Photosynthetic recovery of foliage after wind disturbance activates ecosystem CO2 uptake in cool temperate forests of northern Japan



[1] The effects of wind disturbance on forest dynamics and ecosystem CO2 exchange were examined in cool temperate forests of northern Japan during 2004–2008 using eddy covariance (EC) measurements. One site was a young, even-aged, monoculture, deciduous forest; the other was an uneven-aged mixed forest of evergreen and deciduous overstory tree species, including some over 200 years old. On 8 September 2004, a strong typhoon struck the forests, after which leaf and branch amounts decreased in young growth forest, but foliage showed little change in old growth forest. By 2006, foliage at the young-growth forest had recovered to the 2004 pretyphoon state. Average daily accumulated gross primary production (GPPd), terrestrial ecosystem respiration (TERd), and net ecosystem exchange (NEEd) were assessed for six growth stages annually. After the typhoon, large increases in GPPd were found during the growing stage of overstory tree species with high photosynthetic rates compared to that before the typhoon. Pronounced increases in GPPd and corresponding large reductions in NEEd were detected at the young-growth forest, indicating that NEEd was largely regulated by GPPd throughout the growing stages. Although EC measurements contain uncertainty, our continuous EC measurements revealed that interannual variability in meteorological variables and structural changes in foliage have only small impacts on GPP and NEE, while photosynthetic recovery of foliage from typhoon damage has high potential to increase GPP and enhance NEE as compared with those under nondamage conditions.

1. Introduction

[2] Catastrophic wind disturbances can profoundly affect the foliation, structure, and species composition of terrestrial ecosystems [Foster and Boose, 1992; Everham and Brokaw, 1996; Webb, 1999; Yoshida and Noguchi, 2009]. Defoliation is a common damaging impact of wind disturbance [Ostertag et al., 2003; Li et al., 2007]. Strong winds may also cause structural changes in terrestrial ecosystems, altering gross primary production (GPP) [Cowling and Field, 2003], terrestrial ecosystem respiration (TER), and thus net ecosystem exchange (NEE) [Li et al., 2007]. Many studies have investigated wind disturbance effects on NEE in forest ecosystems [Knohl et al., 2002; Ito et al., 2005; Li et al., 2007]. Ito et al. [2005] evaluated annual variability in NEE in a mountainous forest in a cool temperate area of Japan using a terrestrial carbon dynamics model (the Sim-CYCLE model). Their simulation results demonstrated the importance of considering typhoon damage when estimating annual NEE; however, they did not focus in detail on the regulation of annual NEE after the typhoon. Knohl et al. [2002] investigated windstorm damage and NEE in a Russian boreal forest on the basis of eddy covariance (EC) measurements over 3 months. Their results indicated that frequent windstorms produced large amounts of deadwood and that this coarse woody debris had a long-term influence on NEE in the forest area, with the decomposition of deadwood accounting for about one third of TER. Furthermore, Li et al. [2007] investigated hurricane damage to foliage and NEE in a short statured tree population in the U.S. state of Florida and found that hurricanes produced heavy damage to leaves but little stem damage and thus had negligible effects on NEE. In the Northern Hemisphere, typhoons and hurricanes frequently occur during midsummer to autumn. These periods coincide with important growth periods when many deciduous tree species form leaf primordia on short branches for leaf production in the following year [Kozlowski and Pallardy, 1997; Toda et al., 2007]. Therefore, loss of stems or branches with leaf primordia due to wind damage may affect trees' ability to absorb carbon from the atmosphere via their leaves and thus affect their ability to survive.

[3] Yoshida and Noguchi [2009] examined historical census data to evaluate the vulnerability of major tall-tree species in northern Japan to a severe typhoon in 1954. The damage reached about 25% in terms of basal area and most damaged trees were uprooted. The high proportion of uprooting caused by the wind disturbance has the potential to contribute to sporadic regeneration of several tree species in the recovery process and to dynamic changes in forest size structure. Changes in the size structure of a forest may also result in physical changes in the forest environment, which could in turn affect the annual variability of carbon balance in the forest ecosystem. Few studies have substantially assessed the effect of wind disturbance on NEE [Li et al., 2007], even though wind disturbance is one of the most common natural disturbance factors in terrestrial ecosystems [Turner et al., 2003]. In addition, several reports have assessed the frequency and intensity of hurricanes, tropical cyclones and storms in the present climatic conditions and under a scenario of future global warming. They indicated that warming decreases the globally averaged frequency of hurricanes and tropical cyclones [e.g., Knutson et al., 2008, 2010], whereas a consistent increase in the averaged intensity of tropical cyclones and storms is shown due to warming [e.g., Walsh, 2004; Emanuel, 2005; Knutson et al., 2010]. Therefore, to predict NEE in terrestrial ecosystems under future climate changes, one must evaluate the relationship between the magnitude of wind damage and the recovery of structure, foliage, and NEE after wind disturbances.

[4] EC measurement systems were established in two forest stands of different age and size structure in Hokkaido, northern Japan. Fluxes of energy, water vapor, and CO2 between the atmosphere and the forests have been monitored at these sites since July 2003. A large typhoon (number 18, named Songda) passed over these forests in early September 2004, ripping away large amounts of foliage. This study had two main aims: to investigate the posttyphoon annual carbon balance in the two forests under similar climate conditions and to assess forest recovery and its effects on forest dynamics and carbon balance based on CO2 exchange measured by the EC method from 2004 to 2008 and from forest inventory data.

