Decoupling litter respiration from whole-soil ...

432

ARTICLE Decoupling litter respiration from whole-soil respiration along an elevation gradient in a Rocky Mountain mixed-conifer forest Can. J. For. Res. Downloaded from www.nrcresearchpress.com by U.S. Geological Survey on 01/28/15 For personal use only.

Erin M. Berryman, John D. Marshall, and Kathleen Kavanagh

Abstract: Litter respiration (RL) represents a significant portion of whole-soil respiration (RS) in forests, yet climatic correlations with RL have seldom been examined. Because RL is reduced at low humidities and RS is reduced at low temperatures, these components may show divergent trends with elevation in western North American forests. Using a litter-removal experiment along a forested 750 m elevation gradient in the Rocky Mountains of northern Idaho, USA, we measured RS on soils from which litter had been removed (RNL) and, by difference, RL. Mean RL represented 16% (SE = 2%) of mean RS from July through October of 2007 and 2008. RS was highest at warmer times and sites, and was not suppressed by low soil moisture. In contrast, RL was highest at cooler times, when humidity and gravimetric litter water content were highest. RL was highest at mid-elevations, representing neither the warmest nor wettest sites. Sixty-three percent of variability in site RL was explained by both mean annual temperature (MAT) and mean annual relative humidity (MARH), including a positive interaction effect between MAT and MARH. Our results imply that the equilibration of litter with atmospheric humidity is an important control over litter respiration rates. Key words: relative humidity, carbon, carbon dioxide, temperature, moisture. Résumé : La respiration dans la litière (RL) constitue une portion importante de la respiration globale dans le sol (RS) mais les corrélations entre le climat et RL ont rarement été étudiées. Étant donné que RL diminue lorsque la teneur en humidité est faible et que RS diminue lorsque la température est basse, ces composantes pourraient suivre des tendances divergentes avec l’altitude dans les forêts de l’ouest de l’Amérique du Nord. À l’aide d’une expérience de retrait de la litière le long d’un gradient altitudinal de 750 m en forêt, dans les montagnes Rocheuses du nord de l’Idaho aux États-Unis, nous avons mesuré RS, la respiration dans des sols où la litière avait été enlevée (RNL) et, par différence, RL. La valeur moyenne de RL correspondait a` 16% (ET = 2) de la valeur moyenne de RS de juillet a` octobre en 2007 et 2008. La valeur de RS était maximale dans les stations et durant les périodes plus chaudes et n’était pas réduite par une faible teneur en humidité du sol. À l’inverse, la valeur de RL était maximale durant les périodes plus froides lorsque la teneur en humidité et le contenu en eau gravifique de la litière étaient les plus élevés. La valeur de RL était la plus élevée a` une altitude intermédiaire où les stations étaient ni les plus chaudes, ni les plus humides. La température annuelle moyenne et l’humidité relative annuelle moyenne, incluant une interaction positive entre les deux, expliquaient 63% de la variabilité de RL entre les stations. Nos résultats indiquent que l’équilibre entre l’humidité atmosphérique et celle de la litière est un facteur important qui régit le taux de respiration de la litière. [Traduit par la Rédaction] Mots-clés : humidité relative, carbone, dioxyde de carbone, température, humidité.

Introduction Field measurements of forest soil respiration (RS) often support the theory that respiration proceeds faster at higher temperatures (Arrhenius 1889; Hanson et al. 1993; Lloyd and Taylor 1994). Autotrophic root respiration and heterotrophic respiration both increase under warmer temperatures; however, drought accompanying warmer temperatures may decrease both components (Burton et al. 1998; Suseela et al. 2012). In western North America, dry conditions are common during the growing season, and reduced forest respiration rates during drought support the idea that temperature-driven increases in respiration may be counteracted by moisture stress of respiration (Jassal et al. 2008; Ma et al. 2005). Climate change predictions for western North America call for a combination of increased temperature and reduced snow-derived moisture (Mote et al. 2005); if drought counteracts the effect of increasing temperatures on RS, this complicates the prediction of forest carbon (C) cycle responses to climate change.

