3.1. Tree mortality and changes in canopy radiative properties

Hurricane Wilma resulted in the widespread death of trees >1.5 m in height (Fig. 3). In general, tree mortality decreased with distance from the center of landfall where the highest wind speeds occurred. At the BSC site close to the Gulf shoreline, cumulative tree mortality reached 45% in October 2007. At the SH3 site near the EC system during the first 6 months after the storm, mortality rates reached approximately 25% of the stems >1.5 m in height. Cumulative tree mortality was lowest at the LO3 site, reaching 10% in November 2007.

MODIS EVI data provided evidence of the leaf area changes in the mangrove forest in response to the passage of Hurricane Wilma (Fig. 4). EVI values in the 9-pixel domain around the tower location immediately after the storm were reduced by approximately 50% from pre-storm values. In mid-May 2006, approximately 6 months after the storm, EVI values increased sharply with the onset of the annual wet season. However, average EVI values remained lower following the disturbance. A t-test showed a significant difference (P < 0.001) in the mean (±s.d.) 2004-2005, pre-storm (0.470 ± 0.041) and 2006-2009 post-storm (0.420 ± 0.040) EVI values derived for the 9-pixel domain. The mean (±s.d.) EVI values for the single pixel covering the tower location before the storm (0.446 ± 0.054) were also significantly lower after the disturbance (0.407 ± 0.049; P < 0.001). Intra-annual and inter-annual EVI values were variable and did not exhibit clear seasonal patterns before or after the storm. Both before and after the storm, the forest exhibited strong seasonal trends in albedo synchronized with summer and winter solstices (Fig. 5). On average, the surface albedo was 10% lower during the first year of measurements after Hurricane Wilma compared to pre-storm conditions. Albedo values in 2009 matched the pre-storm, 2004 observations.

Fig. 3. (left) Hurricane Wilma made landfall on October 24, 2005 and caused widespread and delayed mortality (A) of trees >1.5 m in height at these sites. Error bars represent ±1 standard deviation from the mean percentage of dead to total number of stems observed in multiple plots of variable radii at each site. Sediment elevations, relative to the last measurements made before the storm in 2005 (B) declined at the SH3 and BSC sites until the beginning of 2009. [larger image]		Fig. 4. (middle) Enhanced Vegetation Index (EVI) values at the tower location for pre- and post-Hurricane Wilma periods during which eddy-covariance data were also available. Confidence bands represent the minimum and maximum EVI values of the site pixel and 8 adjacent pixels during each 16-day averaging period. [larger image]		Fig. 5. (right) Five-day moving average of mean daily (8:00-16:00 h) albedo during 2004 and 2006 through 2009. [larger image]

3.2. Soil and air temperature profiles

Fig. 6. Mean diurnal NEE trends (A) show net daytime CO2 uptake was lower in March 2007 (after Hurricane Wilma) compared to March 2004. Differences in NEE were attributed to the storm disturbance but not to differences in environmental variables including Kin (B), Rnet (C), and TA (D) and VPD (E). Error bars represent ±1 standard deviation around the mean half-hourly values. [larger image]

Some significant differences in pre- and post storm average T_A values were observed, but these were not sufficient to account for the observed increases in T_S. In the 2006-2007 dry season for example, both daytime and nighttime T_A were significantly lower, while daytime T_S values were relatively unchanged and nighttime T_S values were significantly higher compared to 2004-2005 prestorm values. In the 2007-2008 dry season, the above-canopy T_A values recorded at 27 m and the T_S values at all depths during both daytime and nighttime periods were significantly greater than pre-storm values. However, during this period the temperature gradients between surface soil T_S and above-canopy T_A values were also significantly different compared to the temperature gradients observed before the storm due to the warmer T_S. The increases in T_S after the storm led to significantly smaller daytime and significantly larger nighttime temperature gradients between surface soil T_S and above-canopy T_A in 2006-2008.

Within-canopy air temperature lapse rates between 1.5 m and 20 m were significantly affected by Hurricane Wilma. Following the disturbance, positive lapse rates (i.e., lower T_A with height above the substrate) were observed during the dry and wet seasons of 2007 and 2008. Sensor failure prevented the calculation of lapse rates in 2009. Before the disturbance, solar heating of the intact upper canopy layers generally resulted in negative lapse rates between these two measurement heights. The positive lapse rates after the storm are indicative of increased atmospheric instability inside the forest canopy. During 2007 the bi-monthly average (±1 s.d.) u* threshold of 0.18 ± 0.10 m s^-1 was 25% lower than the monthly average value found in 2004.

