SUMMARY AND CONCLUSIONS
     Oh where did you go...
          I've stepped in the middle of seven sad forests ...
     Oh what did you hear ...
          the roar of a wave that could drown out the world...
     What'll you do now ...
          I'll walk to the depths of the deepest dark forests,
          and reflect from the mountains so all souls can see it...
     It's a hard rain gonna' fall...

     The answer my friend, is blowin' in the wind
                                        Robert Zimmerman

                    SUMMARY OF RESULTS
    This project was intended to examine the vegetation
dynamics associated with hurricane disturbance and recovery in
a wet tropical forest, the LEF.  I collected data on the
vegetation during the first five years of recovery in an effort
to understand the factors that influence the spatial pattern of
damage and the factors that control the path to and rate of
recovery, as quantified by changes in: community composition,
biomass, and canopy structure.  My conclusions regarding
hurricane disturbance and recovery are incorporated into a
computer simulation model, RECOVER.   

     These efforts are based on a direct gradient approach to
understanding the impacts of disturbance.  I use simulated
levels of abiotic factors, measured intensities of hurricane
damage, and topographic position to describe the positions in
gradient space following disturbance and test the predictive
ability of this gradient approach in determining the dynamics
of recovery.

    There are five specific goals to this project:  1) to
describe the spatial patterns of hurricane damage and the
factors that influence these patterns, 2) to predict the
vegetation community response to gradients of hurricane damage,
3) to assess the role of gradients of solar radiation and soil
moisture in predicting rates of biomass accumulation, 4) to
analyze the dynamics of recovery of the canopy structure and
the factors that influence this process, and 5) to develop a
spatially explicit landscape simulation model of the hurricane
disturbance and recovery that incorporates the above concepts. 
These goals are examined in chapters two through six,
respectively.

     Chapter 1: Catastrophic Wind Damage to Forests -  I began
with a review of the literature relating to catastrophic wind
impacts on forests.  With this review I attempted to answer
seven questions:  1) How should catastrophic wind intensity be
quantified?  2) How should catastrophic wind effects be
quantified?  3) Is catastrophic wind damage homogeneous, or are
there spatial patterns of damage, and can these patterns be
predicted?  4) How do biotic factors (e.g. tree size, species,
and stand differences) influence damage?  5) How do abiotic
factors (e.g. topography, soil, and previous disturbance)
influence damage?  6) Can any generalizations be drawn
regarding the dynamics of recovery from catastrophic winds?  7)
Are there consistent differences between tropical and temperate
forests in terms of wind disturbance and response?

        How should catastrophic wind intensity be quantified? I
suggest that all studies of catastrophic wind impacts include
these five measurements: 1) maximum sustained wind, 2) maximum
gusts, 3) storm duration, 4) rainfall total and percent of
average annual rainfall, and 5) distance between the study site
and the site of measurement of the above parameters.

     How should catastrophic wind effects be quantified?
Different measures of hurricane damage result in different
conclusions about severity of disturbance.  To facilitate
comparisons between sites I suggest a dual measure of severity:
1) structural loss - percent basal area lost or percent stems
lost, and 2) compositional loss - percent mortality. 
Regardless of individual research questions, catastrophic wind
researchers should report totals for their sites that include
structural loss and compositional loss.

     Is catastrophic wind damage homogeneous, or are there
spatial patterns of damage, and can these patterns be
predicted?  Catastrophic wind damage does not affect the
landscape uniformly.  Increasing intensity of wind results in a
gap size distribution skewed increasingly to larger gaps, but
the full range of gap sizes from small to large is still
created.  With increasing damage, the landscape changes from a
forest with gaps to one with isolated stands of intact forest.

     How do biotic factors (e.g. tree size, species, and stand
differences) influence damage?  Stem size, species, stand
conditions, and pathogens may all influence the severity of
damage during a windstorm.  I hypothesize that the relationship
between stem size and damage is unimodal and that studies
indicating other patterns are limited in sample size or by the
categories of stem size used in the analysis.  The differential
effects of wind on different species are thoroughly documented. 
In addition, some successional class trends exist.  Pioneer
species, such as Cecropia schreberiana in this study, appear to
be more susceptible to damage and more likely to die.  