At least one-quarter to one-third of annual RS derives from heterotrophic respiration in the litter layer (Bowden et al. 1993; Sulzman et al. 2005). Litter respiration (RL) is reduced under low litter water content and at low relative humidity (Donnelly et al. 1990; Kelliher et al. 2004; Kuehn et al. 2004). In mountainous terrain, relative humidity (RH) may increase with elevation, so RL may be reduced in low elevations compared to high elevations. Because RL represents carbon loss from detritus that would otherwise contribute to soil organic matter formation, reductions in this component may increase soil C storage. Thus, it is critical to know if this C loss is reduced by low moisture, counteracting temperature-associated increases. In a mixed-conifer forest in northern Idaho, USA, we first hypothesized that growing season RS and RL will show divergent trends with elevation. We thought this was due to (i) drier low elevation sites in the northern Rocky Mountains in the summer compared to high elevation sites and (ii) increased sensitivity of heterotrophic respiration to drought compared to root respiration (Scott-Denton et al 2006). We further hypothesized that RS

Received 16 August 2013. Accepted 9 January 2014. E.M. Berryman,* J.D. Marshall, and K. Kavanagh. Department of Forest, Rangeland and Fire Sciences, University of Idaho, CNR 204, Moscow, ID 83843, USA. Corresponding author: Erin Berryman (e-mail: [email protected]). *Present address: Department of Forest and Rangeland Stewardship, Colorado State University, 1472 Campus Delivery, Fort Collins, CO 80524, USA. Can. J. For. Res. 44: 432–440 (2014) dx.doi.org/10.1139/cjfr-2013-0334

Published at www.nrcresearchpress.com/cjfr on 13 January 2014.

Berryman et al.

433

Can. J. For. Res. Downloaded from www.nrcresearchpress.com by U.S. Geological Survey on 01/28/15 For personal use only.

Table 1. Attributes of sites used in this study in the Priest River Experimental Forest. Site ID

Elevation (m asl)

BA(m2·ha−1)

MAT (°C)

MARH (%)

Species composition (% total BA)

Litter layer depth [cm (SD)]

4 37 19 7 43 25 31 54 35

854 888 938 1149 1161 1182 1407 1453 1521

31 61 55 83 73 67 67 58 61

6.2 5.9 6.6 5.2 5.3 6.1 3.7 4.1 3.9

53 61 55 62 60 54 66 63 60

16% ABGR, 33% PSME, 50% THPL 39% PSME, 62% THPL 5% ABGR, 10% LAOC, 10% PIPO, 18% PSME, 58% THPL 4% ABGR, 9% LAOC, 54% THPL, 33% TSHE 10% LAOC, 5% PIMO, 52% PSME, 34% THPL 22% ABGR, 20% PICO, 48% PSME, 11% THPL 21% PIEN, 4% PSME, 45% THPL, 30% TSHE 4% ABGR, 22% ABLA, 57% PICO, 8% PSME, 8% TSHE 7% THPL, 93% TSHE

2.69 (1.36) NA 2.3 (0.91) 3.27 (1.32) NA 2.48 (0.99) 2.34 (1.58) NA 2.73 (1.19)

Note: BA = basal area, MAT = mean annual temperature at 2 m above ground level, MARH = mean annual RH measured 2 m above ground level. Litter layer depth is the mean of one value per subplot (n = 3), obtained as the mean of 8 individual measurements within one 1 m2 area. More site attributes can be found in Duursma et al. (2003). Species codes are: ABGR, grand fir (Abies grandis Donn ex D. Don); THPL, western redcedar (Thuja plicata Donn ex D. Don); PSME, Douglas-fir (Pseudotsuga menziesii (Mirb.) var. glauca (Beissn.) Franco); LAOC, western larch (Larix occidentalis Nutt.); TSHE, western hemlock (Tsuga heterophylla (Raf.) Sarg.); PIEN, Engelmann spruce (Picea engelmannii Parry ex var. engelmannii Engelm.); ABLA, subalpine fir (Abies lasiocarpa (Hook.) Nutt. var. lasiocarpa).

will increase with increasing mean annual temperature (MAT) as in previous studies, but that RL may increase or decrease with MAT, depending on mean annual relative humidity (MARH).