3.3. Solar irradiance and net radiation partitioning

Incoming mid-day solar irradiance K_in values were on average higher during 2004 than years 2005-2009 (data not shown). The R_net/K_in ratios (Table 2) were also generally higher following the disturbance compared to pre-storm values. However, by the dry season of 2008-2009, R_net/K_in ratios were not significantly different than 2004-2005 values.

Significant reductions in daytime sensible heat fluxes (H; W m^-2) above the canopy were observed after the storm. For example, the average-hourly, mid-day H and H/R_net values were significantly lower compared to 2004-2005 values, and relatively stronger reductions were observed in the wet compared to dry season months (Table 2). In both dry and wet seasons, mid-day H and H/R_net values remained significantly lower than pre-disturbance values through 2009. In contrast to the results for H, dry and wet season latent heat fluxes (LE; W m^-2) were higher after the disturbance compared to 2004-2005 values. Monthly total daytime LE and LE/R_net both before and after the storm were highly variable, but the seasonal average values of both variables were significantly higher through 2009 compared to pre-storm values. Applying a constant latent heat of vaporization of 43.74 kJ mol^-1 to these LE fluxes and taking into account seasonal variations suggests that annual evapotranspiration rates increased by as much as 25% (~250 m yr^-1) above 2004-2005 values.

Soil heat fluxes (G; W m^-2) also increased significantly following the hurricane. The largest increases in mid-day G were observed during the 2006-2007 dry season in the first year of measurements after the storm (Table 2). The magnitude of the dry season increases in G from 2004 to 2005 values declined with time after the disturbance, and by 2007-2008 the dry season G was not significantly different from pre-storm values. Average wet season G values returned to pre-disturbance values in 2009.

Table 1. [click on the thumbnail image above to view Table 1]

---

Table 2. [click on the thumbnail image above to view Table 2]

3.4. Net ecosystem exchange of CO₂

Fig. 7. Monthly integrated (A) daytime (-NEEday) and nighttime (NEEnt), and (B) 24-h total (-NEEtot) net ecosystem CO2 exchange. [larger image]

By the time the EC measurements began 1 year after Hurricane Wilma in November 2006, daytime NEE, particularly during the wet seasons following the storm, had largely recovered to pre-storm conditions. The near-complete recovery of daytime NEE is consistent with our observations of substantial regrowth of foliage in the canopy and with the comparisons of pre- and post-storm EVI values (Fig. 4). However, several key differences in pre- and post-storm NEE values were still apparent in the EC data. For example, in March 2007, the mid-day CO₂ uptake rates were reduced by 5-8 µmol (CO₂) m^-2 s^-1 compared to March 2004 (Fig. 6A). These differences in NEE were not attributed to differences in the local climate, as environmental conditions above the canopy were similar during these 2 months (Fig. 6B-E). The differences observed in March 2004 and March 2007 were characteristic of the overall impacts of the storm on dry season CO₂ uptake rates. In contrast, during the 2007 wet season, the average monthly mid-day CO₂ uptake rates differed from pre-storm values by a maximum of 4 µmol (CO₂) m^-2 s^-1. The average nighttime net ecosystem CO₂ exchange rates (i.e. nighttime respiration) during wet and dry seasons following Hurricane Wilma were higher by as much as 0.5 µmol (CO₂) m^-2 s^-1 and 2 µmol (CO₂) m^-2 s^-1, respectively.