2. Materials and Methods

2.1. Climate and Typhoons

[5] This study was conducted in two forests in Hokkaido, northern Japan (Figure 1). The climate of the study area is primarily marine derived, and the annual mean short- and long-wave radiations are 130.0 W m−2 and 306.4 W m−2. The annual mean air temperature, specific humidity and precipitation are 2.7°C, 4.8 g kg−1 and 1237 mm, respectively. The area is widely covered with snow from November to May, and the snow season length averaged 209 days from 1999 to 2005 [Toda et al., 2011]. The plant growing season length is largely affected by the annual snow conditions.

Figure 1.

Map indicating the locations of the forest sites (MSR, 44°20′N, 142°15′E) and Sapporo, the largest city (43°04′N, 141°21′E) in Hokkaido, northern Japan.

[6] In September 1954, a huge typhoon (number 15 named Marie) passed over the island of Hokkaido with maximum recorded wind speeds of >30 m s−1 [Yoshida and Noguchi, 2009]. The strong winds resulted in the most severe damage, killing numerous trees throughout Hokkaido [The Scientific Investigation Group of the Wind-Damaged Forests in Hokkaido, 1959]. In general, only a few typhoons affect the Hokkaido region each year, but most have a low impact on forests. In contrast, the typhoon in 2004 produced the most damage to the forests of Hokkaido since meteorological data acquisition by the Japan Meteorological Agency commenced in 1951.

2.2. Features of the Forest Sites

[7] The study forests were located in a basin surrounded by low mountains in the Uryu Experimental Forest of Hokkaido University, Japan. One of the forests (hereafter referred to as young-growth forest) was established on the flat ridge of a hill with a 5°3′ slope in the northeast part of the Uryu Experimental Forest. In 1979, a nonwooded 72 ha area was scarified with heavy machinery. Since then, the forest has naturally regenerated at the site with no artificial management. The 600 m2 (20 m × 30 m) stand was established in 1998 as a study plot (44°20′N, 142°15′E, 550 m above sea level). The stand was composed of five deciduous species mainly dominated by broad-leaved Betula ermanii (40 years old in 2009) [Toda et al., 2011]. The canopy structure of the stand was homogeneous and a representative canopy height of the young-growth forest was estimated using the cumulative basal area inflection (CuBI) height, defined as the height of the inflection point of the plot between tree height and cumulative basal area [Nakai et al., 2010]. The CuBI height of the stand was 12.1 m.

[8] The other forest (hereafter referred to as old-growth forest) was an uneven-aged mixed forest of evergreen and deciduous overstory tree species, with the oldest trees being more than 200 years old. In the old-growth forest, which dominates the landscape in the Uryu Experimental Forest, we established a 4000 m2 (40 m × 100 m) study stand in 2003. This stand was located approximately 7 km southwest of the young-growth forest stand (44°19′N, 142°15′E, 300 m above sea level (asl)), and the topography around the stand was generally flat. The study stand included two evergreen coniferous and 15 deciduous hardwood species of various ages. The total density of trees in the stand was 2590 ha−1, and the total densities of Viburnum furcatum, Betula ermanii, Phellodendron amurense, Betula platyphylla var. japonica, Abies sachalinensis, Picea glehnii were 613, 413, 350, 218, 208, and 170 ha−1, respectively, based on forest inventory data from 2004. In contrast, the basal area (BA) of evergreen species (P. glehnii (14.3 m2 ha−1) and A. sachalinensis (7.6 m2 ha−1)) accounted for 75.2% of the entire BA (29.0 m2 ha−1) of the stand. We established 160 quadrats (5 m × 5 m in size) in the stand. Of the 160 quadrats, 76 had no trees, and these relatively empty quadrats were scattered heterogeneously throughout the stand. The stand included trees of various heights, due to interspecific and intraspecific competition, natural disturbances, and seedling establishment and regeneration over the long term. The CuBI height of the stand was 24.5 m [Nakai et al., 2010] and the maximum tree height in the stand was 22.9 m for deciduous species (P. amurense) and 33.6 m for evergreen species (P. glehnii). Large trees were located randomly throughout the stand. Thus, the old-growth forest presented a canopy structure with high horizontal and vertical heterogeneity.

[9] In both stands, the forest floor was densely covered with an evergreen dwarf bamboo; Sasa kurilensis Makino et Shibata in the young-growth forest and Sasa senanensis Rehder (SE) in the old-growth forest. These bamboos grow to heights of about 1–4 m and disturb tree seedling recruitment by creating a low-light environment on the forest floor.

2.3. Forest Inventory Assessment in the Young Growth Forest

[10] In the young-growth forest stand, forest inventory assessment has been conducted in late autumn (approximately from the beginning to middle of October) every year since 1998. Tree height, diameter at breast height (dbh) of all individual trees taller than 1.3 m, and number of live and dead individual trees for each tree species in the forest stand have been measured. Individual tree weight w was estimated as the summation of wp and wnp, where wp and wnp were weights of photosynthetic and nonphotosynthetic organs, respectively. Values of wp and wnp were determined based on the allometric relationships between wnp and dbh and tree height h and between wp and the leaf amount and dbh of individual trees [Watanabe et al., 2004]:

equation image

where θ and γ are allometric parameters and SLA is specific leaf area. Four B. ermanii trees of different heights near the stand were cut and these parameters were estimated as γ = 0.025 kg cm−2 m−1 and θ = 0.013 kg cm−2 based on the measured wp, wnp, h, dbh, and SLA (SLA = 15 m2 kg−1) of the four trees [Takahashi et al., 2004; Toda et al., 2011]. The stand biomass was estimated using these parameters and h and dbh for all trees in the stand. In addition, the total dry weight of branches of an individual tree was estimated as a fraction, 0.16, of the maximum leaf amount in each year based on the forest inventory data.

