Sunset on Mars
A Mars rover is a motor vehicle designed to travel on the surface of Mars. Rovers have several advantages over stationary landers: they examine more territory, they can be directed to interesting features, they can place themselves in sunny positions to weather winter months, and they can advance the knowledge of how to perform very remote robotic vehicle control.Curiosity landed in the crater Gale on planet Mars. The landing site coordinates are: 4.5895°S 137.4417°E. The location was named Bradbury Landing on 22 August 2012, in honor of science fiction author Ray Bradbury. Gale, an estimated 3.5 to 3.8 billion-year-old impact crater, is hypothesized to have first been gradually filled in by sediments; first water-deposited, and then wind-deposited, possibly until it was completely covered.
Location of Gale Crater
Curiosity has a lot of instruments onboard. One of them, REMS (Rover Environmental Monitoring Station) measures and provides daily and seasonal reports on atmospheric pressure, humidity, ultraviolet radiation at the Martian surface, air temperature, and ground temperature around the rover. REMS was develeoped in Spain by the Centro de Astrobología (CAB/CSIC-INTA) in collaboration with NASA and JPL-Caltech.
REMS installation. Credit NASA/JPL-Caltech/CAB
The data contained in this project represents the weather conditions on Mars from Sol 1 (August 7, 2012 on Earth) to Sol 1895 (February 27, 2018 on Earth). Sol is equivalent to 1 Martian day (1 Martian day = 24h 40 min).
However, REMS does not take measurements continuously and it takes measurements at different times from one day to another. For different reasons (instrument maintenance, instrument calibration, instrument degradation, etc.), some or all of the magnitudes in this project were not be available.
The main questions I considered are the following:
- How is the weather on Mars?
- How it compares with its twin location on Earth?
- Is it possible to predict the weather with the missing data?
- BONUS: can I obtain pictures from Mars and complement the data?
The base data was extracted from Kaggle, "Mars weather data" by Kannan.K.R. Source: https://www.kaggle.com/datasets/imkrkannan/mars-weather-data
The data was complemented with weather data from Papua New Guinea (twin location of Curiosity on Earth) available on NOAA Global Surface Summary of the Day services, in the same range of dates as provided by Kannan.K.R. Source: https://www.ncei.noaa.gov/access/search/data-search/global-summary-of-the-day
Data prediction and current weather information were extracted using Selenium from the CAB-CSIC/INTA webpage. Source: http://cab.inta-csic.es/rems//
- src/ --> Contains the executable python files of the cleaning and visualization process.
- images/ --> All the necessary pictures
- data/ --> Contains an edited version of the original dataframe
- README --> What you are reading right now.
The first step was to define all the necessary functions to clean my datasets. For example, a function to remove columns would be:
def remove_columns(df, column_name):
"""
This is a function that removes undesired columns. Requires two arguments.
Arguments: dataframe, column name
Input: the current dataframe
Output: the current dataframe without the selected columns
"""
df.drop(columns=f"{column_name}", inplace=True)
return df.sample(2)However, the most important functions are those needed to extract information from the NASA API (exaplained later). These functions are:
def call_Curiosity (date, camera):
"""
This is a function that calls NASA API 'Mars Rover Photos' with two arguments. It returns the url from
a specific camera onboard Curiosity rover.
date: input the desired date in the format YYYY-MM-DD as a STRING,
camera: select between FHAZ, RHAZ, MAST, CHEMCAM, MAHLI, MARDI, NAVCAM, PANCAM, MINITES, as STRING
"""
try:
nasa = os.getenv("token")
url = f"https://api.nasa.gov/mars-photos/api/v1/rovers/curiosity/photos?earth_date={date}&camera={camera}&api_key={nasa}"
request = requests.get(url)
df = pd.DataFrame(request.json())
df_clean = pd.DataFrame(df.values[0][0])
image_url = list(df_clean["img_src"])[0]
display(Image(image_url, width=300, height=200))
return f"Image available for camera {camera} onboard Curiosity rover"
except:
return f"No image available on {date} for camera {camera} onboard Curiosity rover, please select another date"def get_pictures_Curiosity(date):
"""
This is a function that calls call_NASA function with one argument. It returns the url of all the pictures
taken by all the cameras of Curiosity rover from a specific Sol date.
date: input the desired date in the format YYYY-MM-DD as STRING.
"""
cameralist = ["FHAZ", "RHAZ", "MAST", "CHEMCAM", "MAHLI", "MARDI", "NAVCAM", "PANCAM", "MINITES"]
for i in cameralist:
print(call_Curiosity(date, i))
passThe objetive is to extract pictures from Mars given a determinate date.
All functions can be found at Functions.py file.
The second step was to actually clean the databases:
-
Kaggle database:
In the beginning the dataframe looked like this:
With a total shape of 1894 rows x 10 columns. The table contained unnecessary columns and a lot of NaNs. In addition, column titles could be more readable and I also wanted to include more information in the form of new columns.
The cleaning steps were the following:- Remove undesired columns: used "columns_to_remove" custom function
- Created a new column: "Mean_temp"
- Cleaned the atmosphere column: used "clean_atmosphere" custom function
- Renamed column names: used "rename_columns" custom function
- Cleaned the month column: used "clean_month" custom function
- Created a new column: "Season", by importing Month values
Data cleaned:
-
NOAA database:
In the beginning the dataframe looked like this:
The cleaning steps were the following:
- Remove undesired columns: used "columns_to_remove" custom function
- Renamed column names: used "rename_columns" custom function
- Converted temperatures values from Fahrenheit to Celsius: used "FtoC" custom function
- Converted from mBar to Pascals: used "mbartoPa" custom function
- Remove 99999s from Pressure column
- Rounded values of the mean_temp column
Data cleaned:
Both dataset had to be complemented with additional information taken directly from webpages.