The impacts of Hurricane Wilma were less apparent in the seasonal comparisons of average monthly total NEE values before and after the storm. Inter-annual and intra-annual variability in the environmental factors that influence NEE tended to obscure the direct impacts of the disturbance on CO₂ fluxes. For example, standard t-tests showed the differences between the values of average monthly total daytime (-NEE_day), nighttime (NEE_nt) and the 24-h total net ecosystem CO₂ exchange (-NEE_tot) observed in both dry and wet seasons before and after the storm were insignificant (Table 3). Similar results were obtained using the nonparametric Mann-Whitney-Wilcoxon rank-sum test. The most consistent differences in pre- and post-storm average monthly values were found in wet season nighttime respiration (NEE_nt) values. Pre- and post-storm monthly NEE values (including daytime, nighttime, and 24-h total values) showed the most divergence in 2007-2008, possibly due to the delayed peak in mortality of stems which occurred during this period. The general lack of consistent and significant pre- and post-storm differences in dry season, wet season or annual NEE may be partly attributed to the relatively small sample sizes (n = 7, 5, or 12) within each group, and to the magnitude of variability in NEE compared to mean monthly values.

Visual comparisons of the time series of monthly -NEE_day, NEE_nt and -NEE_tot values before and after the storm showed distinct but variable disturbance impacts. Consistent differences in pre- and post-storm monthly -NEE_day were generally not apparent, and exhibited similar seasonal variation before and after Hurricane Wilma (Fig. 7A). Maximum uptake rates occurred in the late dry season both before and after the storm. In two distinct periods after the storm (May-June 2007 and in March-April 2009), sharp increases in daytime CO₂ uptake rates above the wintertime minima were observed which matched or exceeded the corresponding values before the storm. The effects of storm were more apparent in the pre- and post-storm NEE_nt and -NEE_tot comparisons. Post-storm NEE_nt values (Fig. 7A) were consistently higher (i.e., greater total nighttime respiratory CO₂ fluxes), and as a result, the 24-h total CO₂ uptake (-NEE_tot) values were consistently lower (Fig. 7B). The variability in -NEE_tot values also increased after the storm.

f-Tests showed the ridge regression model (Eq. (1)) accurately predicted -NEE_day, NEE_nt, and -NEE_tot for 2004-2005 (P < 0.001 for each predictand). Comparisons between the modeled and observed monthly integrated daytime (-NEE_day), nighttime (NEE_nt), and 24-h total CO₂ exchange (-NEE_tot) for posthurricane periods confirmed our observations that the relative magnitude of the storm's impact on CO₂ exchange depended partly on the season. In this scheme, the average differences between modeled and observed NEE values that were significantly greater or less than zero indicated a disturbance effect that was independent of the effect of other environmental drivers. We found our assessment of the storm's impact on daytime NEE depended on the length of time over which the average differences in modeled and observed values were determined. For example, the modeled, average monthly total daytime CO₂ uptake (-NEE_day,mod) values were often greater than observed values (-NEE_day,obs; Fig. 8A) but the magnitudes of these differences were not consistent across seasons (Table 4) and they did not decline over time as might be expected in a recovering system. The average difference between -NEE_day,mod and -NEE_day,obs determined seasonally was significantly different from zero (P < 0.01) only for the 2007-2008 dry season. The differences in -NEE_day,mod and -NEE_day,obs for the wet seasons of 2007-2009 were not significantly different from zero. However, the average differences between -NEE_day,mod and -NEE_day,obs determined on an annual basis (n = 12) were significantly greater than zero in 2007-2008 and 2008-2009, but not in 2006-2007. This suggests the lack of consistent, significant differences in -NEE_day,mod and -NEE_day,obs for the dry seasons following the storm may have been due in part to the small sample sizes (n = 7) combined with high interannual variability in other environmental drivers. When grouped across years (2006-2009), the average differences between -NEE_day,mod and -NEE_day,obs revealed a significant hurricane effect on the dry season (P < 0.01) but not wet season (P = 0.20) daytime CO₂ uptake rates. The magnitude of the changes in monthly -NEE_day relative to pre-storm values was also generally greater in the dry seasons than the wet seasons.

Modeled values of monthly total nighttime respiratory CO₂ fluxes (NEE_nt,mod) consistently underestimated observations (Fig. 8B), and unlike the results for daytime CO₂ fluxes, the largest differences occurred during the wet rather than the dry seasons after the storm. For example, the values for NEE_nt,mod - NEE_nt,obs were significantly different from zero (P < 0.01) during all the wet seasons through 2009. On the other hand, dry season values of NEE_nt,mod - NEE_nt,obs were significantly different from zero (P = 0.03) only during 2007-2008. Similarly, the average difference between seasonal NEE_nt,mod and NEE_nt,obs pairs grouped across years (2006-2009) revealed a significant hurricane impact on wet season (P < 0.01) but not dry season (P = 0.11) nighttime respiration rates.

