1. Introduction

1.1. Project justification

Fire officials are often concerned that fires in wetlands may lead to extensive smoldering in accumulated organic soils. Many agencies – especially in the Atlantic and Gulf Coastal plains, where forests and grasslands are heavily intermixed with wetlands – expend considerable resources in trying to prevent fires from entering wetlands. Unfortunately, fire control actions (such as plowing firelines along wetland margins) can themselves be disruptive to important ecological and socio-economic wetland functions. We hypothesize that there are instances where the likelihood of smoke management concerns resulting from wetland fires is actually low, while ecological benefits and opportunities to reduce fire agency resource allocation costs are high. This study proposes to begin to identify many of these instances.

Prolonged drought and historic ditching in the southeastern USA have resulted in wetlands being unusually dry. As a result, previously hydrated organic soils in some areas have dried and are now susceptible to burning. Decades of fire exclusion have increased the risk of organic soil fires as invasive upland woody species have encroached on the normally herbaceous plant communities occurring at the periphery of frequently burned wetlands (Wade et al., 1980).

Many wetland fires – even under dry conditions – do not necessarily consume large quantities of organic soil, or require large suppression expenditures or resource commitments. In fact, some fires in wetlands – even under dry conditions – can be ecologically beneficial. We believe it is possible to identify the conditions and situations under which the characteristics of wetland fire are not economically or ecologically detrimental. This information could be used to develop tools to assist fire management agencies in making these decisions.

Consider a highly simplified decision matrix (Table 1) under which wetlands fires are either likely or unlikely to lead to smoldering fire in organic soils, and are unambiguously either ecologically beneficial or harmful. A wetland fire exclusion approach effectively treats all fires in wetlands as though smoldering fire in organic soils is assured, and ecological benefits are either nonexistent or are always outweighed by the need to minimize smoke generation or protect economic interests.

Ecological Effect		Likelihood of detrimental social/economic effects (smoke, lost timber value)
		High	Low
	Beneficial	?	Always permit
	Harmful	Always prevent	?

Table 1. A simplified decision-making scheme for wetland fires. Clearly, wetland fires should always be prevented in the "harmful, high" case, and always permitted in the "beneficial, low" case, though that category is probably not well defined. Decisions in the other two cases depend on relative costs and benefits of each decision.

The Nature Conservancy’s 12,000-acre Disney Wilderness Preserve (DWP) near Kissimmee, Florida affords a unique opportunity to examine wetland fire effects. An existing network of vegetation monitoring transects, peizometers, and organic soil depth measurements was installed and sampled for wetland mitigation monitoring prior to a recent series of significant wildland fires. The fires burned approximately 50% of the installed sampling network. Consequently, we have available a considerable body of data on pre-wildfire conditions and can use the sampling network to acquire comparable data about post-burn conditions. Because the mitigation has involved restoring wetlands by filling ditches, we will also have data on how wetland restoration affects fire.

Our goal is to empirically describe conditions under which consumption of quantities of organic soil did or did not occur in these wetland fires, and some of their associated ecological effects. We hope to establish a basic model as a tool in decision-making on whether to exclude or allow fires to burn in wetlands, including decisions about digging plow lines around wetlands or issuing prescribed fire permits during dry periods.

Several factors will be explored that likely influence wetland fire's ability to consume organic soil and generate smoke (Tasks 5 & 7). There are at least three possible cases in which wetland fire will not automatically result in smoke issues:

1. Some combinations of vegetation and hydrology do not lead to substantial organic matter accumulation. For example, in the Coastal Plain region, many cypress domes and swamps occur on mineral soils, which accumulate little organic material and thus present little risk of extensive smoldering fires. Stand-regenerating fires during severe drought (every 50 -100 years) in these systems may be necessary to maintain species composition, so fire exclusion is actually harmful ecologically.

2. Even wetlands with substantial organic matter, such as bayheads, often have enough moisture that they will not burn, either because water levels are sufficiently high or the wicking action of the peat organic layer keeps the soil moist.