Materials and methods Study area We sampled nine sites that were selected from a larger collection of study sites spanning across an elevation gradient in the Priest River Experimental Forest (PREF) in the northern Rocky Mountains, USA (Duursma et al. 2003; Pocewicz et al. 2004). PREF is a mixed-conifer forest on the western flank of the Selkirk Range in Bonner County, Idaho (N48°21=, W116°50=). Soils are loamy Inceptisols with a volcanic ash cap, at the top of the mineral soil, of about 10 to 20 cm depth. The area has a maritime-continental climate regime, characterized by cold, wet winters and warm, dry summers. Sites were chosen from within the following elevation bands: 800–950 m (low), 1100–1250 m (middle), and 1400–1550 m (high). The sites were further constrained to fall within 500 m of a road to facilitate access. Within each site, three subplots were selected to maximize canopy cover and minimize understory vegetation, aiming for a spatially-continuous litter layer on the forest floor comprised mostly of overstory needles, cones, and twigs. Litterfall occurred sparingly throughout the growing season, peaking during late September through October. Site descriptions, including overstory composition and litter layer depth, are in Table 1. Respiration measurements We measured CO2 efflux from the soil surface when sites were snow-free on the following dates: July, August, and October 2007, and May, June, July, September, and November 2008. The measurements occurred once every four to six weeks at each of three subplots per site, with the exception of when snow was covering the ground surface, leading to only 44 two sampling dates at two of the high elevation plots during 2008. At each subplot, eight collars were placed within a 16 m2 area, four each along two perpendicular bisecting axes at 1.57 m intervals. We used a soil chamber (6400-09, LI-COR Biosciences, Lincoln, Nebraska, USA) in conjunction with a portable infrared gas analyzer (LI-6400, LI-COR Biosciences) to measure respiration. PVC collars (10.16 cm d, 5 cm ht) were pushed 1 to 3 cm into the ground in late spring and left in place throughout the sampling season, except when displaced by animal activity (in which case the collar was replaced but no measurement was taken until the following measurement time). At the time of measurement, the insertion depth of the collar was recorded on the deepest and shallowest side of the soil collar and the mean of these values was used to calculate the true chamber volume. The soil chamber was placed onto the collar with a closedcell foam gasket to seal off the outside air. At the beginning of

each measurement cycle, air inside the chamber was scrubbed of CO2 to concentrations of about 20 to 30 ␮mol·mol−1 below ambient; then, the CO2 concentration inside the chamber was allowed to increase for a minimum of 20 s, which typically increased concentrations by 10 to 20 ␮mol·mol−1 above ambient, while the LI-6400 measured the flux rate over time. The final flux rate recorded for each cycle was calculated for ambient conditions by the instrument software. Three measurement cycles were made at each collar. The measured flux rate tended to decline into an asymptote over the three measurement cycles. We presumed that the observed decline was due to high initial CO2 concentrations in the boundary layer above the soil surface, especially inside the collars where turbulent flow is reduced. These high concentrations would have been eliminated by the measurement process, which would at first create steeper concentration gradients out of the soil and would later lead to a steady-state flux (Davidson et al. 2002). Based on this reasoning, we concluded that the third flux measurement would best describe undisturbed flux rates, rather than a mean of the three cycles. Soil respiration rates (Rraw) were adjusted to represent horizontal area (Rcorr) rather than area along the surface of the slope. This was necessary because accompanying C measurements in our study were expressed on a horizontal area basis, and we wanted to be consistent with them. To the best of our knowledge, this “correction” to horizontal area has not been documented in the literature but is necessary for combining soil respiration estimates with other ecosystem C fluxes and pools as part of a C budget. Projecting a circle with area Araw onto a slanted plane results in an ellipse with an area of: (1)

Acorr ⫽ ␲ × rA × rB

where rA runs across the slope and is constant for each collar, and rB is the product of the slope cosine and rA. The ratio of Rraw to Rcorr is equal to the ratio of Acorr to Araw. Thus, (2)

Rcorr ⫽

Rraw cos(slope)

We corrected the raw flux rate using eq. (2) based on slopes measured inside the collar. This correction resulted in an increase in calculated flux values of up to 20% on the steepest slopes. Litter removal treatments In June 2007, litter was removed from half the respiration collars to separate RS into litter and mineral soil components. Fresh litter and loose duff (Oi horizon) were removed from an area centered at each of four collar locations at each subplot. The area Published by NRC Research Press