2.4. Plant Area Index (PAI)

[11] Seasonal and interannual variations in plant area index (PAI) were measured in both stands with a commonly used nondestructive optical device (LI-2000, LI-COR, Lincoln, NE, USA) once every 2 weeks to detect the plant growth period. At the same time, an alternative PAI estimation was made for the young-growth forest using upward and downward photosynthetically active radiation (PAR) sensors (LI-190, LI-COR) (i.e., the PAR approach) mounted at 2 and 20 m heights on the flux tower and the following equation [Saigusa et al., 2002]:

equation image

where kp is the absorption coefficient. In equation (2), daily averages of PAR2m↓ and PAR20m↓ from 1000 to 1400 local time (LT) were used to estimate PAI. Observed kp has been found to range from 0.3 to 1.5 [Jones, 1992; Saigusa et al., 2002]. In the present study, kp was estimated year-to-year by finding the kp value when PAI calculated by the PAR approach corresponded to the PAI derived from LI-2000 measurements. The estimated kp values ranged from 0.55 to 0.70. PAI estimates by the PAR approach were not conducted in the old-growth forest.

2.5. Litterfall Measurement

[12] Ten litter traps were established at constant intervals of several meters in the young-growth forest to estimate annual variations in litterfall from July to October during 2003–2006. July was chosen as the starting month because leaves start to fall from the dominant B. ermanii species at the end of July. Litter was collected from the traps about once a month and partitioned into leaf and stem (or branch) litter components. Each of these litter components was dried for 24 h at 80°C and the dry weight was measured. The mean value for each of these litter components was then calculated from the ten litter traps. Using data from the ten traps, a rejection test was developed to obtain representative values of the leaf and stem litter components in the forest. Here, only data that satisfied the criterion of P < 0.05 were used in the calculation process.

2.6. Eddy Covariance Measurement

[13] EC measurement systems were mounted at the top of a 20 m tall tower in the young-growth forest and a 30 m tall tower in the old-growth forest. The systems were composed of an open-path infrared gas analyzer (IRGA) (LI-7500, LI-COR) and sonic anemometer (Solent R3, Gill Instruments, Hampshire, UK). Raw turbulence data were recorded at 10 Hz. These turbulence data were then processed and corrected with a series of data quality control checks; i.e., the Webb, Pearman, and Leuning (WPL) correction [Webb et al., 1980], coordinate rotation [McMillen, 1988], humidity correction of sonic temperature [Schotanus et al., 1983], and frequency response correction [Moore, 1986] were applied to mean half-hourly fluxes of CO2. An additional correction for CO2 flux calculation to consider the effect of self-heating of the IRGA on the CO2 flux (i.e., the Burba correction) [Burba et al., 2006, 2008] was not applied to the turbulence data. Hirata et al. [2007] investigated seasonal and interannual variations in NEE of a temperate larch forest with open- and closed-path IRGA. They found that open-path IRGA systematically underestimated CO2 flux not only during the winter period with low temperatures but also during the plant growth period with warmer temperatures, suggesting the importance of applying the Burba correction to the CO2 flux calculation process throughout the year. Wohlfahrt et al. [2008] and Haslwanter et al. [2009], however, assessed the effectiveness of the Burba correction on the flux calculation process using their long-term flux data set for temperate grassland and found that the correction had little effect on CO2 flux. These studies indicate that the application of the Burba correction to the flux calculation process is at the verification stage.

[14] Finally, the half-hourly CO2 flux was derived from the covariance between fluctuations in vertical wind speed and CO2 concentration. The following steps were also conducted to remove poor-quality data. First, measurements during rainfall events were excluded because underestimation of latent heat flux has been found to occur during rainfall events [Watanabe et al., 2005]. Data collected when wind speeds were less than 0.5 m s−1 were also excluded because of sensor accuracy. Finally, data for the lower atmosphere under daily atmospheric conditions had to satisfy the following neutral or unstable range of atmospheric stability [Brutsaert, 1992]: −5/7 < ζ < 16, where ζ = −(zd0)/Lv, z is the measurement height, d0 is the zero-plane displacement, and Lv is the Obukhov length.

[15] EC-measured fluxes of CO2 do not represent actual ecosystem exchange under nonturbulent conditions [Goulden et al., 1996; Kolari et al., 2004]. Thus, fluxes are generally determined by using a regression model under friction velocity (u*) conditions below a given threshold value (u*c) (i.e., different u* threshold values) [Goulden et al., 1996; Falk et al., 2008; Knohl et al., 2009]. Several studies have determined u*c by plotting nighttime NEE against u* [Kolari et al., 2004; Takagi et al., 2009]. Kolari et al. [2009] determined u*c to be 0.2 m s−1 based on the fact that no significant change in measured NEE was observed above the selected u*c. Takagi et al. [2009] calculated average nighttime NEE for each u* class established by splitting the u* range from 0.05 to 0.8 m s−1 at intervals of 0.05 m s−1. They determined u*c by finding u* when the difference between the nighttime NEE under the given u* condition and the average value from nighttime NEE under the highest five u* class conditions was less than 98%.

[16] Procedures for filling gaps in NEE generated by the turbulence criteria mentioned above and for u* filtering were conducted following Kolari et al. [2009]. Half-hourly averaged NEE data meeting the turbulence criteria were used to derive GPP as follows:

equation image

[17] In equation (3), TER was modeled using the following empirical exponential temperature regression:

equation image

where R10 is the base level of respiration (i.e., respiration at 10°C), Q10 is the temperature sensitivity (i.e., the slope of the apparent temperature response of respiration), and T is the mean of air and soil temperature. In the present study, we used Q10 = 2.20 for the young-growth forest [Toda et al., 2011] and Q10 = 2.04 for the old-growth forest, based on tentative soil respiration measurements during the various plant growth stages (Table 1). Of the respiration measurements conducted in the forest stands during 2003–2007, most data from 2003 to 2005 could not be utilized because of poor quality. Instead, we assessed the validity of respiratory Q10 values based on the 2006–2007 measurements by comparing Q10 to that obtained at other forest sites under similar climatic conditions. As a consequence, the respiratory Q10 value for our forests lies among those of cool temperate and subboreal forest sites [Baldocchi et al., 1996; Goulden et al., 1996; Nakai et al., 2003; Saigusa et al., 2005].