3. Fuel loading around and on the edge of the wetland may be low enough that fire will not carry into the wetland. Recent fire history certainly plays an important role in this, as does timber harvesting history, and regional and local hydrologic conditions. Fire suppression often leads to invasion of the wetland edge by woody shrubs [e.g., gallberry, (Ilex glabra), fetterbush (Lyonia sp.)], which suppress the regeneration of wetland herbaceous and canopy species, promote the spread of intense fire into the wetland, and increase the potential for crown fires. Over time, high densities of shrub species and lack of fire in the wetland will lead to the accumulation of greater quantities of organic matter, increasing the risk of hazardous organic soil fire and changing wetland community composition to more hardwood-dominated types.

We will evaluate the ecological effects of a variety of wilderness wetland fires on the study site in drought years to describe conditions under which beneficial and detrimental results occurred (Tasks 1 & 3). Substantial organic matter loss, canopy mortality or potential timber value loss, subsequent invasion by exotic or pest plants, and changes in vegetative composition or structure away from the expected wetland community will be considered harmful effects. Beneficial effects include increase or maintenance of vegetative composition and structure typical of reference wetland conditions, and contribution toward listed species habitat values.

What factors affect the severity and the ecological consequences of fires? Using the data from our transects, we will evaluate the relationship among a number of causal factors – including wetland type and vegetative composition, hydrologic conditions, fire history, timber harvest history, weather parameters, and fuel loading prior to fire – and the severity and ecological consequences of fires.

In this study, we will focus on wildfires in central Florida. Natural fire intervals in wetlands vary quite widely (see Background section below); as with fire in mesic ecosystems, key features of fire and its role in wetlands are likely to vary among regions and habitats. Nevertheless, we believe that the work proposed here has general importance for two reasons. First, because peninsular Florida is a region of major wildfire interest, our work will have regional importance. Second, we believe that the questions asked and the research approach used will be instructive for work in a wide range of settings.

1.2. Project objectives

Our objectives address several overlapping issues related to fire in wetlands: smoke production, socio-economic and ecological costs and benefits, and factors that promote or deter fire in wetlands.

Objective 1. Quantify the change in organic soil depth before and after occurrence of fire, both in wetlands that burned and in wetlands that did not burn, as a comparison.

The specific hypothesis to be tested is:

H₁: There is no significant difference in the amount of organic soil present before and after wetland fires.

Our expectation is that this hypothesis will be supported under some ecological circumstances but not others. Where possible, we will measure this directly. Where measurement is not possible, we will conduct field evaluations to qualitatively evaluate if organic consumption occurred.

Objective 2. Document (from fire records and interviews with fire managers) which fires produced substantial smoke contribution. The specific hypotheses to be tested are:

H_2a: Fires that generated substantial smoke were located where organic consumption was detected.

H_2b: Organic soil fires were least likely in wet prairies and cypress domes

This will indicate which specific wetland fires generated the most smoke concerns and were most likely to result in the combustion of organic matter.

Objective 3. Measure the effect of hydrology on fire characteristics and organic soil consumption. The specific hypotheses to be tested are:

H_3a: Organic soil fires are less likely (and consume less soil if they do occur) in wetlands that have higher water tables and are unditched.

H_3b: The chance of organic soil fires occurring, and the amount of soil consumed if they do occur, decreases as time since wetland restoration increases.

Objective 4. Determine which wetland fires resulted in substantial organic matter loss (from Objective. 1), canopy mortality or potential timber value loss, subsequent invasion by exotic or pest plants, or changes in vegetative composition or structure away from expected wetland community. The specific hypothesis to be tested is:

H₄: Habitat type, time since wetland restoration, ditching, and water table height have no effect on organic matter loss, canopy mortality, density of invasive species, or changes in community structure.

Objective 5. Document fire intensity characteristics for fires in wetlands on the study site. The specific hypotheses to be tested are:

H_5a: Fire spatial extent, percent consumption of standing fuel, canopy scorch height, and canopy mortality are less in areas with thinner organic soil, such as wet prairies and cypress domes.

H_5b:Fire intensity characteristics are less in areas with lower fuel loads (including live shrub cover) in and around wetland.

