Volume2, Issue 9, September 2024 International Journal of Modern Science and Research Technology

ISSN No: 2584-2706

Predicting Variation in Daily Martian

Atmospheric Temperature Using Machine
Learning Techniques.
Gabriel Tosin Ayodele
Faculty of Engineering and Informatics,
Department of Computer Science,
University of Bradford

Abstract
According to research and explorations other six models used in this work with
conducted on mars. It was believed that the scores as follows: RMSE=1.5525,
planet once supported life. This resulted MSE=2.4105, MAE=1.1939 and
from some evidence indicating the presence R2=0.8283. Finally, the scores obtained
of water and a thick warm atmosphere on were visualised.
mars in the past. Due to this, man became
curious to explore the planet as the red Keywords: Mars, space exploration,
planet is believed to be similar to earth. Models, Python, NASA, EMIRS, EMM,
Many technological tools have evolved evaluation, pre-processing .
with advancements in science and
technology, making the entire exploration Introduction:
mission easy. Machine learning standout to Scientists oldest guess is that life existed on
be an essential technical tool today that is in this planet because of its nature resembling
use for mars exploration. Although the field Earth (Chang K., 2013). In a quest for
is new, little work has been done in knowledge, mankind has used diverse space
exploring mars' weather patterns for exploration technology to conduct research
possible habitation of life. This work relied on other planets to determine the presence
on the strength of machine learning or otherwise of life. To this exploration,
techniques to analyse the mars data planet mars are deemed to be a planet that
obtained from the EMM EMIRS instrument has potential life. Planet mars can in fact
database. In this work, various techniques viewed directly from the earth as it is
such as exploration, pre-processing, model colored red hence being referred to as the
development, and evaluations were Red Planet (Charles et al. al 2022). The
performed on the dataset to predict the phenomenon which makes Mars an
variation in daily atmospheric temperatures appearance of reddish hue is due to the
of mars present within the dataset. Seven existence of oxidized iron on the surface of
machine learning models were developed in the planet. However, current programs like
this work: Random Forest, Linear sending rovers to mars and getting the data
regression, SVM, ANN, CNN, Lasso about the Martian weather are effective but
Regression and Decision tree. The one of the most important factors – variable
developed models were subjected to Martian weather – still has to be further
hyperparameters fine-tuning their investigated. The goal of this study is to
accuracies. Finally, each model was build a model that could forecast
evaluated using RMSE, MAE, and R2 value fluctuations in atmospheric temperature on
evaluation techniques. Following the Mars during a day. The model will be
evaluation, random forest outperformed all trained and the accuracy tested using data

