Snow Depth & Soil Freezing as a Regulator of Microbial Processes

In the northern hardwood forests in New Hampshire we are analyzing how soil freezing events cause root and microbial mortality, which can lead to increased rates of N and P mineralization and loss.

Project Summary

In this Project

Project Summary

Plots and treatment

We established eight new 10 m x 10 m plots (4 treatment, 4 reference) to provide a range of freezing intensity. Plots were located in two high elevation, north facing and two low elevation, south facing maple-beech-birch stands at HBEF. This design provided significant variation in soil freezing and a strong basis for extrapolating from our experimental plots to larger scales.

To establish these new plots, we cleared minor amounts of understory vegetation from all plots (to facilitate shoveling) and installed soil and snow thermistors, frost tubes, time domain reflectometry probes (for soil moisture), trace gas flux chambers, lysimeters and minirhizotron root observation tubes. We also set up and conducted a level survey (9 points per plot) in each of the treatment plots prior to soil freezing to monitor frost heave.

Our snow manipulation treatment was carried out in the winters of 2002/2003 and 2003/2004. During the winter (November through April), weather and soil condition data were collected at intervals ranging from every 30-minutes (weather data) to every 2 hours (soil moisture). Treatment plots were kept snow free, or close to snow-free by shoveling, until late January when snow was allowed to accumulate on the plots. Manual measurements of snow depth, snow water equivalence, and frost depth (from frost tubes) were obtained approximately biweekly.

At the approximate time of maximum frost extent, we again conducted a level survey to measure soil frost heave in the treated plots (2003) and in treated and reference plots (2004).

The winters of 2002/2003 and 2003/2004 were characterized by below normal temperatures. The shoveling treatment was very effective in inducing soil freezing in all treated plots, in both years (Figure 1). Interestingly, we observed more natural soil frost (the reference plots) at our lower elevation (lower valley and upper valley) locations due to early winter storms that brought insulating snow to the higher elevation plots and not to the lower elevation plots. Thus our experimental design, which takes advantage of the natural climate gradient in the Hubbard Brook valley, provided an additional test of our main hypothesis that there will be "colder soils in a warmer world."

Figure 1. Frost depths (as measured in frost tubes) in low (UV, LV) and high (EK, WK) elevation plots on February 22, 2003 and February 27, 2004. Plots were kept snow free until the end of January.

Frost heave measurements (Figure 2) show that the physical disturbance caused by soil freezing is significant and variable. This variability is controlled by soil moisture and the presence/absence of a perched water table, i.e. higher water availability leads to more frost heaving.

Figure 2. Relative frost heave measured at the maximum frost extent in low (UV, LV) and high (EK, WK) elevation plots, February, 2004.


In 2003, we collected two soil cores from the treatment plots (with a diamond-coated drill used for sampling concrete) at the time of maximum frost extent, and one soil core from the reference plots. These cores were transported back to the cold rooms at the U.S. Army Cold Regions Research and Engineering Laboratory (CRREL) and examined for ice lensing and root damage. Root damage was assessed visually and by a comparison of relative electrolyte leakage from the roots. In April of 2003 and 2004, when soil temperatures were between 5 - 10 oC, we removed two soil blocks and a set of soil cores to compare root vitality by TTC staining (tetrazolium triphenyl chloride reduction to formazan), by relative electrolyte leakage and by visual assessment. In May, we transplanted one sugar maple and one red spruce sapling into each plot. These saplings were used for whole root system assessments at the end of the experiment. Roots were also observed with minirhiztorons, with monthly filming during the growing season.

In November 2003, a complementary root damage experiment was established for sugar maple (adjacent to one of the low elevation sites) and red spruce (adjacent to one of the high elevation sites). In this experiment we manipulated the type of damage (none, abrasion, stretching) and the snow cover (for same period as Freeze plots). These roots were assessed with TTC and sugar maple roots were subsampled for analysis of differences in mycorrhizal infection. In January 2004, sugar maple and red spruce roots that had undergone winter hardening were obtained when the frost heave surveys were redone. Using roots obtained in January, a material testing protocol was developed and several tests were completed on the effects of freezing only versus freezing and stretching on root structural integrity.

