Capacity Building Workshop - Python for Climate Data Analysis

Capacity Building WorkshopPython for Climate Data Analysis

Christoph Menz

November 26, 2019

Module Overview

• Lecture: Introduction to Python• Hands-On: Python for Beginners• Exercise: Python• Lecture: Introduction to Python Libraries• Hands-On: Access and analysis of netCDF Data with Python• Lecture: Python Libraries for Data Visualization• Hands-On: Visualization of Scientific Data with matplotlib• Hands-On: Visualization of Geospatial Data with cartopy• Exercise: Analysis and Visualization of netCDF Data with

python

,Christoph Menz RD II: Climate Resilience 2

Introduction to Python


Introduction to Python

• High-level general-purpose programming language• Emerges in the late 80s and early 90s (first release 1991)• Based on teaching/prototyping language ABC• Freely available under Python Software Foundation License• Major design philosophy: readability and performance• Important features:

• Dynamic types (type automatically declared and checked atruntime)

• Automatized memory management• Objects, Loops, Functions• Easily extendible by various libraries (numpy, netCDF4,

scikit-learn, ...)


Scope of this Course

• Basic programming and scripting with python• Read, preparation, statistics and visualization of netCDF

based data• Focus on python version 3.x using Anaconda platform• Online Tutorials:

Anaconda Tutorialshttps://docs.python.org/3/tutorialhttps://www.tutorialspoint.com/python3https://scipy.org & http://scikit-learn.orghttps://matplotlib.org/http://scitools.org.uk/cartopy


https://docs.python.org/3/tutorial

https://www.tutorialspoint.com/python3

https://scipy.org

http://scikit-learn.org

https://matplotlib.org/

http://scitools.org.uk/cartopy

Anaconda - Data Science Platform


Jupyter - Interactive Computing Notebook


Spyder - Interactive Computing Notebook


Variables in python

• Variable types are automatically defined at runtime

variable_name = value

• Uses dynamic and static type casting:• 5*5.0 is a float and "Hello world" is a string• str(5) is a string and int("9") is a integer

• python got 13 different built-in types:• bool, int, float, str, list, tuple, dict, bytearray, bytes,

complex, ellipsis, frozenset, set• Possibility to create your own type for object-oriented

programming ( class statement)


Variable Types• Basic variable types

Boolean - boolIn [1]: x = TrueIn [2]: y = FalseIn [3]: Y = True

Int, Float and Complex - int, float, complexIn [1]: x = 5In [2]: y = 5.0In [3]: z = 5.0+2.0j

Characters and Strings - strIn [1]: char = "a"In [2]: string = 'python'In [3]: SeNtEnCe = "This is a sentence."In [4]: x = """This is a sentence

...: across multiple lines"""In [5]: string[0:2]In [5]: py


Variable Types - Lists• Lists are sequences of variables of arbitrary type

(also lists of lists of lists ... possible)• Lists are mutable• Single elements of lists can be accessed by indexing

(from 0 to length - 1)

ListIn [1]: List = [2.0, 5.0, True, 7.0, "text"]In [2]: ListList = [[2.0, 5.0], [True, 7.0, "more text"]]

In [3]: ListList[0] = List[4]In [4]: ListListOut [4]: ["text", [True, 7.0, "more text"]]


Variable Types - Tuples• Tuples are similar to lists• But tuples are immutable

TupleIn [1]: Tuple = (2.0, 5.0, True, 7.0, "text")In [2]: TupleTuple = ((2.0, 5.0), (True, 7.0, "more text"))

In [3]: TupleTuple[0] = Tuple[4]----------------------------------------------------------------TypeError Traceback (most recent call last)<ipython-input-52-862d5dc2e8bb> in <module> ()----> 1 TupleTuple [0] = Tuple [4]

TypeError : 'tuple' object does not support item assignment


Variable Types - Dictionaries• Dictionaries are unordered collections of arbitrary variables• Dictionaries are mutable• Elements of dictionaries get accessed by keys instead of

indices• Keys in dictionaries are unique

DictionaryIn [1]: my_dict = {"a":2.0, "b":5.0, "zee":[True, True]}In [2]: my_dict["b"] = 23.0In [3]: my_dictOut [3]: {'a':2.0, 'b':23, 'zee':[True,True]}In [4]: {"a":2.0, "b":5.0, "zee":[True, True], "a":7}Out [4]: {'a':7, 'b':5.0, 'zee':[True,True]}