H_5c: Fire intensity characteristics are less in wetlands with fairly intact hydrologic regimes (i.e., unditched or restored for several years)

Objective 6. Quantify the composition and percent cover of invasive (woody non-wetland) species before and after fire in different types of wetlands. We believe that fire reduces the prevalence (coverage) of upland woody vegetation that invades the periphery of wetlands resulting from prolonged fire exclusion. The specific hypothesis to be tested is:

H₆: Fire reduces the percent cover of upland woody vegetation that invades the periphery of wetlands.

Objective 7. Quantify herbaceous species composition in different types of wetlands before and after fire. We predict that fire increases herbaceous species richness while decreasing the structure of wetland and invasive upland woody vegetation, and that species composition of fire-dependent wetland types (wet prairie, marsh, cypress) will move toward reference compositions (from literature). The specific hypothesis to be tested is:

H₇: Wetland herbaceous species richness and percent cover increases with the occurrence of fire.

Objective 8. Quantify changes in wetland canopy species density and size structure before and after fire. We expect that fire will increase recruitment of wetland canopy species (e.g., cypress, loblolly bay) on thin organic soils, but not on deep organic deposits.

H₈: Canopy species recruitment in forested wetland types increases following fire on thin organic soils (e.g., cypress domes), but the reverse is true on deep organic soils (e.g., bayheads).

1.3. Products

We will provide the following information or products by the end of this project:

· Articles discussing the scientific results of these studies in journals such as Wetlands, Ecology, Natural Areas Journal, and International Journal of Wildland Fire, as well as presentations at scientific conferences.

· A set of initial recommendations for wetland fire policies and practices. The fire suppression costs associated with protecting wetlands may be reduced where certain wetland types present minimal risk of organic soil fire ignition. Prescribed burning of wetlands at appropriate times is likely to reduce erratic fire behavior and increase the utility of wetlands as natural fire barriers. We will publish a set of initial recommendations on the conditions under which burning wetlands present low risk of substantial organic soil fire. These recommendations will also consider the ecological costs and benefits of permitting or suppressing fires. The publication will be presented and discussed at the workshop, and suggestions integrated into final production.

We recognize that the research proposed here is the first step towards an improved understanding of wetland fires, which would include a formal decision-making tool on when to permit or suppress fires. Toward that end, we believe it is important not only to make available our data, but also to develop an environment for communication about wetland fires. This motivates the following products:

· A web site to disseminate information about the project. We will allow the transfer of data and results. Associated with the web site will be an email listserver to facilitate discussion among interested managers and scientists.

· Natural Areas Training Workshop to be conducted at the Disney Wilderness Preserve through TNC’s Natural Areas Training Academy, a institute designed to provide applicable information and training opportunities to natural resource professionals and practitioners. We will invite fire management agency personnel to discuss the results of our project and development of potential tools, including the recommendations publication.

While outside the scope of our present proposal, we recognize that cost-effective wildfire suppression strategies may benefit from a decision-making tool that evaluates the likelihood of organic soil ignition and erratic fire behavior in particular wetland types under different conditions. A predictive model will allow for the classification and mapping of such probabilities for specific wetlands based on "semi-static" variables (soil type, vegetation composition and structure, topography, hydrologic condition), which could be combined with "dynamic" field variables (weather, fuel moisture, fire characteristics) to assess the risk of deleterious fire effects for particular circumstances. We intend to pursue development of such a tool in the future, using the data acquired under this proposal as well as data from future designed experiments to provide parameter estimates.

1.4. Background

Fire plays an important role in many wetland ecosystems; literature on this role dates to at least the 1930’s (Kirby et al., 1988). A useful discussion of trends in this literature is by Kirby et al. (1988).

Fire return intervals in wetlands in the presettlement southeastern United States varied between nearly annual to about 300 years (Frost 1995). This suggests that natural wildfires have played quite varying roles in this region’s wetlands.

