Seth Guikema and The Guikema Research Group have recently had two papers accepted for publication. The first, accepted by Scientia Horticulturae, is “Predictive Models in Horticulture: A Case Study with Royal Gala Apples,” written by Guikema, Tom Logan (IOE PhD student), and Stella McLeod. The second paper is titled “Statistical Analysis of the Effectiveness of Seawalls and Coastal Forests in Mitigating Tsunami Impacts in Iwate and Miyagi Prefectures” and was written by Guikema, Roshi Nateghi, and Jeremy Bricker. It will be published in PLOS One.
Dr. Guikema moved to the University of Michigan from Johns Hopkins University in August 2015. The Guikema Research Group currently consists of one postdoc, two PhD students, and approximately eighteen undergraduate and masters research students at the University of Michigan and two PhD students and one undergraduate student at Johns Hopkins University.
“Statistical Analysis of the Effectiveness of Seawalls and Coastal Forests in Mitigating Tsunami Impacts in Iwate and Miyagi Prefectures” came about due to a NSF grant. One of The Guikema Research Group’s partners on the grant is Tohoku University in Japan. Guikema says there’s disagreement in Japan about whether seawalls help or hurt in the event of a tsunami. “One argument is, you put up a big seawall and it keeps water out. Good, so it helped. The other argument is, you put a big seawall up, people move in behind it, and you’ve got a lot of people behind it and you’ve got an even bigger event and a lot of people die and more houses are destroyed,” he explains.
Because Japan is so good at archiving data, Guikema says, “We were able to find data at the city level on the number of deaths, the acres of forest destroyed, and, in some cases, the number of cows killed by tsunamis going back to 1896.” Once that data was translated and put into Excel, the group was able to look at whether death rates went up or down with higher seawalls based on the historical data.
While tsunamis may not seem that closely related to the subject of his other paper, Guikema says the two papers actually use similar methods of predictive data modeling. “You have a data set on something that you’re interested in knowing about or predicting. For example, with tsunamis, we actually had two models – one for death rates, so the fraction of people in the impacted area that died, and the damage rates, so the fraction of the houses in the affected area that were damaged. So we have data at the city level of the damage rate in each of those past events, and then we also have things like how high was the tsunami, how high were the seawalls — both maximum and minimum height because they’re not uniform across the city –, how big was the flooded area. We also had a bunch more stuff like time of day because when it comes at night it’s more dangerous than when it comes during the day because people aren’t awake to see it. So you’ve got all this data and then it’s an issue of finding the right model that relates that response variable, the thing you’re interested in, to all these other variables.”
The same approach can be utilized in predicting apple size, for instance. Guikema says “Predictive Models in Horticulture: A Case Study with Royal Gala Apples” was the first, and may be the only, time he’s co-authored a paper with one of his PhD students and that student’s mother. The idea came about because Guikema’s PhD student Tom Logan hails from New Zealand and his mum, Stella McLeod, is the research and development manager for one of New Zealand largest apple exporters and had a lot of data on apple size. What started as Logan’s class project (in the equivalent to Guikema’s new IOE 691), turned into a research paper.
Apple exporters have to choose markets to sell to based on fruit size and negotiations begin months prior to harvest. Correctly predicting the size of a harvest is important. The orchard previously had an informal procedure for estimating apple size. The research group came up with a model that predicts apple size at harvest for Royal Gala apples in New Zealand.
Guikema says the two papers are, “good examples of the breadth of interdisciplinary work that we do in [The Guikema Research Group]…. We do a lot related to risk analysis, natural hazards, terrorism, and disasters…. We work on national resource management, which involves applying IOE decision analysis to things like Great Lakes restoration…. We also do a lot with power outage forecasting for storms…. A lot of it is data driven and predictive modeling across the range of things…. So a lot of it is coming up with new methods to try and analyze these data sets, using existing methods when we can, and just doing fun, diverse stuff.”