a type of weather modification treatment used to increase precipitation – to provide water on...
TRANSCRIPT
The Use of Neural Network in Determining the Ideality of
Day for Cloud Seedingby
Cuesta, Yvanne Christine R.Uy, Ma. Ro-anne R.
CLOUD SEEDING
A type of weather modification
Treatment used to increase precipitation – to provide water on dams and agricultural fields during drought seasons
CLOUD SEEDING OPERATION
Preflight ObservationProceed? Yes or No
In-Flight Observation
Post flight ObservationMagnitude of precipitation?Drizzle/Light/Moderate/ Heavy
ISSUES
Requires large amount of funds Doesn’t produce enough rain Rain do not fall on the right location Some believe that it does not work successfully
at all…
In other words… A large amount of money is wasted
There is no established quantitative way of pursuing cloud seeding.
Neural Network
information processing model inspired by the human brain
used in classification through pattern based learning
Process Flowchart
Training of program
Testing and Validation of Program
Acquisition and Compilation of
Cloud Seeding Data
Conversion of data into data input
Acquisition and Programming of Implementing
Software
Time Cover Speed Humidity Pressure Output14 50 5 62 1010 212 50 8 61 1004 1
15.3 20 5 67 1002 210.3 25 2 79 1009.5 210.3 30 3 74 1002 313 20 4 64 1008 39.2 20 10 64 1006 210 30 5 69.4 1006 315 25 8 64 1008 315 25 8 66 1006 3
14.49 30 8 71 1006 29.3 25 5 67 1010 19.2 25 7 64 1007 2
11.15 25 6 66 1006.8 313 20 4 50 1008 2
Cloud Seeding Data
Generated Values of Training PhaseRow Id. Predicte
d ValueActual Value Residual Time Cover Speed Humidity Pressure
1 2.649967 2
-0.64996
714 50 5 62 1010
4 2.65435 2 -0.65435 10.3 25 2 79 1009.5
5 2.648228 3 0.35177
2 10.3 30 3 74 1002
6 2.663887 3 0.33611
3 13 20 4 64 1008
9 2.653325 3 0.34667
5 15 25 8 64 1008
10 2.651824 3 0.34817
6 15 25 8 66 1006
11 2.645084 2
-0.64508
414.49 30 8 71 1006
12 2.652481 1
-1.65248
19.3 25 5 67 1010
Generated Values of Training Phase
0 1 2 3 4 5 6 7 8 90
2
4
6
8
10
12
14
16
18
20
Lift chart (training dataset)
Cumulative Output when sorted using predicted valuesCumulative Output us-ing average
# cases
Cu
mu
lati
ve
Generated Values of Validation Phase
Row Id. Predicted Value
Actual Value Residual Time Cover Speed Humidity Pressure
3 2.899099 2
-0.89909
915.3 20 5 67 1002
14 2.442311 3 0.55768
9 11.15 25 6 66 1006.8
15 2.578017 2
-0.57801
713 20 4 50 1008
Generated Values of Validation Phase
0.5 1 1.5 2 2.5 3 3.50
1
2
3
4
5
6
7
8
Lift chart (validation dataset)
Cumulative Output when sorted using predicted valuesCumulative Output us-ing average
# cases
Cu
mu
lati
ve
Generated Values of Testing Phase
Row Id. Predicted Value
Actual Value Residual Time Cover Speed Humidity
2 2.58298 1 -1.58298 12 50 8 61
7 2.111562 2 -0.111562 9.2 20 10 64
8 2.395388 3 0.604612 10 30 5 69.4
13 2.129739 2 -0.129739 9.2 25 7 64
Generated Values of Testing Phase
0.5 1 1.5 2 2.5 3 3.5 4 4.50
1
2
3
4
5
6
7
8
9
Lift chart (test dataset)
Cumulative Output when sorted using predicted valuesCumulative Output us-ing average
# cases
Cu
mu
lati
ve
Summary
PhaseTotal Sum of
Squared ErrorsRMS (Root
Mean Square) Error
Average Error
Training 4.475583417 0.747962517 -0.27739325
Validation 1.006171177 0.579128995 -0.32756267
Testing 3.63599732 0.953414564 -0.643184
t-Test
Phase Mean Standard Deviation
Significant Difference
Training
Validation
Testing
Conclusion The ranges in which cloud seeding operations are
likely to be successful
Time Observed 8-10 A.M. / 3 -5 P.M.
Cloud Cover 50-75% (4 – 6 Oktas)
Wind Speed (and Direction)
2 – 19 km/h (1-10 Knots) / (Direction varies based on the location)
Humidity 60% +
Barometric Pressure 980-1010 mb
It is possible to establish a more accurate and precise way of predicting the ideality of Day for cloud seeding.
Recommendations
Wide range of cloud seeding data wherein taking into consideration other factors besides from the given factors
Usage of Backpropagation Neural Network
Bibliography American Friends of Tel Aviv University (2010, November 1). 'Cloud seeding' not
effective at producing rain as once thought, new research shows. ScienceDaily. Retrieved December 10, 2010, from http://www.sciencedaily.com /releases/2010/11/101101125949.htm
Hagan, M.T. (1996). Neural Network Design. Boston, MA: PWS Publishing. Macfarlane, M. (2009, February 3). Major study proves cloud seeding effective.
Cosmomagazine. Retrieved July 21, 2010 from http://www.cosmosmagazine.com/news/2514/major- study-proves-cloud-seeding-effective.
Matthew, J. (2000). An Introduction to Neural Networks. Retrieved July 21, 2010 from http://www.generation5.org/content/2000/nnintro.asp.
Moseman, A. (2009, February 19). Does cloud seeding work?. Scientific American. Retrieved August 11, 2010 from http://www.scientificamerican.com/article.cfm?id=cloud-seeding-china- snow
Shoukat, U. & Zakia, H. (2005). Improve an efficiency of feedforward multilayer perceptrons by Serial training. Journal of Theoretical and Applied Information Technology, 6(1), 017 – 020. Retrieved August 11, 2010 from http://www.jatit.org/volumes/research- papers/Vol6No1/2Vol6No1.pdf.
XLMiner (n.d.). Online Help. Retrieved January 22, 2011 from http://www.resample.com/xlminer/help/Index.htm