A multi-institutional research team, led by researchers at the Virginia Institute of Marine Science, set out to characterize some of the uncertainties associated with projected changes in low-oxygen conditions in the Chesapeake Bay. Their study analyzed the methodological application of a future climate scenario to see whether a certain approach could ultimately impact estimates of future estuarine health, and found some surprising results.
Low-oxygen, or hypoxic, conditions are a regular occurrence in this large, shallow estuary on the U.S. East Coast, and while recent efforts to undo past environmental harm have shown measurable improvements, these hard-fought gains may be at risk due to a combination of factors driven by anthropogenic climate change. Increasing air temperatures can directly decrease oxygen levels by reducing the solubility of the dissolved gas in water, and can indirectly decrease oxygen levels by speeding up the metabolism of microbes that use oxygen to consume decaying organic matter. Additionally, increasing precipitation in the Chesapeake Bay watershed can amplify runoff of nutrients from agricultural and urban sources, helping to fuel the growth of estuarine algae that eventually die, sink to the bottom, and are decomposed, reducing dissolved oxygen even further.
Hypoxia has been well studied for decades in the Chesapeake Bay, and several recent modeling papers have also analyzed how changes in temperature, runoff, and sea-level rise might affect future oxygen levels. However, all previous estimates that applied outputs from climate models utilized what is known as a delta approach, wherein the change of a variable like temperature is added onto a historical scenario to get an estimate of how future scenarios may differ. This method is much less computationally demanding than running a continuous simulation for multiple decades, but may provide an inaccurate picture if the future interannual variability and frequency of weather events look substantially different from historical observations.
The findings of this study sought to answer this question directly, asking how big of a difference the choice of method made for the same climate projection applied to the Chesapeake Bay. The authors first ran a simulation of watershed and estuarine physics and biogeochemistry continuously for more than 80 years (taking more than 40 days in real-world time), for the first time ever in this region. This experiment was used as the baseline gold standard to compare against two shorter experiments, one applying the previously mentioned delta method and a similar simulation using a “time slice” methodology. Instead of adding a perturbation to a baseline scenario as in the delta method, this time slice approach simply applies the same future conditions of air temperature, precipitation, etc… as in the continuous simulation, but without simulating the intervening years.
Comparison of the three different methodologies (Delta, Continuous, and Time Slice) compared in this research.
Given that all three experiments had the same baseline conditions, and the future delta method scenario was based on inputs from the continuous experiment, the researchers expected to find only minor differences among these model simulations. However, the results showed that the delta method increased hypoxia by an amount double to what was projected by the continuous and time slice experiments. The authors were able to rule out the possibility of physical changes causing this discrepancy, as all experiments increased water temperatures, salinity, and sea level by approximately the same amount. Digging further into the results, the researchers found that inputs of terrestrial sources of nitrate were to blame for this dramatic change in downstream effects of hypoxia.
Climate models in this region tend to forecast an increase in the number of intense precipitation events in the future, meaning that a continuous or time slice method will have more intense stormy events but lower levels of precipitation on drier days. The delta method, by comparison, increases levels of precipitation across all days, tending to more consistently increase soil moisture while leaving the underlying climatology of downpours unchanged. This modification to runoff and soil moisture levels then led to enhanced biogeochemical activity, increasing the amount of nitrate released from the watershed to the Bay.
Simplified cartoon of terrestrial and aquatic changes that occur in the Continuous (left) and Delta (right) experiments.
These results demonstrate the importance of considering long-term ecosystem memory in future projections. The continuous experiment showed lower levels of nitrate concentration than either the delta or time slice experiments, as the dynamic terrestrial model slowly but surely converted much of this source of inorganic nitrogen to less bio-available organic nitrogen. This result also contributes to the large difference in levels of hypoxia, and highlights the need to consider long-term terrestrial processes when estimating biogeochemical impacts on coastal water bodies.
The paper’s findings call into question the reliability of future scenarios that use the delta methodology for estimating marine biogeochemical processes with a long ecosystem memory. In this study, changes in nutrient inputs have been shown to affect levels of algal growth and hypoxia, but this result will also affect how carbon is taken up and released from coastal water bodies. Increases in carbon uptake may help to acidify coastal waters, further stressing fish and shellfish populations that have less oxygen-rich water to use as habitat. Future estimates of ecological interactions that are reliant upon projections of what occurs at the base of the food web are also likely to be affected.
In the future, studies of the Chesapeake Bay may benefit from a more collaborative, multi-institutional approach that uses computational resources to simulate a broader ensemble of future continuous scenarios. Such simulations could then be further explored to identify some of the sensitivities of different numerical models and narrow the range of uncertainty in future results. This approach has already been undertaken over the past 15 years among researchers in the Baltic Sea region, and could be applied in multiple other areas worldwide where anthropogenic nutrient runoff has expanded the spatial range of low-oxygen waters. Going forward, scientists should use appropriate caution when choosing a climate scenario methodology, as the seemingly trivial decision can have broad impacts on coastal biogeochemical responses due to changing conditions in the marine and terrestrial environments.