Table 1. Description of the Six Plant Growth Periods From Stage A to Stage Fa
StageThe Period of the Plant Growth Stage
  • a

    Herein, we divided each year into these six stages to evaluate annual variability in carbon balance components. PAI and Sdp indicate the plant leaf area index and snow depth, respectively.

ADOY 1 to DOY 90 (or DOY 91 in a leap year)
BDOY 91 (or 92 in a leap year) to the day when PAI in the forest ecosystem is greater than 1 m2 m−2
CFrom the day when PAI in the forest ecosystem is greater than 1 m2 m−2 to DOY 250
DFrom DOY 251 (DOY 252 in a leap year) to the day when PAI ≤ 1 m2 m−2
EFrom the day when PAI < 1m2 m−2 to the day when Sdp > 0.1 m
FFrom the day when Sdp > 0.1 m to the end of the year

[18] In the case of missing or rejected NEE, the GPP was modeled empirically as follows:

equation image

where Pmax is the rate of saturated photosynthesis, ϕ is a parameter defining the convexity of the light response curve, α is the initial slope of the curve [Kolari et al., 2009], and PARtop is PAR measured at the top of the flux tower. Parameter ϕ was given a value 0.85 in the present study. The other parameters in equations (4) and (5) (i.e., R10, α, θ) were estimated using a 31 day time-moving window of the accepted NEE and TER fluxes. More detailed information on the parameter determination has been described by Kolari et al. [2009].

[19] Using the same approach as Takagi et al. [2009], we obtained annual variability in u*c from 0.25 to 0.55 m s−1 for both forests during the study period. For long-term calculation, it is useful to estimate the average value of u*c in a forest during the study period. However, actual u*c is likely to vary from year to year [Takagi et al., 2009] and thus annual NEE might be overestimated or underestimated depending on the selected u*c. Recent studies have reported that different u* thresholds cause large uncertainty in ecosystem CO2 exchange estimations for forest ecosystems [Miller et al., 2004; Kutsch et al., 2008]. Therefore, in the present study, we evaluated the annual carbon balance (i.e., GPP, TER, and thus NEE) for two separate u*c values (0.25 and 0.55 m s−1) so that we could consider the possible range of uncertainty in the annual carbon balance created by the selected u*c values.

2.7. Auxiliary Data

[20] Meteorological and hydrological measurements were conducted together with the EC measurements in both forests. Downward and upward shortwave radiation (CNR1, Kipp & Zonen, Delft, The Netherlands) was measured at the top of each flux tower. Wind speed (010C; Met One Instruments, Rowlett, TX, USA), air temperature, and relative humidity sensors (HMP45D, Vaisala, Vantaa, Finland) were mounted at six heights on the tower, and specific humidity was calculated using these data. Volumetric soil water content was measured at a depth of 0.05 m with a platinum resistance temperature sensor (CS615, Campbell Scientific, Logan, UT, USA). In addition, a rugged acoustic distance sensor (C-SR50, Campbell Scientific) was mounted on the tower at a height of 6 m and attached to the tip of a 2 m long cross arm to measure snow depth. The snow depth derived from the acoustic distance sensor was corrected by air temperature at the same height. These variables were recorded at half-hour intervals using a dedicated data logger (CR-23x, Campbell Scientific). In addition, precipitation was measured using a tipping bucket rain gauge (RH-50, Ikeda-keiki, Tokyo, Japan) with an independent data logger (H21-002, Onset Computer Corp., Bourne, MA, USA).

3. Results

3.1. Annual Variation in Weather

[21] Solar radiation (Sd) and specific humidity (q) showed similar seasonal changes from 2004 to 2008 in the young-growth and old-growth forests (Figures 2a and 2c). The respective annual averages and standard deviations of Sd and q during the 5 years were 137.8 ± 5.2 W m−2 and 5.18 ± 0.12 g kg−1 in the young-growth forest and 133.6 ± 3.7 W m−2 and 5.48 ± 0.18 g kg−1 in the old-growth forest. In contrast, air temperature (Ta) and precipitation (P) differed between the forests (Figures 2b and 2d). The respective average Ta and annual P were 4.0 ± 0.44°C and 1170.0 ± 76.9 mm in the young-growth forest and 5.27 ± 0.26°C and 1097.0 ± 89.6 mm in the old-growth forest. Annual differences were observed in P in both forests during the 5 study years; the difference between the maximum and minimum values of annual P was 208.0 mm in the young-growth forest and 190.5 mm in the old-growth forest (Figure 2d). Measured volumetric soil water content (Wm) reached maximum values on approximately day of year (DOY) 130–150 for each year, followed by rapid decrease (Figure 2e). Minima of Wm occurred during DOY 180–240 and then gradually increased during DOY 210–240. The average minimum and maximum values of Wm were 17.6% and 43.9% at the young-growth forest and 24.2% and 51.3% at the old-growth forest, respectively, during the study period, although seasonality in Wm varied more in the young-growth forest than in the old-growth forest. In 2004, an earlier timing of Wm increase in autumn was detected at the young-growth forest. In the young-growth forest, Wm had reached 41.0% at the end of 2004, the highest value in the study period; at the end of the other years in the study period, Wm values were almost the same at about 32.9%. In the old-growth forest, Wm averaged 36.7% and did not differ greatly among the study years. Seasonal changes in wind speed (Ws) in the young-growth forest varied from 2 to 6 m s−1 on average and were greater than those in the old-growth forest, where a small range of 2 to 3 m s−1 was found (Figure 2f).