434

cleared was large relative to the collar (0.09 m2, or 20 times the collar area) to minimize the influence of lateral diffusion from adjacent soil with intact litter layers. We avoided removing compacted duff into which fine roots had penetrated; we refer to the resulting respiration rates as RNL to denote soil respiration with no litter. A sheet of nylon mesh screen (pore size ⬃1.7 mm) was then secured over the exposed soil area and respiration collar to prevent litterfall from entering the area. The protective screening was temporarily removed to allow for soil respiration measurements, but otherwise left intact over the study period. We then calculated the litter component (RL) of forest floor respiration by subtracting the subplot mean (n = 4) RNL from the subplot mean RS (n = 4). Proportional litter respiration (PL) was determined by dividing RL by RS. Our method for determining the contribution of litter to respiration sometimes led to negative values for RL (19% of all measurements), which might have resulted from a high variability in respiration rates relative to a low treatment effect. Negative litter respiration values were retained in the data sets and incorporated into the statistical models. Environmental measurements At each site, respiration component measurements were associated with key environmental variables thought to influence respiration rates: air and soil temperature, soil moisture, and air humidity. During each soil respiration measurement, soil temperature was measured at 5 cm below the litter surface adjacent to each collar using a thermocouple (Omega Engineering, Stamford, Connecticut, USA), which we call instantaneous soil temperature (Tsoil). Soil water content (SWC) was logged using EC-10 probes (Decagon Devices, Pullman, Washington, USA) at depths of 10, 20, and 40 cm from the litter surface at a central location at each site once every 2 hours (SWC10, SWC20, SWC40, respectively). Air temperature (Tair) and relative humidity (RH) was monitored at 2 m above ground level using shielded integrated-circuit temperature/ capacitive-type humidity sensors (iButtons, Maxim, Dallas, Texas, USA). Humidity measures were averaged into daily and annual means (from 1 July 2007 through 30 June 2008) for relating to growing season mean respiration rates. Soil moisture sensors were calibrated for each site and depth by collecting gravimetric soil moisture samples at the same depths near the location of the sensors at wet and dry times of the year and generating linear calibration equations. Precipitation was measured by daily dipstick measurements from a standard 8 inch diameter National Weather Service storage canister located on one location within PREF at an elevation of 725 m. Because precipitation was only sampled at one location in the forest, we only correlated precipitation with seasonal trends in respiration rather than trends across the elevation gradient. To determine whether changes in litter respiration and relative humidity were linked to changes in litter moisture, we determined gravimetric litter water content (LWC) from each site on three sample dates (August 2007, October 2007, and June 2008). We destructively collected one litter layer sample from an area of ⬃80 cm2 from each subplot, dried the samples at 60 °C, and weighed them to determine moisture loss. Removing the litter layer may have altered soil temperature and moisture with potential impacts on soil respiration rates. However, soil temperatures at the time of respiration measurements were not different between the litter-removed areas and the litter-intact areas (Student’s t test; P = 0.50). To determine effects of litter removal on soil moisture, we compared gravimetric soil moisture (0–10 cm) from litter-removal treatments to soil at similar depth from untreated areas 30 cm outside the sites. The litter removal treatment affected 0–10 cm soil moisture only on one date and five sites, when it increased soil moisture in October 2007 by 0.25 g-water·g soil−1 (Student’s t-test; P < 0.01). However,