IJMSRT24SEP032 www.ijmsrt.com 201

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

used. All this contributes to existing data Mars were investigated by Kass et al.
available on Martian atmosphere implying (2020) with Mars Climate Sounder, while
that the results will play a significant role in Mars’ weather using ML models was
offering fresh insight on climate on the red examined by Ishaani and Puri (2021); the
planet. Research in the subject and best model was identified as LSTM.
conditions of the Martian climate has been Alejandro de C. G. et al. (2019) applied
a matter of interest to scientist and ANN in order to estimate the sol pressure
researchers for several decades now. Mars and air temperature on Mars per hour with
is exceptional in that it provides vary great accuracy. They include Charalambous
climate and climate is important for manned et al. (2021) and Olsen et al. (2021), who
and robotic missions that may occur in the researched seismic and geophysical activity
planet’s future as well as to examine for any towards assessment of environmental
effects. Mars like other planets has different information on Mars. Also, P.N. Timothy
kinds of weather and one of it is diurnal (2022) pointed the significance of the ML in
variation of temperature. This variation, Martian science and called for data to be
which happens over a day, is an important driven. However, the review finds some of
part of weather on Mars and could give these questions still remaining unanswered
valuable information about the forces that including the need to develop advanced
drive this planet’s weather. For these models of weather prediction on Mars. This
insights, the variations have been provided paper concludes that more research and the
by means of machine learning techniques. use of Ml are needed to enhance awareness
This can be realised through the generation and readiness for future Mars landings. To
of models. achieve this goal, this research uses several
Even though much has been written about machine learning algorithms to make
the Martian climate and research has gone predictions of the atmospheric conditions
far in identifying this variation in daily on Mars especially atmospheric
atmospheric temperature the current models temperature. These include: Random
are not precise enough. The current models Forest: A technique that builds a set of
used in the prediction of weather on Mars decision trees and makes use of them to
are from empirical correlations, and are make the class/label predictions for
dependent on limited data (Ishaani P. and classification and numerical values for
Puri V. 2021; Ali M 2021). Therefore, these regression. Specially to deal with the
models lack the sophistication and problem of data imbalance, it works as
versatility needed to describe how multiple policies and make predictions by
atmospheric parameters are interconnected calculating the mode of each tree for
and how they affect the surface of a planet. classification and mean for regression.
Lasso Regression: A regularization
Literature review: technique that put a penalty on the model
This article focuses on the literature review parameters, collapsing regression
of utilizing machine learning (ML) for the coefficients for certain variables towards
prediction of weather on Mars by zero, and therefore, selects the most
highlighting the ML application’s extent significant predictors. Decision Tree: Tree-
because of the exhaustive Martian weather based structures, where data is modelled by
conditions and the consequences it will nodes and involves the impactful
have on future Martian missions. Scientists correlation of features and target variables.
have rampantly employed ML in areas of It is easily understandable and applicable
weather, Bellutta (2017) independently for interval or ratio level and nominal or
implemented ANN in defining the opacity ordinal data.
of Martian atmosphere. Gravity waves on

IJMSRT24SEP032 www.ijmsrt.com 202

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Artificial Neural Network (ANN): A which is commonly used in astronomy to

computational model containing a set of store data in array-like or table-like
algorithms based on the human brain and structures. To utilize the data for analysis, a
task-specific patterns or features extracted Python script was written to extract the
from input data. ANNs are learning tools FITS files and convert them into CSV
that work from example data and are format for easier processing and analysis.
specifically good at making predictions
based on some input data. Strategies used to conduct analysis
Support Vector Machine (SVM): A type The figure below shows the workflow
of learning algorithm that has a supervisory followed to conduct proper analysis on the
signal in the training data or feature set and dataset to perform proper analysis on the
falls under the classification category. It Martian dataset used:
makes an nth dimensional plane or decision
surface so as to classify the data into
different classes/regions.
Linear Regression: A type of supervised
learning method that predicts the dependent
variables by making a regression of straight
line.
Convolutional Neural Network (CNN):
This type of CNN was originally developed
with image classification in mind but can
nevertheless be used in non-image inputs as
well. Cortical Computing Corp has
developed technology to search for patterns
in large sets of data using a deep learning
architecture mimicking brain connectivity.
Also, the features of the Emirates Mars
Infrared Spectrometer (EMIRS) are
described. Spearheaded in 2020, EMIRS
will investigate the atmosphere of Mars
particularly the thermal, dust, water vapor,
and ice. This was developed in partnership
with Arizona State University and the
Mohammed Bin Rashid Space Centre to
help study Mars’ lower and middle
atmospheric layer during EMS.
Fig. 1: workflow for Martian analysis
Methodology Data Preprocessing:
Data preprocessing improves the quality of
3.1 Dataset Used
raw data, making it suitable for machine
The dataset for this research was obtained
learning models. It involves steps like data
from the Emirates Mars Mission (EMM)
cleaning (removing or imputing missing
official website. Specifically, data from the
values), outlier removal, normalization or
Emirates Mars Infrared Spectrometer
(EMIRS) instrument was downloaded from standardization, and feature transformation.
In this work, Python's Pandas library was
the following source: EMM EMIRS Data.
used to apply these techniques.
Preprocessing ensures that the model
The downloaded dataset was in Flexible
Image Transport System (FITS) format,