Figure 2.

Seasonal variations in meteorological variables at the young-growth forest (YGF) and old-growth forest (OGF) study sites from 2004 to 2008. Ten day averages of (a) shortwave solar radiation (W m−2), (b) air temperature (Ta) (°C), and (c) specific humidity (q) (g kg−1); (d) cumulative value of precipitation (P) from day of year (DOY) 1 for each year (mm); and 10 day averages of (e) volumetric soil water content (Wm) (%) and (f) wind speed (Ws) (m s−1).

[22] In early September 2004 (8 September, DOY 252), a large typhoon struck the area. The forest sites experienced strong typhoon winds of more than 10 m s−1 for 12 h (1000–2100 LT) in the young-growth forest and for 7 h (1000–1800 LT) in the old-growth forest. The maximum values of hourly mean Ws were 22.1 m s−1 in the young-growth forest and 15.3 m s−1 in the old-growth forest (Figure 3). The hourly mean and maximum values of Ws were also recorded at an automatic weather station (44°21′44″N, 142°15′54″E, 287 m asl) located between the forests. These data indicated strong typhoon winds of more than 10 m s−1 for 5 h (1200–1700 LT), including a short maximum Ws of 59.2 m s−1 (at 1500 LT) (Figure 3).

Figure 3.

Hourly mean and maximum values of wind speed (Ws) (m s−1) during September 2004 around the study area. Wind speed was sampled at 10 min intervals at the weather station in the Uryu Experimental Forest. The maximum of the 10 min mean Ws observations (Wsmax) was obtained by finding the highest value out of the six observations per hour. Hourly mean Ws was measured in the young-growth (solid line) and old-growth (dotted line) sites, as well as the weather station (solid gray line). The arrow indicates the day the typhoon struck.

3.2. Interannual Variability in Forest Dynamics

[23] In the young-growth forest, the number of trees per plot gradually decreased from 344 in 1998 to 277 in 2008 due to interspecific and intraspecific competition between individual trees, equivalent to a tree density (ρ) decrease from 0.58 m−2 in 1998 to 0.35 m−2 in 2008 (Figure 4a). Stand level dbh and stand biomass (Bw) showed increases of 2.5 cm (from 5.9 cm in 1998 to 8.4 cm in 2008) and 2.1 kg m−2 (from 4.0 kg m−2 in 1998 to 6.1 kg m−2 in 2008), respectively (Figures 4b and 4c).

Figure 4.

Annual variations in (a) the number of trees (closed circle, y axis on the left side) and tree density (ρ) (m−2) (open circle, y axis on the right side), (b) average diameter at breast height (dbh) (cm), and (c) stand biomass (Bw) (kg m−2) in the young-growth forest (YGF) from 1998 to 2008.

[24] The seasonal pattern of the plant area index (PAI) in the young-growth forest varied within the study period (Figure 5a). The maximum PAI in 2004 was approximately 3.3 m2 m−2 during mid-August (DOY 229). About 3 weeks later, the PAI drastically decreased to 1.7 m2 m−2 on DOY 252. The maximum PAI in 2005 was the smallest in the study period. However, the PAI gradually increased in the later growth stages in 2005 to an average of 2.2 m2 m−2. The PAI after 2006 showed the same seasonal trend as found in the earlier growth stages in 2004. In addition, from continuous seasonal PAI data from the young-growth forest, we assessed annual variability in the initial timing when PAI > 1 m2 m−2 in spring during the study period. The initial timings differed slightly by year (DOY 147 in 2004, DOY 151 in 2005, DOY 152 in 2006, DOY 155 in 2007, and DOY 153 in 2008).

Figure 5.

Seasonal and annual variations of plant area index (PAI) in (a) the young-growth forest (YGF) and (b) the old-growth forest (OGF) from 2004 to 2008. Each circle indicates a value of PAI obtained using a nondestructive device (LI-2000, LICOR), while each line shows PAI derived using photosynthetically active radiometers (LI-190, LICOR) attached to the flux towers.

[25] The PAI in 2004 in the old-growth forest was hardly reduced by the typhoon (Figure 5b). The duration of PAI > 1 m2 m−2 was longer in the old-growth forest than in the young-growth forest and values of PAI > 1 m2 m−2 were detected even after DOY 300. In the 5 years, PAI showed the same trend of maximum PAI during DOY 180–240 in summer and initial timing when PAI > 1 m2 m−2 in spring (Figure 5b).

[26] Leaf and branch litter components in litterfall in the young-growth forest were measured from 2004 to 2006, with respective amounts of 146.5 and 23.1 g m−2 yr−1 in 2004, 117.5 and 6.3 g m−2 yr−1 in 2005, and 147.9 and 6.3 g m−2 yr−1 in 2006. Leaf litter exceeded branch litter in all 3 years and was lower in 2005 than in 2004 and 2006. The amount of branch litter in 2004 was significantly higher than in the other years (P ≤ 0.05).