Operations

Addition & SubtractionIn [1]: 3 + 5.0Out [1]: 8.0In [2]: 3 - 5Out [2]: -2

Multiplication & DivisionIn [1]: 4 * 4Out [1]: 16In [2]: 8 / 2Out [2]: 4.0In [3]: 7 // 3Out [3]: 2In [4]: 7 % 3Out [4]: 1

• python supports the usualmathematical operations onfloat, int and complex

• Dynamic casting depends onoperator and variable type

Power & RootIn [1]: 4**2Out [1]: 16In [2]: 4**2.5Out [2]: 32.0In [3]: 16**0.5Out [3]: 4.0


Boolean Operations

ComparisonsIn [1]: 5 > 3Out [1]: TrueIn [2]: 5 >= 3Out [2]: TrueIn [3]: 5 < 3Out [3]: FalseIn [4]: 5 <= 3Out [4]: FalseIn [5]: 5 == 3Out [5]: FalseIn [6]: 5 != 3Out [6]: True

• Python uses usual comparisonoperations

• in -Operator permits an easy searchfunctionality

in-OperatorIn [7]: 7 in [1, 2, 3, 4, 5]Out [7]: FalseIn [8]: "b" in {"a":4, "b":6, "c":8}Out [8]: True


Boolean Operations

ComparisonsIn [1]: 5 > 3Out [1]: TrueIn [2]: 5 >= 3Out [2]: TrueIn [3]: 5 < 3Out [3]: FalseIn [4]: 5 <= 3Out [4]: FalseIn [5]: 5 == 3Out [5]: FalseIn [6]: 5 != 3Out [6]: True

• Python uses usual comparisonoperations

• in -Operator permits an easy searchfunctionality

in-OperatorIn [7]: 7 in [1, 2, 3, 4, 5]Out [7]: FalseIn [8]: "b" in {"a":4, "b":6, "c":8}Out [8]: True


Boolean Operators

• python supports the basic logicaloperators to combine booleans

Logical NOTOperator Results

not True Falsenot False True

Logical ANDx Operator y Results

True and True TrueTrue and False False

False and True FalseFalse and False False

Logical ORx Operator y Results

True or True TrueTrue or False TrueFalse or True TrueFalse or False False


Methods of Objects/Variables

• Python variables are notjust atomic variables

• Python variables areobjects by themself

• Each variable alreadycomes with associatedmethods

• Syntax:variable.method

Object methodsIn [1]: x = []In [2]: x.append(3)In [3]: x.append(5)In [4]: print(x)[3,5]In [5]: y = {"a":1,"b":2,"c":3}In [6]: print(y.keys())dict_keys(['a','b','c'])In [7]: "This is a sentence".split(" ")In [7]: ['This','is','a','sentence']In [8]: " ".join(["This","is","a","list"])In [8]: 'This is a list'

You can use the dir() function to get an overview of all methodsavailable for a given variable.


Condition and Indentation• Condition start with if and ends with : (equivalent to

”then” in other languages)• Syntax: if expression :

statement

• Python uses indentation (leading whitespaces) instead ofbrackets to seperate code blocks

ifIn [1]: x = 7In [2]: if x >= 5 and x <= 10:

...: print("x is above 5")

...: print("x is below 10")

...:x is above 5x is below 10

INDENTATION

INDENTATION




statement• Python uses indentation (leading whitespaces) instead of

brackets to seperate code blocks

ifIn [1]: x = 7In [2]: if x >= 5 and x <= 10:




INDENTATION

INDENTATION




statement• Python uses indentation (leading whitespaces) instead of

brackets to seperate code blocks

ifIn [1]: x = 7In [2]: if x >= 5 and x <= 10:




INDENTATION

INDENTATION


Condition and Indentation• if conditions support arbitrary number of elif conditions and

one possible else condition

if ... elif ... elseIn [1]: x = 20In [2]: if x >= 5 and x <= 10:

...: print("x is between 5 and 10")

...: elif x < 5:


...: elif x in [15,20,25]:

...: print("x is 15, 20 or 25")

...: else:

...: print("x is out of bound")

