Abstract
A random 10% sample of communities in highly-impacted counties from the 2004 and 2005 Florida, U.S. hurricane seasons were used to study tree debris generation and damage from seven hurricanes. Woody debris amounts in cubic meters (m3) (cubic yards; yd3) as well as rates and costs for cleanup, tree removal and pruning were obtained for these events. Average debris generation per 30.5 m (100 ft) of street segment ranged from 0.59 m3 (0.77 yd3) for low damage, 3.4 m3 (4.44 yd3) for moderate damage, and 17.47 m3 (22.85 yd3) for high damage levels; cost of removal and disposal averaged $28.25 per m3 ($21.47 per yd3). Most communities used unit costing to account for tree removal and pruning rates, averaging 2% and 28% of street trees, respectively. Tree canopy, wind speed, and percentage of urban developed land all had a significant effect on debris amount. Florida communities with a greater tree density generated decreased amounts of debris during the 2004 and 2005 hurricane season. These results can be used to help communities plan for hurricane management activities and estimate potential damages to their tree resource.
Studies in Florida (U.S.) and elsewhere have documented the impact of hurricane force winds on urban forest structure, damage to woody vegetation, harm to forested landscapes, and of the varying resistance of different tree species (Duryea 1997; Duryea et al. 2007a; Duryea et al. 2007b; Everham and Brokaw 1996; Francis and Gillespie 1993; Gresham et al. 1991; Mayer et al. 2007; Oswalt and Oswalt 2008). Studies have also demonstrated that hurricane winds and storm surges can severely damage individual trees and landscapes, causing trees to defoliate, partially break, drop branches, topple, or uproot (Duryea et al. 2007a; Francis and Gillespie 1993; Gresham et al. 1991). Hurricanes can also alter the structure and function of ecosystems (Kupfer et al. 2008; Oswalt and Oswalt 2008). To effectively plan and manage hurricane-prone urban forests in Florida, the effects of wind storms must be taken into consideration (Escobedo et al. 2007).
The aftermath of a hurricane often includes a large amount of debris from infrastructure or landscape damage, which may result in the need to prune or remove hurricane-damaged trees (FEMA, 2007a). The cost of cleaning up debris and the remediation of tree damage is borne by local communities, but may be offset if a declaration of a federal disaster is made by the President of the United States, which hedges costs through reimbursement programs as designated by the Stafford Act (FEMA, 2007b). The ability to quantify tree debris to be removed and other tree-related costs quickly and accurately after a storm impacts the reimbursement monies requested by the Federal Emergency Management Agency (FEMA) and, indirectly, the communities’ relations with FEMA during the recovery operation. Methods that estimate tree-related debris and damage from ice storms have been used to develop protocols and tools to assist communities with post-storm recovery activities (Bloniarz et al. 2001).
Non-sediment hurricane debris is generally categorized, and in certain cases reimbursed, in two broad categories: clean wood, and construction and demolition (FEMA, 2007a). Although models from the United States Army Corps of Engineers (USACE) estimate that hurricane-generated debris consists of 30% clean woody debris, state emergency services reported that up to 70% of the debris removed after Hurricane Floyd in North Carolina was composed of vegetation (COES, 2005). Following Hurricane Ivan in September 2004, Escambia County, Florida reported that almost 90% of all hurricane debris removed was composed of vegetation (Escambia County, 2004).
Debris removal and disposal is an enormous problem for storm-affected communities. For example in 1992, 11.5 million cubic meters (15 million cubic yards) were reported following Hurricane Andrew (Ward 2002). A report from mid-2006 stated that more than $1 billion had been spent in the state of Florida for Hurricane Wilma recovery efforts alone, and that $2.3 billion had been sent to Florida up to that point for the 2004 storm season (FEMA, 2006a). Typically, about half of such recovery costs were likely due to debris management (Ward 2002).
There are few methods in existence for accurately estimating urban forest debris following hurricanes. Currently, the USACE debris estimation model produces a debris estimate for an affected area based on the number of households and categories for heavy, medium, and low amounts of vegetation (FEMA, 2007a). This model is acknowledged to have an error of nearly 30%. Elsewhere, Broward County, Florida developed a Geographical Information Systems (GIS) based Hurricane Debris Estimation Tool that accounts for tree and palm debris (Umpiere and Margoles 2005). The tool uses tree and palm cover and assumes a tree to palm ratio and debris volume per affected tree, per unit area.