     How do abiotic factors (e.g. topography, soil, and
previous disturbance) influence damage?  Topography,
disturbance history, and soil conditions all influence severity
of damage.  The role of topography in channeling wind is
complex.  Valleys are not always protected sites, even when
their aspects do not align with the storm winds.  Of particular
interest is the influence of lee slopes, which may provide
protection, and may be sites of turbulent airflow and therefore
increased damage.  Ridges are more exposed to acute wind
conditions, but may be pre-conditioned by chronic wind and
therefore experience less damage.  The relationship between
soil and wind damage relates to root growth, which may be
restricted by: shallow soils, high water table, a shallow
impermeable soil layer, or soil texture.  Previous disturbances
may influence subsequent wind damage by: a) increasing
turbulence by opening the canopy, b) selectively removing
susceptible trees, and c) shifting the vegetation composition
toward more wind resistant species.

     Can any generalizations be drawn regarding the dynamics of
recovery from catastrophic winds?  Four distinct paths of
response are identifiable: regrowth, recruitment, release, and
repression.  But the influence of disturbance severity and
environmental gradients on these paths is little understood.  I
propose that the dual damage parameter of structural and
compositional damage holds promise toward predicting the
principal path of response to wind disturbance.

     Are there consistent differences between tropical and
temperate forests in terms of wind disturbance and response?
There is a latitudinal gradient of increasing frequency and
intensity of catastrophic wind toward the tropics and
corresponding differences in the rate and path to recovery.
Regions that are more frequently impacted by these storms
exhibit lower levels of mortality, although structure damage
may be high, and recover at a faster rate relative to temperate
regions.

     Chapter 2: Factors Influencing the Spatial Pattern of
Hurricane Damage - I examined three aspects of the spatial
pattern of hurricane disturbance:  1) how do the spatial
patterns vary with different measures of hurricane damage, 2)
what is the gap size of hurricane disturbance, and 3) what are
the relative roles of abiotic environmental factors
(topography, substrate features, and disturbance history) and
biotic factors (stem density, basal area, and community
structure) in influencing patterns of disturbance.  Relative to
the last issue, I tested this hypothesis:

Hypothesis 1 -   Hurricane damage is more highly correlated to
     abiotic environmental factors than to biotic factors.

     Different measures of hurricane damage lead to different
perceptions of patterns of damage over the landscape.  The more
general measures of damage, such as basal area lost or canopy
damage, tend to minimize the species-specific differences and
result in more complex patterns of damage with increasing
proportions of larger patches in the scales examined in this
study (tens to hundreds of meters).

     Hurricane damage results in distinct patches of damage. 
Again, with more general measures of damage, more larger
patches (> 0.04 ha) are identifiable, and a greater proportion
of the total damaged area is in large patches.  Although more
large patches are created, the majority of patches are small
(0.0025 ha) and isolated.

     At scales of tens of meters, the pattern of damage is
controlled principally by biotic factors, specifically the
distribution of species that respond differently to hurricane
winds, therefore the hypothesis is rejected.  A trend appears
to exist with increasing importance of abiotic factors
(topography and storm intensity) at large spatial scales,
shifting to control by biotic factors at finer spatial scales.

     Chapter 3: Hurricane Damage Gradients and Vegetation
Community Dynamics -  The use of a two-dimensional gradient
space of structural damage and compositional damage is
effective in differentiating sites whose recovery is dominated
by recruitment of early successional species as opposed to
those sites whose recovery is dominated by regrowth of
surviving primary forest species.  These two appear to be the
principal vectors of recovery of the LEF following Hurricane
Hugo: regrowth and recruitment.  This predictive model is
effective on the largest scales, tens of hectares, in this
study, and down to the scale of 20 m plots, but fails if the
plot data is taken from a finer resolution (5 m).  At this fine
a spatial scale, plots can be impacted by damage that occurred
outside the plot.  The failure of the two-dimensional damage
gradient space is most often related to plots that recover
through recruitment, but are classified as undamaged.  