...:x is 15, 20 or 25


Loops• For loops iterate only a specific number of times• Syntax: for variable in iterable :

statement• Iterable are objects you can iterate over (list, tuple, dict,

iterators, etc.)

for-LoopIn [1]: for x in [2,4,6,8]:

...: print(x*2)

...:481216


Built-In Functions

• Python ships with several built-in functions for daily usage• Syntax: function(arguments)• Function arguments are comma seperated values

print() FunctionIn [1]: print("123")123

In [2]: print(123)123In [3]: print(1,2,3,"123")1 2 3 123

len() FunctionIn [1]: len("123456")Out [1]: 6

In [2]: len([3, 5, 8])Out [2]: 3In [3]: len({"a":13,"b":21})Out [3]: 2


Type Related Built-In Functions

• Use the type() function toget the type of any variable

• Type conversion can bedone using one of thefollowing functions:bool(), int(), float(), str(),list(), tuple(), dict()

type() FunctionIn [1]: type("PyThOn")Out [1]: strIn [2]: type(3)Out [2]: intIn [3]: type(3.0)Out [3]: floatIn [4]: type({"a":13,"b":21})Out [4]: dict

Type Conversion IIn [1]: bool(0)Out [1]: FalseIn [2]: bool(2.2)Out [2]: TrueIn [3]: int(2.8)Out [3]: 2

Type Conversion IIIn [1]: list((2,3,5))Out [1]: [2, 3, 5]In [2]: tuble([2,3,5])Out [2]: (2, 3, 5)In [3]: float("3.14")Out [3]: 3.14


Mathematical Built-In Functions• Python supports basic mathematical operations• Work on numbers: abs and round• Work on list and tuples: min, max, sum and sorted

abs() and round()In [1]: abs(-5)Out [1]: 5In [2]: round(24.03198)Out [2]: 24In [3]: round(24.03198,3)Out [3]: 24.032

min(), max(), sum() and sorted()In [1]: min([55,89,144,233])Out [1]: 55In [2]: max([55,89,144,233])Out [2]: 233In [3]: sum([55,89,144,233])Out [3]: 521In [4]: sorted([12,3,17,3])Out [4]: [3, 3, 12, 17]In [5]: sorted(["b","aca","aaa","cd"])Out [5]: ['aaa', 'aca', 'b', 'cd']


Help Built-In Function• The most important built-in function is help()• Gives you a short description on the given argument

(variables or other functions)

help()In [1]: help(max)Help on built-in function max in module builtins:

max(...)max(iterable, *[, default=obj, key=func]) -> valuemax(arg1, arg2, *args, *[, key=func]) -> value

With a single iterable argument, return its biggest item. Thedefault keyword-only argument specifies an object to return ifthe provided iterable is empty.With two or more arguments, return the largest argument.


User-Defined Functions• Python supports also user-defined functions• Arbitrary number of function parameters (also optional

arguments possible)

User-Defined Function: my_functionIn [1]: def my_function(x, y, opt_arg1 = 1, opt_arg2 = "abc"):

...: out = x + y

...: print(opt_arg1)


...: return outIn [2]: z = my_function( 2, 3, opt_arg1 = "cba")cbaabcIn [3]: print(z)5

mandatoryparameters

optionalparameters


User-Defined Functions• Python supports also user-defined functions• Arbitrary number of function parameters (also optional

arguments possible)

User-Defined Function: my_functionIn [1]: def my_function(x, y, opt_arg1 = 1, opt_arg2 = "abc"):

...: out = x + y



...: return outIn [2]: z = my_function( 2, 3, opt_arg1 = "cba")cbaabcIn [3]: print(z)5

mandatoryparameters

optionalparameters


Hands-On: Python forBeginners


Exercise: Python


Exercise

1. Test if the following operations between various types arepossible: float*int, bool*int, bool*float, bool+bool,string*bool, string*int, string*float, string+int

2. What is the result of the following operations:["a","b","c"]*3, (1,2,3)*3 and{"a":1,"b":2,"c":3}*3. Could you explain why the lastoperation isn’t working?

3. Print all even numbers between 0 and 100 to the screen (hint:use a for loop and if condition).


Exercise4. Write a function that calculates the mean of a given list of

floats (hint: use sum() and len()).5. Write a function that calculates the median of a given list of

floats (hint: use sorted() and len() to determine thecentral value of the sorted list, use if condition to distinguishbetween even and odd length lists).

