On the problem of crowd estimation for the Aug. 4, 2012 EDSA prayer rally: an interview for CBCP News
August 7, 2012 1 Comment
My estimate of the crowd size during the Aug. 4 EDSA Prayer Rally was featured in CBCP News. I was then asked by CBCP News to answer a few follow-up questions. But since I tend to answer in paragraphs and not in sentences, I think my response would not fit into a regular news column. So I’ll post my responses here and CBCP News can simply copy parts of it or repost the whole thing:
1.) Why did you feel you needed to come out with this crowd estimate, considering that other groups had come out with their figures?
After coming from the Aug. 4 EDSA rally, I read in Facebook about the estimates published in newspapers which give figures of 7,000 and 10,000 persons for the rally. My hunch is that newspaper writers have a deadline for sending their articles before 3 pm, so that it can be part of tomorrow’s headlines. Thus, the crowd present during the 5 p.m. mass was not counted. So I made my own estimates and came up with 45,000 to 60,000 persons
2) What is your field of expertise and how long have you been with the Manila Observatory?
My expertise is in theoretical physics, particularly in the use of Clifford (geometric) algebra in many branches of physics: mechanics, optics, and electromagnetics. I am an Assistant Professor of Physics at Ateneo de Manila University. I do my research on ionosphere and magnetosphere at Manila Observatory’s Ionosphere Research Building, now known as ICSWSE (International Center for Space Weather Science and Education) Subcenter. I was with MO since 2008 when I was still writing my Ph.D. dissertation. But I do not speak in behalf of the Ateneo Physics Department or of Manila Observatory. I speak only on my own as a theoretical physicist.
3) Are there other methods of crowd estimation? What limitation could these methods have?
Ideally, there should be a camera at the top of Robinson’s Galleria or aboard a plane or a satellite, so that we can get pictures at different times and determine the exact extent of the crowd in time. Here is a good example of how crowd estimation is done from wired.com:
At President Barack Obama’s 2009 inauguration ceremony, the high-resolution, Earth-orbiting GeoEye-1 satellite took pictures from 423 miles away, and another camera was hanging from a balloon 700 feet off the ground. After examining pictures from both of these sources, researchers put crowd estimates at anywhere from 1 to 2 million.
In the manual method of crowd estimation, you can mark out the areas with similar crowd densities by encircling the areas with a colored pen or by subdividing the areas into a regular grid of square boxes. Areas with similar crowd densities we can refer to as clusters. You can then zoom in to one part of the cluster, count the number of persons per square meter, and multiply this by the area of the cluster. The result is the number of persons per cluster. Then you add all the number of persons in each cluster to get the size of the crowd. The only difficulty is to determine which group of people belongs to in a particular cluster. The more cluster types you use, the more precise your estimate becomes, but it also makes distinguishing one cluster from another more difficult. The fewer clusters you use, the easier it is to distinguish each cluster, but the margin of errors in crowd size estimates would be bigger.
In the computer method of crowd estimation, one way is to get the total area of the black parts and divide it by the average area of each black head in the image. The principle is straightforward and there are computer programs that can do this, depending on the threshold level for the gray scale. But what makes this method difficult is the possibility of counting black shadows and black shirts, too, which would increase the crowd estimate. Furthermore, umbrellas and blondes would make the method useless. There is also the problem image distortion due to perspective (areas closer to the camera appear larger) and camera lens imaging (straight lines becomes curved due to pincushion and barrel distortion). And as your camera goes higher and higher to see the whole crowd, image resolution deteriorates, making it difficult for the computer and even for human crowd estimators to distinguish one person in the crowd from another. To write a computer algorithm for crowd estimation that can handle all these problems is a very difficult challenge.
I am using the manual method. Since I don’t have a picture of an aerial view of the whole crowd, I have to make estimates on the extent of the crowd based on the pictures available, and assume there is only one cluster for the whole crowd for simplicity–an assumption which I think is a valid if you look at the pictures by Anna Cosio in Carlos Palad’s blog, Catholic Position vs the RH Bill. I computed the total estimated area covered by the crowd by dividing the area into strips with the same 17 m width, and added the area of each strip. The I used some rules of thumb in wired.com. I verified these rules by drawing on the floor a square with one meter on each side. I stood inside the square and found that 4 people can fit there with enough elbow room as I saw in the pictures. So I used 4 persons/sq.m. and came up with 60,000 persons. Even if I assume only 3 persons/sq.m., that is still 45,000 persons. I doubt that the crowd density is only 2 persons/sq.m., but even that gives 30,000 persons, which is still three times the estimate of 10,000 in newspapers.
