Assignment #05

Unit 05

Until next unit, work through the material provided in Unit 5 and solve the following exercise blocks in separate notebooks.

Notebook 1

#05-01: Looping vs vectorizing

Consider this:

import numpy as np
count = np.arange(10, 20)
v1a = np.full(count.shape, False)
v1b = v1a.copy()
v2a = v1a.copy()
v2b = v1a.copy()

Write a loop that achieves the same result as the following line of code, so that v1a and v1b are equal again:

v1a[count > 16] = True

Write a vectorized statement using logical indexing that achieves the same result as the following line of code, so that v2a and v2b are equal again:

for i, ct in enumerate(count):
    if ct > 13 and ct < 17:
        v2a[i] = True

Notebook 2

For this exercise block, you will work with a real data set, nowcast data from a numerical weather prediction model in Glacier National Park, Canada, WX_GNP.csv.

Here is a description of the spread sheet variables:

  datetime: datetime in the form YYYY-MM-DD HH:MM:SS
station_id: ID of virtual weather station (i.e., weather model grid point)
        hs: Snow height (cm)
      hn24: Height of new snow within last 24 hours (cm)
      hn72: Height of new snow within last 72 hours (cm)
      rain: Liquid water accumulation within last 24 hours (mm)
      iswr: Incoming shortwave radiation (also refered to as irradiance) (W/m2)
      ilwr: Incoming longave radiation (W/m2)
        ta: Air temperature (degrees Celsius)
        rh: Relative humidity (%)
        vw: Wind speed (m/s)
        dw: Wind direction (degrees)
      elev: Station elevation (m asl)

#05-02: Explore WX

Read the comma separated spreadsheet WX_GNP.csv
How many different virtual stations are included in the data frame?
How many unique time stamps does the data frame contain?
What are the earliest and latest time records? Which time records are displayed in the first and last rows?
How many stations are located above 2000m but below 2200m?
Characterize the elevation distribution of the stations with different percentiles

Tip

The methods .unique(), .min(), .max(), .describe(), and .quantile() will be handy for these tasks.

#05-03: Subset WX and compute more summary stats

Create another data frame as subset of WX, which contains all the data from one station_id of your choice.
What is the average air temperature and its standard deviation?
What is the median relative humidity rh when either hn24 is greater than 10 cm or rain is greater than 2 mm? What about the median rh during the opposite conditions?
Compute a new column hn72_check that should conceptually be identical to hn72. Use only hn24 to derive hn72_check.
Test whether hn72_check is indeed equal to hn72. Why not?
Store the new data frame in a csv file.

highlight-style: github

Solutions

Solution #05-01: Looping vs vectorizing

import numpy as np
count = np.arange(10, 20)
v1a = np.full(count.shape, False)
v1b = v1a.copy()
v2a = v1a.copy()
v2b = v1a.copy()

v1a[count > 16] = True

for i, ct in enumerate(count):
    if ct > 16:
        v1b[i] = True

print(f"v1a is equal to v1b: {np.array_equal(v1a, v1b)}")

for i, ct in enumerate(count):
    if ct > 13 and ct < 17:
        v2a[i] = True

v2b[(count > 13) & (count < 17)] = True

print(f"v2a is equal to v2b: {np.array_equal(v2a, v2b)}")

v1a is equal to v1b: True
v2a is equal to v2b: True

Solution #05-02 & #05-03: a real dataset

Download the notebook: 05_tasks_solved.ipynb