6. Test your mean and median function with the following lists:

list mean median[4,7,3,2,7,4,2] 4.143 4.0[2,6,3,1,8,5,4] 4.143 4.0[2,1,4,5,7,9] 4.667 4.5[2,7,4,8,5,1] 4.500 4.5


Introduction to PythonLibraries


Libraries• Basic functionality of python is limited• Libraries extend the functionality of python to various fields

(I/O of various formats, math/statistics, visualization, etc.)• Import syntax: import <library>• Sublibrary/Function import:

from <library> import <sublibrary/function>• Use syntax: <library>.<sublibrary/function>

LibrariesIn [1]: import osIn [2]: from os import listdirIn [3]: listdir("/")In [3]: ['root','etc','usr','bin', ... ,'srv','tmp','mnt']In [4]: import numpy as npIn [4]: np.sqrt(2)In [3]: 1.4142135623730951


Python Package Index

• Search for libraries on the web• Short description, install instructions and source files

https://pypi.org


https://pypi.org

Install with Anaconda Navigator• Anaconda Navigator can install libraries (→ Environment)• You can install multiple environments with different libraries


Important Libraries

os OS routines implementation in pythoncftime Implementation of date and time objects

numpy Fast general-purpose processing of multi-dimensionalarrays

scikit-learn Machine-learning routines in python

pandas Easy and intuitive handling of structured and timeseries data

netCDF4 I/O of netCDF filesmatplotlib Basic 2D visualization in python

cartopy Draw geospatial data in python


Introduction topython-numpy


Introduction to python-numpyFast general-purpose processing for large multidimensional arrays

• Implements a powerful N-dimensional array type(huge improvement over lists/tuples)

• Basic linear algebra, Fourier transform, and random numbercapabilities

• I/O of formated and unformated data• Based on C and FORTRAN77 routines in the background• Requirement for most scientific python libraries (matplotlib,

pandas, netCDF4, etc.)

Import NumpyIn [1]: import numpy as np


Numpy ndarray• Key element of numpy is the

new variable class:ndarray

Create ndarrayIn [2]: x = np.array([1,2,3])In [3]: type(x)Out [3]: numpy.ndarray

• Ndarrays implement a couple of new methods.ndim get number of dimensions

.shape get shape of array.size get total size of array

.reshape change shape of array.flatten make array flat

.swapaxes swap dimensions

Ndarray methodsIn [4]: y = np.array([[1,2,3],[4,5,6]])In [5]: y.shapeOut [5]: (2,3)In [6]: y.flatten()Out [6]: array([1,2,3,4,5,6])




















Numpy Functions• Numpy implements several array functions

arange numpy version of rangerepeat repeat n times

linspace vector from intervalrandom random numbers

zeros array filled with 0ones array filled with 1

meshgrid matrices from vectors

Numpy FunctionsIn [1]: np.linspace(1, 3, 5)Out [1]: array([1., 1.5, 2., 2.5, 3.])In [2]: np.random.randint(1, 100, 5)Out [2]: array([52, 75, 29, 52, 24])In [3]: np.zeros([3, 5])Out [3]:array([[0., 0., 0., 0., 0.],

[0., 0., 0., 0., 0.],[0., 0., 0., 0., 0.]])


Math Functions• Mathematical functions for elementwise evaluation

ExponentialIn [1]: np.exp([0, 1, np.log(2)])Out [1]: array([1. , 2.71828183, 2. ])In [2]: np.log([0, np.e, np.e**0.5])Out [2]: array([0. , 1. , 0.5])

exp and log are defined asnatural exponential andlogarithm (base e)

log is invers to exp

Trigonometric FunctionsIn [1]: x = np.array([0, np.pi, 0.5*np.pi])In [2]: np.sin(x)Out [2]: array([0., 0., 1.])In [3]: np.cos(x)Out [3]: array([1., -1., 0.])In [4]: np.tan(x)Out [4]: array([0., 0., 0.])

Further Functionsarcsin, arccos, arctan,deg2rad, rad2deg, sinh,cosh, tanh, arcsinh,arccosh, arctanh, sqrt,log2, log10, exp2, ...