IJMSRT24SEP032 www.ijmsrt.com 203

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

performs optimally by transforming and

cleaning the dataset. This step is essential to finding the best
Data Exploration: parameters that deliver the most accurate
Data exploration is the initial phase of predictions.
analysis, using statistical and visualization
techniques to uncover patterns in the Model Evaluation:
dataset. Tools like Pandas, NumPy, To evaluate model performance, techniques
Matplotlib, and Seaborn were utilized in like Mean Squared Error (MSE), Root
this report to understand Martian Mean Squared Error (RMSE), Mean
atmospheric temperature variations. Absolute Error (MAE), and R-squared were
used. These metrics help measure how well
Feature Selection: the model predicts the target variable and
Selecting relevant features is critical for identify underperformance.
model performance. This report selected
features such as pressure and various Result Analysis And Discussion
temperature metrics based on their 4.1 Results obtained from Experiment
correlation with the target variable conducted on variation in temperature.
(atmospheric temperature). Proper feature
selection enhances the accuracy of
predictions.

Models Deployment:
Seven machine learning models were
deployed: Linear Regression, Random
Forest, Decision Tree, Artificial Neural
Networks, Lasso, Convolutional Neural
Network (CNN), and Support Vector
Machine (SVM). These models were Fig. 19a: Atmospheric Temperature for
trained and tested using an 80/20 data split 5th July 2021 (Bar chart)
to identify the best-performing model.

Data Splitting:
The dataset was divided into an 80%
training set and a 20% testing set using
Python’s train_test_split function from the
sklearn library. This step ensures that the
models are trained and validated
effectively.

Hyperparameter Fine-Tuning: Fig. 19b: Atmospheric Temperature for

Hyperparameter tuning was performed 21st July 2021 (Bar chart)
using grid search to optimize model
performance.

IJMSRT24SEP032 www.ijmsrt.com 204

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Fig. 19c: Atmospheric Temperature for 5th July 2021 (Scatter plot)

Fig. 19d: Atmospheric Temperature for 20th July 2021 (Scatter plot)

Fig. 19f: Atmospheric Temperature for 30th July 2021 (Scatter plot)

IJMSRT24SEP032 www.ijmsrt.com 205

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Fig. 19g: Atmospheric Temperature for 15th July 2021 (Scatter plot)

Fig. 19h: Atmospheric Temperature for 27th July 2021 (Scatter plot)

The above results were obtained from presented by Dimitra A. et. al., 2023 and
taking 6 days (I.e. sols) within the dataset Michael D et. al, 2022.
and visualizing their atmospheric This variation in mars temperature are
temperatures. From the above figures 16a- largely sponsored by the amount of dust
16h, it can be clearly seen that there is a present in its
variation in the daily temperature value for atmosphere. According to Royal Belgian
mars. From the figures, it can be observed institute for space Aeronomy (2022),
that within a martian day, the temperature variation in day and night temperature of
seems to be lower at early hours and higher mars is mainly due to the weakness of
at late hours. Although some days greenhouse effect which contributes to the
experience higher temperature values early Martian soil storing very few amounts of
hours compare to late hours e.g 21st and 15th energy.
July 2021. This is a clearly indication of 4.2 Results obtained from Experiment
diurnal variation in temperature of mars as conducted for similarities among mars
temperatures

IJMSRT24SEP032 www.ijmsrt.com 206

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Fig. 20a: Relationship between AT, ST for CO2 and ST

Fig. 20b: Linear Relationship between AT and ST outside CO2

IJMSRT24SEP032 www.ijmsrt.com 207

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Fig. 20c: Linear Relationship between AT and ST for H2O

Fig. 20d: Linear Relationship between AT and ST in Mars

IJMSRT24SEP032 www.ijmsrt.com 208

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Fig. 20e: Relationship between AT, ST for CO2 and ST for H2O

Fig. 20f: Relationship between ST, ST for CO2 and ST for H2O

IJMSRT24SEP032 www.ijmsrt.com 209

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Fig. 20g: Relationship between AT, ST for H2O and ST.