The freeze treatment significantly affected the vitality of fine tree roots (F1, 174 = 12.36, p = 0.001) and there were significant differences in root vitality between the organic and mineral horizons (F1, 174 = 19.32, p < 0.001). First and second order roots in the organic horizon were the group most damaged by the treatment (Figure 3; treatment by soil horizon interaction: F1, 1 = 8.12, p = 0.005; treatment by root order interaction: F1, 1 = 6.28, p = 0.013). Our hypothesis that frost heave could be severing entire root systems at the level of higher order roots was not supported by the spring 2003 and 2004 TTC results. Additional evidence against this possibility came from root tensile tests that found no significant difference (t = 0.878, p = 0.43) in tensile strength of sugar maple roots that had been frozen (mean maximum load sustained 4.913 ± 1.697 lbf) versus those that had been frozen and stretched while frozen (4.521 ± 1.307 lbf). These material tester results designed to mimic formation of an ice lens directly beneath a root suggest that no significant structural damage was incurred by stretching (2 mm displacement) of the frozen roots.

Figure 3. Comparison of root vitality in spring by TTC vitality test in control and shoveled plots in organic and mineral horizons for 1st and 2nd order (closed circles) and 3rd and higher order (open circles) fine tree roots. Higher absorbance (520 nm) relates to greater reduction of TTC to formazan and represents respiration potential of the cells. Data from 2003 and 2004 were pooled here as there was no significant difference between years.

Soil microbial biomass and activity

Microbial biomass C and N content were measured using the chloroform fumigation-incubation method. In this method, sediments are fumigated to kill and lyse microbial cells in the sample. The fumigated samples are inoculated with fresh sediment, and microorganisms from the fresh sediment grow vigorously using the killed cells as substrate. The flushes of carbon dioxide (CO2) and 2 M KCl extractable inorganic N (NH4+ and NO3-) released by the actively growing cells during a 10 day incubation at field moisture content are assumed to be directly proportional to the amount of C and N in the microbial biomass of the original sample. A proportionality constant (0.45) is used to calculate biomass C from the CO2 flush. Inorganic N flush data are not corrected with a proportionality constant.

We also measured inorganic N and CO2 production in unfumigated "control" samples. These incubations provided estimates of soil microbial respiration and potential net N mineralization and nitrification. Respiration was quantified from the amount of CO2 evolved over the 10 day incubation. Potential net N mineralization and nitrification were quantified from the accumulation of NH4+ plus NO3- and NO3- alone during the 10 day incubation.

Denitrification potential was measured using the denitrification enzyme activity (DEA) assay developed by Smith and Tiedje (1979) as described by Groffman et al. (1999). In this assay, sieved sediments are amended with NO3-, dextrose, chloramphenicol and acetylene, and incubated under anaerobic conditions for 90 minutes. Gas samples are taken at 30 and 90 minutes, stored in evacuated glass tubes and analyzed for N2O by electron capture gas chromatography.

Net N mineralization and nitrification were also measured using an in situ intact core method. During the growing season, 10, 2-cm diameter intact cores were removed from each plot each month. Five of the cores were returned to the laboratory for extraction (2 N KCl) of inorganic N (NH4+ and NO3-) and five were returned to the plot for in situ incubation. Cores were incubated for approximately 4 weeks before harvesting and extraction. Overwinter rates were estimated by incubating cores from December through April. Inorganic N was quantified colorometrically using a PerstorpTM 3000 or Lachat Quikchem 8100 flow injection analyzer. Net N mineralization rates were calculated as the accumulation of total inorganic N over the course of the incubation. Net nitrification rates were calculated as the accumulation of NO3- over the course of the incubation.

Soil:atmosphere fluxes of carbon dioxide (CO2), nitrous oxide (N2O) and methane (CH4) were measured using an in situ chamber design. Chambers (four per plot) consisted of 287-mm diameter (ID) by 40-mm high polyvinyl chloride (PVC) cylinders which were placed on permanently installed PVC base rings immediately prior to measurement. At sampling intervals of approximately 0, 10, 20 and 30 min following placement of the chamber on the base, 9-mL gas samples were collected from gas sampling ports in the center of the chamber top using fine-needle polypropylene syringes. Samples were transferred to evacuated glass vials which were stored at room temperature prior to analysis by gas chromatography with electron capture (N2O), thermal conductivity (CO2) or flame ionization (CH4) detection. Fluxes were calculated from the linear rate of change in gas concentration, the chamber internal volume and soil surface area.

