ABOUT THIS PROGRAM
This program uses data from the NOAA Global Historical Climatology Network (GCHN) to
analyze historical snowfall and snowpack at North American weather stations and looks 
for correlations between these measures and historical El Nino/La Nina episodes between 
the years 1950 and 2010.  The Nino Index is a measurement of the variation from a 30 
year average. This program calculates similar variations from the snowfall and snowdepth 
30 year averages and looks for correlations between these values and the Nino Index. 
These variations are referred to as "delta" in this program. More information about 
the data used in this analysis can be found here: 
http://www1.ncdc.noaa.gov/pub/data/ghcn/daily/readme.txt

More detailed information about the work done in this program can be found in the 
project report: 'IS 602 Final Project.pdf' also found in the project tar file.

This program will provide prompts to the user to select the statistical significance 
level for the correlation analysis, and to then select a weather station to view a 
scatter plot of the data for that weather station.

This program was created as final project for a Data Science Python programming class. 
As such, the emphasis was mostly on the code and less on the scientific process, so 
there are some potential scientific issues with this analysis. For example, in order 
to simplify the coding for this project this program ignores missing daily values when 
calculating monthly averages.

HOW TO USE:
Download the 'Python Weather Project.tar' and extract all files, this file contains all 
files needed EXCEPT the ghcd_all.tar.gz file which was too large to include in this 
file. The ghcnd_all.tar.gz file must be downloaded from the NOAA ftp site:
ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/daily/

The user will be presented with two options 
Select 0 to run all of the extraction and processing functions to create the 
processed data files

or

Select 1 to use the pre-processed file, which is the tar file: ghcnd_processed.tar


If 0 is selected, at least 30GB of free space will be required for extracting and 
copying files, and the extraction and processing can take up to an hour to complete.

You must have all of the following files in your directory to run option 0:
ghcnd_all.tar, ghcnd-inventory.txt, ghcnd-stations.txt,Nino_index.csv

If 1 is selected the ghcnd_processed.tar file will be used. This archive contains 
37 files, which contain complete data records (all years, all months between October
and April) for snowfall and snowdepth for the years 1950 through 2010, which also 
have significant correlations to the Nino Index when considering average monthly 
variations from the 30 year monthly average.

You must have all of the following files in your directory to run option 1
ghcnd_processed.tar, ghcnd-inventory.txt, ghcnd-stations.txt, Nino_index.csv

METHODOLOGY AND CODING
Importing and Extracting the data: 
Data (“.dly” files) are extracted from the TAR file using the tarfile module.
In subsequent steps using the metadata files "ghcnd-stations" and "ghcnd-inventory" 
are used to pare down the list of all station files to just those that are North 
American stations where there there is data for 1950 and 2010.

To further reduce the files to just those that are relevant a station by station
analysis is completed, to first look for stations with complete data records 
(see above) and finally correlations with the Nino Index as described briefly above.

STATION ANALYSIS
Note: Each “.dly” file has an identical structure

Monthly Average: 
Computes a total snowfall and a snowpack average for for the “snow months” defined in 
this project as October – April for all years (1950 to the present).

30 Year Monthly Average: 
Computes an average monthly snowfall (an average of the monthly snow totals) and 
snowpack for the snow months (October – April) for the 30 year period 1981 – 2010. This
is the most recently completed 30 year period and is commonly used to calculate 
climatological “normals” in meteorology. It should be noted that the ONI data is based
on multiple centered thirty year periods; however, that is not the standard practice 
in meteorology and it would significantly complicate the analysis for this project to 
follow the ONI method.  Therefore only the 1981-2010 thirty year period will be used to 
compute snowfall and snowpack averages.

Monthly Snowfall vs 30 Year Average: 
Calculates the snowfall/snowpack deltas between the values in step 1 (monthly averages)
and the values in step 2 (30 year averages). This step will create an array of monthly
delta values (October-April) for the weather station, for each year in the ONI record 
(1950 – present).

Correlation Analysis: 
Using scipy.stats.pearsonr to compute the correlation between each array (snowfall 
and snowpack) in step 3 and the ONI data set.(Each data point in the ONI table is a 
three month rolling average so a January-February-March 2010 value would be compared 
to a February 2010 value from the snowfall/snowpack arrays).  

Scatter Plot and Regression analysis: :
A scatter plot with a linear regression line for snowfall and snowpack can be displayed 
for stations that show statistically significant correlations. For the purpose of this 
project, Pearson correlation p-value = 0.10 or less.