Skip to main content

Post by Uma Shankar 


Uma Shankar is a Research Associate with the UNC Institute for the Environment’s Center for Environmental Modeling and Policy Development

The current severe drought conditions in western North Carolina are resulting in significant environmental and economic impacts in this region.  In particular, the widespread uncontrolled wildfires in our state, and the Southeastern region as a whole, are of serious concern.

In North Carolina more than 40,000 acres have burned, while estimates put the number at more than 120,000 acres for all southeastern states.  These wildfires are endangering lives, threatening public property, and impacting air quality. Over 200 hospital admissions were reported in Tennessee due to respiratory illnesses since Nov. 1, and a number of deaths reported in the recent Gatlinburg, TN wildfires.


Recent wildfire near Gatlinburg and Pigeon Forge, Tenn. (Jessica Tezak/Knoxville News Sentinel via AP)

For more than a decade I have been partnering with the U.S. Forest Service and other agencies to develop integrated computer modeling approaches that can help predict the location and frequency of wildfires in the Southeast. Much of that research involved the development of statistical models to forecast what forest conditions would look like over the next fifty years accounting for such factors as climate change, and population and economic growth. This work has been a true collaborative effort among federal agencies and academic scientists. An important outcome of this work has been the estimates shown in Figure 1 of the annual areas burned in the Southeast at the country level due to lightning-caused and human-caused wildfires from 2015 – 2060 (Prestemon et al., 2016).

picture2 picture1

Figure 1. Projections of lightning-ignited (L) and human-ignited (R) wildfires, 2006 and 2010-2060, aggregated over all counties for the Southeastern U.S., including upper and lower 90% bounds of 2,250 Monte Carlo iterations of 9 climate model realizations (from Prestemon et al., 2016)

Figure 1 shows that while climate change could drive up lightning-ignited burned areas in the future, the areas burned in human-caused wildfires will likely decrease or stay nearly constant in the future. The latter trend is partly because of expected socioeconomic growth in the region allowing greater availability of resources to suppress and manage fires, as well as breakup of forested land through development. Overall, this research published in Prestemon et al. (2016) shows that total wildfires (a) increase by 4% in the Southeast from 2056-60 compared to 2016-20, and (b) there is wide variability state by state in these total wildfire changes, e.g., -47% to +30%.

The long-term forecasting of wildfires frequency and intensity is complex due to the inherent randomness of wildfires, the variation in vegetation and fuel loads, and climate variability.  Specifically factoring in climate variability has its own set of challenges, as global climate models are not able to represent with sufficient detail the regional differences and unique weather factors, such as rainfall trends and patterns in the Southeast.

To address these challenges our research team has used various methods to downscale global climate model outputs to understand climate change impacts at the regional level. Our use of an ensemble of model realizations rather than a single scenario simulation reduces the uncertainty (and increases the reliability and accuracy) in regional climate change projections.

We have used statistically derived projections of annual areas burned from 2015-60 to constrain stochastic model estimates of wildfire ignitions and daily areas burned, based on flammability criteria and amount of available fuel. Figure 1 shows the projected annual areas burned due to lightning and human-caused ignitions. Figure 2 shows how these burned area estimates then translate into emissions of fine particulate matter (termed PM2.5) in 2043, 2048, 2053, and 2058, relative to their 2010 levels, assuming no changes in the amount of fuel available. They can then be used to project PM air quality in these future years.


Figure 2. Comparison of future projections of annual Southeast-wide PM2.5 emissions against the (constant) 2010 EPA National Emissions Inventory level (circles) using statistically downscaled meteorology (squares), and dynamically downscaled meteorology with the Weather Research and Forecasting model (diamonds), for a constant fuel load

Building on this work, our research is now focusing more specifically on evolving fuel loads, i.e., leaves, litter and dead wood that burn during a wildfire. The recent wildfires in the western part of the state, as well as in Georgia and Tennessee have been fueled by the abundance of dry and flammable materials lying thick on the forest floor.  One of the factors being considered in our current research is the variability of fuels in Southeastern forests and how that may change in future climate scenarios. Forest Service experts consider fuel loads to be the biggest source of uncertainty in predicting wildfire occurrence and magnitude, and we expect this area of research to be of importance to land managers.

This new research is intended to enhance and make the results of our model simulations more reliable.  This research has in the past and will in the future assist and inform land managers and air planners by providing improved assessments of wildfires and their impacts to guide their resource management decisions under evolving climate and socioeconomic conditions in the region.

This research was highlighted in the Washington Post, Nov. 30: The raging wildfires in the Southeast could be a glimpse of what’s to come

You can learn more about this research at the following journal article:  Prestemon, J. P., U. Shankar, A. Xiu, K. Talgo, D. Yang, E. Dixon IV, D. McKenzie, and K. L. Abt, 2016: Projecting wildfire area burned in the south-eastern United States, 2011-60, Int. J. Wildland Fire,




Comments are closed.

Follow this website

Get every new post delivered right to your inbox.