The above results were generated by
executing an experiment that was designed The first experiment was conducted without
to test if correlation between all asserted removing outliers present within the dataset
temperature values of the EMIRS while the second experiment removed
instrument holds true. Analyzing the results outliers present within the dataset before
presented in Fig. 20a – Fig. 20g it is possible model development. The results obtained
to identify the linear dependency of all the from these two experiments are presented
sorts of temperatures within the dataset below:
under consideration. The EMIR instrument 4.3.1 Results from experiment without
captured 4 temperature readings which are, removal of outliers
Atmospheric temperature (AT), Surface The results below were obtained after
Temperature(ST), Surface temperaure out models were developed without removal of
of Water region (ST out of H2O) and outliers present within the dataset:
Surface temp outside CO2 region(ST out of
CO2). From the above results, these
temperatures have an aspect of mutual
coexistence as the various figures showed
the existence of a linear relationship where
data from one affects the other, as
evidenced by the part where the different
figures formed a linear relationship.
4.3 Results obtained from model
development
For model development, two different
experiments were conducted on the dataset.

IJMSRT24SEP032 www.ijmsrt.com 210

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Model MAE MSE RMSE R_2 score

Random forest 1.3179 4.1967 2.0486 0.9448
Linear Regression 3.4310 50.0191 7.0724 0.3423
Support Vector Machine 3.3729 45.1238 6.7174 0.4067
(SVM)
Artificial Neural Network 149.4879 22704.6600 150.6807 -297.5380
(ANN)
Decision Tree 1.4242 3.6380 1.9073 0.9521
LASSO regression 3.0438 18.8760 4.3447 0.7518
Convolutional neural 166.4401 27810.7507 166.7656 -364.6769
Network (CNN)

Table 1: Result after removal of outliers of used machine learning models without
The above table presents the results handling outliers present within the
obtained from development and evaluation dataset.

4.3.2 Results after handling outliers present within dataset

Model MAE MSE RMSE R_2 score

Random forest 1.1188 2.4228 1.5566 0.8275

Linear Regression 1.5275 3.5820 1.8926 0.7449

Support Vector Machine 1.3887 3.9814 1.9953 0.7165

(SVM)

Artificial Neural 31.6600 1530.9090 39.1268 -108.0029

Network (ANN)

Decision Tree 1.5839 5.1040 2.2592 0.6366

LASSO regression 2.0469 7.2166 2.6864 0.4862

Convolutional neural 169.4396 28725.9820 169.4874 -2044.3299

Network (CNN)

Table 3: Result after removal of outliers the importance of outlier removal in

The results from two experiments on the improving model accuracy. Neural
EMIRS dataset are summarized in Tables 1 networks (ANN and CNN) produced
and 2. In Table 1, where outliers were not negative R2 scores and high errors,
addressed, Random Forest and Decision demonstrating their unsuitability for the
Tree models achieved high R2 scores but dataset, as their predictions were worse than
exhibited large errors, indicating that the a simple mean prediction. Among the
high R2 scores were influenced by models tested, Random Forest
overfitting. In contrast, Table 2 shows lower outperformed the others by achieving
R2 scores but significantly reduced errors minimal error values and a realistic, high R2
after outliers were removed, highlighting score, making it the best model for this

IJMSRT24SEP032 www.ijmsrt.com 211

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

dataset. The comparison of model

performance is illustrated in the following
figures.

Fig. 21a: RMSE for the 5 models with high performance

Fig. 21b: MSE for the 5 models with high performance

IJMSRT24SEP032 www.ijmsrt.com 212

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Fig. 21c: MAE for the 5 models with high performance

Fig. 21d: R2 Values for the 5 models with Due to this performance the 5 models were
high performance subject to hyper parameter fine tuning using
The high performance of Random Forest GridsearhCV (Joseph R., 2018) and the
can be clearly seen across all 4 figures. results obtained are presented in Table 4
below:

Model MAE MSE RMSE R_2 score

Random forest 1.1939 2.4105 1.5525 0.8283
Linear Regression 1.5275 3.5819 1.8926 0.7449
Support Vector Machine 1.5038 3.6944 1.9221 0.7369
(SVM)
Artificial Neural Network
(ANN)
Decision Tree 1.5898 5.0146 2.2393 0.6429
LASSO regression 1.5271 3.5809 1.8923 0.7450
Convolutional neural -------------- ------------- -------------- --------------
Network (CNN)