As in our previous studies, soil freezing did not affect microbial biomass and activity (Table 1). Of particular interest is the lack of response of nitrification (Table 1, Figure 7), a key driver of hydrologic N losses. These results support conclusions from our previous study that N losses associated with soil freezing are associated with root damage and a reduction in plant uptake. Rates of nitrification were low in winter relative to during the growing season, but were significantly greater than zero. Nitrification rates were higher at the high elevation sites than the low elevation sites, likely due to higher soil moisture at the high elevation sites.

Table 1. Microbial biomass C and N in the forest floor of treatment and control stands before (2002) and after (2003 and 2004) treatment. Values are the mean (with standard error) of four treatment and four reference plots sampled in Spring 2002, Winter, Spring and Summer 2003 and Spring and Summer 2004.

 2002 (pre-treatment)2003/2004 (post-treatment)
Microbial biomass C
(mg C kg-1)
2093 (77)2359 (74)2976 (244)3220 (265)
Microbial biomass N
(mg N kg-1)
401 (19)444 (26)391 (34)414 (26)
Potential net N mineralization
(mg N kg-1 d-1)
8.2 (1.9)7.6 (1.0)8.2 (0.9)9.0 (0.9)
Potential net nitrification
(mg N kg-1 d-1)
3.5 (0.8)2.1 (0.6)3.2 (0.4)2.6 (0.3)
Soil nitrate pool
(mg N kg-1)
6.3 (1.6)4.4 (1.6)6.4 (0.9)4.6 (0.9)
Soil ammonium pool
(mg N kg-1)
45 (6)64 (8)53 (6)51 (5)

Figure7. Potential net nitrification in low (UV, LV) and high (EK, WK) elevation plots from summer 2002 (pre-treatment) through summer 2004.

Analysis of soil solution samples collected in lysimeters show increased leaching losses of N in the treatment plots in some locations but not in others (Figure 8). The variation in response was not related to the intensity of freeze disturbance (frost depth, frost heave) on the plots. Plots dominated by sugar maple (UV, WK, in Figure 8) appear to be most susceptible to freeze disturbance, a result that is consistent with our previous studies, and with our root damage data.

Figure 8. Mean nitrate concentrations in lysimeters beneath the Oa horizon in low (UV, LV) and high (EK, WK) elevation plots from winter 2002/2003 through summer 2004

While there were no strong treatment effects on soil:atmosphere fluxes of CO2, N2O and CH, we did observe marked variation in fluxes with landscape position. We consistently observed higher CO2 production and higher CH4 uptake in the warmer, drier valley locations (LV, UV) than in the colder, wetter Kineo locations (WK, EK) (Figure 9). Interestingly, the average temperature difference between these locations (approximately 2 o C) is similar to what is predicted to occur with global warming. These results suggest that soil respiration and methane uptake will likely both be higher in a future, warmer world, yielding both negative and positive warming feedback effects.

Figure 9. Soil:atmosphere fluxes of carbon dioxide (CO2), nitrous oxide (N2O) and methane CH4) in two low elevation, south facing locations (UV, LV) and two high elevation, north facing locations (WK, EK) at the Hubbard Brook Experimental Forest from 2001 - 2003. These are data from reference plots only. Values are mean (standard error) of four chambers per plot.

Soil freezing and maple leaf litter decomposition

To test the physical effects of freezing on the mineralization of litter, we produced 15N labeled sugar maple litter. A 10 cm dbh sugar maple tree was injected with a 15N solution in May 2002. The senescent leaf litter was collected, air dried and a sub-sample was analyzed for 15N content. The average atom% of five sub-samples was 0.491, well above background levels of 0.366 atom%. This litter was placed in all eight field plots in October 2002 and 2003 . The litter was contained within open topped, 0.25 m2 screened 'corrals' covered with 0.60 cm mesh to keep litter in the plots and exclude additional native litter from entering the plots. Litter that was present in the corrals was removed and 50 g of labeled litter was added to each corral. These plots were sampled in spring (after snowmelt) and summer of 2003 and 2004.

To determine the fate of the 15N in the sugar maple litter, the available, microbial, mineralizable and total soil N pools were measured. To determine potential availability of N in litter to herbaceous plants under different freeze regimes, green leaf tissue of plants growing in the litter corals was collected in August 2003 and 2004. The green leaf tissue was dried at 60oC, ground with a KLECO pulverizer to a fine powder and sent to the stable isotope laboratory at UC Davis for 15N analysis.