Grouping the dry and wet season values showed the average differences between modeled and observed daytime and nighttime CO₂ fluxes were significant (P < 0.04) in all years with the exception of 2006-2007. The apparent lack of a distinct hurricane impact on 2006-2007 CO₂ fluxes in the first year of our measurements may have been due to the delayed mortality of stems observed across the impact zone and at the EC site. Modeled values of monthly integrated, total 24-h CO₂ fluxes (-NEE_tot,mod) significantly overestimated observations (Fig. 8C) during all periods with the exception of the 2006-2007 dry season (P = 0.08).

The impacts of Hurricane Wilma were most apparent in the comparisons of annual CO₂ uptake rates before and after the storm. Annual total CO₂ uptake in the first year of measurements following the storm exhibited an approximate 30% reduction compared to average annual uptake rates in 2004-2005 (Table 5). In 2008, annual CO₂ uptake rates were again approximately 30% lower compared to 2004-2005 values and in 2009, the annual CO₂ uptake rates were 21% lower than pre-storm values.

Table 3. [click on the thumbnail image above to view Table 3]

---

Table 4. [click on the thumbnail image above to view Table 4]

---

Table 5. [click on the thumbnail image above to view Table 5]

3.5. Discussion

3.5.1. Changes in canopy radiative properties and energy balance

Fig. 8. Comparison of ridge-regression model estimates versus observed, monthly integrated (A) daytime (-NEEday), (B) nighttime (NEEnt), and (C) 24-h total (-NEEtot) net ecosystem CO2 exchange during January 2004 to August 2005 (04-05), and the three, 12-month (November-October) hydrologic years following disturbance. [larger image]

Field observations of the impact zone immediately after Hurricane Wilma revealed substantial and widespread defoliation as well as significant numbers of downed stems and large branches (Smith et al., 2009). Estimates of mortality of stems caused by the storm (Fig. 3) are comparable to mortality rates measured as a result of insect outbreaks and severe fires in other forests (Amiro et al., 2010). Although measurements of the changes in leaf area in the forest before and after the storm were not available, the distinct reductions in EVI (Fig. 4) and albedo (Fig. 5) after the storm confirmed the severity of the impact on the forest. EVI decreases as light absorption through photosynthesis decreases (Huete et al., 2002), and the decreases in surface albedo following the storm may have been due to a greater proportion of solar radiation being absorbed by surface water and the moist dark soils beneath the damaged canopy. Both EVI and the albedo data indicated that the most severe effects of Hurricane Wilma occurred within the first year following the disturbance prior to the resumption of our EC measurements, but the two measures produced different estimates for the total duration of the storm's impact. The EVI data suggest the effects of the storm persisted through 2009, while the albedo data suggest recovery in canopy radiative properties was complete by mid-2008. The discrepancy may be due to spatially variable impacts of the storm and to the different measurement areas and sensitivity of these two indices.

We suggest the differences in EC and energy balance data before and after Hurricane Wilma through at least mid-2008 can be attributed to several linked mechanisms originating from wind damage to the canopy. First, we ascribe the decreases in the u* threshold to changes in canopy roughness caused by the loss of large limbs and the toppling of stems. Gu et al. (2005) also showed the u* threshold can change over time and space depending on the leaf area index, stem density, and canopy height. Second, we suggest the reductions in biomass and light interception in the upper canopy led to decreases in both the radiative fluxes from the upper canopy to the atmosphere and to the significant decreases observed in H (Table 2). The reduction in heating in the upper canopy layers would also have contributed to the observed shift from predominately negative to positive lapse rates between 1.5 and 20 m, and in the shift from statically stable to statically unstable conditions within the canopy. These lapse rates contributed to more efficient mixing of air masses between the lower and upper-canopy layers and the atmosphere. Lastly, we also suggest that the relatively higher soil temperatures and higher soil heat fluxes after the storm (Table 1) may have been the result of the increased penetration of solar irradiance through the damaged canopy. In turn, the greater amount of available energy in the lower canopy appears to have enhanced direct evaporation from the open water and saturated soil surfaces. In support of this hypothesis we note that after the storm, LE/R_net ratios increased by up to 40%, while H/R_net ratios decreased by 10-30% (Table 2). The pre- and post-storm differences in the partitioning of available energy into LE and H declined between 2006 and 2009 as the canopy recovered towards the predisturbance state.

