Project Instructions
User Manual:
Open the PDF directly: View PDF 
.
Page Count: 10
| Download | |
| Open PDF In Browser | View PDF |
Project Instructions
1
Introduction
The Swedish Meteorological and Hydrological Institute (SMHI) routinely
records the temperature at various locations in Sweden. In some places, this
has been going on for hundreds of years.
In 2013, the institute published
its oldest dataset with average daily temperatures of Uppsala starting from
January 12:th 1722 and ending in 2013.
More recent datasets for other
locations are also available as part of the SMHI OpenData initiative.
A
collection of SMHI datasets are made available on the course webpage. There
is clearly a wealth of information hiding in these datasets! Use your skills as a
programmer and as a scientist in order to make some interesting observations
regarding the Swedish climate!
2
Project description
Select one of the available SMHI datasets and write a program that extracts
information from the Swedish climate data. Use plots in order to visualize
your results. Use ROOT to make plots and eventual statistical analysis: a
small code skeleton (see Section 4) is provided to get you started. Section 3
contains a few examples of what kind of information you could get from the
data. Decide on at least three interesting results to produce. At least one of
them should be your own idea and should not come from Section 3. Use your
imagination! Start by looking at the data and understanding the content of
the les and the format, pay attention to descriptive metadata. Note that
the data within one dataset can be of dierent quality, so you might consider
skipping "bad" records.
In the real world, scientists share code and collaborate. In this project,
you will work together with your colleagues as a team. All team members
must have a
github.com account (the same we used for the course) and their
contribution must be seen as authorship on github commits. Each group will
appoint one team member as the release manager. The release managers will
1
create a repository on their
github.com account and de ne the policies with
which the other team members will add their code to the repository.
One policy (let's call it
A
) could follow the way we did during the course,
where Florido was the release manager and all students would fork its up-
stream repository into their own remote origin master and send pull requests
to release manager's remote upstream. In this way the release manager is the
only one to decide what changes go in the main upstream repository. This
is the best way to keep the code and changes clean.
Alternatively (policy
B
), the release manager can add all the team mem-
bers as collaborators to the created repository, and everyone can clone, pull
and push to the same main repository as their remote origin. In this case
all the team members are responsible to keep the code tidy, and the release
manager must be able to rollback changes when these are not acceptable.
This can be complicated if there is no deep understanding of git by all team
members, so use it only if you know what you're doing.
Either way, make sure you create a branch for each big change or contribution you make, and then ask the release manager to merge the branch to
the upstream master (policy
A
), or do the merge yourself in your working
directory before pushing to the remote origin (policy
B
). It might be helpful
to repeat the interactive tutorial we did during the course to understand
these concepts.
Add your code, draft reports and other notes in the
github.com
reposi-
tory. It is better if those are in separate folders.
We will only accept projects that can be checked by reaching the respective
github URL. You should provide us with such URL1 . For example,
Florido's repository for the course had such URL:
https://github.com/floridop/MNXB01-2018
By using
git you and your team members can work independently, com-
mitting the changes to their own fork repository as you go. When working on
this kind of project, it is customary to keep a log. The
contain a le called
ChangeLog,
git repository should
where you document your progress. Add a
date and a short descriptive comment to this log for each major change that
you make to the code.
The examiners should be able to understand how
did the code develop by reading such a log. As a rst step of the project
work, every team should prepare a short
Workplan document containing the
planned distribution of sub-tasks that have to be done in order to achieve the
goal. Discuss and agree with your colleagues early on what sort of code needs
1
Report it by e-mail to Oxana, mention your group letter
2
to be written and agree preliminarily on who should write what in order to
avoid unnecessary collisions. Record the task distribution in the
document in the
git repository and,
Workplan
if necessary, update it regularly. If you
get stuck or if you realize that the code needs to be restructured, talk to
your colleagues. This is a team project!
Naturally, you need to document not only your code but also your re-
AT X to write a scienti c report that describes the results that
sults. Use L
E
you managed to produce and the approaches that you used, such as short
description of classes, functions, methods etc. The report should be complete
with plots to back up your claims. Scientists collaborate on their reports in
very much the same way as they collaborate on their code. Begin by splitting
the report into many
.tex
.tex
les that you
\input
from a main le. Use one
le for each result that you decided to produce. Then put all of the
report code in
git.
Even for very big reports, if everyone is working just on
their own part of the report using the appropriate
.tex
le, problems with
collisions are rare.
3
Example results
This section contains a few examples of what you could do with the data.
Use them as guidelines and as inspiration. You don't have to do everything
the examples say, but more complete implementations earn you both brownie
points and (probably) a better grade. Remember to comment your code!
