Monday, April 29, 2013

Dr. Delmelle is Using Engineering To Solve the World's Problems

The spread of disease has always plagued mankind. Great epidemics have ended millions of lives in the past, making the study of such diseases all the more important. But new problems also plague us as we grow and expand in the world. Overpopulation in many regions of the world has led to transportation challenges and complicated the optimal siting of facilities or the delivery of services such as hospitals and police. Then there are smaller problems in the world, such as how useful the new bike-sharing system in Charlotte actually is. All of these seem so unrelated, but Dr. Eric Delmelle sees the connection: geography. And, using engineering techniques, he and his team plan to help solve all of these problems, one by one.

When it comes to epidemiology, Delmelle's interests are in the modeling of vector-based diseases, specifically dengue fever, which is spread by mosquitos: "These diseases have a particular spatial and temporal signature” ," Delmelle noted.  "These diseases create geographical patterns and clusters in specific areas, which is critical for prevention purposes" says Delmelle. All of this mass of data comes from the Health Ministry of Columbia, specifically for the city of Cali. "It is a large, dynamicmetropolitan area that has seen a lot of migrants moving in with poor sanitation infrastructure, which makes it a particularly interesting environment," he noted.

 Figure 1 displays dengue fever cases in 2010, which was considered an outbreak in dengue fever. Spatial and space-time smoothing techniques are used to extract meaningful patterns of dengue intensity.

Dengue fever cases in the city of Cali, Colombia (a), spatial patterns in (b) and space-time patterns in (c).

The purpose of this research is to better understand the spatial patterns of infectious disease and predict where they will reemerge. Though we may not see dengue fever as a critical problem in the United States, other infectious diseases are at alarming levels, such as West Nile Virus. With a changing climate, the risk of similar diseases only goes up, especially in the southern states, such as Florida. Along those lines, Delmelle also collaborates with Dr. Eastin, an associate professor of meteorology in the Geography and Earth Sciences Department in order to help predict dengue fever outbreaks in the city of Cali. For instance, weather forecasts may help predict accurate counts of dengue fever, helping to increase awareness among population of an imminent risk. Using an autoregressive model, the team predicted a significantly high number of dengue cases in 2013. By mid-February 2013, the city of Cali had reported 1339 cases, with three individuals dying of the disease.

Another Diagram Explaining the Space-Time patterns of the dengue fever
This research stream is not just for research sake. Delmelle and his team take all of their findings back to the individuals they have worked with in Columbia. "We have the chance to visit numerous times on-site, receive critical feedback from local authorities on the techniques we use and to validate our results." Establishing a connection with local decision makers is a very important part of the process to Delmelle. "We do not tell people what they should do, but we do inform them on the risks posed by infectious diseases. It's important to see that the results of our models can be beneficial to the local communities for prevention purposes. ”

Vector born illnesses are not the only research problem Delmelle is interested in. Using similar spatial modeling techniques, Delmelle also tackles general problems in urban areas."Several problems we study are of geographic nature" Delmelle tells me. By utilizing engineering techniques –such as operations research -, Delmelle and his team look to optimize particular problems. In the case of infectious diseases, the optimal space-time allocation of spraying efforts may help reduce the magnitude of the infection.

Yet, the same techniques can be used for other, more basic problems. Delmelle brings his research approaches to look at local issues, helping the Charlotte-Mecklenburg region with multiple dilemmas. One such major issue is the question of where to site schools in a city that is only growing in population, but not evenly across the county. By attacking the problem as a quantitative optimization problem -- a number problem -- Delmelle and his team are able to utilize data to find the most optimal solution. "We were able to analyze data up to 2008 and predict where schools should be closed in the region and we've been fairly accurate," says Delmelle. By assigning an optimization value to every school, the team was able to predict, for the most part, which schools would be closing and where new ones should be built. His team was also able to help recommend adding more modular classrooms to help meet the increasing school demand, to keep classroom size stable while the new schools are built.
A Graphic Depicting the Work Delmelle Has Done to Help Place Schools in the Best Locations

A similar approach was conducted to identify bus stops redundancy in Charlotte. Delmelle collaborated with the Charlotte Area Transit System (CATS) helping to tackle where the best locations for each bus stop would be, using the optimization of  individual accessibility and low operating costs as their main objectives. "The city took great care to listen to our research," Delmelle said. Currently, Delmelle and his students are also evaluating accessibility to public parks with Mecklenburg County by different modes of transportation (car, public transit, biking, walking). These results will inform park decision makers on where to locate new parks and which ones are in immediate need to upgrading its amenities.  The next venture into help the Queen City would be to help optimize the new bike racks the city has placed. "We just received some of the data and it's all fairly new, so now we can run simulations to see where the best new places to expand would be," says Delmelle, rather excited by the problem.

No comments:

Post a Comment