Project Background
Fire has played a major role in determining the distribution of plants across the Coastal Plain of the southeastern US. The extent of fire dependent ecosystems such as the pond pine/ high pocosin forest type at the Alligator River National Wildlife Refuge have been reduced as a result of fire exclusion and land conversion. Wildland fire fuel loading in this area has become a hazard to life, property, ecosystem health, and the habitat of threatened and endangered species as a result of past fire exclusion policies and practices. The President's National Fire Plan is concerned that fuel loads are reaching hazardous levels that can lead to widespread catastrophic wildfires in forest ecosystems and the wildland/urban interface. This wildland fire risk is currently impacting ecosystem management planning throughout the region, and specifically on the Alligator River National Wildlife Refuge and Air Force Dare County Bombing Range in Dare County, North Carolina.
|
|
| Pond pine cane pocosin. The stand on the left had not been treated with fire, the stand on the right had been burned two years prior to the taking of the picture. Click the image to see a larger picture. | |
As areas around wildlife refuges and military facilities continue to be developed, federal and state agencies will be challenged to develop management plans to lower wildland fire risk, while implementing ecosystem management, threatened and endangered species (TES) management and recovery plans, remaining in compliance with federal and state air quality standards, and protecting the public's heath and safety. Wildlife refuges and military training ranges in remote areas, like the Alligator River National Wildlife Refuge and the Air Force Dare County Bombing Range comprise thousands of acres that are experiencing encroachment and fragmentation by coastal residential development. These areas are critical to maintaining and protecting species diversity.
Measurement and Modeling of Down Woody Debris and Fuels
Fuel classification during the last 75 years has evolved from a fire control planning focus to the beginning of predictive fire behavior modeling in the 1970s. Current fuel classification models have focused on the rate of spread, resistance to control, and the flame length of fires in surface fuels. Fire behavior is predicted by land managers with thirteen stylized fuel models (Rothermel, 1972; Albini, 1976). Decision support systems such as FARSITE and the National Fire Danger Rating system are based on the Rothermel's fire spread model and are the basis of predicting fire behavior today. Land managers recognize that these models are limited in their ability to predict extreme fire behavior, persistent fires, and fuel consumption. Some of these limitations are currently being addressed by a fuel characteristic classification (FCC) research project funded by the Joint Fire Science Program (JFSP - http://jfsp.nifc.gov/) (Sandberg et al., 2001). But of the 53-fuelbed types with detailed or general information currently in the FCC, only one forest type found in Dare County has been identified for inclusion in the FCC database.
The availability of fire-spread models has increased the need for quantitative fuel field data. A line-intersect method developed by Brown (1974) has been widely adopted to quantify fuel-loading inputs. The USDA Forest Service Forest Inventory and Analysis (FIA) program recognized the need for extensive information on fuels across the landscape. Fuel field protocols were piloted by the former Forest Health Monitoring Program between 1998 and 2000, and implemented in 2001 on a 1/16th subset of the standard base FIA grid plots (http://fia.fs.fed.us/library/field-guides-methods-proc/). These FIA methods generally partition the forest ecosystem into pools for live trees, down deadwood, standing dead trees, understory vegetation, forest floor materials, and soil. Estimating site-specific fuels from this database has been particularly problematic. The data is not consistently available from the largest inventory data source, FIA, and there is little data on fuel pools in the scientific literature. Additionally the biomass algorithms are based nationally on data collected primarily on western US tree, shrub, and herbaceous species and associated wood density for decay classes.
Land managers in the Coastal Plain of the eastern US recognize four general fuel types on organic soils (i.e., low pocosin, high pocosin, open cane, and overstoried cane). Past fuel and fire behavior research has resulted in only qualitative measures of fuel loads and rates of spread. A more detailed fuel classification based on species composition, standing dead and down deadwood, fuel size classification, understory vegetation, and vertical distribution of fuels would have much more utility than the broad fuel model classification system now in use. Fire in the organic soil areas of the Coastal Plain centers around the frequent and costly blowup wildfires occurring there and the use of fire as a fuel reduction and habitat management tool. Wildfires in this area can under certain combinations of fuel and weather, grow from a low intensity burn to a virtually uncontrollable burn until weather conditions change or the fire has run out of fuel. Control efforts are often hampered by inaccessibility, poor soil trafficability on wet organic soils in the area, and fires that tend to burn deeply into the organic soils. A better understanding of the behavior of fires and the role of fuel loading in fire behavior in the pocosins, especially the factors that contribute to the occurrence of major fires, will contribute to the control of wildfires and the use of prescribed fire as a management tool in the region.