Table 4: Result after Hyper parameter fine tuning

IJMSRT24SEP032 www.ijmsrt.com 213

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Fig. 22a: RMSE for the 5 models with high performance (After hyper parameter fine
tuning)

Fig. 22b: MSE for the 5 models with high performance (After hyper parameter fine
tuning)

IJMSRT24SEP032 www.ijmsrt.com 214

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Fig. 22c: MAE for the 5 models with high performance (After hyper parameter fine
tuning)

Fig. 22d: R2 Values for the 5 models with could have stemmed from over fitting, as
high performance (After hyper captured by inflated errors amounts
parameter fine tuning) (RMSE, MSE, MAE). However, I found
Table 4 and Figures 22a-22d illustrate the smaller error scores in the second
improved performance of the five models experiment when outliers were excluded,
after hyperparameter fine-tuning. Random which indicated better fit of the model to
Forest, the best-performing model, showed data. Hyperparameter tuning was done to
notable improvements in terms of MSE, further improve the performance of each
RMSE, and R2 scores. Compared to the model and Random Forest performed better
earlier results in Figures 21a-21b and Table than other models in the two experiments.
3, the new results in Figures 22a-22b and For the Random Forest classifier its R-
Table 4 demonstrated better accuracies. Squared scores were 83%, 82% along with
Specifically, the MSE of the Random Forest 82.5% in consecutive assessments which
model decreased by 0.0123, the RMSE make it quite reliable. This concurs with the
decreased by 0.0041, and the R2 score study made by Joanna W. (2022) in which
increased by 0.0008. These improvements Random Forest outperformed neural
indicate that the hyperparameter tuning networks and SVM algorithms in predicting
process was effective in optimizing the Mars weather. The present work differed
models' performance. from the prior research done by Ishaani &
Puri (2021), Ali M. (2021) and Alejandro
Critical Discussion Summary (2019) which relied on pressure and
In the present research, to model the daily temperature predictions using
Martian atmosphere temperature comparatively smaller datasets of REMS
fluctuations, the EMM EMIRS dataset was excluding EMIRS data but in this study,
analyzed employing seven machine daily temperature fluctuations has been
learning algorithms in two experiments. predicted using the dataset received from
While in the first experiment, models were EMIRS. ANN and CNN models that were
developed with outliers still in the dataset, used in the Martian weather forecasting
in the second experiment, the above defined showed poor results on this dataset which
outliers were removed from the dataset. The indicates that the variability of datasets
results obtained in the first experiment considerably influences the stability and
yielded higher R-squared values, which efficiency of the models.
The study also confirmed the daily Martian
temperature variation observed by Michael