Can. J. For. Res. Vol. 44, 2014

we do not believe these effects would have compromised most of our respiration measurements. Statistical analyses We first tested our hypothesis about different elevation trends of instantaneous RS and RL using a one-way analysis of variance (ANOVA) on a balanced data set, including only dates at which each site was sampled (July–October for both 2007 and 2008), testing for differences among individual elevation classes using Tukey’s honestly significant difference (HSD) test. To explain patterns using climatic variables, we then used least-squares linear regression to relate (i) seasonal changes in site-level climatic variables (RH, soil temperature measured at each collar, soil chamber air temperature, and litter water content) and precipitation (measured at one site in the study area) to instantaneous RS and RL (“instantaneous models”), and (ii) elevational changes in mean site temperature and humidity to growing season mean RS and RL (“cross-site models”). For instantaneous models, we isolated seasonal climate effects from effects of climatic change along the elevation gradient by applying an analysis of covariance (ANCOVA), testing for the effect of climatic variables after accounting for covariance with the elevation class (low, moderate, and high elevation). For RS, instantaneous models related simultaneous soil temperature and the nearest bi-hourly soil moisture reading to RS. SWC10, SWC20, and SWC40 (rather than a profile-level mean) were included separately as explanatory variables for RS, and only the depth that yielded the most significant regressions (based on r2) are presented. For instantaneous RL models, we accounted for time lags between RL and RH for up to 4 days prior to measurement of RL. The seasonal influence of precipitation was considered by summing precipitation from the 7 days prior to the respiration measurement (Precip7), reasoning that precipitation may have a lagged effect on respiration components that are moisture-dependent. Cross-site models related MAT and MARH to growing season respiration means at each site. For goodness-of-fit, we report adjusted r2 and P values, and we judged significance at P levels below 0.05, unless otherwise indicated. ANOVA were conducted using PROC ANOVA and ANCOVA were conducted using PROC GLM in SAS software version 9.3. Non-parametric Kruskal–Wallis tests (kruskal.test), correlations (cor.test), and regressions (lm) were conducted using R (R Core Team 2013). All data were tested for assumptions of normality and homegeneity of variance (Shapiro–Wilks test and visual assessment); if assumption violations occurred, appropriate nonparametric tests were used and are reported in the Results. Mean values are reported ± 1 SE.

®

Results The elevation gradient produced a range of MAT and MARH (Table 1; Fig. 1d). As hypothesized, RS and RL differed by elevation class, with divergent trends from each other (Fig. 1; Table 2). RL was non-normally distributed for the moderate elevation class, so a Kruskal–Wallis test was used to demonstrate effect of elevation class differences (H = 13.9, df = 2, P < 0.001). RS was similar for the low and moderate elevations and dropped off in the high elevations (Fig. 1a). In contrast, RL was lowest at both low and high elevations, peaking in the moderate elevations (Fig. 1b). PL displayed similar elevation trends as RL (Fig. 1c). Litter respiration varied seasonally, peaking in June 2008 or October 2007, depending on the site (Fig. 2). LWC was lowest in August 2007 (when precipitation was sparse) and highest during June 2008, when LWC appeared to increase with increasing elevation class (Fig. 3). Both LWC and RL were best correlated with mean daily RH from 2 days prior to measurement (RH2), out of the 5 day time window examined (Fig. 4). After accounting for covariance of climatic variables with elevation, seasonal changes in Precip7 and RH2 were positively correlated with seasonal changes in instantaneous RL and PL; LWC was not correlated to either RL or PL. Tair was Published by NRC Research Press

Berryman et al.

435


Fig. 1. Box-and-whisker plots showing the trend in growing season (a) soil respiration (RS), (b) litter respiration (RL), (c) proportional litter respiration calculated via RL/RS (PL), and (d) site mean annual temperature (MAT, black line) and mean annual relative humidity (MARH, grey line) with elevation class at nine sites in Priest River Experimental Forest in Idaho, USA. The center line in each box represents the median value, the box boundaries represent the 25th and 75th percentile values, and the error bars represent the 10th and the 90th percentile values, with outliers indicated by circle symbols. Notches represent the lower and upper bounds of 95% confidence intervals. Only data from dates at which every site was sampled are included. Different letters denote statistical difference at P < 0.05 (Tukey). ANOVA results are presented in Table 2; RL was tested using a Kruskal–Wallis test (H = 13.9, df = 2, P < 0.001). Lines of MAT/MARH trends with elevation class are linear leastsquares fits. Low elevation sites ranged from 800 to 950 m a.s.l., moderate elevation sites ranged from 1100 to 1250 m a.s.l., and high elevation sites ranged from 1400 to 1550 m a.s.l.

Table 2. Analysis of variance showing that RS varies by elevation class (low, moderate, and high). Variable

Source

DF

SS

MS

F

P of F

RS

Model Error Corrected Total

2 42 44

44.8 50.2 95.0

22.4 1.20

18.7