Results suggest that soil freezing slows decomposition (Figure 4) but accelerates the movement of N from fresh litter into soil microbial biomass. Treatment plots had higher (p < 0.05) atom% 15N in the microbial biomass in both organic and mineral soils (Figure 5). Total soil N pools of the treatment plot mineral soil also had higher (p < 0.05) amounts of 15N (0.3684 atom%) than control plots (0.3655 atom%).

Figure 4. Fraction of mass remaining in treatment and reference plots treated with 15 N labeled litter after two years of decomposition, from 2002 to 2004. Litter was applied in fall 2002 and microbial biomass was sampled in May and August 2003 and 2004.

Figure 5. 15 N atom % in microbial biomass in treatment and reference plots treated with 15 N labeled litter. Litter was applied in fall 2002 and microbial biomass was sampled in May and August 2003 and 2004. Microbial biomass in mineral soil was significantly enriched in treatment relative to reference plots.

Freeze effects on soil structure

We evaluated the effects of freezing on macroaggregation and on free- versus protected POM using a sequential wet-sieving and flotation procedure. Soil cores were collected in late spring after thaw, separated into mineral soil and forest floor horizons, and air-dried prior to analysis. Samples were separated into 4 size classes using a wet sieving procedure. Prior to sieving, each sample was placed on a piece of wet filter paper to wet up by capillary action. Slaking, or rapid wetting of dry soil particles, exerts enough force to disrupt some macroaggregates (aggregates > 250 µm). We chose capillary wetting rather than slaking to avoid disrupting those macroaggregates that were not water-stable. Therefore we examined freeze effects on the total pool of water-stable and unstable macroaggregates. Following wetting, the sample was placed on a 2000µm sieve in a basin of water and gently moved up and down in a rotating motion, 50 times in 2 minutes. The soil remaining on the sieve (>2000µm fraction) was dried for 48 hrs at 100oC and the <2000µm fraction was transferred to a 250µm sieve. The 250-2000µm and 53-250µm fractions were separated, dried, and weighed following the same procedure. The <53µm fraction remained in water following the final sieving and was dried to constant mass at 100oC and weighed.

The organic matter concentration of each size fraction was determined by loss-on-ignition (450oC for 3 hours). Potential mineralization of N and C were quantified in laboratory incubations of each size fraction. One g subsamples from each size fraction were mixed with an equal mass of a uniform, field-moist, mineral soil mixture to provide a source of moisture and microbial inoculum. Two subsamples were pre-incubated for 24 h at 20 oC in 40-mL beakers sealed inside Mason jars. Each jar contained a base trap (20 mL glass vial containing 10 mL 0.01 M NaOH) to remove any CO2 from the jar headspace. Following pre-incubation, one subsample of each pair (initial) was extracted in 15 mL 2M KCl by shaking for 1 hour. Extracts were filter-sterilized through 0.2 µm filter membranes and NH4+ and NO3- concentrations were quantified using a Perstorp flow-injection analyzer. Base traps were replaced at 48 hour intervals for the remaining (final) subsamples for 12 days, after which these subsamples were subjected to the same KCl extraction and NH4+ and NO3- analysis as initial samples. Base traps were titrated with 0.01 M HCl in the presence of 2 M BaCl2 to quantify CO2 evolution from soils. One-g samples of the mineral soil inoculum mixture were incubated independently to quantify their contributions in the mixtures, which varied from 1 - 10% of the total for net N mineralization and from 30-60% of the total for C mineralization.

Light and heavy fractions were separated from the 250-2000µm size class of soils collected in the second year of the study using a flotation method. We defined the light fraction as material that floats in 1.85 g/cm3 sodium metatungstate, and the heavy fraction was defined as material that sinks in 1.85 g/cm3 sodium metatugstate. Light fraction in this density class consists of free particulate organic matter (POM), and some mineral-associated POM. Two replicate 2.5g subsamples of each 250-2000µm fraction were placed in conical centrifuge tubes with 20mL of 1.85 g/cm3 sodium metatungstate. Additional sodium metatungstate was added for a total volume of 40mL. Tubes were placed uncapped in a vacuum at 18.3 Hg for 10 minutes, left for 12-15 hrs, and centrifuged for 1 hr at 2000 rpm. Following centrifugation, the light fraction (defined here as any particles floating in the surface 30 mL) was vacuumed off, rinsed, dried at 100oC for 24hrs, and weighed. The heavy fraction, (material remaining in the bottom 10 mL of each tube) was filtered, rinsed, dried at 100oC for 24 hrs, and weighed.