Following the storm, the percent closure of the surface energy budget ((H + LE)/(R_net - G)) during the daytime improved from 67-70% to 79-83% during 2004-2005 and 2006-2009, respectively (Table 2). This degree of closure is within the range of values reported for FluxNet sites across 50-site years (Wilson et al., 2002) and boreal forests (Barr et al., 2006; Sanchez et al., 2010). For the mangrove forest, one potential cause of the surface energy imbalance includes the lateral transport of heat associated with high and low tides. Studies in other tidally influenced ecosystems (e.g., salt marshes) have reported daytime energy closure as low as 49% (e.g. Moffett et al., 2010).

3.5.2. Impacts on daytime and nighttime NEE

The ridge-regression modeling indicates that the inter-annual variability in non-hurricane drivers of NEE, such as T_A, K_in, and salinity (Barr et al., 2010) was not sufficient to explain the full range of post-storm differences we observed in daytime and nighttime NEE, particularly when these differences were calculated across years (Table 4, Fig. 8). However, in contrast to the soil temperature and energy balance data that suggest the impacts of Hurricane Wilma persisted until 2008, the EC measurements show that the values for average monthly total daytime NEE (-NEE_day) after Hurricane Wilma at times approximated pre-disturbance values by late 2006 and early 2007. This was particularly evident in the comparisons of pre- and post-storm wet season daytime CO₂ uptake rates. During the first year after the storm, we observed widespread emergence of new leaves on clumped, epicormic shoots on both L. racemosa and A. germinans beneath the damaged upper canopy. Over time, branch elongation and leaf re-growth in the upper canopy occurred along with a gradual loss of leaves in the lower canopy layers, possibly as a result of light competition. These observations match with those reported following Hurricane Andrew in 1992 (Baldwin et al., 2001; Doyle et al., 1995). We speculate that the initial rapid recovery of leaf area, especially on the epicormic shoots and surviving juveniles in the lower canopy layers resulted in the relatively quick recovery of total CO₂ uptake capacity in this forest. This matches with the rapid recovery in EVI and may also explain why we did not find significant differences in average monthly -NEE_day during the first dry season measurements in 2006-2007. Our results also suggest that the impact of the storm on energy balance and soil temperatures persisted through the initial leaf-out and subsequent changes in leaf area distribution into 2008-2009. The relatively rapid recovery of daytime CO₂ uptake after Hurricane Wilma reflects an adaptation to the frequent disturbance from hurricanes and lightning damage in the Florida Everglades region (Smith et al., 1994; Zhang et al., 2008).

Some other aspects of canopy recovery from this disturbance also likely had an influence on our EC data. For example, using airborne LIDAR measurements, Zhang et al. (2008) showed that Hurricane Wilma, and to a lesser extent Hurricane Katrina, caused 8-fold increases in both canopy gap density and total gap area from 1.1% to 12.1% along the Shark River. Hurricane Wilma also created new gaps (roughly 10-100 m^-2) in the canopy immediately surrounding the EC site. Since the disturbance, substantial re-growth has occurred in such gaps, including the growth of juvenile trees that survived the storm and widespread emergence of new propagules. Variable disturbance responses among the dominant species at this site were also noted and may have influenced NEE values. R. mangle, one of the three co-dominant species at the EC site, does not regenerate if the plumular apex is lost or severely damaged as often occurs during disturbance events (Gill and Tomlinson, 1971; Tomlinson, 1986). Crowns from the other two dominant species at the site, L. racemosa and A. germinans, are capable of regenerating from meristem tissue (Snedaker et al., 1992). After the initial impact from Hurricane Wilma, damaged R. mangle trees continued to produce leaves on surviving branches even though mortality in these trees continued to rise (Fig. 3).

