10:00 to 11:15 am, Friday November 19, 2004
On Friday morning, Nov 19, 2004, I visited the office of a Professor and Post-Doc who are working on studying animals in the Jasper Ridge Biological Preserve. They study how resources and terrain affect population of mammals. How do river locations affect deer population? How does the deer population affect mountain lions?
Jasper Ridge Experiments
They are designing experiments for the next couple years. One experiment will be run in Jasper Ridge, and at the moment they are evaluating camera technology that will help them do their studies. They will most likely use a digital setup, because that allows them the least pain in transferring photos from the capture site to their computers, where they will do the identification. The digital capture sensor is a housing with a motion sensor and digital camera (as we’ve seen in the visit to Jasper Ridge).
One issue is that for some reason, with the current technology, digital is much slower than 35mm. It takes .6 seconds for the motion sensor to tell the 35 mm camera to snap. It takes about 2 seconds for the motion sensor to tell the digital camera to snap. They are worried that the animal may be gone from the field of view of the camera by the time the camera snaps.
They are using a standard hexagonal sampling pattern in JRBP. This pattern is used by many scientists, but at a larger scale. At JRBP, the distance between sites 1 and 2 might be 360m, but in larger experiments, it might be several km. They plan to work with a larger community of scientists, and JRBP itself might be a single node in a larger sampling pattern.
One site is a pair of cameras, so that you can get more angles on the animal, and hopefully be able to identify it better. All the data will eventually be put into these huge matrices… N animals by 360 days (by other dimensions too), showing how often and where the animals appeared. As can be expected, these matrices will be pretty sparse. There will be other data to correlate with this appearance information. Data such as temperature, humidity, time, etc from met stations (meteorological) on the JRBP sites will be correlated with the data, to find meaningful relations.
Another way to track animals (instead of capturing them on film), is to physically trap them, mark them, and rerelease them. Then, you can sample the population by setting traps again, and seeing which of the ones you marked before are in this new sampling. That is, if most of the ones you caught before are caught again, then the population is probably small. But if you catch a whole new batch, then the population is likely large. You can also track animals by tagging them with radio tags (like a gps transmitter or LoJack on each animal).
They are interested in creating useful data visualizations out of their data. Visualizations might include maps, time series data, anything that can help them reason about their data.
Summary: Their next experiment will generate a LOT of data, and they would be interested in novel data visualizations and human interfaces that can help them navigate and solve problems in this vast sea of data.
Tropics of Mexico
In the next several years they will be running a large scale experiment in Mexico’s tropical jungle area. This area is very dense with trees. Their experiment will span 50 ha (500 000 m2). They’ll put sensors at each tree, and visit periodically (on the order of months and years) to check on the progress of trees, leaves, etc. Think of it as Conway’s Game of Life (http://en.wikipedia.org/wiki/Conway%27s_Game_of_Life), but on a huge scale. They plan to tag each tree in this area, and keep track of many variables, resulting in analysis of th mortality rates depending on density of neighbors.
While their long-term experiment is running, they will allow scientists from all over the world to come into their plot and do experiments within the context of the larger experiment. Thus, one could track a particular species of insect based on tree density.
Summary: They are planning a long term experiment that has the issues of data sharing/integration, and huge data sets.
One way to check how often animals visit an area is to see how eaten the leaves are. The biologists explained that they can visit a plot of land, tag the leaves, and then come back later to see how much of each leaf was eaten. They use software to help them determine the area of the leaves that were nibbled on. These software use some human intervention, but basically calculate a contour of the leaf, and how much of the area is green.
Summary: Better software could possibly be made for leaf analysis.
Long Term Data
A big issue that faces biologists (including our friends at Cal Academy of Sciences) is the integration and sharing of data sets. The two biologists that I spoke with are working within a community to make sure their work in the Tropics will be shared and accessible by scientists world-wide.
Summary: Like most interesting problems with large data sets, collaboration and standards are big issues.