The FEMA HAZUS hurricane module uses tree density and tree height data to derive tree-loss curves for estimating downed tree debris following hurricanes (FEMA, 2006b). However, the tree-loss curves only consider trees greater than 9.14 m (30 ft) in height. Validation using actual tree debris estimates showed that this model over-estimated tree debris by approximately 90% in Virginia and 41% in North Carolina states (FEMA, 2006b). A standardized method for estimating ice storm damage and debris costs is available as the Storm Damage Assessment Protocol, a component of the i-Tree program (www.iTreeTools.org). However, this protocol does not specifically account for hurricane damage profile and management.
Urban forests in Florida encompass a wide variety of species, urban morphologies, and land cover types. Any protocol that is developed must be applicable over this range of variability. Most studies on hurricane effects on urban and natural forest are based on single tree, plot or landscape-scale based assessments (Duryea et al. 2007a, Duryea et al. 2007b; Francis and Gillespie 1991; Kupfer et al. 2008; Oswalt and Oswalt 2008). Many debris assessments use empirical studies such as these as inputs for broad-based, often single-parameter models and approaches for quantifying debris (FEMA, 2007a).
This study takes a different approach, and uses on-ground data reported from hurricane impacted communities to assess and characterize debris and damage at the urban and community forest level. Its objectives are to assess, compile, and analyze urban forest debris and tree damage data from the 2004 and 2005 hurricane seasons in Florida. These estimates can then be used to develop a hurricane debris assessment protocol that estimates potential damage before a storm and report tree debris and damage after a hurricane in Florida.
MATERIALS AND METHODS
Debris and Tree Damage Data Collection
We examined recent hurricanes making landfall in Florida: Charley, Francis, Ivan, and Jeanne in 2004; and Dennis, Katrina, and Wilma in 2005. An initial sample of hurricane-affected coastal and inland communities was used to determine the utility of FEMA Project Worksheets (PWs) for quantifying tree debris, damage, and costs.
Project Worksheets are compiled by each community requesting public, monetary assistance for hurricane damage, and constitute a dynamic record of information and data needed for justification for reimbursement in an emergency management project (FEMA, 2007c). Project Worksheets are categorized by the type of hurricane damage. We sampled Category A, the debris damage category, which contains relevant debris data as well as hazard tree pruning and removal reimbursement; and Category G, which is a general category that often contains debris data. Project Worksheets were obtained from the Florida Division of Emergency Management. The initial sample of PWs from 15 communities from the 2004 and 2005 hurricane seasons in Florida provided enough data to justify their use in obtaining debris quantity and cleanup costs.
A list of Florida communities impacted by each of the seven hurricanes during 2004 and 2005 with PW applications in Category A was compiled and each community was then assigned a unique number. Using the unique number and a random number generator, a 10% sample of communities impacted as depicted in maps (FEMA, 2004) by each hurricane was then selected for use in this study. Entities other than communities (e.g., counties, tribes, official departments, hospitals) were excluded from the sample. Stump removal, for which insufficient data were available, was also excluded from the data collection.
Data for debris amounts and reported costs were taken directly from the PWs. Individual cost line items were not specifically listed by many communities. In addition to removal and reduction of debris, individual costs might have included debris monitoring, site preparation, tipping fees, equipment or other costs that were directly related to debris elimination. Debris quantity and costs were commonly reported in several different PWs for a hurricane by a single community. Therefore, data from all PWs for a single hurricane event were summarized along with community specific data as presented in Table 1. Data on both tree removals and hazard prunings were less consistently reported in PWs. Data was obtained by direct request from the communities.
Hurricane debris amounts and costs, urban forest tree density and cover data, street mileage, and percent of urban developed land were derived from a number of sources (Table 1). Maximum sustained wind speed data for each community were obtained from tabular data (NOAA, 2004), and reported for selected communities by the National Hurricane Center. Where wind speed data were not available for a sample community, wind speeds were assigned to sample communities based on nearest community with a recorded wind speed.