3.1
The temperature of a given day
Use the data to make a histogram of the temperature of a given day of the
year. An example of such a histogram is shown in Figure 1. If you choose
to implement this example, uncomment either of the two functions called
tempOnDay.
One of the functions accepts a month and a day as arguments.
The other one accepts a date (which is an integer in the range 1 to 365, or
366 if you design your code to understand leap years).
Implement one or
both. If you have one, the other should be simple!
•
Create a histogram like the one shown in Figure 1.
•
What is the mean temperature of the given day?
•
What is the standard deviation of the temperature?
•
Can you predict the probability of observing a particular temperature?
3
Entries
14
Temperature on 23/8
12
10
8
6
4
2
0
-20
-10
0
10
20
30
40
Temperature [°C]
Figure 1: Histogram showing the temperature on August 23:rd every year
since 1722.
Hints
In ROOT, you often write functions that create one or more plots.
If you want to look at those plots after the function completes, you have to
put the histograms on the heap.
If you put them on the stack, they will
go out of scope and be deleted automatically once the function completes.
Once a histogram is deleted it disappears from all plots. This can lead to
a situation where you create histograms on the heap that you never delete.
Technically, this ts the description of a memory leak. It's a design aw of
ROOT that it encourages memory leaks in this way. But since you will only
be creating a handful of histograms for plotting, you shouldn't worry about
it. A handful of histograms won't cause the program to run out of memory.
The example code below shows how to create a histogram of integers with 365
bins between 1 and 366. The title of the axes can be speci ed directly in the
constructor using the trick shown. We then set the ll color to a darker red,
increment a bin and calculate some properties of the distribution. Finally,
we create a new canvas and draw the histogram.
4
TH1I* hist = new TH1I("temperature", "Temperature;Temperature
[#circC];Entries", 300, -20, 40);
hist->SetFillColor(kRed + 1);
hist->Fill(-3.2); //Increment the bin corresponding to -3.2 C
double mean = hist->GetMean(); //The mean of the distribution
double stdev = hist->GetRMS(); //The standard deviation
TCanvas* can = new TCanvas();
hist->Draw();
3.2
The temperature for every day of the year
If we can create a histogram showing the temperature of one day, why not
do it for every day of the year? Uncomment the
tempPerDay
function and
try it out! Figure 2 shows how the mean temperature varies throughout the
year. Of course, just knowing the mean is not all that interesting. In order
to make useful predictions, we also need to know the standard deviation of
the temperature of each day.
•
Plot the mean temperature of each day of the year.
•
Use error bars in order to visualize the standard deviation.
Hints
The example code shows how to loop over every bin in a histogram
and explicitly set the contents and error of each one. In ROOT, bin 0 is the
under ow bin. This bin is not shown on plots. Bins 1 to
see when drawing the histogram. Finally, bin
nBins + 1
nBins are what you
is the over ow bin.
for(int bin = 1; bin <= hist->GetNbinsX(); ++bin) {
hist->SetBinContent(bin, 5);
hist->SetBinError(bin, sqrt(5.));
}
3.3
The warmest and coldest day of each year
In Sweden, summers and cold and winters are colder.
actually the coldest? Implement the
But which day is
hotCold function and nd out!
Figure 3
shows when the warmest and coldest days have typically occurred during
the last few hundred years. To make things more interesting, note that each
dataset may contain readings from dierent places (for example, the Uppsala
5
Temperature [°C]
20
15
10
5
0
-5
-10
50
100
150
200
250
300
350
Day of year
Figure 2: Histogram showing the mean temperature on each day of the year.
dataset has a few readings from Stockholm). If you use the Uppsala dataset,
write your function such that all readings from a region other than Uppsala
are ignored.
•
Create histograms of the warmest and coldest day each year.
•
Predict when the warmest and coldest day is most likely to occur.
•
Can you show both histograms in the same plot?
•
Can you make a t function that wraps around as in Figure 3, or
suggest a data presentation that is easier to t?
Hints
The mean of a histogram could be obtained with the
GetMean mem-
ber function. A more fancy, and sometimes more useful, way of extracting
information from a distribution is to t a mathematical model to it. The example code shows how to de ne a custom function (a Gaussian in this case)
and then t that function to a distribution contained in a histogram. The
function is de ned in the range
[1, 366] and takes three parameters.
6
They are
Entries
9
Warmest day
8
Coldest day
7
6
5
4
3
2
1
0
50
100
150
200
250
300
350
Day of year
Figure 3: The plot shows how often a given day of the year was the warmest
or coldest for every year since 1722. Lines show Gaussian ts of the distributions.
