Service using persistence and the central measurements I got were: High(Original) Low(Original) High(Fixed) Low(Fixed) Mean Error 1 -2 0 -2 Deviation 4 7 8 7 Mode 3 2 3 -2 Median 1 -2 1 -2 The persistence forecast primarily changes the spread of the data because the standard deviation for the high increases to 8. The hypotheses that one television station is more accurate than the others does not hold a lot of value from the statistics we collected throughout the semester. Because of various gaps in the data pool and different amounts of spread in data trends, it is not of great value to say the National weather service has the most accurate forecast with a mean error of 1 because there are so many holes in the data. Likewise it is not very valuable to say that channel 5 is more accurate because they have a mean error of 0 because the spread of the data is so wide with a standard deviation of 8. There is a reason to believe that an accurate forecast from a local channel is more valuable than one from the local National Weather Service. The local channel is invested in the community, whereas the National Weather Service is not. That investment gives weight to the forecasts they make. Local partnerships can add to a weather channels repour. For example, Chris Higgins is the meteorologist at fox2now. He is an alum from high school and I went to school with his son and so I know Higgins personally, hence his repour with me is increased and I give more value to his forecasts. The cost of the forecasts are felt by small businesses and the community at large. In class, we mentioned that some businesses need to know what the weather will be or it could cost them a lot of money. If something needs to be done about the power lines because of incoming weather, people need to know that, and if they do not get an accurate forecast for the weather then they have a big problem....