Debris data were converted to a total cost per cubed meter by summing all reported debris volume disposed and its associated costs from PWs. Estimates of debris generation per unit street segment length were made with i-Tree’s “Sample Street Segment Generator” using ArcGIS 9.1 and street mileage obtained from U. S. Census Bureau’s TIGER/Line data (U.S. Census Bureau, 2004), in an ESRI shapefile (i-Tree 2008).
Relationships between tree cover, density, and debris amounts collected from Street Rights-of-Way were investigated using Renewable Natural Resources Planning Act (RPA) assessment urban forest cover data tabulated by Census Designated Place (Nowak et al. 2008). These data are based on 30 m resolution Landsat TM and ETM imagery that is part of the 2001 National Land Cover Database. RPA data included percent tree canopy and developed urban land for each sampled community (Table 1). Percent canopy was converted to community-specific tree density using representative mean tree density factors from similar geographic and physiographic communities (Escobedo et al. 2008). Communities in north Florida were assigned Pensacola’s tree density of 315 trees/ha (127 trees/ac), Tampa Bay’s tree density of 257 trees/ha (104 trees/ac) was used for the central Florida gulf coast, Miami-Dade Counties’ 83 trees/ha (34 trees/ac) for south Florida and lower Atlantic coast, and Gainesville’s 242 trees/ha (98 trees/ac) were used for communities in central Florida’s non-coastal areas.
The distribution of cubic meters of debris per linear street segment data was determined to be significantly different from normal. To most appropriately characterize the positively skewed data, cubic yards of debris were transformed with the natural logarithm, and the number of linear street miles was used as an offset in the linear predictor function. As offsets are akin to exposure rates in Poisson regression with unequal time intervals, their use in spatial models is frequently used to adjust the response to varying levels of coverage. Using a log-normal distribution to characterize the response variable and taking into account the potential spatial correlation between adjacent cities, a generalized linear mixed model was fit with the SAS procedure PROC GLIMMIX (SAS, 2006), using sustained maximum wind speed, tree density, tree canopy, and the amount of developed urban land as predictor variables. City center location was obtained using Google Earth (2007) (UTM; Universal Transverse Mercator, Northing and Easting) and used to determine spatial context in the analysis.
The model results were examined using information criteria and P values associated with each independent value. A type I error level of 0.10 was used to eliminate non-significant effects and their interactions, together with the corrected Akaike’s information criteria (AICC), which is a small sample bias-corrected version of the AIC fit statistic. The AICC quantifies the residual variance, with a penalty term for additional independent variables. The final estimated model included only significant effects and also had the lowest AICC, indicating substantial evidence that the data arose from this model.
RESULTS
Data Collection
A random 10% sample of 680 PWs, representing the communities reporting debris data in FEMA category A or G in the 2004 and 2005 hurricane seasons, were reviewed to compile and assess debris amounts and cost, as well as tree hazard pruning and removal data. From the 68 PWs, we obtained usable debris amounts and cost data from 43 communities. However, tree pruning and removal data were considerably less available, with only 11 communities specifically referencing tree work in their PWs, and only 5 communities providing usable pruning and removal cost estimates. No usable data were obtained from any PW for community for tree removal or pruning rates (removal or pruning quantities/total tree population). These data were obtained from direct contact with the communities of Orlando, Pensacola, and Winter Park. Averages for the sampled communities are presented in Table 2.
Debris Removal Costs and Production
Total debris management costs averaged $28.11 per m3 ($21.50 per yd3) in the 43 communities that provided usable data. Debris production per 30.5 m street segment ranged in values from 0.15 m3 (0.2 yd3) in Sanford, to 31 m3 (40.6 yd3) and 46.4 m3 (60.7 yd3) in Golf and Gulf Breeze respectively. Thus, rates of debris production were averaged into low, moderate, and high categories. These averages, per street segment, were 0.59 m3 (0.77 yd3) for low damage, 3.40 m3 (4.45 yd3) for moderate damage, and 17.47 m3 (22.85 yd3) for high damage levels.