     Chapter 4:  Biomass Production in Response to Post-
Hurricane Environmental Gradients - I assessed the predictive
power of the primary gradients of soil moisture and solar
radiation and the secondary gradient of hurricane damage, in
determining rate of biomass accumulation.  I examine the
question: are energy profits maximized at the center of
gradients of abiotic environmental factors?

     My results show a distinct sensitivity of post-disturbance
growth to gradients of solar radiation and soil water.  The
relationship developed in the BEW seems robust enough to
transfer to the HRP, resulting in 78% of plots correctly
categorized in terms of growth rate.

     Maximum growth rates after disturbance are concentrated in
intermediate soil moistures, but at the higher solar radiation
levels.  It is possible that solar radiation levels higher that
those simulated for this study would result in lower growth
levels, but these results to do not support the assumption that
net energy profits, as reflected by growth rate, will be
maximized at the center of gradients of abiotic environmental
factors.

     Chapter 5:  Canopy Restructuring in Response to Hurricane
Disturbance - I described canopy structure using the foliage
profile technique, and combined sampling layers into five: 1)
herb (0-1 m), 2) shrub (1-4 m), subcanopy (4-12 m), canopy (12-
20 m), and surviving (20-30 m).  I evaluated the role of two
secondary gradients in influencing the dynamics of canopy
restructuring: 1) community dynamics (driven by gradients of
hurricane damage) and 2) topographic position.  I tested two
specific hypotheses:

Hypothesis 2 - Recovering forest stands have significantly
     different canopy structures when dominated by recruitment
     (early successional species) as opposed to dominated by
     regrowth (surviving late successional species). 

Hypothesis 3 - Canopy structure is significantly different
     between sheltered valleys and exposed ridges.

     Hypothesis 2 - Canopy structure differences in recovery
vectors - Tree community differences (quantified by API)
resulting from different severities of hurricane disturbance,
result in two distinct patterns of canopy recovery.  In plots
dominated by recruitment of pioneer species, the canopy
recovers as a homogeneous layer that grows upward, shading
lower levels.  In plots dominated by regrowth of surviving
trees, the canopy grows outward from these trees, creating a
foliage profile that is more heterogenous, and with more
structure at lower levels.  The statistical tests applied to
these two canopy structures did not detect significant
differences in the overall distribution or in the variance of
the canopy height.  However, in both vectors, the canopy layers
are significantly different from each other, but the maximum
cover occurs in the shrub layer of plots dominated by regrowth,
and in the subcanopy layer of plots dominated by recruitment. 
The herb layer, predicted to have the highest percent cover in
the regrowing plots, had less cover than the shrub layer,
possibly because of shading from the dense shrub layer.

     Hypothesis 3 - Canopy structure differences in topographic
position - The difference in solar radiation available to lower
canopy levels is amplified by topographic position.  Sheltered
valleys receive less solar radiation due to shading of the
terrain, so the lower canopy levels have less structure and the
difference between these two vectors of recovery is increased. 
More diffuse solar radiation penetrates to the lower canopy
levels on exposed ridges, so the  vegetation structure at these
levels is denser and the difference between the two vectors of
recovery is diminished.  The statistical tests applied to these
two canopy structures did not detect significant differences in
the overall distribution.  However, the percent cover of the
shrub layer is significantly lower in the valley sites that on
the ridge sites.  The herb layer, again, showed no difference. 
This lowest layer of the canopy may be responding to a moisture
gradient rather than solar radiation levels.

     Chapter 6:  A Simulation Model of the Response of
Vegetation to Hurricane Disturbance - I synthesized all of
these findings into a computer model, driven by secondary
gradients of topography and hurricane damage, and simulated
primary gradients of solar radiation and soil moisture.  This
model predicts patterns of recovery as quantified by community
change, biomass production, and structure of the canopy.