Statistical Functions

• Numpy implements usualstatistical functions

• Implementation as function(np.mean) and arraymethod (x.mean)

mean: mean(x, axis = <axis>)sum: sum(x, axis = <axis>)

median: median(x, axis = <axis>)maximum: max(x, axis = <axis>)minimum: min(x, axis = <axis>)

<axis>: dimensions along to evaluate(int or tuple of ints)

Statistic Functions IIn [1]: x = np.random.random((4,2,8))In [2]: np.mean(x)Out [2]: 0.46376In [3]: x.sum(axis = (0,2))Out [3]: array([15.59966 , 14.08082])In [4]: np.median(x)Out [4]: 0.38988

Statistic Functions IIIn [5]: np.min(x, axis = 2)Out [5]: array([[0.0381, 0.2301],

[0.0220, 0.1045],[0.1903, 0.2746],[0.0539, 0.0203]])

In [6]: x.max()Out [6]: 0.9788


Statistical Functions• Specific percentile of a given array:

percentile(x, q = <percentile>, axis = <axis>)<percentile>: percentile in [0,100]

Statistic Functions IVIn [1]: x = np.random.normal(0,1,1000)In [2]: np.percentile(x, q = 15)Out [2]: -1.07467In [3]: np.percentile(x, q = 85)Out [3]: 1.04699In [4]: np.percentile(x, q = (2.5, 97.5))Out [4]: array([-1.85338831, 2.011201 ])


















What else is numpy capable of

• Logical functions:isfinite(x), isnan(x), equal(x, y),all(b, axis = <axis>), any(b, axis = <axis>), ...

• Various functions for linear algebra (ordinary matrixmultiplication, matrix decomposision, eigenvalues andeigenvectors, determinant, solve linear equations)

• I/O functions to read and write formated ascii or unformated(raw binary) files

• Draw random numbers from various distributions (uniform,Gaussian, binomial, Poisson, chi-square, ...)

• ...https://docs.scipy.org/doc/numpy/reference/routines.html


https://docs.scipy.org/doc/numpy/reference/routines.html

Introduction topython-netCDF4


Introduction to python-netCDF4

• Read and write netCDF4 files in python• Based on Unidata group netCDF4-C libraries• Uses python-numpy arrays to store data in python• We will cover only the read-functionality in this course

python-netCDF4In [1]: from netCDF4 import DatasetIn [2]: from cftime import num2date

Dataset Main object to read and write netCDF filesnum2date Contains functions to translate the dates


Read a netCDF Dataset• netCDF files can be accessed by: Dataset(<file name>)• New object type netCDF4._netCDF4.Dataset• Can access every detail of the netCDF file (dimensions,

variables, attributes)

python-netCDF4In [1]: from netCDF4 import DatasetIn [2]: nc = Dataset("some_netcdf_file_name.nc")

...

In [999]: nc.close()

• New object nc implements various object-methodsnc.<method> to access the netCDF file

• nc needs to be closed after everything is done: nc.close()


Access Global Attributes• Get list of all global attributes: nc.ncattrs()• Get value of specific attribute: nc.getncattr(”<attribute>”)

Access Global AttributesIn [3]: nc.ncattrs()In [3]:['institution','institute_id','experiment_id',...'cmor_version']In [4]: nc.getncattr("institution")In [4]: 'Max Planck Institute for Meteorology'In [5]: nc.getncattr("experiment")In [5]: 'RCP8.5'


Access Dimensions• Get a dictionary of all dimension: nc.dimensions

(not a function)

Access DimensionsIn [3]: nc.dimensions.keys()In [3]: odict_keys(['time', 'lat', 'lon', 'bnds'])In [4]: nc.dimensions["time"].nameIn [4]: 'time'In [5]: nc.dimensions["time"].sizeIn [5]: 1461In [6]: nc.dimensions["time"].isunlimitedIn [6]: True

nc.dimensions[”<dim>”].name name of <dim>nc.dimensions[”<dim>”].size size of <dim>

nc.dimensions[”<dim>”].isunlimited() True if <dim> is record(size of record dimensions (time) can increase unlimited)


Access Variables• Get a dictionary of all variables: nc.variables

(not a function)