Some ecological effects of fire in wetlands are similar to those in mesic ecosystems. For example, many populations of wetland plants depend on fire for persistence – either by suppressing competing species that are less fire-resistant, or for regeneration (e.g., pond pine (Pinus serotina), black titi (Cliftonia monophylla), and titi (Cyrilla racemiflora) – Ewel, 1990). Cypress (Taxodium sp.) forests are clear examples (note that there is disagreement about the taxonomy of cypress (Ewel 1995); here we discuss bald cypress and pond cypress together). Wade et al. (1980) argued that in many cypress domes, there is sufficient soil to support invasions of hardwood trees, but fire prevents these. Indeed, cypress are relatively fire resistant – their bark is a good insulator (Hare 1961) and surface fires usually lead to much greater mortality among other species than cypress (Ewel and Mitsch 1978). This said, it is clear that cypress’ fire tolerance is relative only: in the study by Ewel and Mitsch, just over half the cypress individuals survived the wildfire. Fires both above and below-ground kill cypress; the important point is that fires exact a substantially larger toll from other species.

There is evidence that fire also acts to maintain other types of wetlands as fire dis-climaxes (Christensen 1981, Nyman and Chabrek 1995). For example, many freshwater marshes appear to be maintained in this way (Wade et al., 1985). Moreover, there is evidence that wetlands fires sometimes act to increase species diversity within the habitat type (Christensen 1981, Wade et al., 1985). Thus, at least under some circumstances, fires in wetlands can increase the diversity of habitats as well as the diversity of species within habitats. This is not universally the case; some studies have found no change in diversity following fire, or even reductions in diversity (e.g., Kirkman and Sharitz 1993, Lee et al., 1995).

It should be clear that fire in wetlands can have negative ecological consequences. Cypress swamps may be fire-maintained, but loss of half the trees may sometimes be unacceptable. Fire in bayheads often ignites their deep organic soil. Not only does this lead to smoke problems; it can consume much of the soil, thus destroying the bayhead (Wade et al. 1985).

Soil conditions and the height of the water table are key in determining the outcome of fires in wetlands. There is substantial evidence (e.g., Bacchus 1995) that the outcome of prescribed burns in wetlands depends strongly on groundwater levels. Studies in the Okeefenokee Swamp suggest that this is also true for wildfires (Yin 1993). Most importantly, there is evidence (Wade et al. 1985, Bacchus 1995) that wetland ditching and groundwater pumping have led to a situation in which many organic soils are no longer hydrated, thus increasing the chance of wildfire, and especially of smoldering fire. Flammability of these organic soils appears to depend most strongly on their bulk density (Hungerford et al., 1995). It is important to note that many wetlands occur on soils with low mineral content (Wade et al. 1985) or high bulk density (Hungerford et al., 1995), where the danger of smoldering fire is low.

2. Methods and materials

2.1. The Disney Wilderness Preserve

Work will be conducted at The Nature Conservancy’s Disney Wilderness Preserve (DWP) near Kissimmee, Florida.FL. DWP (Figure 1) is a mostly roadless 12,000-acre Nature Conservancy project located in the pine flatwoods ecosystem of central Florida. It was created in 1992 as an offsite mitigation project for wetland impacts at Walt Disney World Resort, the Orlando International Airport and various small developments in Orlando. The mitigation involves restoring hydrology in over 4,000 acres of wetlands on a former cattle ranch, primarily through ditch-filling and re-establishing natural processes, primarily fire. The Nature Conservancy has expanded the mission of the project to include an ecological research program to meet restoration- and management-information needs in pine flatwoods systems. In addition, TNC has established the Natural Areas Training Academy to reach and teach professional natural-areas managers.

There are 20 natural communities on the Preserve, ranging from lakeshore to xeric oak hammock. The mitigation monitoring network includes 104 vegetation sampling transects, 448 shallow ground water wells, a main and five remote weather stations, 38 continuous surface and ground water level recorders, and 180 photopoint stations. Site facilities include a 5,400-ft² Conservation Learning Center with classroom, office and dry lab areas. Onsite housing is provided for interns, visiting researchers and technicians. Because much of the site is not accessible to street vehicles, infrastructure includes a sizable fleet of all-terrain vehicles and field radios for communications.

Figure 1. Fires (hatched areas) occurring on the Disney

Wilderness Preserve since 1998 and transect network (red lines).