In the first year of sampling, the only significant effect of soil freezing on particle size distribution was an increase in the smallest (<53µm) size class. In the second year, freezing increased organic matter content of all size fractions, except for the largest (>2000µm). This change corresponded to an increase in the light fraction and an increase in both net N mineralization potential and the ratio of N mineralized:C mineralized of the 250-2000 m fraction (Figure 6). However, we could not detect effects of freezing on C mineralization from any size fraction. These results suggest that freezing fragments root and leaf litter, increasing particulate organic matter (POM), and that low N immobilization potential of this POM contributes to inorganic N production following freeze events.

Figure 6. Net N mineralization potentials among size fractions expressed per g whole soil, following freezing treatment in (A) 2003 and (B) 2004. Significant differences between control and freeze treatment are indicated by stars (*p<0.05, **p<0.001, ***p<0.0001). Error bars are standard error of the mean, n=12 in 2003 and n=20 in 2004.


A one dimensional energy balance snow-soil process model, SLTHERM, was adapted to depict spatially explicit snowpack and soil freezing dynamics for the 3,160 ha HBEF valley. The valley ranges in elevation from 180 to 1000m and snowpack variation is due to solar exposure variations related to hilly terrain, a variable forest canopy and soil characteristics. The model is driven by locally collected temperature and precipitation data modified by elevation (3 classes), forest cover type (5 classes), aspect (24 classes) and slope (12 classes) factors. A total of 56 total classes were present in the HBEF valley. The model depicts development and melting of the snowpack and temperatures at the snow-soil interface.

ECC activities. During 2004 two workshops were held at the Institute of Ecosystem Studies (IES) with funding coming from the supplement to this grant and from the new collaborative grant that funds the ECC (DEB-0404801). Both centered around the collaborative development of a model on mercury fluxes and transformations. These workshops brought together the student participants in the ECC with mercury expert Steven Lindberg (Oak Ridge National Lab) and atmospheric deposition specialist Kathleen Weathers (IES). Together the group worked on the structure and coding of the model, and the mining of literature for parameterization values.

The SLTERM model was clearly capable of depicting snow depth and soil temperatures at our sites (Figure 10). More importantly, it is capable of providing spatially-explicit depictions of how snow depth and soil temperatures change in space and time across the Hubbard Brook valley (Figure 11). Of particular interest is the spatial variation at the end of the season; as snowmelt begins, there is marked variation in snow cover and soil temperature across the valley. Analysis of these spatial and temporal patterns points to particular locations in the landscape that have likely naturally been subjected to significantly more soil freezing over time. These locations could serve as a basis for future "natural experiments" to evaluate the long-term effects of increased freeze frequency (as is likely to occur with climate change) on soil and vegetation properties.

Figure 10. Modeled and measured snow depth (top) and soil temperature (bottom) at the lower valley experimental site.


Day 30 - Snow depth
Day 30 - Soil temperature
Day 130 - Snow depth

Day 130 - Soil temperature

Figure 11. Simulated snow depth (left) and soil temperature (right) at Day 30 (top) and Day 130 of 2003.