(as de ned in the
Gaussian
function) an amplitude, a mean and a standard
deviation. The meaning of the parameters passed to the
Fit
function is to
t quietly (Q), to not automatically plot the function when the histogram
is plotted (0) and to t only within the range in which the function is de ned (R). After tting, we print the mean of the Gaussian as well as the
uncertainty of that mean. Finally, the example shows how to create a legend
containing two histograms and then draw those histograms in the same plot.
double Gaussian(double* x, double* par) { //A custom function
return par[0]*exp(-0.5*(x[0]*x[0] - 2*x[0]*par[1] +
par[1]*par[1])/(par[2]*par[2]));
}
TF1* func = new TF1("Gaussian", Gaussian, 1, 366, 3);
func->SetParameters(5, 200, 50); //Starting values for fitting
hist->Fit(func, "Q0R");
cout << "The mean is " << func->GetParameter(1) << endl;
7
cout << "Its uncertainty is " << func->GetParError(1) << endl;
TLegend *leg = new TLegend(0.65, 0.75, 0.92, 0.92, "", "NDC");
leg->SetFillStyle(0); //Hollow fill (transparent)
leg->SetBorderSize(0); //Get rid of the border
leg->AddEntry(hist, "", "F"); //Use object title, draw fill
leg->AddEntry(anotherHist, "A title", "F"); //Use custom title
hist->Draw();
anotherHist->Draw("SAME"); //Draw on top of the existing plot
leg->Draw(); //Legends are automatically drawn with "SAME"
3.4
The mean temperature of each year
SMHI has produced some very nice plots that show the mean temperature of
each year since the temperature measurement started. An example is shown
in Figure 4 (a). Can we make something similar in ROOT? Of course we can.
Figure 4 (b) shows one such attempt. Try to make your own plot with the
tempPerYear function.
The function accepts one argument. This is a future
(or past) year for which to predict the temperature using extrapolation.
•
Plot the mean temperature of each year in your dataset.
•
Fit the data with a model of your choice.
Use the model to predict
the temperature of a given year in the future.
Can you draw any
conclusions when it comes to global warming?
•
Can you plot the deviation from the average as in Figure 4?
•
Can you plot the moving averages?
Hints
The neat deviation-from-the-average eect is actually obtained by
plotting several histograms in dierent colors on top of each other. Try it!
A moving average can be done using, for example, a graph. Add one point
to the graph for every few years in the histogram. The
y -value
of the point
should be the average of the surrounding years. Then plot the graph with
a smooth line between the points. The example shows how to create a new
graph and ll it with one point for each bin in a histogram.
Expand
The call to
increases the capacity of the graph by 100 if adding a new point
would make the graph run out of space. Finally, the example shows how to
draw the graph (on top of the current plot) with a smooth curve.
8
Temperature [°C]
(a)
9
8
7
6
5
4
3
2
1750
1800
1850
1900
1950
2000
Year
(b)
Figure 4: The plots show the mean temperature of each year since 1722.
Positive and negative deviations from the mean are indicated by dierent
colors. Lines show moving averages. Both a plot from SMHI (a) and one
from ROOT (b) are shown.
TGraph* graph = new TGraph();
for(int bin = 1; bin < hist->GetNbinsX(); ++bin) {
graph->Expand(graph->GetN() + 1, 100);
graph->SetPoint(graph->GetN(), hist->GetBinCenter(bin),
hist->GetBinContent(bin));
}
graph->Draw("SAME C");
9
4
Code skeleton
rootlogon.C. If ROOT
rootlogon.C, it will execute all
The code skeleton contains a le called
is started
from a directory that contains a
the state-
ments in the
rootlogon
function automatically.
This is a convenient way
to set up the environment, e.g. by executing some style scripts that determine how plots should look. The provided
changes to the default plots.
called
tempTrender
rootlogon.C
makes some basic
The code skeleton also comes with a class
project.cpp. These les are comrootlogon function. As you can see
and a script called
piled and loaded automatically by the
by looking at the les, they are almost empty. It is your job to implement
the
tempTrender
class and any other classes or functions that you might
need. The datasets from SMHI are in the directory called
datasets.
Each
group should pick a dataset (or a couple) to work with. Good luck with the
project!
10
Source Exif Data:
File Type : PDF File Type Extension : pdf MIME Type : application/pdf PDF Version : 1.5 Linearized : No Page Count : 10 Page Mode : UseOutlines Author : Title : Subject : Creator : LaTeX with hyperref package Producer : pdfTeX-1.40.18 Create Date : 2018:10:23 19:21:23+02:00 Modify Date : 2018:10:23 19:21:23+02:00 Trapped : False PTEX Fullbanner : This is MiKTeX-pdfTeX 2.9.6354 (1.40.18)EXIF Metadata provided by EXIF.tools