Mr.

Hanna’s IB Math Analysis IA -

An analysis of the August 21st, 2017 Solar Eclipse

Intro:

On August 21st, 2017 a total solar eclipse occurred,

starting on the central coast of Oregon in the United States. I
happen to live in Newport, Oregon and had the opportunity to
witness the entire event first hand. I knew it would be exciting to
watch, but I was truly blown away by the absence of light and the
drop in temperature that I experienced.

https://www.statesmanjournal.com/videos/news/local/eclipse/2017

Not wanting to miss an opportunity to do some math and science with the event, I set up and
recorded both temperature and light intensity over the course of about two hours so that I could
analyze the data at a future time. I used two different temperature measuring devices; one was a
Vernier Temperature probe and the other was an Arduino powered thermistor. I also used an Arduino
to record the light intensity on a light dependent resistor and recorded it as a percentage of the
maximum intensity it could measure. My purpose in this paper is to investigate how the temperature
was altered because of the eclipse.

Math Background

I investigated the typical rise in

temperature over a typical day in Newport,
Oregon for several times of the year and under
various weather conditions.(***include a citation)
I found that there is a reliable linear increase in
the temperature from about 9am until about 1pm
for almost any day of the year, and particularly
during the summer months. (Consider putting in
an equation to model this!)

Temperature data from timeanddate.com for Dec 3rd, 2020

On the morning of the eclipse however,

the temperature dropped while the sun was
hidden behind the moon and it took awhile for the
temperature to recover. I anticipated that the
picture in my mind would have looked like the
sketch at right; a roughly linear increase until the eclipse followed by a dip in the temperature and then
a return to the increase.
I decided to model the drop in temperature mathematically from the data I collected and then
compare what actually happened to what would typically have happened to the temperature rise.
As I looked at the raw data it appeared like the dip in temperature could be modeled by a
domain-restricted trigonometric function. I decided to use a cosine model since the temperature
started at a peak, dipped to a minimum and then rose once again. Unfortunately, I don’t have enough
data to see if the temperature was going to completely recover or not. (Look to see if I can find the
actual temperature data for the day).

At right is a graph of the raw data that I

collected with my two different temperature
sensors. I started collecting data at 9:05 am
and I took data points every 2 seconds for
7208 seconds, or just over two hours.
Although the data is a bit jumpy because of the
sensitivity of the equipment, it is easy to see a
general trend where the temperature fell and
then rose again after the eclipse passed. The
blue line is the data from the Arduino sensor
and the red line is from the Vernier
Temperature probe. I was pleased that
although they were slightly offset from each
other, which could have indicated a difference Temperature(Celsius) vs Time(seconds) for August 21, 2017, Newport,
OR
in their zero value (systematic error), they both
tracked the rise and fall of temperature variation fairly closely.

In order to come up with a mathematical model for my data, I needed to choose a couple of
representative data points that would allow me to compute the parameters I need for my equation. As
mentioned, I am choosing to use a cosine model of the form:

T (t)= Acos (B (t−C))+ D

where T is the temperature in degrees Celsius,

t is the time in seconds from the start of the
data collection, and A, B, C, and D are
parameters that I need to calculate. I marked
the coordinates that I chose from the blue
Arduino sensor line as P, Q, and R and labeled
and marked their values on the graph at right.
I chose to use the blue Arduino line since it
appeared to have less sensitivity to small
fluctuations. This doesn’t mean that I know whether it is more precise, just that there was less noise in
the signal.

Showing the Math Section

The first order of business is to determine the parameter A in my model. The amplitude can be
found by calculating ½ the total height of the curve. A is related to the amplitude of a cosine or sine
function in the following way: Amplitude = |A|.
Since my curve starts and ends at a maximum value, I do not need to change the sign of A to a
negative, hence I will drop the absolute value.

My next calculation is for the parameter B, which is related to the period of the function
according to the equation:

I can calculate the period by determining the difference in time betweens points P and R.

R−P=6935 s−1400 s=5535 s

This is the period of the temperature change, which yields the following for B:

The units for the parameter B, are Hertz (Hz), as it is s-1.

The parameter C gives the horizontal shift for the cosine function. In this case I simply write
down that C = 1400, since the peak is shifted horizontally from the Temperature axis by 1400 seconds.
The last parameter, D, is the value of the midline of the graph, which is simply the average of
the maximum and minimum values.