Tree density, tree canopy cover, wind speed, and percentage of urban developed land all had an effect on the amount of cubic meters of debris. The model which included all significant effects and best supported the data based on the AICC (169) included a third order interaction of tree canopy cover, wind speed, and percentage of urban developed land (α < 0.10) (Table 3).
There was a marginally significant negative correlation between tree density and debris generation (P = 0.08), indicating that as tree density in communities increased there was a slight decrease in debris (Table 3, Figure 1a). Because of the significant interactions among variables, the effects of tree canopy cover, wind speed, and percentage of urban developed land were examined in conjunction with the other variables. Tree canopy cover had a positive relationship with debris when the interactions among variables were taken into account and all other predictor variables were at their averages; as canopy cover increased in a community, debris generation increased (Figure 1b). Sustained wind speed and amount of developed urban land had a more complex effect, and when considered with all other interactions their effect on debris was relatively flat (Figure 1c, Figure 1d).
Figure 2a and Figure 2b display the continuous two-way interaction of wind speed when percent developed urban and tree canopy cover were held at constant “low” and “high” values. Debris generation strongly decreased with increasing storm strength when communities had a high tree canopy cover (Figure 2a). For communities with low tree canopy cover, stronger storms produced more debris, but with a much weaker relationship. The amount of urban development also had a differential effect when considered with wind speed. The positive correlation indicated greater debris generation with increased wind speed for communities with low urban development (Figure 2b). For communities with high urban development, however, debris generation was lower with increasing winds.
Tree Removal and Hazard Pruning Costs and Rates
All communities that reported tree removal and pruning data in PWs or were contacted directly, used unit costing rather than time and material for individual trees to account for these costs. Reported unit tree removal and pruning costs averaged $447 for tree removals and $147 for hazard tree pruning. Data from Orlando indicated that the current per-man hour rate for a fully equipped tree crew was $50. Post-hurricane street tree removal and pruning rates were obtained using Pensacola and Winter Park data. Removal rates averaged 0.65% for Pensacola and 3.6% for Winter Park, for an average of 2%. Pruning rates averaged 34% for Pensacola and 23% for Winter Park, for an average of 28%. Winter Park data represented the average of three hurricanes that occurred sequentially in the 2004 season.
DISCUSSION AND CONCLUSION
Characterization of hurricane debris and damage is scale-dependent and can vary from the individual tree to the landscape level (Duryea et al. 2007a; Duryea et al. 2007b; Everham and Brokaw 1996; Francis and Gillespie 1991; Kupfer et al. 2008; Oswalt and Oswalt 2008). Differences in the range and variability of weather, urban forest structure, tree species characteristics, landscape morphology, and community values can also confound findings (Kupfer et al. 2008). However, this complexity might assist in understanding the recurrent contradictory or ambiguous results from this and other studies of hurricane-caused damage to treed landscapes.
The majority of communities that were sampled for this project did not measure hurricane-level winds greater than 121 km/h (75 mph; 65.2 kt) based on the sustained wind speeds reported by NOAA (2004); wind speeds ranged from 48 km/h (30 mph; 26 kt) to 115 km/h (72 mph; 63 kt). Monitors measuring winds greater than 121 km/h (75 mph; 65.2 kt) might have been rendered inoperative by hurricane force winds. Additionally hurricane dynamics are complex and tornadoes, wind bursts or antecedent rain might have more of an influence on debris production than does maximum sustained wind speed (Everham and Brokaw 1996). Also, as hurricanes move inland, wind speeds are reduced but are still capable of causing substantial levels of damage to trees and infrastructure. A substantial increase in tree damage from branch failures has been reported by Luley et al. (2002) to begin around 81 km/h (50 mph; 43.4 kt). Francis and Gillespie (1991) report tree damage to begin around 60 km/h (37mph; 33 kt) with damage increasing rapidly to approximately 130 km/h (81 mph; 70 kt), but not worsening at higher gust speeds.
Results of urban forest damage as expressed by debris generation from this study are consistent with other studies, which found that tree and stand characteristics were the best predictors of damage while storm meteorology was of secondary importance (Kupfer et al. 2008; Oswalt and Oswalt 2008). The relationship between wind speed and debris generation was found to be significantly related, while there are clearly other factors, in addition to wind speed, that are important in debris generation. Assuming that all other effects are at their average over the range of the independent variables, this study indicates that tree canopy cover and density, wind speed, and percent of urban development had complex relationships with debris generation (Figure 1a, Figure 1b).