IJMSRT24SEP032 www.ijmsrt.com 215

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

D. et al. (2022), as seen in the up to August 2022. Furthermore, more

visualizations, supporting the findings of extensive hyperparameter tuning should be
previous research. performed to optimize model performance,
addressing the time and resource constraints
Challenges, Conclusion, and Future faced in this study.
Work Summary
References:
5.1 Challenges: Abhishek Sheshnath (2022). 'Importance of
Missing Values: The EMM EMIRS dataset Data Exploration in Data Analysis A
had many missing values, which posed Review Paper'. International Journal of
challenges during model development. Advanced Research in Science,
Significant pre-processing was required, Communication and Technology
and assumptions were made that may have (IJARSCT) 2(2). pp. 1-5.
impacted the models' accuracy. DOI: 10.48175/568 362
Time and Data Size: The large size of the Akshita C. (2020). 'MAE, MSE, RMSE,
dataset caused delays in importing the data Coefficient of Determination, Adjusted R
and executing analyses. Hyperparameter Squared — Which Metric is Better?'.
fine-tuning for models took significant Availableat: https://medium.com/analytics-
time, up to 30 minutes per model, due to the vidhya/ (Accessed: 15 January, 2023)
dataset's size. Ali Momennasab (November 2021).
‘Machine Learning for Mars Exploration’.
5.2 Conclusion pp. 1- 11.
This study achieved its objectives by Alejandro de Cabo Garcia, Alfonso
exploring the diurnal variation in Martian Degado-Bonai, Maria Bele Perez Lancho,
atmospheric temperatures and examining Jorge Garcia and German Martinez. (May
the linear relationship among the four 27,2019). ‘MARSchine Learning
temperature types in the dataset. Prediction: prediction of mars Metrological
Preprocessing steps were conducted to variables using Artificial Neural Networks’.
prepare the data for model development. pp. 1-2.
Two experiments were performed: the first Arizona University (2022). “mars
with outliers present and the second with atmosphere”. Available at:
outliers removed. The second experiment, https://marsed.asu.edu/mep/atmosphere.
followed by hyperparameter tuning, (Accessed: 11 December, 2022)
resulted in better performance. Among the Arizona University (2022). “Wind”.
models, Random Forest outperformed Available at:
others with RMSE=1.5525, MSE=2.4105, https://marsed.asu.edu/mep/wind
MAE=1.1939, and R²=0.8283, (Accessed: 5 February, 2023)
demonstrating its suitability for Martian Ayat H., Sahar A., Basim A. and
temperature prediction. Mohammed N. (2023). ‘Impact of feature
reduction techniques on classification
5.3 Future Work accuracy of machine learning techniques in
Future research could explore other models, leg rehabilitation’. Measurement:
such as RNN, LSTM, and CCNN, which have Sensors, 25, 100544. Available at:
been used for Mars weather prediction in the https://doi.org/10.1016/j.measen.2022.100
past. Additionally, future experiments should 544
employ K-fold cross-validation for more Christopher Lee (16 April 2019).
reliable results. The dataset could also be ‘Automated crater detection on mars using
expanded using newly released data from the deep earning’. pp. 2-9.
EMM, which includes Martian weather data

IJMSRT24SEP032 www.ijmsrt.com 216

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Charalambous, A. E. Stott, W. T. Pike, J. B. us/articles/202529139-Why-am-I-seeing-a-

