Big Data and Machine Learning in Earth Sciences

If big data challenges can be summed up as “to cost-effectively scale computation and storage in the face of ever-increasing data volumes and varieties with an ever-escalating demand for velocity”, we posit that these challenges have been present since the dawn of digital computing for Earth science. Our desire for better fidelity from predictions urges us to incorporate into numerical models evermore comprehensive physical interactions with evermore refined intricacy, which compels evermore extensive and expansive observations with evermore detailed focus, further intensifying the challenge. As a result, another type of challenge arises, i.e., the wish to realize the full value from the deluge of data generated by models and observations. Unfortunately, until recently, we have had to rely mostly on human beings’ cognitive faculty. Machine learning promises to address this challenge. Since Earth is a complex, nonlinear system rife with processes spanning a broad spectrum of spatiotemporal scales, we can better constrain our hypotheses and direct our investigations when leveraging dissimilar data featuring complementary strengths. Similarly, the performance and, especially, machine learning models’ generalizability improve with increasing volume and variety of training data. Thus, these two types of challenges are linked: machine learning, and more recently, deep learning techniques, which are to automate the analysis and interpretation of big data, will not be very effective if the data cannot be wrangled and processed using parallel and scalable methods. We aim for this Special Issue to review the progress and explore the prospect of addressing these two interconnected challenges in the context of their history.

Dear Colleagues,

We Earth scientists are certainly no strangers to the challenges of big data. However, improvements in model and observation resolutions, the introduction of exotic simulation grids, and the employment of novel observation technologies and strategies have conspired to aggravate the acuity of the challenges, especially with the end of Moore’s law approaching. Thus, scalable parallel processing has attained paramount importance in addressing these challenges. The most challenging step in optimizing scalable parallel processing is perhaps data preparation, as attested by the 80/20 rule: data analysts or scientists devote only 20% of their time to actual analysis, with the rest, i.e., 80%, spent in preparing the data for analysis.

Moreover, even if we were able to process the volume and variety of our big data in a timely fashion in the traditional manner, it would not be sufficient because, relying purely on human effort, the analysis and interpretation of the processed results may still overwhelm us. We need help from artificial intelligence and machine learning (AIML), which have demonstrated remarkable advancement recently, often achieving comparable or even better performance than humans in specific or less advanced cognitive functions. AIML techniques are usually computationally intensive, especially in their training/learning stage, again accentuating the importance of scalable parallel processing.

This Special Issue welcomes contributions from all Earth science domains on the description, evolution, and solutions (or proposed solutions) of the aforementioned challenges.

Dr. Kwo-Sen Kuo
Dr. Rahul Ramachandran
Prof. Dr. Morris Riedel
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Applied Sciences is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2400 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

• machine learning
• big data
• earth science datasets
• scalable parallel processing
• data management challenges
• data mining
• data fusion
• cloud computing pipelines
• data augmentation techniques
• high-performance computing
• data processing challenges
• in-memory processing
• deep learning
• apache hadoop/spark stacks

Potential contributors you would like to invite to submit:

• processing approaches for large quantities of earth science datasets
• scaling up high-performance computing machine learning algorithms
• processing pipelines using cloud computing for earth observation data
• innovative deep learning networks for earth science data analysis
• harnessing open source frameworks of hadoop/spark ecosystem