Field data collection programs include hydrologic, climate, photopoint, organic soil and vegetation monitoring programs for the restoration, as well as adaptive management monitoring programs for 16 listed species populations and fire management and exotic species control programs. Outside collaborators, including graduate students and government agencies, conduct onsite research projects. The research program supports a substantial database of ecological information, including fire and land-use history, hydrology, topography, soils, plant communities, vegetation, aquatic fauna, wildlife populations size and distribution, climate data, and archaeological resources. In addition, we have an advanced Geographic Information System (GIS) and an extensive spatial database, including aerial photographs from every decade since 1940. As mentioned above, a series of wildfires has occurred at DWP since the onset of the present severe drought in 1998. About half the area of our monitoring network has burned in that time (Figure 1).

2.2. Organic soil depth change

As part of our regular monitoring prior to recent fires, we sampled organic soil depths using a soil probe and extension rods, prior to recent fires at points every 10 meters along all 104 transects. We took these measurements adjacent to the cleared paths to avoid sampling disturbed substrates. We will repeat this sampling along all 52 burned transects to estimate the amount of loss of organic soil. We will also sample the unburned transects to test the hypothesis that soil loss is due to fire. In cases where re-measurement of organic depth is likely to be inaccurate or not depict fire effects (e.g., some other disturbance has substantially disturbed soil), we will either drop that transect or use field evaluation (e.g., exposed tree root balls) to determine if and how much organic soil was consumed.

2.3. Fire characteristics

We will collect information on characteristics of recent fires from field documentation and personal communications with fire personnel. Descriptions of which fires produced dense smoke that did not disperse will be qualitative, derived from fire crew interviews. Immediately following each fire, we mapped the perimeter of each burn with corrected global position system (GPS) technology, and generated a coverage for GIS analysis. The burn area layer overlays registered low-level aerial photographs and transect location coverages. Post-burn field evaluations document percent consumption and canopy scorch height. Canopy mortality will be determined from changes over time in vegetation sampling data for canopy stratum and aerial photo analyses (best for pines on edge of wetlands).

2.4. Fuel loads

Standing woody and live fine fuel loading in and around each wetland is described from vegetation sampling data collection prior to the fires. No information on non-live fuel (litter, woody debris) is collected. Also, weekly live fuel moisture content measurements have been collected during the past year and will continue. Additionally, pre-burn photopoint information (see below) will provide qualitative documentation of standing live fuel loads. Information on history of timber harvesting, which often stimulates woody species regeneration and increases fuel loading around wetlands, will be provided by TNC. Aerial photographs from each year since 1993 will be examined to document changes to extent of dense shrub canopy on edges of wetlands where fire occurred.

2.5. Vegetation data

As previously mentioned, a network of 104 permanent transects has been established on the study site for restoration monitoring. Vegetation data have been collected annually along a subset of them since 1993; at least one year of pre-burn data exist for all transects. These transects are the framework for almost all aspects of our field data collection.

Restoration monitoring has been completed along many of the transects that burned. This proposal requests funding to collect data necessary for the study along these transects, for two field seasons. We expect additional wildfires to occur this year (because the drought conditions of the past two years are continuing). We will include such burns in our study.

Transects originate from the center of each wetland, the “zero point”, and continue into the highest point of adjacent upland, perpendicular to the slope of the wetland. Several sampling plots are randomly stratified within each plant community zone or band that crosses the transect; peizometers (shallow ground water wells) have been installed in the middle of each zone. Rebar defines each 2-m line intercept plot for ground and shrub strata sampling, and belt transects are located 1 m off the transect centerline, which is cleared annually. We sample three strata of vegetation: ground (all vascular plants < 1 m high), shrub (woody plants > 1 m high but < 10 m high), and canopy (woody plants > 10 m high).

2.5.1. Ground vegetation

The line-intercept method is used to estimate vegetative cover. The 2-m samples are located 1 m from, and parallel to, the centerline of each transect. Cover is measured in 1-cm increments (to the nearest cm) with a demarcated 200-cm PVC pole. For individuals whose total height exceeds 1 m, only that portion below 1 m is measured as ground stratum. The area not occupied by live vegetation is listed as "open ground/water" (bare ground, litter, exposed roots, dead plant material, or water). All species present directly below the pole receive a minimum length of 1 cm. Although no individual plant measurement can exceed 200 cm, it is possible to have the sum of all cover values (species, open ground, etc.) exceed 200 cm for a sampling unit where species overlap. Notations are made as to whether or not the sampling plot burned in the past year.