McClean, T. Warren, A. Spiga, D. Banfield, negative-R-2. (Accessed: 1 February,
R. F. Garcia, J. Clinton, S. Stähler, S. 2023).
Navarro, P. Lognonné, J.-R. Scholz, T. Dimitra A., Nour A., Dattaraj B. and
Kawamura, M. van Driel, M. Böse, S. Mathilde S. (2023). 'Diurnal variation of the
Ceylan, A. Khan, A. Horleston, G. surface temperature of Mars with the
Orhand-Mainsant, L. M. Sotomayor, N. Emirates Mars Mission: a comparison with
Murdoch, D. Giardini, B. Curiosity and Perseverance rover
Banerdt (2021). ‘A Comodulation Analysis measurements'. Monthly Notices of the
of Atmospheric Energy Injection Into Royal Astronomical Society: Letters 518
the Ground Motion at InSight, Mars’. (1).pp. L1–L6. Available at
JGRPlanets 126(4). Available at: https://doi.org/10.1093/mnrasl/slac094
https://doi.org/10.1029/2020JE006538. Doran G., Francis R., Estlin T., Johnstone
Chang K. (2013). 'Mars Could Have S., Gaines D., Verma V., Burl M.,
Supported Life Long Ago, NASA Says'. Frydenvang J., Montaño S., Wiens
Available at: R.C.(2017). ‘AEGIS Autonomous
https://www.nytimes.com/2013/03/13/scie Targeting for ChemCam on Mars Science
nce/space/mars-could-have- Laboratory: Deployment and Results of
supported.(Accessed: 21 January, 2023). Initial Science Team Use’. Sci. Robot., 2,
Charles Q., Mike W., Elizabeth H., Daisy eaan4582.
Dobrijevic(2022).'Mars: Everything you Dundar M, Ehlmann B. L., and Leask E.
need to know about the Red Planet'. (2019). “Machine-Learning-Driven New
Available at: https://www.space.com/47- Geologic Discoveries at Mars Rover Land-
mars-the-red (Accessed: 23 December, ing Sites: Jezero Crater and NE Syrtis.”
2022). Wiley. doi: 10.1002/es- soar.10501294.1.
Cloud Architecture Center (2022). 'Data Edwards C., Christensen P., Mehall G. et al.
preprocessing for ML: options and (2021). ‘The Emirates Mars Mission
recommendations'. Available at: (EMM) Emirates Mars InfraRed
https://cloud.google.com/architecture/data- Spectrometer (EMIRS) Instrument’. Space
preprocessing-for-ml-with-tf- transform Sci Rev 217 (77). Available at:
(Accessed: 11 January, 2023). https://doi.org/10.1007/s11214-021-00848-
Crisp J., Grotzinger J. P., Vasavada A. R., 1
Anderson R. C., Baker C. J., Barry R., EMM (2020). 'EMM Over view'. Available
BlakeD. F., Conrad P., Edgett K. S., at:https://www.emiratesmarsmission.ae/mi
Ferdowski B., Gellert R., Gilbert J. B., sion/about-emm/(Accessed:5 January,
Golombek M., Gómez-Elvira J., Hassler D. 2023).
M., Jandura L., Litvak M., Mahaffy P., Fan C., Chen M., Wang X., Wang J., and
Maki J. and Wiens R. C. (2012). ‘Mars Huang, B. (2021). 'A Review on Data
Science Laboratory Mission Preprocessing Techniques Toward Efficient
and Science Investigation’. Space Science and Reliable Knowledge Discovery From
Reviews 170(4). pp.5–56. Available Building Operational Data'. Frontiers in
at: https://doi.org/10.1007/s11214-012- Energy Research, 9. Available at:
9892-2. https://doi.org/10.3389/fenrg.2021.652801
Daniele Bellutta (2017). ‘The Fog War: Fendall M., Denkins P., Fisher S. and
A Machine Learning Approach to Oluseyi H. (2014). “Human mars
forecasting weather on Mars’. pp. 2-5. Exploration Research Objectives”. pp. 1-
Desmos (2022). ‘Why am I seeing a 30
negative R^2 value?’. Available at: Frost J. (2017). 'Overfitting Regression
https://help.desmos.com/hc/en- Models: Problems, Detection, and

IJMSRT24SEP032 www.ijmsrt.com 217

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Avoidance'. Available at: air quality index by different geographical