This information is crucial for the last phase of the project, "Mars Today", in order to predict the weather of Mars beyond the limit date of the Kaggle dataframe.
The scrapping was performed on the REMS widget from the CAB-REMS webpage, http://cab.inta-csic.es/rems/es/ using Selenium.
Similar operations were performed with the database originated.
And the final database:
At this point, one might think why was import to extract data from NOAA. The data itself from Mars doesn't give many information unless it's compared with the conditions on Earth.
Thus, the coordinates from the Gale crater were extrapolated on Earth. The twin location on our planet is next to Papua New Guinea (5º south from the equator)
Info about temperature and pressure were related an visualized using Plotly.
Please remember, the third planet from the Sun have a very thin atmosphere, approximately 100 times less dense than ours. This fact has direct implications on the surface temperature: without magnetic field or atmosphere to thermoregulate and distribute heat, the planet has to deal of temperatures from -30 ºC to -60 ºC. Meanwhile on Earth, the range of temperatures is much warmer and less prominent as on Mars.
Take also a closer look to the subtle shift from Papua New Guinea. Since 2016, the temperature baseline has been displaced upwards! Could be an indicative of an anomaly, such as climate change. On the other hand, the temperature pattern on Mars remains constant, with very little deviations from one year to another. This can give us the key to predict future values beyond 2018.
It is impossible to not realize the weird pattern on Mars, which has no common points with Earth's. Why so?
According to J.A.Rodríguez-Manfredi et al. on their paper published in JGR Planets, "Mars Surface Pressure Oscillations as Precursors of Large Dust Storms Reaching Gale" (DOI: https://doi.org/10.1029/2021JE007005), Martian dust storms strongly interfere with global circulation patterns and change the diurnal and semidiurnal pressure variability as well as oscillations with periods greater than one sol associated with planetary waves. The specific pressure oscillations preceding each storm period are likely to be signatures of the large-scale circulation patterns that enable the growth and propagation of the storm fronts.
Another paper from N. Rennó et al., titled "Pressure observations by the Curiosity rover: Initial results" (DOI: https://doi.org/10.1002/2013JE004423) suggest that it is possible to witness atmospheric features at various spatial and temporal scales, e.g., the gradually increasing pressure due to the advancing Martian season, diurnal tides, thermal vortices, and other local atmospheric phenomena.
Pressure and Temperature are strictly related, being directly proportional: the more temperature, the more pressure. That is why when meteorologists anounce anticyclones (regions of the atmosphera with high of high pressure) the temperatures increase.
Some kind of direct proportion can be seen on Earth, with all the values distributed along the X and Y axis. However, this phenomena cannot be seen on Mars, leaving a "hole" where certain values of Pressure and Temperature can not coexist.
Solar heating on the day side and radiative cooling on the night side of a planet can induce pressure difference. Thermal tides, which are the wind circulation and waves driven by such a daily-varying pressure field, can explain a lot of variability of the Martian atmosphere. Compared to Earth's atmosphere, thermal tides have a larger influence on the Martian atmosphere because of the stronger diurnal temperature contrast. The surface pressure measured by Mars rovers showed clear signals of thermal tides, although the variation also depends on the shape of the planet's surface and the amount of suspended dust in the atmosphere.The atmospheric waves can also travel vertically and affect the temperature and water-ice content in the middle atmosphere of Mars.
Mars and the distribution of dust storms on its Surface.
Mars before and after a dust storm
As seen before, the temperature and pressure profiles, although being very abrupt, are very stable. This can allow anyone to extrapolate the values and predict the data.
Disclaimer: REMS is currently active and has been reporting weather data until the present day. Newer data can be obtained from the Planetary Data System from NASA (https://pds.nasa.gov/). The purpose of this process is to practice new skills.
MarsToday() function works as follows: an user inputs a given date beyond 2018-02-28 in the format YYYY-MM-DD. MarsToday calls the scrapped database on CAB/REMS "widget" and the initial database "mars", looks for the date on both places and retrieves all the measurements made in years before, but same month and day. Then, the data is stored in a dataframe with Pandas and a series of basic statistics are performed: it takes all values and shows to average and the standard deviation. The values are the extrapolation of the weather given a future date.
Once the prediction is made, MarsToday calls "widget" database again and shows the information of the given information.
By comparing the two outputs, the user can realize if the approximation was good enough or not.
As a bonus (and for a cooler effect), the minitool calls NASI API by the two functions "get_pictures_Curiosity" and "call_NASA" and looks for all images taken from any camera onboard Curiosity rover in the selected date.
Here is an example of a call to MarsToday function:
Mars has shown us the effects of a planet with no atmosphere and high amounts of CO2 in its atmosphere. Temperature and pressure variates significantly due to the harsh atmospheric changes, including large heat waves and dust storms. Earth shows milder temperatures thanks to the atmospheric density and magnetosfere.
It is possible to predict the weather on Mars with pretty good results using basics statistics and rough methods. Nevertheless, with a good tool such as Machine Learning it could be possible to predict those values with more accuracy.
As a personal opinion, this project has been a major challenge and I had to use resources never explored before, but that have had an incredible great impact on my behalf.
- https://api.nasa.gov/
- http://cab.inta-csic.es/rems/es/
- https://mars.nasa.gov/msl/weather/
- https://pds.nasa.gov/
- https://www.ncei.noaa.gov/access/search/data-search/global-summary-of-the-day
- https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2021JE007005
- https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2013JE004423