Now that I have each of the parameters, my equation can be written as:

This equation is only valid, however, within the domain restriction of { t | 1400s ≤ t
≤ 6935s}. Applying my equation to anything outside of these t-values is not valid
because it would be outside the time frame of the eclipse.
 I now need to graph this equation over the top of my data and see how well it fits. For ease of
viewing I eliminated the red graph and superimposed my curve, in black, on top of the actual data
below. My graphing software would not let me easily restrict the domain of my function, so I
superimposed red brackets to indicate the limits instead.

Analysis of my model

As I look at the graph above several things stand out to me.

First, the maximum and minimum look to be generally well aligned with the data, at least within
the domain restriction I used. I am also pleased that the minimum matches up well horizontally with
the data. In general, there is about as much of the data above my line as below my line, meaning that
my curve is finding a good balance between being too high and too low, at least in the region where
the domain restriction applies.

I accessed the temperature records for the local Hatfield Marine science Center Weather
Station for the date in question so that I could compare my results to an official record. When I plotted
the data however, I discovered that since they only record the temperature every 5 minutes, they were
not able to catch the significant drop in temperature during the eclipse. There was also intermittent fog
over the bay leading up to the eclipse, which also may have prevented their temperature sensor from
detecting the drop due to the eclipse.
 Further Math analysis

I was a bit disappointed that I was not

able to gather more official data from the day of
the eclipse, but I wanted to model what I think
should or could have happened anyway. As I
considered the temperature as a function of time
according to my earlier prediction (see sketch at
right) I decided to develop an equation to model
it.
A piecewise function seems to me to be
the most practical approach. I will first come up
with an equation for the linear increase in temperature and then connect that to a cosine function and
then end with another linear function.
The first piece of the function will have a domain restriction that starts at zero
(corresponding to 9:00 am) and ends when the eclipse begins. Specifically: { t | 0s ≤ t ≤
2700s} the second piece will have the following domain: { t | 2700s ≤ t ≤ 5500s} and
the third { t | 5500s ≤ t ≤ 7200s}. (check the limits of the domain!!)

I will use actual temperature data from the morning of August 21st before the eclipse and then
from several hours after the eclipse to create my function. I looked up the temperature recordings for
the day and times in question from the HMSC weather station archives.
(http://weather.hmsc.oregonstate.edu/weather/weatherproject/archive/2017/HMSC_201708.dat)
The following summarizes the information:
9:00 am: Temperature = 57.1oF = 13.9oC
1:00 pm: Temperature = 65.4oF = 18.6oC
I needed to convert these temperatures to Celsius according to the equation:
C = (F - 32)*(5/9),
where C is Celsius temperature and F is the corresponding Fahrenheit temperature.

For the first part of the piecewise function I will create the linear equation using the following
data points:
(0s, 13.9oC) and (14400s, 18.6oC) where 0s corresponds to 9:00 am and 14400s corresponds
to 1:00 pm.

The slope of my line will be given by:

18.6 C−13.9 C −1❑
m= =0.000326 Cs ❑ .
14400 s

The equation then becomes: T =0.000326 t +13.9. I will use this equation for sections 1 and 3
of my piecewise function.
 The middle function will include the cosine term. I used the same amplitude and period for the
function I created in the first part of this paper. I needed to change the horizontal shift to match the
new domain restriction I was using and also the vertical shift because the graph continued upwards
instead of falling off. My new equation for this section is:
T =0.00326 t +13.9+3.83 cos (0.00106 t−1.25)+ 0.350

I chose not to simplify the equation so that it would be easier to see the two parts; the linear increase
and the cosine perturbation caused by the eclipse. I need to check that the endpoints of each
piecewise section allow the function to be continuous.

Model for the temperature drop for the Newport, OR August 17th 2017 Solar Eclipse (from 9:00-11:00am)

(Analysis of my model and comparison to actual data)

The data provided by Hatfield Marine Science Center was not frequent enough to catch the
temperature drop by the eclipse in an obvious or meaningful way. This means that I don’t have
anything to compare my actual model against.

Conclusions:

My model has a few inherent limitations.

First, the actual data was more sensitive to small temperature fluctuations. The precision of my
measurement devices allowed for these measurements, while my model does not. That does not
necessarily mean the model is incorrect, but rather that it assumes those small fluctuations are
insignificant with respect to the broader trend of the data.
The connecting points between the linear functions and the middle cosine function are rather
abrupt. In real life these transitions would be smoother. Our class has not learned about calculus
however, and so I didn’t choose to make the slopes of the transitions match. The mathematical model
could be improved by including this analysis.
 **Investigate when the next eclipse will take place and make a plan for gathering accurate
data.**