2.5.2. Shrub vegetation

We will use digital photographic analysis to evaluate the cover in the shrub stratum. Photographic monitoring stations developed to capture change in shrub and canopy cover will use permanent camera points located on or around the sampling transects. These stations will focus on the high fuel loads (invasive, woody species) existing on wetland borders. To provide scale and quantitative measure of vegetative cover, a 30 cm X by 2.5 m density board is placed vertically at the target point. The board is painted alternately bright orange and white in 0.5 m segments. Multiple photographs will be taken to target the recommended light exposure. All photographs will be digitally analyzed using ImagePro Plus.

2.5.3. Canopy vegetation

Canopy species are sampled using a belt transect method. Only woody species that have the potential to reach canopy height (10 m or more) are measured (woody shrub species are characterized in the ground and shrub strata sampling). All living individuals of canopy species are recorded, from seedlings and root sprouts to mature trees.

Data are recorded for all canopy species within 2-m wide belt plots located at 1 m from the centerline on both sides of the transect. A 3-m long PVC sampling pole is guided perpendicular to the centerline of the transect, and stems of all canopy species in the 1-3 meter section of the pole are counted, thus excluding the cleared path and areas immediately adjacent. Each side of the transect is sampled in 10 meter segments, resulting in 40 m² plots when both 2x10m sides are added together.

Each stem is counted and placed into a size class. Individuals with diameters less than 1 cm dbh are classified by height (i.e., greater or less than 30 cm). The next size class is 1-5 cm dbh. Larger individuals are divided into 5-cm dbh size classes (1-5, 5-10 and so on).

2.6. Supplemental information

2.6.1. Exotic or pest plant invasion

A survey program is in place on the study site to comprehensively search for, locate (GPS), identify, and rapidly treat invasions of pest plants. Occurrences documented during vegetation sampling are also included. A large database that organizes information from this program will be used to evaluate if invasions occurred in any of the burned wetlands following the fire.

2.6.2. Hydrology

Monthly water level information from the network of about 40 continuous recorders and 300 peizometers located along all transects will be provided by TNC.

2.6.3. Weather

Weather parameters, including continuous rainfall, wind speed, RH, and daily min/max temperature from 5 weather stations throughout DWP will be provided by TNC.

2.6.4. Historic information

Information about the site conditions in and around each wetland, including wetland type (natural community), restoration status (time since ditch filling), fire history (when wetland and edge burned within last 20 years), and timber harvest history (pine and cypress harvesting history within last 20 years) will be provided by TNC.

2.7. Statistical analyses

The analyses for this work vary with the types of data and hypotheses, but are all standard statistical methods. To test whether there are losses in organic matter (H₁), we will use heterogeneity of slopes models and analysis of covariance (ANCOVA), which will allow us to remove variance due to confounding factors such as wetland type. To test the hypotheses about smoke generation (H_2a and H_2b), we will use chi-square tests and log-linear models, respectively. The hypotheses under Objective 3 lead to three kinds of analyses. To examine how the chance of organic soil fire changes with water table height and time since wetland restoration, we will use logistic regression. To test for differences in the amount of organic matter consumption, given that a fire occurred, we will use analysis of variance (ANOVA). Finally, to test for the effects of ditching on organic soil fire, we will use log-linear models. The analyses for Objective 4 (characteristics of wetlands that lead to negative burn outcomes) can all be done with ANOVA; we will also examine multivariate interactions of these characteristics with multivariate ANOVA (MANOVA). ANOVA and MANOVA are also appropriate for the hypotheses under Objective 5 (relationship between fire intensity characteristics and wetland characteristics), Objective 6 (effect of fire on invasive upland species), Objective 7 (effect of fire on herbaceous vegetation), and Objective 8 (effect of fire on canopy recruitment, as a function of soil type).

We will conduct statistical analyses using SAS and S-Plus software, which is already available in our labs.