In the mangrove stands of south Florida, leaf production rates are lower during the dry season (Gill and Tomlinson, 1971; Arreola- Lizárraga et al., 2004; Parkinson et al., 1999). In addition, leaf longevity increases with time after a hurricane disturbance (Ross et al., 2001). The combination of low leaf production rates in the dry season plus reductions in the average life-span of emerging leaves may have been partly responsible for the relatively stronger impact of the storm on dry season compared to wet season CO₂ uptake rates. We note that the impacts of the storm on albedo were also relatively stronger in the dry seasons compared to the wet seasons of 2007 and 2008. The largest differences in pre- and post-storm -NEE_day were observed in the dry season of 2007-2008, and this coincided with the peak in cumulative tree mortality after the storm at the nearby SH3 site. However, we do not fully understand why the impacts of Hurricane Wilma seemed to more conspicuously effect dry season compared to wet season CO₂ uptake rates.

The significant and persistent increase in nighttime respiratory CO₂ fluxes (NEE_nt) was the most evident effect of Hurricane Wilma on forest-atmosphere exchanges. Several factors likely contributed to the increases in NEE_nt, such as the decomposition of storm-generated litter and coarse woody debris (CWD). The statically unstable conditions which characterized the canopy air mass after the disturbance (Table 1) also facilitated the transport of respiratory CO₂ fluxes from soil, litter, and woody debris substrates to the EC system. Smith et al. (1994) estimated that hurricane Andrew in 1992 contributed 141 Mg ha^-1 of CWD in the mangrove forests on the southwest coast of Florida. Turnover times of the labile fraction of mangrove CWD in Florida range from 4 months to 23 years, and represent 10-20% of CWD by dry weight (Romero et al., 2005). Decomposition rates of the refractory components of CWD are two orders of magnitude smaller than those of the labile fraction (Romero et al., 2005) and likely contributed little to the observed increases in nighttime respiration at the tower site.

In contrast to the results for daytime CO₂ fluxes, our observations and the model results show the impact of Hurricane Wilma on nighttime CO₂ fluxes was generally larger in the wet seasons than in the dry seasons. The reason behind this trend is uncertain, though higher T_A and higher T_S during the wet seasons may have resulted in a relative increase in decomposition rates of litter and CWD generated by the storm. Across seasons, the EC data suggest the long-term effect of hurricanes on mangrove forest - atmosphere exchanges of CO₂ is manifest largely through their impact on ecosystem respiration (Table 5). The impact of hurricanes on the fluxes of particulate and dissolved inorganic and organic carbon in tidal mangrove forests should also be investigated further. The fluxes of dissolved and particulate carbon are considered to be significant components of the carbon balance in these systems (see Bouillon et al., 2008; Barr et al., 2009).

3.5.3. Soil surface elevations and ecosystem trajectories

The storm surge associated with Hurricane Wilma added approximately 43 mm of sediment elevation at SH3 and LO3 (Fig. 3B) through deposition of calcitic material. At the SH3 site, the sediment deposition was followed by a more than 20 mm decline in soil surface elevation in a span of 3 years after the storm. In a shorter-term study using SET measurements at SH3 before and after Hurricane Wilma, Whelan et al. (2009) found that the initial loss of 10.5 mm in substrate elevation at this site 6 months after the storm could be attributed to -4.9 mm of erosion in the surface layer, -6.3 mm loss due to compaction in shallow and middle soil layers where roots are found, and a +0.8 mm expansion in the deep soil zone. Between 6 and 12 months following the storm, Whelan et al. (2009) report that erosion slowed (-3.6 mm), and the shallow soil zone expanded (6.8 mm) at SH3 as new rootlets stabilized the soil matrix and invaded the calcitic deposits. At the LO3 site, little or no elevation loss occurred during the 2 years following the storm. The trees at LO3 were defoliated during the storm but suffered much less structural damage and mortality compared to the SH3 and BSC coastal margin sites (Smith et al., 2009). The combined data from the LO3, BSC, and LO3 sites through 2009 showed that the progressive decline in soil elevation following the storm was strongly correlated (R² = 0.72) to the magnitude of cumulative tree mortality. This relationship leads to the hypothesis that root death and decomposition followed by soil compaction continued to be a primary mechanism in the loss of soil elevation after the storm. The loss of soil surface elevations in the damaged portions of this forest slowed substantially in 2009 and coincided with a leveling of cumulative mortality rates in stems. It is not known what role erosion played in the decline in soil surfaces after the storm.