     This model reasonably represents the dynamics of recovery
during the first four years of recovery.  When applied to
longer simulations, questions regarding the dynamics of
recovery can be examined.  Simulated results indicate little
impact of including a reorganization phase in the recovery
process.  Increasing the estimated recovery time from 60 to 100
years also has little impact, except due to the elevated
biomass assumed to be associated with a longer recovery time. 
Since recovery is modeled as an asymptotic curve, the fastest
rate of recovery occurs in the first years following
disturbance, so lengthening the recovery time has little
effect.  Changes in the disturbance regime do impact the
variables used to track recovery.  Disturbance frequency seems
to have a greater impact on community composition, biomass
levels, and canopy structure, than does changing the intensity
of storms.

MULTIPLE-DIMENSION PRIMARY GRADIENT SPACE
     The Multiple-Dimension Primary Gradient Approach (MDPGA)
attempts to quantify the impacts of disturbance through the
affects on gradients of solar radiation, water, temperature,
and nutrient availability, which I refer to as primary
gradients.  Before disturbance, any point in geographical space
can be placed in a region of multiple-dimension ecological
space represented by the variation of these primary gradients. 
Disturbance results in a displacement from this original
position along these abiotic gradients to a new position.  The
location of this new position in the multiple-dimension
gradient space controls the path to recovery, or the path to a
new and different 'stable' point.  If this theoretical
framework is correct, it holds promise for:  allowing
comparison among different disturbance types, facilitating
comparison among different ecosystems impacted by disturbance,
insights into the synergistic affects of multiple disturbances,
and identifying the critical differences between anthropogenic
and natural disturbances.

     This study did not test this theoretical framework
directly.  I did not measure the primary gradients, before or
after the hurricane.  Instead, I used simulated levels of
primary gradients of solar radiation and water, and gradients
of topography and hurricane damage, which I refer to as
secondary gradients.  Secondary, because each influence primary
gradients.  These simulated primary gradients and additional
secondary gradients do appear to influence the dynamics of
recovery.  However, a true test of the MDPGA requires: direct
measurement of the primary gradients before and after
disturbance, a demonstration that different disturbances can
cause different vectors of displacement, and that these
differences are reflected in the dynamics of recovery in a
predictable manner.

          RESEARCH GOALS FOR THE FUTURE
     Therefore, the first goal of future work in this area must
be to quantify abiotic gradients over the landscape of the LEF. 
This will serve to track changes in these gradients as the
forest continues to recover from Hurricane Hugo.  I suggest a
sampling methodology that focuses on several catenas of
differing aspects.  A single remote automatic weather station
(RAWS) could be moved from one sampling point to another on a
regular basis.  It would be preferable to have several RAWSs in
place and continuously recording.  At least one other RAWS must
be available to place in disturbed sites, new treefalls or
landslides, as opportunities arise.

     In addition to tracking the changes in the primary
gradients during recovery from hurricane Hugo, we must continue
to track the vegetation as it recovers, both to relate this
recovery to changes in the abiotic gradients and to test the
validity of the extrapolations of the data in the model
RECOVER.

     A third goal for future research is to be prepared for the
inevitable next hurricane.  In part, this preparation is
accomplished by continuing to monitor abiotic gradients and
vegetation dynamics in the inter-hurricane period.  If the
simulated results are accurate, the systems may never 'return
to a stable point', but rather be in a constant state of
recovery until the next disturbance.  If this is the case, the
MDPGA is still applicable toward determining the vector of
recovery and the position along this path when the next
disturbance occurs.  Rather than a system of stable states,
displacement, and return to the stable state, the LEF may
exhibit predictable paths through the gradient space, with
disturbance induced displacements of these paths.  Only long-
term monitoring through several hurricane disturbance and
recovery cycles will illuminate this issue.  

     Finally, the MDPGA must be tested for other disturbances
at other sites, particularly anthropogenic disturbances.  This
analysis must be applied to other hurricane-impacted forests,
and to ecosystems subjected to different disturbance regimes. 
Only then can it be determined if this approach can help to
unify our knowledge of the dynamics of ecosystem disturbance
and recovery.

Return to the Table of Contents