4. Does the Manila Observatory do crowd estimation regularly? When?
No, Manila Observatory as an institution does not do crowd estimation, because its focus is primarily on geophysics and disaster science–earthquakes, typhoons, pollution, and space weather–and how these disasters can be quantified, predicted, mitigated, and avoided to save more lives. Some of my colleagues at the observatory–three of them also my fellow physics faculty in Ateneo–are working on satellite and ground data to map out climate change, rainfall patterns, and land use. But the techniques in satellite and ground data processing can easily be applied to crowd estimation, provided sufficient data such as aerial and street level photos are available. In Ateneo de Manila University, there are undergraduate students who are writing software for monitoring pedestrians and for counting fish fingerlings. There are also researchers working with cameras on toy planes to map out flooded areas. Many of these researchers are members of the Ateneo Innovation Center under Dr. Greg Tangonan, who is also the Director for the Congressional Commission on Science and Technology and Engineering (COMSTE). In short, there is expertise in Manila Observatory and Ateneo de Manila University to do crowd estimation. It is only a question whether they are interested to do it for street rallies and whether they have the manpower to do the research. The harvest is great but the laborers are few.
As a theoretical physicist, I only do crowd estimation using pen and paper, and the Aug. 4 EDSA Prayer Rally was my first work. I am willing to do crowd estimation regularly as a service for the Church, provided I am given sufficient data consisting of time-stamped pictures in aerial and street level views. The results of the analysis can be published in the web, i.e. in my blog. Other researchers can then challenge the methodology and assumptions, and come up with their estimates using the same or more comprehensive data set. If there are more researchers working on this problem, we can create a Philippine Journal on Crowd Estimation. The results can be applied to any type of crowd–armies of ants, schools of fish, flocks of birds, herds of cattle–even if they would be as numerous as the stars in the heavens or as the sands in the sea. For this is how science is done: a continuous dialogue in search for truth.
5) Do you think your personal convictions affected your scientific work on this particular crowd estimation? Why or why not?
I am a Catholic who loves the Church in the same way as Faramir loves Gondor: “And I would have her loved for her memory, her ancientry, her beauty, and her present wisdom. Not feared, save as men may fear the dignity of a man, old and wise” (Two Towers, p. 314-315). I read the Bible, the Catechism, the lives and writings of saints, and the history of the Church. I organize Latin masses and promote the rosary. In the case of the RH Bill, and of all other issues such as women ordination, same-sex marriage, and human evolution, I only follow what St. Ignatius of Loyola laid down in his Spiritual Exercises–The Rules for Thinking, Judging, and Feeling with the Church:
Rule 1: With all judgment of our own put aside, we ought to keep our
minds disposed and ready to be obedient in everything to the true
Spouse of Christ our Lord, which is our Holy Mother, the hierarchical
Rule 13: To keep ourselves right in all things, we ought to hold fast
to this principle: What I see as white, I will believe to be black if
the hierarchical Church thus determines it. For we believe that
between Christ our Lord, the Bridegroom, and the Church, his Spouse,
there is the one same Spirit who governs and guides us for the
salvation of our souls. For it is by the same Spirit and Lord of ours
who gave the ten commandments that our holy Mother Church is guided
But I am also a physicist with a passion for precision as the data allows. My model for a scientist is St. Ignatius who counts the number of times he fell into a particular fault per day by writing dots in a paper and observing how the number of dots decrease as the days go by. St. Ignatius is one great observer of the motions of his soul that the Society of Jesus he founded became one great network of observatories for observing the motions of the world–the oceans and winds, the moon and stars. The Jesuits are the pioneers in many branches of physics because their mission is to go to the frontiers of knowledge and the crossroads of cultures in order to convert the world for Christ. Seismology was dominated Jesuits during its early development and Padre Faura of Manila Observatory made the first prediction of typhoon tracks in the country. As a tribute to their scientific work, 35 lunar craters are named after Jesuits, with one of the largest named after Fr. Christopher Clavius, SJ, the architect of the Gregorian calendar we now use and a scientist who was treated with great respect by Galileo.
As a Jesuit-trained lay physicist, I am married to my profession, so to speak, and I am faithful to my craft. What I write as a physicist, the others can verify even if they are not Catholics. What I compute is to the best of my knowledge using the available data and the time constraint–I have to publish my estimate the next day. More precise estimates require days or weeks of work. I hope somebody can correct me and present a more precise estimate of crowd size during the Aug. Anti-RH Bill rally using more accurate data and better methodology.