https://statisticsbyjim.com/regression/overf areas’. pp. 1347–1357.
itting-regression (Accessed: 11 January, Kristin R., Olivier G., Candice C., Erwin
2023). D., Jeremie L., David G., Travis G.,
Gera P.(2022). 'What are Hyper Parameter Sylvestre M. and Roger C. (2021).
Tuning Techniques in Machine Learning?'. “Clustering Supported Classification of
Available at: ChemCam Data From Gale Crater, Mars”.
https://codedamn.com/news/machine- Earth and Space science 8(12). pp. 1-
learning/what- are-hyper-parameter-tuning- 10.https://doi.org/10.1029/2021EA001903
techniques-in-machine-learning Laura Dattaro (2013). ‘Do mars Have
(Accessed:2 January, 2023). seasons?’. Available at:
Gramigna E. (2020). ‘Calibration https://weather.com/science/news/does-
techniques for studying Venus and Mars mars-have-seasons-20130910 (Accessed:
atmospheres’. AerotecnicaMissili&Spazio, 12 September,2022).
99(4), pp.255–261. Michael D., Badri, K., Atwood, S.A.,
Ishaani P. and Puri V. (2021). ‘Mars Edwards, C. S., Christensen, P. R., Wolff,
weather data analysis using machine M. J., et al. (2022). ‘EMIRS observations
learning techniques’. Earth Science of the aphelion-season Mars atmosphere’.
Informatics 14. pp. 1885–1898. Geophysical Research Letters, 49,
Available e2022GL099636.
at:https://doi.org/10.1007/s12145-021- Available at: https://doi.
00643-0 org/10.1029/2022GL099636.
Issam El Naqa and Ruijiang Li and Murphy Mohammad N., Mohammad A., Moses K.
M. J. (Eds) (2015). ‘Machine Learning in and Mehrdad K. (2020). 'Water on mars- A
radiation oncology’. pp. 1-4. literature review'.Galaxies, 8(2), pp.1-40.
Joanna Wendel (2022). 'Machine learning Available at:
algorithms help scientists explore Mars'. https://doi.org/10.3390/galaxies8020040.
Available at:https://phys.org/news/2022- NASA (2021). “Mars”. Available at:
01-machine-algorithms-scientists-explore- https://www.nasa.gov/mission_pages/mer/i
mars.html (Accessed: 2 January, 2023). ndex.html (Accessed:1 January, 2023)
Joaquin Rojas (2022).’ Mars Weather’. National Weather Service. (2022). ‘The
Available at: Planet Mars’. Available at:
https://public.tableau.com/app/profile/joaq https://www.weather.gov/fsd/mars#:%7E:t
uin.rojas/viz/mars_16588821428780/Dash ext=Temperatures%20on%20Mars%20ave
boar1 (Accessed: 26 September, 2022). rage%20about,lower%20latitudes%20in%
Joseph R. (2018). 'Grid search for model 20the%20summer(Accessed:26 September,
tuning'. Available 2022).
athttps://towardsdatascience.com/grid- Olsen KS ,Korablev O, Trokhimovskiy A,
search-for-model-tuning(Accessed:12 Lefevre F, Montmessin,FFedorova A,
January, 2023). Toplis M, Alday J, Belyaev D, Patrakeev
Kass D., Heavens N., Kleinbohl A. and A, Ignatiev N, Shakun A, Grigoriev A,
Schofield J. (2020). “A multiannual record Baggio L, Abdenour (2021). ‘NASA Mars
of gravity wave activity in Mars’s Exploration’.Availableat:https://mars.nasa.
lower atmosphere from on-planet gov/msl/mission/overview/(Accessed:27
observations by the Mars Climate October, 2022).
Sounder”. Icarus 341:113630. Priyadarshini I, Puri V., Cotton C. (2021).
Kumar R., Sharma, R., Sharma, D., ‘A novel LSTM–CNN–grid search based
Priyadarshini, I., Pham, B., Bui, D. T., & deep neural network for sentiment
Rai, S. (2019).‘Inferring air pollution from analysis’. J Supercomput. pp. 1-22.

IJMSRT24SEP032 www.ijmsrt.com 218

DOI: https://doi.org/10.5281/zenodo.13846044
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Qiu D. , Rothrock B.,Tanvir Islam , Didier

A., Vivian Z. Sun , Mattmann C. and Ono
M. (2021). ‘SCOTI: Science Captioning of
Terrain Images for data prioritization
and local image search’. pp. 1-9. Available
at:https://doi.org/10.1016/j.pss.2020.10494
3.
Rahul A. (2022). ' Covolutional Neural
Network (CNN)'. Available
athttps://www.techtarget.com/searchenterp
riseai/definition/convolutional-neural-
. (Accessed: 3 January, 2023).
Rong, Shen Bao-wen and Zhang. (2018).
‘The research of regression model in
machine learning field’. MATEC Web of
Conferences. 176. 01033.
10.1051/matecconf/201817601033.
Royal Belgian Institute for Space
Aeronomy (2022). 'Mars climate, important
temperature difference between day and
night'. Available at:
https://www.aeronomie.be/en/encyclopedia
/mars-climate (Accessed: 5
January, 2023.).
Timothy Paul N., Louis S. and
NicholasErickson (2022). ‘Squeezing Data
from a Rock: Machine Learning for
Martian Science’. pp. 2-14.
DOI:10.3390/geosciences12060248.
Xiao M., Wang D., Wu M., Liu K., Xiong
H., Fellow, IEEE, Zhou Y., and Fu Y.
(2022). 'Self-Optimizing Feature
Transformation'. pp. 1-13. Available at:
https://arxiv.org/pdf/2209.08044.pdf#:~:te
xt=Abstract

IJMSRT24SEP032 www.ijmsrt.com 219

DOI: https://doi.org/10.5281/zenodo.13846044