Access VariablesIn [3]: nc.variables.keys()In [3]: odict_keys(['lon', 'lat', 'time', 'time_bnds', 'pr'])In [4]: nc.variables["pr"].ncattrs()In [4]:['standard_name','long_name','units',...'comment']In [5]: nc.variables["pr"].getncattr("standard_name")In [5]: 'precipitation_flux'


Access Variables

• Access data of given variable:nc.variables[”<variable name>”][:]

• Data is represented by a numpy-array

Access VariablesIn [3]: data = nc.variables["pr"][:]In [4]: type(data)In [4]: numpy.arrayIn [5]: data.mean()In [5]: 0.5545673


Access Time

• Time variable is usually saved as numerical value in givenunits and calendar

• Function num2date can be used to translate numerical valueto datetime

Access TimeIn [3]: time = nc.variables["time"][:]In [4]: type(time)In [4]: numpy.arrayIn [5]: timeIn [5]:array([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10.,

11., 12., 13., 14., 15., 16., 17., 18., 19., 20., 21.,...


Access Time• Function num2date can be used to translate numerical value

to datetime• Returns a numpy-array of datetime-objects (containing: year,

month, day, ...)

Convert TimeIn [3]: time = nc.variables["time"][:]In [4]: units = nc.variables["time"].unitsIn [5]: calendar = nc.variables["time"].calendarIn [5]: cftime.num2date(time, units = units, calendar = calendar)array([

cftime.datetime(1979, 1, 1, 0, 0),cftime.datetime(1979, 1, 2, 0, 0),cftime.datetime(1979, 1, 3, 0, 0),...

units: ’days since 1979-1-1 00:00:00’ calendar: ’standard’


Hands-On: Access andanalysis of netCDF Data

with Python


Python Libraries for DataVisualization


Introduction topython-matplotlib


Introduction to python-matplotlib• Library for 2D plotting in python• Originates in emulating MATLAB graphics commands• Produce nice looking plots fast and easy, but user still have

the power to change every detail (line properties, font, ticks,colors, etc.)

https://matplotlib.org


https://matplotlib.org

Basic Code LayoutImport plotting module from matplotlib:from matplotlib import pyplot

Creating plotting environment:fig = pyplot.figure( figsize = (4,4) )ax = fig.add_subplot(1,1,1)

<plotting and layout>

fig = pyplot.figure( figsize = (4,4) )

ax = fig.add_subplot(1,1,1)

Saving to file and closing plotting environment:fig.savefig(”<output filename>”)pyplot.close(fig)


Line Plot• Line plot y versus x (w/wo point markers):

ax.plot(x, y, color = <color>, ls = <ls>, lw = <lw>,marker = <marker>, ms = <ms>, ...)

Optional Parameters<color> color of plot

(string, hex, ...)<lw> line width

(float)<ls> line style

(”-”, ”–”, ...)<marker> marker style

(”x”, ”o”, ”.”, ...)<ms> marker size

(float),

Christoph Menz RD II: Climate Resilience 57

Bar Plot• Bar plot height versus x (w/wo point markers):

ax.bar(x, height, width = <width>, yerr = <yerr>,fc = <facecolor>, ec = <edgecolor>, ...)

Optional Parameters<width> width of each bar

(float or array)<yerr> optional error

(float or array)<facecolor> color of bar faces

(string, hex, ...)<edgecolor> color of bar edges

(string, hex, ...)


Histogram• Histogram plot of given values:

ax.hist(x, bins = <bins>, density = <density>,histtype = <type>, fc = <facecolor>, ec = ...)

Optional Parameters<bins> bins of histogram

(integer or vector)<density> count or density

(True/False)<type> type of histogram

(’bar’, ’barstacked’,’step’, ’stepfilled’)


Plot Layout

ax.set_xlim()

ax.set_ylim()

ax.set_xticks()ax.set_xticklabels()

ax.set_yticks()ax.set_yticklabels()

ax.set_ylabel()

ax.set_ylabel()

ax.set_title()


Plot Layout

ax.set_xlim()

ax.set_ylim()



ax.set_ylabel()

ax.set_ylabel()

ax.set_title()


Plot Layout

ax.set_xlim()

ax.set_ylim()



ax.set_ylabel()

ax.set_ylabel()

ax.set_title()


Plot Layout

ax.set_xlim()

ax.set_ylim()



ax.set_ylabel()

ax.set_ylabel()

ax.set_title()


Plot Layout

ax.set_xlim()

ax.set_ylim()



ax.set_ylabel()

ax.set_ylabel()

ax.set_title()


Hands-On: Visualizationof Scientific Data with

matplotlib


Introduction topython-cartopy


Mesh Plot• Plot a colored map with given longitude, latitude and data:ax.pcolormesh(lon, lat, data, cmap = <colormap>,

vmin = <vmin>, vmax = <vmax>, ...)