|
| Ground fire at Pocosin Lake NWR, Pungo unit |
Prescribed Fire Emissions Monitoring and Modeling
Biomass burning in the southeastern US can be a potentially significant source of photochemically active and radiatively important trace gases as well as particulate matter (PM) (Vose et al., 1997). Areas burned in the region vary annually, but are typically several million acres per year, resulting in trace gas and PM emissions that range from 2 to 15% of total emissions from other sources. Little data on emissions from prescribed burning is currently available, and this fire type in particular is projected to increase in the southeastern U.S. Emissions of reduced compounds, many of which are air toxins, are thought to be lower during prescribed fires compared to wildfires covering the same area. This is suspected largely because it is known that wildfires occur typically during excessively dry periods when much of the forest floor is dry and susceptible to smoldering incomplete combustion, the source of many toxic compounds.
 |
| Smoke from a prescribed fire impacts a neighborhood adjacent to Croatan National Forest. |
Continuous monitoring of ozone (O3), PM, and oxides of nitrogen (NOx) have shown that air pollutant concentrations are enhanced by forest fire emissions. In the rural environment, the influence of the forest fire on air quality can be detected, and significantly higher (50-150%) pollution levels than seasonal median values have been documented (Cheng et al., 1998). While fire events can cause high transient air pollutant concentrations, for most criteria pollutants, the fire emissions are a relatively small fraction of the annual emission inventory. For fine particulate matter, however, the annual emission estimates from biomass burning represent a significant fraction of many southern States' emission inventories, especially in the counties where the emissions are concentrated (Dennis et al., 2002). Given the current emphasis by the EPA on particles, it is imperative that real-world emission data from open burning sources be developed.
It is generally thought that emission factors or pollutants are among the more consistent and reliable components of biomass burning emission models. However, comparisons of recent studies suggest that under some conditions, especially where smoldering combustion is important, emission factors (EF) are still quite uncertain (Andreae and Merlet, 2001; Hays et al., 2002). Residual smoldering combustion (RSC) emissions from forest floor burns can be produced for up to several weeks after the passage of a flame front and they are mostly unaffected by flames. Fuels prone to RSC include downed logs, duff, and organic soils. These fuels are very important in our proposed study area. Limited observations suggest that RSC is a globally significant source of emissions to the troposphere (Bertschi et al., 2003). These authors used a model which predicts trace gas EF for fires in a wide variety of aboveground fine fuels. It failed to predict emission factors for RSC. For many compounds, the EF for RSC-prone fuels is very different from the EF for the same compounds measured in fire convection columns above forest ecosystems. Some large changes resulted in estimates of biomass fire emissions with the inclusion of RSC. For instance, EF increases by a factor of 2.5 even when RSC accounts for only 10% of fuel consumption. This shows that many more measurements of fuel consumption and emission factors for RSC are needed to improve estimates of biomass burning emissions. The ecosystems we propose to study are particularly susceptible to this type of combustion. Fine particulate matter estimates from this type of combustion in southern ecosystems is non-existent and needs to be developed.
Smoke Modeling
Smoke emissions from wildland fires is one the most important constraints on land mangers conducting prescribed burns. The quantity, duration, time of day, and spatial dispersion of smoke must all be considered when assessing the impacts on human health and safety. Existing smoke models do a poor job of estimating smoke production and duration. This is especially true on the deep organic soils found in the Coastal Plain of the southeastern US. Many of the dispersion models in use by wildland mangers today (SASEM (Sestak and Riebau, 1988), VALBOX (Sestak et al., 1989), VSMOKE (Lavdas, 1994), NFSpuff (Harison, 1995), TSARS (Hummel and Rafsnider, 1995), and CALPUFF (Scire et al., 1994)) have been adapted from industrial stack models for use in wildland fires. Smoke models for prescribed burning differ from point-source industrial models due to additional data requirements for pattern of ignition, fuel moisture by size, fuel loading by size, fuel distribution, and local weather that influences burn rates and dispersion. The FARSITE (Finney, 1998) model was developed to address these data requirements and is used to model forest fire behavior in variable fuels, terrain, and changing local weather conditions. A recent JFSP project (Finney et al.) has improved FARSITE's ability to model smoke production by incorporating a fuel consumption model (BURNUP). FARSITE does not model smoke dispersion, but output from the combined models can now be used in smoke dispersion models.
|
| Smoke plume from prescribed fire at Pocosin Lakes NWR. |
Approximately six million acres of forest and agricultural land are burned each year in the southern US, an area defined roughly by lands located south of the Ohio River and from Texas eastward. Most burning is done during the period from January through March. Although the vast majority of prescribed fires are burned without incident, there are occasions when smoke from smoldering fuels persists after sunset and becomes entrapped within slow-moving drainage flows. Entrapped smoke can drift into populated areas and impact residents, particularly those with respiratory problems. Smoke-laden air masses have drifted across roadways and have contributed to poor visibility. Smoke and associated fog has been implicated in multiple-car pile-ups that have caused numerous physical injuries, heavy property damage, and fatalities.