Project publications
  • Campbell, J.L., M.J. Mitchell, P.M. Groffman and L.M. Christenson. 2005. Winter in northeastern North America: An often overlooked but critical period for ecological processes. Frontiers in Ecology and Environment 3:314-322.
  • Campbell, J.L., M. J. Mitchell, B. Mayer, P.M. Groffman and L.M. Christenson. 2007. Fate of 15N-labeled nitrate and 34S-labeled sulfate applied to snow at the Hubbard Brook Experimental Forest, New Hampshire. Soil Science Society of America Journal 71:1934-1944.
  • Cleavitt, N.L., T. J. Fahey P.M. Groffman, J.P. Hardy, K.S. Henry, and C.T. Driscoll. 2008. Effects of soil freezing on fine roots in a northern hardwood forest. Canadian Journal of Forest Research 38:82-91.
  • Fitzhugh, R.D., C.T. Driscoll, P.M. Groffman, G.L. Tierney, T.J. Fahey and J.P. Hardy. 2001. Effects of soil freezing disturbance on soil solution nitrogen, phosphorus, and carbon chemistry in a northern hardwood ecosystem. Biogeochemistry 56:215-238.
  • Fitzhugh, R.D., C.T. Driscoll, P.M. Groffman, G.L. Tierney, T.J. Fahey and J.P. Hardy. 2003. Soil freezing and the acid-base chemistry of soil solutions in a northern hardwood forest. Soil Science Society of America Journal 67:1897-1908.
  • Fitzhugh, R.D., G.E. Likens, C.T. Driscoll, P.M. Groffman, T.J. Fahey, J.P. Hardy and M.J. Mitchell. 2003. Role of soil freezing events in interannual patterns of stream chemistry at the Hubbard Brook Experimental Forest, New Hampshire. Environmental Science & Technology 37:1575-1580.
  • Groffman, P.M., J. P. Hardy, S. Nolan, C.T. Driscoll and T.J. Fahey. 1999. Snow depth, soil frost and nutrient loss in a northern hardwood forest. Hydrological Processes 13:2275-2286.
  • Groffman, P.M., C.T. Driscoll, T.J. Fahey, J.P. Hardy, R.D. Fitzhugh and G.L. Tierney. 2001. Colder soils in a warmer world: A snow manipulation study in a northern hardwood forest ecosystem. Biogeochemistry 56:135-150.
  • Groffman, P.M., C.T. Driscoll, T.J. Fahey, J.P. Hardy, R.D. Fitzhugh and G.L. Tierney. 2001. Effects of mild winter freezing on soil nitrogen and carbon dynamics in a northern hardwood forest. Biogeochemistry 56:191-213.
  • Groffman, P.M., C.T. Driscoll, G.E. Likens, T.J. Fahey, R.T. Holmes, C. Eagar and J.D. Aber. 2004. Nor gloom of night: A new conceptual model for the Hubbard Brook ecosystem study. BioScience 54:139-148.
  • Groffman, P.M., J.P Hardy, C.T. Drisoll and T.J. Fahey. 2006. Snow depth, soil freezing and trace gas fluxes in a northern hardwood forest. Global Change Biology 12:1748-1760.
  • Hardy, J.P., R. Jordan, P. Groffman, S. Nolan, C. Driscoll and T. Fahey. 1999. Snow depth and soil frost modeling in a northern hardwood forest. 55th Eastern Snow Conference Proceedings, Fredericton, New Brunswick., p. 211-214.
  • Hardy, J.P., P.M. Groffman, R.D. Fitzhugh, K.S. Henry, T.A. Welman, J.D. Demers, T.J. Fahey, C.T. Driscoll, G.L. Tierney and S. Nolan. 2001. Snow depth, soil frost and water dynamics in a northern hardwood forest. Biogeochemistry 56:151-174.
  • Judd, K.E., G.E. Likens and P.M. Groffman. 2007. High nitrate retention during winter in soils of the Hubbard Brook Experimental Forest. Ecosystems 10:217-225.
  • Nielsen, C.B., P.M. Groffman, S.P. Hamburg, C.T. Driscoll, T.J. Fahey and J.P. Hardy. 2001. Freezing effects on carbon and nitrogen cycling in soils from a northern hardwood forest. Soil Science Society of America Journal 65:1723-1730.
  • Steinweg, J.M., M.C. Fisk, B. McAlexander, P.M. Groffman and J. P. Hardy. Effects of soil freezing on soil particle size distributions and organic matter mineralization in a northern hardwood forest. Biology and Fertility of Soils. In press.
  • Tierney, G.L., T.J. Fahey, P.M. Groffman, J.P. Hardy, R.D. Fitzhugh and C.T. Driscoll. 2001. Soil freezing alters fine root dynamics in a northern hardwood forest. Biogeochemistry 56:175-190.
  • Tierney, G.L., T.J. Fahey, P.M. Groffman, J. P. Hardy, R.D. Fitzhugh, C.T. Driscoll and J. B. Yavitt. 2003. Environmental control of fine root dynamics in a northern hardwood forest. Global Change Biology 9:670-679.

Cary Institute of Ecosystem Studies | Millbrook, New York 12545 | Tel (845) 677-5343

Privacy Policy Copyright © 2014