Group Project Profile: Ragweed in a Changing Climate

by Linn Jennings, Laura Hancock, and Samuel Safran

Ambrosia artemisiifolia, better known as common ragweed, is a leading cause of hay fever allergies. It grows in disturbed areas, like roadsides and abandoned fields. Increased atmospheric CO₂ has been shown to increase the pollen production and growth of ragweed. Thus, with predicted changes in land use and climate, pollen production of common ragweed is likely to increase. Our team carried out three experiments – a presence/absence study, a demographic study, and greenhouse experiment – to collect data that will be used to develop maps of allergy risk under both current and future environmental conditions.

Greenhouse Experiment - Laura

Hoop houses were used to create three CO₂ treatments– 400 ppm, 600 ppm, and 800 ppm. These levels correspond to current carbon dioxide concentrations and concentrations predicted to occur over the next 100 years. Growth, morphology, and reproduction characteristics of 1248 individuals from 24 total populations in New York, Massachusetts, and Vermont were analyzed to show how, and if, these characteristics differed between treatments and between ecotypes from the different states.  We found that there were distinct qualities to these three different ecotypes and that they reacted to the CO₂ treatments differently. Specifically the Vermont populations, which are from cooler, less urbanized environments, had distinct characteristics when compared to the more similar New York and Massachusetts populations.

Ragweed Demography - Linn

Our research focuses on the current phenology and life cycle of A. artemisiifolia in 24 populations located along rural to urban and warm to cool gradients from Boston to the Berkshires in Massachusetts.

We collected data on growth rates and flowering for each population. Other data, such as percent cover in each plot (3 to 5 plots per population) and land cover, were collected to evaluate the impact of surrounding vegetation and fine scale land cover on plant phenology. The data demonstrate a trend of taller plants in the cooler sites, less urbanized sites, which might indicate a plant selection for shorter plants in regions that are mown more frequently. Further, the most important predictor variables found in the presence-absence survey were not significant predictors of plant size and flowering. When modeling future allergy hotspots, different predictor variables will be needed when modeling the presence and absence of ragweed and when modeling the plant size and flowering time of ragweed.

I set up 24 sites across the state with the help of my group partners, Sam and Laura, and with the help of my mentor, Dr. Sydne Record. Each week, starting after 4-5 weeks of building and working in the greenhouses at Harvard Forest, I drove to each of the sites to collect growth, abundance and flowering data. I spent about 3-4 days a week in the field, and I would enter and analyze my data on Thursdays and Fridays. I spent lots of time out on the road, but I had a great time traveling around Massachusetts.

Presence-Absence Survey - Sam

I set off (GPS in hand) to Vermont, New York, and Massachusetts in search of the pesky plant. My mission: to sample hundreds of randomly generated points for ragweed abundance and local land cover characteristics. Being random, the points took me to all sorts of interesting locations. Trudging through freshly manured corn fields, tiptoeing across forested swamps, forging mountain rivers, and pacing back and forth in front of Brooklyn store-fronts all afforded me much time to ponder the sorts of habitats and climates that ragweed favors. Thankfully, once all the data were collected, I also had the power of dynamic geographic and statistical tools to help answer this very question. 

By pulling bio-climatic data about each point from WorldClim layers, demography data from Census layers, and other information from my own layers (distance to nearest road, slope, etc.), with GIS, I was able to build a large data set for the ~200 points sampled to date. To analyze these data, I ran a bagged classification and regression tree (CART) analysis to identify the most important variables for predicting ragweed presence/absence. The analysis indicates strong positive correlations between observed edge habitat and ragweed presence as well as between observed forest habitat and its absence.  The strongest predictor of ragweed presence was found to be distance to nearest road, a variable derived remotely with GIS. Some clear climate effects—including average precipitation and temperature during the growing season—were also shown to be highly relevant predictors of ragweed presence.  

In the end, this fancy regression gives us preliminary framework for modeling ragweed’s distribution across the whole New England landscape. By holding a portion of the data I collected out of the initial analysis, I was able to run it through the model after the classification tree was developed.  This test tells us our model correctly predicts ragweed presence/absence more than 70% of the time.  Exciting stuff! This dataset will be used to test the accuracy of free and simple methods for identifying ragweed habitat, to model its distribution under current climate conditions, and to map regional allergy hotspots under future climate scenarios at a scale relevant to an individual’s exposure to pollen.