Biology Climate Change

O Vibrio, Vibrio, wherefore art thou Vibrio?

Article: Urquhart EA, Zaitchik BF, Waugh DW, Guikema SD, Del Castillo CE (2014) Uncertainty in Model Predictions of Vibrio vulnificus Response to Climate Variability and Change: A Chesapeake Bay Case Study. PLoS ONE 9(5): e98256. DOI: 10.1371/journal.pone.0098256

Why do we care about Vibrio vulnificus?

Figure 1: Vibrio vulnificus under a microscope. Source: safeoyster.org
Figure 1: Vibrio vulnificus under a microscope. Source: safeoyster.org

Vibrio vulnificus (hereafter just called Vibrio) is a bacteria naturally found in estuarine and coastal environments. Although it may vary place to place, you can likely come across this bacteria if this simple environmental recipe is in place: a salinity between 5-25 practical salinity units and a temperature above 15°C. Many scientists have been working on models to predict when Vibrio will be present in coastal waters using salinity and temperature in those ranges.

So why do we care about modelling Vibrio’s presence in our beach water? The answer is simple: it is a potentially deadly pathogen! Someone who becomes infected with Vibrio is likely to get an infection, ranging from a painful skin infection upwards to a life threatening blood infection. In fact, Vibrio vulnificus is a relative of cholera. How do you get infected with Vibrio? The two main ways are by eating contaminated shellfish or by swimming with an open wound when the bacteria is present. As Figure 2 shows, the number of reported Vibrio cases have been increasing. This increase in Vibrio cases could be due to a number of reasons (for example, more people going to the beach), but scientists are beginning to believe that climate change is to blame.

Figure 2: The reported cases of Vibrio in Maryland from 2001 to 2008. Source: MD Dept. of Health and Mental Hygiene
Figure 2: The reported Vibrio cases in Maryland from 2001 to 2008. Source: MD Dept. of Health and Mental Hygiene

In Chesapeake Bay, the water has warmed by an average of 0.3-0.4°C per decade and changes in precipitation could decrease the winter and spring salinity by 7%. What this means is that climate change may be creating conditions which allow Vibrio to exist in the water for a longer amount of time.

In this study, Urquhart et al. used three different statistical models to predict the probability that Vibrio would be present in the upper Chesapeake Bay. Ultimately, the creation of an accurate model, or combination of models, will help us determine how climate change will affect the trends and distribution of this dangerous pathogen.

The Models and Data

 Three different probability models were used to investigate how climate change is affecting the presence of Vibrio in the upper portions of Chesapeake Bay. All three models used temperature and salinity as the only indicators for Vibrio’s likelihood of appearance, but each varied either structurally or with the data used to “train” the model.

Figure 3. Map of the study area, showing contours of average surface water salinity. Dark markers represent in situ monitoring stations used for each of the sub-regions in this study: upper (star), mid (circle), and lower (square).
Figure 3. Map of the study area, showing contours of average surface water salinity. Dark markers represent in situ monitoring stations used for each of the sub-regions in this study: upper (star), mid (circle), and lower (square).

The first model used was named the NOAA_GLM; this model is a generalized linear model (hence the GLM acronym) and was trained with 235 samples collected during July and October of 2007 and April, July, and October of 2008. Simply put, a generalized linear model uses an equation of a line (remember y=mx+b from algebra?) using surface water temperate and salinity as the independent variables. The model is evaluated using the collected samples (which include Vibrio) to train it. Models are always better when real data is incorporated.

The second model had the same structure as the NOAA_GLM in that it is also a generalized linear model. The JHU_GLM (where JHU is John Hopkins University) differs because it is trained with 148 samples collected during July and September 2011 and March through June of 2012. The last model is a general additive model called JHU_GAM. This model uses the same sample data as JHU_GLM, but has a smoothing function which makes it more flexible; it also incorporates non-linear responses, or relationships that do not follow that y-mx+b.

So, the NOAA_GLM and JHU_GML models both had the same structure while the JHU_GLM and JHU_GAM both used the same data for training.

The environmental data (surface water temperature and salinity) used to run the models came from bimonthly samples from 16 monitoring stations in Chesapeake Bay (Figure 3). This data set is from the historical environmental data collected from 1985-2013 by the Chesapeake Bay Data Program. This data helps to provide the current climate conditions in the upper Chesapeake Bay that will help predict how future changes in temperature and salinity will effect Vibrio presence.

The Findings

Unfortunately, the models all gave different results on the predicted distribution of Vibrio with the changing environmental data (temperature and salinity). Figure 4 shows this wonderfully. In the month of June, the JHU_GAM would predict that the presence of Vibrio is highly likely while the NOAA_GLM model would suggest Vibrio would be less likely to be in the bay. Because there is so little data available on Vibrio’s current patterns and distribution in Chesapeake Bay, it is impossible to tell which model is the most accurate.

Figure 3. Upper Chesapeake Bay monthly Vibrio probability hind-casts for April through October 2012, for NOAA_GLM, JHU_GLM, and JHU_GAM methods.
Figure 4. Upper Chesapeake Bay monthly Vibrio probability hind-casts for April – October 2012, for all model predictions. The warm colors indicative that it is highly likely that Vibrio would be present  while the cool colors suggest it is less likely that Vibrio would be present.

Another big difference in the model predictions is the importance of temperature on Vibrio’s probability of presence (Figure 5). For example, the NOAA_GLM suggests that the presence of Vibrio is largely controlled by temperature increases. So a warmer year, as predicted by NOAA_GLM, would result in peak Vibrio concentrations while the JHU_GLM predicted the opposite trend. In fact, all three models had a different response to temperature increases.

Figure 5. Monthly climatology of temperature and V. vulnificus probability for each method in the upper (a), mid (b), and lower (c) regions of the Chesapeake Bay. Peak temperature observations by year versus V. vulnificus probability for each method in the upper (d), mid (e), and lower (f) regions of the Chesapeake Bay. Trend lines are included for each method’s observations.
Figure 5. Monthly climatology of temperature and V. vulnificus probability for each method in the upper (a), mid (b), and lower (c) regions of the Chesapeake Bay. Peak temperature observations by year versus V. vulnificus probability for each method in the upper (d), mid (e), and lower (f) regions of the Chesapeake Bay. Trend lines are included for each method’s observations.

With these conflicting model results, the authors caution at the present on the use of a model to definitively predict how Vibrio presence will change alongside climate change. Ultimately, more Vibrio samples, over a longer time span, is needed to continue to improve these statistical and predictive models.

Significance

 Although the three models gave different, and sometimes the completely contradictory results, one thing is clear: we do not know a whole lot about how climate change will affect the presence of Vibrio in estuarine waters. More field samples of Vibrio, surface water temperature, and salinity are needed to make the necessary improvements on these models. Moving forward, it is important to use more than one statistical model to predict the presence of Vibrio in the bay since model comparison is needed before any validation can be made!

Leave a Reply

Your email address will not be published.