4. Q&A with Guido Cervone


Guido Cervone sails

Guido Cervone circumnavigates the Maryland-Delaware-Virginia (DELMARVA) peninsula,

part of the 2013 Offshore Sailing Training Squadron (OSTS) training. 

Photo taken approximately at  37.4479463,–74.8201732 onboard the

U.S. Naval Academy  sailing vessel “Fearless.”  Photo provided by Guido Cervone.


How did you first get interested in applying your research methods to the problems of environmental hazards?


GC: While growing up in Rome, Italy when I was a child, I was exposed to many environmental hazards including several strong earthquakes—Italy is very seismic—and extreme weather events.  I particularly remember living through the nuclear radioactive contamination from the 1986 Chernobyl accident, which extended all the way to parts of Southern Europe.  Living first hand these potentially life-threatening events sparked an interest that I retain to this day.
     Throughout my studies I have always considered studying hazards paramount for the development and sustainment of our society.  Perhaps, the turning moment in my career was in 2002 during my second year Ph.D. studies, when I attended a NASA presentation about remote sensing.  
     Up until that moment, I worked on the development of spatio-temporal machine learning algorithms, but predominantly applied to synthetic computer-generated data.   Learning that it was possible to observe the entire planet with high spatial and spectral resolution daily was love at first sight!  From that moment I started studying remote sensing, and using my machine learning background to find anomalies and similarities in the data.

Imagine the future of how geoinformatics and machine learning would be used in managing environmental hazards: what does it look like? 

GC: In recent years, the advances in our ability to observe Earth and its environment through the use of air-, space-, and ground-based sensors has led to the generation of large, dynamic, and geographically distributed spatiotemporal data. The rate at which geospatial data are being generated exceeds our ability to organize and analyze them to extract patterns critical for understanding our dynamically changing world. New challenges arise from an unprecedented access to massive amounts of Earth science data that can be used to study the complementary nature of different parameters. These developments are quickly leading towards a data-rich but knowledge-poor environment.
     Geoinformatics algorithms are needed to address these scientific and computational challenges and provide innovative and effective solutions to analyze these large, often multi-modal, spatiotemporal datasets. The ability to generate knowledge in near-real time, analyzing massive amounts of data, can help at all stages related to natural hazards. At the planning stage, geoinformatics algorithms can  help quantify the risk of constructing a particular facility, or determine where to locate sensors to detect signs of an accident or an impending event.  Geoinformatics can also be used during an event, to provide real-time knowledge to first responders. Data fusion is a particularly active area of research, where observations  from different sensors are processed to give better estimate of a developing event.  Finally, geoinformatics can help after the occurrence of disasters to improve upon predictive models that can help protect people properties and the environment.

Can you give a specific example of how volunteered geographic information fills in the gaps or mediates the weaknesses in remote sensing data and vice versa?

GC: Remote sensing is the de-facto standard to observe and study Earth.  However, observations are limited due to revisiting time of satellites and physical constraints dictated by Earth’s atmosphere.  Using polar satellites— those that orbit closer to Earth’s surface and are particularly indicated for high resolution observations— it is impossible to acquire a continuous data feed for a particular location.  Furthermore, atmospheric opacity may cause observations to contain missing data relative to the surface of Earth.  Therefore temporal gaps in the observations are inevitable.  
     Volunteered Geographical Information (VGI) can be used to augment the satellite observations, and fill the temporal gaps where data was not collected. Merging VGI and remote sensing data is a difficult data fusion problem, where the high-spatial and low-temporal resolution observations from satellites are fused with the low-spatial and high-temporal resolution from VGI. 
     The difficulty in fusing the data is further exacerbated by the fact we are merging two very different datasets: one objective and based on physical observations from an electronic instrument, and another one subjective and based on the perception of people of a particular event.

In your career so far, what deliverable or outcome are you most proud of?

GC: Perhaps what I am most proud of is the reconstruction of the non-steady source release from the Fukushima nuclear disaster.  I have created with my colleague Pasquale Franzese a new machine-learning-based methodology that uses satellite, in-situ observations and numerical atmospheric modeling to reconstruct the nuclear release. We have published our work, and based on it, I received in 2013 the Medaglia di Rappresentanza from the President of Italy and the Italian Scientists and Scholars of North Amerca Foundation (ISNAAF) award.

Can you talk more about the use of geoinformatics for the optimization of numerical model forecasts for wind energy power production?

GC: Optimizing weather forecasts for energy production is a very important and exciting research topic which I have recently started investigating as part of my association with the National Center for Atmospheric Research (NCAR) in Boulder, CO. Large uncertainty in model predictions for parameters such as solar irradiance or wind speed can cause a deficit between the demand and supply of energy from solar and wind power plants.  It is very important to reduce this uncertainty by improving the models, but also output probabilistic measures associated with the predictions. The analog ensemble methodology developed at NCAR was designed precisely to output probabilistic forecasts that the power company can use to assess the risk of under- or over-producing electricity.

Is there a relationship between your passion for sailing and your work?

GC: I believe sailing is the most fun activity known to man! I have been sailing ever since I was very young, and especially in recent years I have been active in racing on big and small boats.  Whereas I considered sailing a hobby and often an escape, a lot of the concepts that I research and teach about are very handy when racing and cruising. 
     On some occasions I have run my own atmospheric wind models before major races, and performed sensitivity studies to understand the likeliness of wind shifts in power and direction. However, the most relevant work I did for which I merged my passion of sailing with my research activity is in an article that appeared in 2013 in the International Journal of Remote Sensing. I investigated the relationship of Sea Surface Temperature (SST) acquired by different satellites and models with in situ measurements I performed while transporting a boat from Bermuda to the USA. 
My goal was to determine if it is possible to have better estimation of the location of the Gulf Stream by merging in situ data with satellite observations. 
     Crossing the Gulf Stream in a small boat can be very dangerous, and thus it is very important to have accurate predictions.  For the future, I would like to study the possibility of using high-resolution satellite SST observations as an alternative way of navigation in case of GPS failure.

What are your future research plans?

GC: Joining the Department of Geography and Institute for CyberScience at Penn State allows me to continue my line of research while expanding it to new horizons.
     I plan to continue working on the development and implementation of geoinformatics algorithms for my two major projects, namely the fusion of remote sensing data and VGI for the study of environmental hazards, and the optimization of numerical models for alternative energy production. For this latter project, I hope to strengthen the collaboration between NCAR and the Department of Geography, hopefully giving opportunities to students to visit NCAR, and take advantage of their impressive computational facilities.
     I believe my research is very complementary to that of my colleagues, and I expect to be able to apply some of these algorithms, along with developing new ones, to solve new classes of problems.