Increased tree density in a community and high tree canopy cover with increasing winds were two interactions that resulted in decreased debris generation (Figure 1a, Figure 2a). This can possibly be a result of urban forest structure, management, and the effects of past hurricanes. Everham and Brokaw (1996) indicate that uniform canopy heights of wind-impacted forests are less susceptible to windstorms. Because of this, tree canopies collectively reduce turbulences and shed winds. Additionally, the authors report that past and current management activities might selectively remove susceptible trees and change species and size composition to a more wind-resistant structure.
This study’s results also suggest that models and methods that use tree cover or wind speed alone might not be considering important factors that influence tree debris generation (FEMA, 2007a). However, other factors such as urban forest structure and impact of prior hurricane events warrant further investigation and could improve the reliability of multivariate models (Everham and Brokaw 1996; Kupfer et al. 2008).
This study found that debris generation varied from a low average of 0.59 m3 (0.77 yd3), per 30.5 m of street segment for low damage, 3.4 m3 (4.44 yd3) for moderate damage, and 17.47 m3 (22.85 yd3) for high damage levels. Given the variability of cubic meter debris production in the sampled communities, these damage level averages are plausible and within range of past post-storm tree debris production events (Rankin 2000). These averages can be applied by a Florida community, who can use street lengths to obtain a preliminary estimate of potential debris generation from their urban forest.
It is interesting to note that tree removal rates were surprisingly low in the communities that reported these data. One reason may be that trees needing removal were treated as debris and removed during debris removal activities. Alternatively, it is possible that standing trees were not removed regardless of the level of damage. Tree pruning rates were more consistent with ice storm events in the northeast United States (Bloniarz et al. 2001).
An important difference between ice storm damage cleanup in the northeast US and hurricane storm damage clean up in our sample is the use of unit costing for tree removals and pruning, or the inclusion of tree removal and pruning costs in debris cleanup costs. This is suggested by the absence of reported tree removal and pruning data in PWs, the absence of itemized (e.g., diameter class) data for these costs when calling communities, and the common use of a unit costing approach in PWs. We concluded from our limited data that itemization of costs for tree pruning and removal is rare for hurricanes in Florida, despite FEMA’s requesting this approach in their formal documentation (FEMA, 2007a).
Limitations of this study included the few data that were reported in PWs or available from communities are related to the direct exploration of any relationships between specific costs, tree removal and hazard pruning rates, and wind speed. In fact, we only obtained usable pruning and removal data from two communities, which had data by tree as required by FEMA. Since communities apparently use unit costing or report removal and pruning costs with debris clean up, these data will be more difficult to obtain than debris removal estimates. Additionally, recent and state-wide Florida urban forest cover and structure data are unavailable. Clearly, it would be desirable to find additional data such as these so that more robust numbers may be used for vegetation debris estimates.
Urban forests in Florida are prone to hurricanes, and as a result, a substantial portion of post-hurricane debris is from trees. Hurricane and tree debris must be considered in urban forest management plans and emergency management activities (Escobedo et al. 2007). Thus, there is a need to support local and regional governments with timely information on the extent and location of damage to urban forests, as well as tools to help plan for and manage the debris generated, resources for clean up, and eventual restoration of urban forest structure and function. This study found significant differences in Florida hurricane debris and its management from what the literature suggests. These results need to be communicated to concerned officials and researchers to improve our confidence in the estimation of hurricane damage to urban forests and the associated costs.
Acknowledgements
This study was funded by the Florida DACS-Division of Forestry. We thank Andrew Kittsley, David Metzker, Cecilia Meeks, Steve Simpson, Wayne Zipperer, David Nowak, and Eric Greenfield for providing data as well as Benjamin Thompson, Eric Kuehler, Dudley Hartel, Charlie Marcus and two anonymous reviewers for helpful comments on earlier drafts of this manuscript.
- Copyright © 2009, International Society of Arboriculture. All rights reserved.