The loss of sediment elevation following tree mortality at the SH3 site is concurrent with higher nighttime respiration rates measured at the nearby EC site, leading us to further hypothesize that enhanced soil carbon loss due to elevated T_S may have also contributed to the decline in soil mass (and elevation) observed at SH3 and BSC sites following the storm. Our observations show that the timing of the recovery in ecosystem respiration rates and soil temperatures to near pre-storm values in 2009 coincides with the period when declining soil surface elevations and tree mortality rates also stabilized, and this is considered as support for this hypothesis. However, since soil temperatures increased by 1-2 °C following the storm, it is likely that any increases in soil CO₂ flux caused by the higher T_S represented a relatively small fraction of the total increase in ecosystem respiration observed after the storm (Lovelock, 2008).

Despite the subsequent loss of soil surface elevations following the storm, the calcitic deposits brought by the Hurricane Wilma storm surge resulted in net gains in soil surface elevations in inland areas of the mangrove forest. However, in the areas located closest to the coast, such as those represented by the BSC site where the storm caused significant erosion, sediment elevations remained lower than pre-storm conditions. Surveys of aerial photographs from the early 1900s showed that similar hurricane impacts in the Everglades coastline resulted in permanent transitions from mangrove forests to intertidal mudflats along tidal channels draining to the Gulf of Mexico (Smith et al., 2009). Our findings also agree with the findings of Cahoon et al. (2003) who found Honduran mangroves forests that suffered partial mortality after Hurricane Mitch (1998) gained elevation with root production while the basin mangrove forests that suffered total mortality experienced irreversible peat collapse.

3.5.4. Summary and conclusions

This study produced new information to quantify the impacts of tropical storm disturbance on carbon cycling and energy balance in the mangrove forests of the Florida Everglades. The tropical storm disturbance caused structural changes in the forest that resulted in nearly 100% defoliation in the upper canopy and widespread tree mortality. Immediately after the storm, surface radiative attributes such as EVI declined by 50% compared to pre-storm values. One year after the disturbance, surface albedo remained 10% lower than before the storm. The changes in canopy architecture and leaf area after the storm appear to have allowed for greater solar irradiance transfer to the lower canopy layers and to the substrate. As a result of this disturbance, significant increases in both daytime and particularly nighttime soil temperatures down to -50 cm were observed. Within-canopy conditions also changed from statically stable to statically unstable which contributed to efficient transport of water vapor and CO₂ from the substrate and lower canopy to the upper canopy layers. These changes contributed to greater partitioning of the available energy into latent heat fluxes and a decrease in sensible heat fluxes. Annual evapotranspiration rates were 25% higher after the storm. The pre- and post-hurricane differences in the partitioning of available energy into latent and sensible heat fluxes declined over time as the ecosystem structure approached its undisturbed state in 2009.

The most evident impact of Hurricane Wilma was observed in the persistent increase in nighttime ecosystem respiration, possibly due to decomposition of the new litter and coarse woody debris generated by the storm. Increases in belowground respiration due to the higher soil temperatures may also have contributed to higher rates of respiratory CO₂ fluxes measured after the storm. The effects on nighttime respiration were most pronounced in the wet seasons. In contrast, the damage from the storm had a more pronounced effect on CO₂ uptake during the dry seasons compared to the wet seasons. However, because much of the recovery in active leaf area occurred prior to the resumption of our EC measurements, we were often unable to consistently differentiate the independent effects of the storm from non-hurricane drivers of CO₂ uptake on seasonal time scales. This may have been due to the delay in maximum tree mortality after the storm and to complex leaf regeneration patterns on surviving stems and juveniles. The impact of the storm on annual CO₂ uptake rates was more apparent, and net carbon assimilation rates remained approximately 250 g C m^-2 yr^-1 lower in 2009 compared to the average annual values determined for 2004-2005.

Across the hurricane-impacted region, cumulative tree mortality rates were correlated with declines in peat surface elevation. In the most-disturbed zones, soil surface elevations remain substantially lower 4 years after the storm. Less-impacted stands exhibited a high degree of resilience and rapid recovery following Hurricane Wilma and the net sediment deposition caused by the storm may help to maintain the viability of the forests in these areas in this era of sea level rise. These new results confirm previous work in the Florida Everglades that show tropical storm disturbance is a primary factor regulating plant-soil carbon cycles and mangrove ecosystem trajectories in this region.