Several attempts to compile records of smoke-implicated highway accidents have been made. For the 10-year period from 1979-1988, Mobley (1989) reported 28 fatalities, over 60 serious injuries, numerous minor injuries, and millions of dollars in lawsuits. In their study of the relationship between fog and highway accidents in Florida, Lavdas and Achtemeier (1995) compared three years of accident reports that mentioned fog with fog reports at nearby National Weather Service stations. Radiation fogs occur during the same conditions that entrap smoke - clear skies and light winds. Southern states are becoming more urban, and the numbers of tourists driving to resort areas along the Gulf coast, the Atlantic coast, and the Florida peninsula are increasing. Therefore, the number of accidents related to smoke and fog could be expected to increase.
|
| Smoke covering NC hwy 306 from a prescribed fire at the Croatan National Forest. |
As the risk of encountering smoke on roadways increases, land managers involved with prescribed fire need to be able to better predict where smoke trapped near the ground is transported upon leaving burn sites. Smoke at night is invisible unless illuminated by reflected moonlight or headlights from motor vehicles. The only working strategy in current use is for land managers to dispatch aircraft to observe the smoke plume and sentries to drive the roads surrounding burn sites to post warnings in areas where smoke does cross a roadway.
PB-Coastal Plain is a sister smoke model to PB-Piedmont (Achtemier, 2000), which was developed to simulate local smoke movement at night over terrain typical of that of the Piedmont of the southeastern US. The two Prescribed Burn (PB) models, PB-Piedmont and PB-Coastal Plain, are designed to run on laptop computers at faster-than-real-time. PB-Piedmont simulates the movement of night wind as it drifts through the shallow gaps in ridges and down road and stream cuts of the Piedmont, while PB-Coastal Plain tracks smoke movement over the flat Coastal Plain and nearby water surfaces. The models are currently connected with hourly weather forecasts and can predict smoke movement in a range of about half an hour. PB-Coastal Plain is intended to extend the theory of PB-Piedmont to model for coastal land/water interfaces and to include differences in vegetative land use. PB-Coastal Plain is more complex than its sister model for several reasons. First, the forces that move smoke in the Coastal Plain are not always local. Interactions between land and water surfaces 100 km or more away from a burn site can steer the winds at the burn site during a typical nighttime period. PB-Coastal Plain must be capable of modeling land/water interactions over large areas. Second, there exist few stream basins of sufficient depth to channel smoke. Smoke over the Coastal Plain can be channeled by differences in vegetative land use.
Because of the complexity of the meteorology of the Coastal Plain and the number of assumptions that must be made to develop a model that will run on a PC-environment in faster than real time, PB-Coastal Plain must be extensively validated with observations. Gaining the required observations is also difficult because of the following reasons; 1.) Coastal burn locations are more inaccessible than inland sites - there are fewer roads, 2.) The meteorology of the Coastal Plain is highly complex - data from distant weather stations will not represent local weather, and 3.) PB-Coastal Plain requires validation data usually not available through standard measurements.
The BlueSky smoke prediction system (Ferguson et al. 2001; O'Neill et al. 2003) is an automated centralized framework for predicting cumulative concentrations of smoke from wildland and agricultural fire. It integrates tools and knowledge about fire location, fuel loads, moisture conditions, combustion processes, fire behavior and spread, weather, and smoke dispersion. The National Fire Plan (01.PNW.A.1) is funding the scientific development of BlueSky. Also, an automated system for validating BlueSky with monitoring data in the northwestern U.S. is being funded by the Joint Fire Science Program. A successful prototype has been operating in many western states for nearly 2 years, and we have a pending proposal with the Environmental Protection Agency to implement the BlueSky Rapid Access Information System (BlueSkyRAINS) in the southeastern U.S. The topography, weather, fuel, and moisture conditions in the eastern U.S. are sufficiently different than the west, however. Therefore, we are seeking additional support to evaluate the system in North Carolina so we can improve it before implementation in that region. In addition, we hope that additional funding for the collaborative experiments will provide the vehicle needed to begin steps toward integrating the PB-Coastal Plain model into the BlueSky system.
|
| A BlueSky run for the Croatan National Forest in southeast North Carolina |
Technology Transfer
Land managers in the southeastern coastal plain have few decision support tools available to them for implementing fire and air quality management programs at the local level. Those they do have are either still under development, difficult to use for field station personnel ("user hostile"), or do not operate in the near real-time environment that managers and practitioners require. Some of the difficulty in use stems at least in part from a lack of involvement by potential end-users in beta testing and development of prototypes. Recent improvements in desktop computing speed and capacity, along with dramatically improved data transmission speeds over long distances present an opportunity to provide results that are timely enough to influence program implementation decisions.