3.3. Seasonal and Annual Variability in GPP, TER, and NEE

[27] The daily GPP estimated with the u*c values increased with increasing PAR under lower PAR conditions and approached a constant value under higher PAR conditions at stages C and D of plant growth in both forests (Figure 6). The following empirical equation represented the relationship between PAR and GPP for each year:

equation image

where GPPmax is the daily average maximum photosynthetic rate (gC m−2 d−1) and αP is a parameter representing the initial slope of light sensitivity of the GPP. Values of GPPmax and αP were determined from the PAR and GPP values for each stage. In stage C in the young-growth forest, GPP was smaller in 2005 than in 2004 for the same PAR values. The higher GPP values in 2006 to 2008 outweighed the GPP in 2004 under the same PAR conditions (Figure 6a). The GPPmax in stage C ranged from 11.34 gC m−2 d−1 in 2005 to 17.71 gC m−2 d−1 in 2008. The trend of the PAR-GPP relationship in stage D was also different from that in stage C in the young-growth forest, and the GPPmax values were detected for the 5 study years in the range of 5.66 gC m−2 d−1 in 2004 to 10.07 gC m−2 d−1 in 2007 (Figure 6b). In contrast, the PAR-GPP relationship had the lower GPP values in 2004 and 2006 and the highest GPP values in 2008 under the same PAR conditions in stages C and D in the old-growth forest (Figures 6c and 6d). GPPmax for the 5 study years ranged from 13.90 (in 2004) to 16.35 (in 2008) gC m−2 d−1 in stage C and 9.15 (in 2006) to 12.01 (in 2008) gC m−2 d−1 in stage D.

Figure 6.

The relationship between daily mean photosynthetically active radiation (PAR) and daily accumulated gross primary production (GPP) for (a) stage C and (b) stage D in the young-growth forest (YGF) and for (c) stage C and (d) stage D in the old-growth forest (OGF). Data were obtained at hourly intervals and averaged using the GPP values calculated using u* threshold values (u*c) of 0.25 and 0.55 m s−1. Lines represent the empirical relationship between PAR and GPP expressed in equation (6) for each year from 2004 to 2008.

[28] We additionally assessed annual variability in the average daily accumulated GPP (GPPd), TER (TERd), and NEE (NEEd) for each of the six plant growth stages in both forests (Figures 7 and 8). In the young-growth forest, GPPd was smaller than TERd in stages A, E, and F during the study period (Figure 6). The resulting NEEd was positive. In stages B, C, and D, GPPd was larger than TERd for all years, and therefore NEEd showed negative values. In stage C, GPPd after 2006 was larger than GPPd in 2004, while TERd was lower in 2007 and 2008 than in 2004. Values of NEEd in 2007 and 2008 were about 1.6 gC m−2 d−1 lower than in 2004. Hence, we found a remarkable annual difference in NEEd in the growth stages before and after the typhoon. The smallest GPPd and largest TERd in 2004 were recorded in stage D (P ≤ 0.10 for GPPd and P ≤ 0.02 for TERd). Accordingly, NEEd in 2004 was positive (P ≤ 0.01). After 2004, GPPd was always larger than TERd in stage D. NEEd maintained a consistent average value of −2.1 gC m−2 d−1 after 2004.

Figure 7.

Daily accumulated gross primary production (GPPd) (gC m−2 d−1), terrestrial ecosystem respiration (TERd) (gC m−2 d−1), and net ecosystem exchange (NEEd) (gC m−2 d−1) for each of the six growth stages during 20042008 in the young-growth forest (YGF). Data used here indicate the daily averages of GPPd, TERd, and NEEd calculated using u* threshold values (u*c) of 0.25 and 0.55 m s−1. Error bars represent the deviation from the average.

Figure 8.

Daily accumulated gross primary production (GPPd) (gC m−2 d−1), terrestrial ecosystem respiration (TERd) (gC m−2 d−1), and net ecosystem exchange (NEEd) (gC m−2 d−1) for each of the six growth stages during 2004–2008 in the old-growth forest (OGF). The notation is the same as in Figure 7.

[29] Meanwhile, GPPd was larger than TERd in all stages in the old-growth forest (Figure 8). NEEd was entirely negative except in stage A. In stages B and C, GPPd was larger in 2005 than in 2004. Values of GPPd in these stages were also always larger than those of the young-growth forest (Figures 7 and 8). In stage B, NEEd was almost constant although annual variability in GPPd and TERd was detected. Values of GPPd in stage C were slightly higher in 2007 and 2008 than in 2004, but the GPPd increase in the old-growth forest was smaller than that in the young-growth forest. In stage D, GPPd was almost constant except in 2008. The annual variability in GPPd in stages B and D indicated that the 2004 typhoon had little effect on GPPd in the old-growth forest. Accordingly, annual variability in NEEd was regulated by both GPPd and TERd in stages B and D and by GPPd in stage C.

4. Discussion

4.1. Seasonal and Annual Variability in the Carbon Balance

[30] Meteorological variables such as Sd, Ta, and q had quite similar seasonal variability during the study period (Figure 2). In the young-growth forest, we found an earlier timing of Wm increase in autumn and larger Wm at the end of 2004 compared to other years (Figure 2e). This result suggests smaller water uptake by roots in plants affected by defoliation damage caused by the 2004 typhoon. Furthermore, with fewer leaves remaining, trees may have reduced their photosynthetic activity. Wm was at its lowest levels in 2007 and 2008 in both forests (Figure 2e). However, these reductions had little effect on GPP (Figures 7 and 8), indicating the low impact of water stress on tree photosynthesis. Accordingly, our results showed only small effects of annual variability in meteorological variables on GPP and thus NEE for both forests.

4.2. Posttyphoon Annual Carbon Balance in Stages B and D

[31] We found not only small annual carbon balance components but also small annual variability in these components in stage B in the young-growth forest (Figure 7). During the study period, the timing of leaf opening was almost constant among years (see section 3.2). This suggested that small carbon balance components, especially GPPd in stage B, were regulated not by overstory tree species but by understory vegetation. In stage D, we also found smaller carbon balance components compared to those in stage C and small annual variability in carbon balance components except in 2004 (Figure 7). The typhoon in 2004 caused a remarkable decrease in GPPd and increase in TERd. The amount of fresh leaf litter fell to the ground due to the effect of the typhoon might have accelerated litter decomposition by microbial activity during early September, when temperatures were still high. In contrast, the amount of leaf litter in 2006 was the same as in 2004, but TERd was smaller in 2006 than in 2004. Without the strong wind disturbance of the typhoon, most leaves fell in October, when air temperatures around the stand are lower. Thus, the difference in temperature conditions between September and October might have yielded the annual difference in stage D TERd (Figure 7 and Table 2).