Optional Parameters<cmap> color definition

of the map(Colormap)

<vmin> minimum valuefor color bounds(float)

<vmax> maximum valuefor color bounds(float)


Introduction to python-cartopy

• Matplotlib can only plot raw data without referencingunderlying geographical information (no countries, no lakes,no projection, ...)

• Cartopy builds on matplotlib and implements advancedmapping features

• Developed by UK Met Office• Added features:

• Boundaries of continents, countries and states• Adding rivers and lakes to map• Adding content from shape file to map• Relate map to a projections and translate between different

projections


Basic Code LayoutImport matplotlib.pyplot and coordinatereference system (crs) from cartopy:from matplotlib import pyplotimport cartopy.crs as ccrsCreating figure environment:fig = pyplot.figure( figsize = (4,4) )Creating axes with specific map projection:proj_map = ccrs.Robinson()ax = fig.add_subplot(1,1,1, projection = proj_map)

Adding mesh plot with projection of given data:proj_data = ccrs.PlateCarree()ax.pcolormesh(lon, lat, data, cmap = cm.jet, transform = proj_data)


Projections: OverviewM

apP

roje

ctio

n

ccrs.PlateCarree() ccrs.Robinson() ccrs.Orthographic()

...

Dat

aTr

ansf

orm

atio

n Transformation between projections:

proj_cyl = ccrs.PlateCarree()proj_rot = ccrs.RotatedPole(77, 43)lon = [-170, 170, 170, -170, -170]lat = [-30, -30, 30, 30, -30]ax.fill(lon, lat, transform = proj_cyl)ax.fill(lon, lat, transform = proj_rot)


Adding Features to MapCartopy implements variousmap features

import cartopy.feature as cfeature

coastline = cfeature.COASTLINEborders = cfeature.BORDERSlakes = cfeature.LAKESrivers = cfeature.RIVERS

ax.add_feature(<feature>)

• Features in 3 different resolutions (110 m, 50 m and 10 m)from www.natrualearthdata.com

• External shapefiles can also be plotted


Colorbar• Add a colorbar to an existing map plot:

map = ax.pcolormesh(lon, lat, data)fig.colorbar(map, ax = <ax>, label = <label>,

orientation = <orientation>, ...)

Parameters<ax> parent axes

to add colorbar<label> label to add

to colorbar(string)

<orientation> colorbar orientation(”horizontal” or”vertical”)


Further Plotting Routines• Besides pcolormesh matplotlib/cartopy supports additional

plotting routines

Contour Plotax.contour(lon, lat, data)

Filled Contour Plotax.contourf(lon, lat, data)

Wind Vector Plotax.quiver(lon, lat, U, V)

U and V are zonal andmeridional wind components


Hands-On: Visualizationof Geospatial Data with

cartopy


Exercise: Analysis andVisualization of netCDF

Data with python


Exercise

1. Create a line plot showing the annual temperature anomalytimeseries of observation and GCM model simulation of theManila grid box. The anomaly is defined as the temperatureof each year minus the average of 1981 to 2000.Hints:

• Use read_single_data() to read the data from file.• Use ilon = 12; ilat = 19 as coordinates of Manila.• Select the timeframe 1981 to 2000 using get_yindex().• Calculate the average using either np.mean() function ordata.mean() method.

• Use create_lineplot() and save_plot() to create andsave the plot.


Exercise

2. Create a map plot of the GCM temperature bias (for theperiod 1981 to 2000). Here the bias is defined as thedifference of the long term averages (1981 to 2000) betweenGCM simulation and observation (GCM minus observation).Hints:

• Use read_single_data() to read the data from file.• Select the timeframe 1981 to 2000 using get_yindex().• Use create_mapplot() and save_plot() to create and save

the plot.


Capacity Building Workshop - Python for Climate Data Analysis

Documents