For Bohrer, Climate Modeling Like Forecasting Grocery Use



By Claire Cusick



Gil BohrerGil Bohrer is pursuing a doctorate in environmental engineering, and working on a climate modeling project in the Panamanian rainforest, but his job has as much to do with atmosphere physics as with ecology.

He works on a research team that is studying wind dispersal of seeds in the Panamanian rainforest. He is creating a computer program that will enable a regional meteorological model – one that covers hundreds of square kilometers -- to simulate wind dispersal of seed in a resolution of just a few square meters. But this 33-year-old spends his days with algorithms, not with trees.

Bohrer’s modeling contribution will incorporate data from small areas into a regional model. Others in the research group can then apply the model to climatological questions - how changes to the land surface – including trees, fields and cities -- affect an area’s climate, wind, clouds, rainfall and trees.

"The current representation of the land surface – whatever covers the soil – used by the regional model is that of a single layer of 'plants,'"he said. "This works when the resolution covers hundreds of meters, but when the resolution is just a few meters, you cannot approximate a forest that is much taller than the lowest layer.

"I will write the part of the program that vertically splits the representation of plants and their contribution to the atmosphere dynamics – adding humidity, intercepting sunlight and heat, and blocking wind flow – to thinner layers from the top of the canopy to the ground, and thus allow the model to work in higher resolutions than it could before,"Bohrer said.

So instead of studying one layer that is fifty meters vertical – which might include the entire forest canopy -- the model could support many thinner layers within the forest canopy.

To study a smaller area in higher resolution takes more computation resources in order to take into account all of the interactions taking place. It’s all a balance, Bohrer said, making decisions about "what you approximate, what you simulate, and what you neglect."

"Then, once you have the model, you never trust it without validation,"he said. It is always a work in progress – it can always be improved. So once the model is built, he will constantly compare what his model predicted for a given day with actual data from that day.

"If there is no match, my model is wrong,"he said, although wrong may not be the best way to describe what’s happening. The mismatch may because the model does not yet incorporate enough variables to simulate a real day.

The desired outcome of this research is a better high-resolution model. Such a model has application far beyond the seed dispersal. Models can help city planners understand how air pollution spreads and make decisions about how to shape, control or regulate industrialization. Emergency response teams could use such a model to simulate a chemical or biological flume from an unconventional bomb, and figure out how best to respond to such an event. Farmers would gain insight into how to most effectively irrigate a field.

Bohrer also hopes for a personal outcome from his research–— namely an academic or faculty position. He’s not locked into environmental engineering, either. He likes any kind of complex modeling because such work has a multidisciplinary aspect to it.

In the meantime, this research experience combines his background in ecology, love of the outdoors, and interest in how things work.

Before coming to Duke, Bohrer built computer models for a private company in his home country of Israel that monitored storage and inventory for large retailers.

In other words, he tried to predict, for example, how much fat-free cream cheese shoppers would buy on a typical day. But what makes a typical day in the dairy aisle? And what makes people buy more or less fat-free cream cheese – price, freshness, flavors? He had to take all that data and balance those factors, to build a model that helped retailers decide how much product to keep in stock. Then he had to test the model: does a coming ice storm affect the purchase of fat-free cream cheese the way it would bread and milk?

"The system will observe last year’s holiday sales and automatically account for last year’s behavior,"Bohrer said. "The trick is exactly how much more to order, so you don’t have too much and it goes bad."

Bohrer remembers some interesting tidbits from that job. During the Israeli holiday of Shavuo’t, cream (not cream cheese) is a top seller, because many people use it to make cheesecake. Hard cheeses – the ones that be used on an appetizer cheese tray – also have brisk sales. But not milk.

"Shavuo’t for the dairy retailer is like Christmas for Hallmark,"he said.

To celebrate the Jewish New Year in September, it is traditional to eat apple and honey, he said. He noticed that prior to the holiday, there was a huge increase in apple baby food, which meant that people were getting their apples in different ways.

Then, his modeling work was for a "consumer climate."Now, the application is the Panamanian rainforest.

When deciding on where to pursue his doctoral work, Bohrer heard about the work of Professor Roni Avissar, who is now his dissertation supervisor, from colleagues at Ben Gurion University in Israel, where he earned a M.Sc. in evolutionary ecology.

Coming to Duke provided "a good opportunity to combine the environment and the science,"Bohrer said. Having specialized in ecology, he wanted to change his emphasis.

"In biology, the approach is more descriptive, and they try less to fit reality to a mathematical model,"he said. "I was more interested in the modeling."

His first year here reflected the choice of switching emphasis. He had to improve his knowledge in various areas, so he took classes in higher math and fluid physics. In his second year, he finished all his graduate classes, and could then begin working hard on the research. He hopes to spend a total of two to three years working on the project, and graduate in 2007.

Bohrer’s M.Sc. thesis work focused on fungi in the Kalahari Desert. But that experience taught him that he was interested more about the underpinnings of the science – taking the reality and putting it into the computer. Having traveled in Asia, Europe and Africa, he also wanted to experience both living and working in another new place. He arrived in Durham in August 2002.

Now, his research will take him to the tropical climates of Panama, and to the still-pretty-humid climate of Duke Forest. "It’s diversity,"he said. "It’s fun."