Table 2. The Percentage of Total Annual GPP, Total Annual TER, and Total Annual NEE in Each Plant Growth Stage in the Young-Growth (YGF) and Old-Growth (OGF) Forestsa
YearYGF StagesOGF Stages
  • a

    Dash indicates that the conditions for that stage (see Table 1) were not satisfied or that NEE in the plant growth stage was positive.


[32] On the other hand, each of the carbon balance components had a relatively large value in stage B in the old-growth forest (Figure 8 and Table 2). During stage B, deciduous trees had not yet opened their leaves. Thus, these components might have been largely controlled by evergreen species because of the low impact of dwarf bamboo on carbon balance components as shown in the corresponding stage in the young-growth forest (Figures 7 and 8). In stage D, annual values of GPPd and TERd were second largest among the growing stages (Table 2). The large values of both GPPd and TERd resulted in NEEd values similar to that of the young-growth forest (Figures 7 and 8). Thus, together, stages B and D in the old-growth forest represented large parts of the annual GPP, TER, and NEE (31.1%, 29.0%, and 33.3%, respectively) compared to their contribution in the young-growth forest (Figure 8). In contrast, annual variability in GPPd was small except in 2008. This result indicates that the small structural damage to LAI caused by the typhoon did not affect annual variability in GPP in stage D (Figures 5 and 8), but led to small annual variability in components in the old-growth forest.

4.3. Posttyphoon Annual Carbon Balance in Stage C

4.3.1. Young-Growth Forest Stand

[33] In the young-growth forest, stand variables such as dbh and Bw showed annual increases (Figures 4b and 4c), while others such as ρ showed annual decreases with forest stand development (Figure 4a). The gradual decrease in ρ in the young-growth forest means that the 2004 typhoon did not have a large impact on ρ. In contrast to annual variations in ρ, PAI decreased drastically in the young-growth forest in 2004 (Figure 5); in particular, large amounts of branches fell and broke during the typhoon. Generally, foliage appears to recover rapidly from typhoon damage if the damage is limited to defoliation with negligible stem damage [Li et al., 2007]; stem and branch recovery, however, can take several years to decades [Merrens and Peart, 1992]. The period from mid-August to September every year is important for deciduous trees. During this time, leaf primordia form on branches to act as photosynthetic organs in the following year [Kozlowski and Pallardy, 1997]. In this study, the typhoon caused intensive structural damage to branches in the young-growth forest, although the amount of branch litter was smaller than the amount of leaf litter. The branch damage in 2004 slowed foliage recovery and led to large reductions in PAI in 2005 (Figure 5a) compared to pretyphoon PAI in 2004; this thus contributed to a reduction in GPP and an increase in NEE in 2005 (Figure 7).

[34] Foliage recovery in the young-growth forest was found after 2006. In addition, average PAI showed a slight increase of at most 0.1 m2 m−2 before and after the typhoon (i.e., the averaged PAI in stage C was 2.73 m2 m−2 in 2004, 1.97 m2 m−2 in 2005, 2.67 m2 m−2 in 2006, 2.83 m2 m−2 in 2007, and 2.60 m2 m−2 in 2008) in the early growth stage (i.e., stage C) (Figure 5a). In contrast, the average value of GPPd increased by 11.8% from 2004 to 2008 (i.e., this percentage was obtained using a GPPd of 8.49 gC m−2 d−1 in 2004 and 9.50 gC m−2 d−1 in 2008 during stage C) (Figure 7). Annual variability in GPPd and PAI in stage C revealed that the ratio of GPPd to PAI was clearly larger in 2008 than in 2004 (i.e., GPPd/PAI values in 2004 and 2008 were 3.11 and 3.65 gC m−2 d−1, respectively). Furthermore, GPPd showed higher sensitivity to PAR in 2008 compared to 2004 in stage C (Figure 6a). Therefore, annual variability in GPPd in stage C both before and after the typhoon was little affected by structural changes in foliage or climate conditions, suggesting that leaf physiological responses recovered quickly from damage, leading to high GPPd. Furthermore, TERd values were smaller in 2007 and 2008 than in 2004 (Figure 7). Previous studies reported reduced soil respiration following hurricane disturbance [Steudler et al., 1991; Li et al., 2007]. Our findings agree with their results, even though these earlier studies focused on different species and climatic conditions from those of our study. The reduced TERd in 2007 and 2008 compared to 2004 in stage C might have been due to reduced precipitation and thus soil water content, which may have reduced microbial activities. Accordingly, GPPd had a large impact on NEEd compared to TERd, indicating that foliage photosynthetic recovery resulted in a large annual difference in NEEd in stage C (Figure 7).

4.3.2. Old-Growth Forest Stand

[35] Contrary to the seasonal trend in PAI with large typhoon-induced foliage damage in the young-growth forest, the seasonal PAI trend showed little foliage damage in the old-growth forest in stage C during the study period (Figure 5b). Therefore, the average PAI in stage C was almost the same in all study years (3.37 m2 m−2 in 2004, 3.11 m2 m−2 in 2005, 3.26 m2 m−2 in 2006, 2.98 m2 m−2 in 2007, and 2.99 m2 m−2 in 2008, estimated using PAI measured during DOY 210–240). We also found small variability in pretyphoon and posttyphoon GPPd compared to the variability in the young-growth forest (Figure 8), indicating that the typhoon had little impact on GPPd in the old-growth forest. However, GPPd values in 2007 and 2008 were larger than those in 2004. Therefore, GPPd/PAI in stage C also increased in 2007 and 2008 compared to that in 2004 (GPPd/PAI was 2.79 gC m−2 d−1 in 2004, 3.40 gC m−2 d−1 in 2007, and 3.41 gC m−2 d−1 in 2008). High sensitivity of GPPd to PAR was found in 2008 compared to that in 2004 in stage C in the old-growth forest as well as in the GPPd-PAR relationship in the young-growth forest (Figure 6c). These results suggest that photosynthetic recovery due to physiological responses of tree leaves during stage C enhanced GPPd not only in the young-growth forest but also in the old-growth forest.

4.4. Uncertainty of the EC-Derived NEE

[36] The EC-derived NEE was partitioned into GPP and TER using a gap-filling approach with two u*c threshold values to include the effect of difference in u*c values on GPP, TER and NEE. We found only a slight difference between GPP values when u*c = 0.25 m s−1 and 0.55 m s−1 in the young-growth forest and the old-growth forest, respectively. In contrast, TER differences were somewhat larger than GPP differences under these u*c conditions in the young-growth forest. The u* filtering was applied to fluxes measured under nonturbulent conditions, which mainly occur at night. Therefore, TER under nonturbulent conditions was affected by the selected u*c value. A slight difference was also observed between NEEs calculated based on the two u*c conditions, reflecting the strong regulation of NEE by GPP. The average maximum discrepancy values associated with the u*c values for the young-growth and old-growth forests in the stage C were 5.06 gCm−2 d−1 and 5.43 gCm−2 d−1 for GPPd, 2.72 gCm−2 d−1 and 3.23 gCm−2 d−1 for TERd, and 2.91 gCm−2 d−1 and 2.46 gCm−2 d−1 for NEEd, respectively (Figures 7 and 8). The maximum discrepancy of GPPd, TERd and NEEd caused by the u*c values was smaller in stage D than in stage C in both forests; 2.78 gCm−2 d−1 and 3.43 gCm−2 d−1 for GPPd, 2.36 gCm−2 d−1 and 2.10 gCm−2 d−1 for TERd, and 0.45 gCm−2 d−1 and 1.42 gCm−2 d−1 for NEEd in the young-growth and old-growth forests, respectively (Figures 7 and 8). The results indicate the effect of the selected value of u*c when evaluating GPPd, TERd and NEEd in a plant growth stage. In contrast, each forest showed similar trends each year in the discrepancy of the carbon balance components in both stages (i.e., C and D). Thus, because of this consistent discrepancy in the carbon balance components, we believe that one can evaluate the effect of typhoon damage on carbon balance.

5. Summary and Conclusions

[37] Eddy covariance measurements revealed that photosynthetic recovery after a typhoon activated carbon uptake by forests. These are believed to be the first quantitative findings to demonstrate that changes in physiological tree responses after wind disturbance have a large potential to alter ecosystem CO2 exchange. Several reports have presented carbon annual NEP (i.e., equals -NEE in the present study) for representative terrestrial ecosystems across tropical to boreal climate regions worldwide [Schlesinger, 1997; Bonan, 2008]. The annual NEE estimate for our study area presented by Bonan [2008] was consistent with our results when wind disturbance effects on annual NEE were negligible. In contrast, the annual NEE estimates that reflected annual variability in plant photosynthetic activity and terrestrial ecosystem respiration after the typhoon damage were beyond the range of errors of annual NEE that were estimated by Bonan [2008]. This suggests the importance of considering wind disturbance effects on annual NEE when predicting annual variability in NEE in terrestrial ecosystems. Amiro et al. [2010] investigated terrestrial CO2 fluxes after various disturbances in numerous North American forests and found that most forests shifted from a carbon source to a carbon sink during the first 20 years after the disturbance. Little, however, is known about how long the changes in the ecophysiological responses of trees to climate conditions will enhance or reduce NEE following large wind disturbance-induced foliage damage. Most recently developed global vegetation dynamics models have included disturbance effects. Plant ecophysiological feedback to the climate following disturbance, however, remains to be resolved. Therefore, further long-term EC measurements are needed for use in process-based modeling to predict ecosystem CO2 exchange after a typhoon that produces heavy branch damage and defoliation. Advances in modeling may provide some answers to help eliminate the large discrepancy in simulated global carbon uptake for terrestrial ecosystems [Knohl et al., 2002; Friedlingstein et al., 2006; Purves and Pacala, 2008; Toda et al., 2009].


[38] We wish to thank Kentaro Takagi and Thomas Powell for their helpful comments and Takeshi Ohta for providing us with the forest meteorological data used in this article, and the data are available at (∼wecnof/jp/index.html). We would like to thank the members of Environmental Biology Division and the staff of Uryu Experimental Forest, Field Science Center for Northern Biosphere, Hokkaido University, for their support in the field study. The Uryu Experimental Forest is operated as part of northern observational field core sites in Japan Long-term Ecological Research (JaLTER) network ( = en), Japan Flux (JapanFlux) network ( Partial financial support was given by the 21st Century Center of Excellence (COE) Program (E-01) (principal investigator, M. Ikeda), Global COE program (principal investigator, Y. Yamanaka) funded by the Ministry of Education, Culture, Sports, Science and Technology. This study was supported partly by the Grant for Joint Research Program of the Institute of Low Temperature Science, Hokkaido University (principal investigator, T. Hara). This research was also supported and financed, in part, by Grant-in-Aid for Young Scientists (B) Scientific Research to one of the authors (M.T.).