Phenology of fine roots and leaves in forest and grassland


†Correspondence author. E-mail:


  • 1The phenology of temperate vegetation is advancing in association with climate warming. Most phenology data, however, comes from flowers and tree leaves. We tested the generality of results from shoot phenology by expanding data collection in two new directions. We related forest leaf phenology to root phenology, and to phenology in a second habitat, grassland.
  • 2We measured leaf and root phenology simultaneously in aspen forest and adjacent native grassland. Root growth accounts for 80–90% of productivity in these habitats. Seasonal variation in soil moisture and temperature were also measured.
  • 3Forest leaf production was greatest about 45 days before peak root production, resulting in a significant negative correlation between leaf and root production in forest. Grassland leaf production was greatest about 15 days before peak root production, and grassland leaf and root production were significantly positively correlated. The duration of root production was 40% greater than that of shoot production.
  • 4Forest leaf production increased significantly with increasing soil moisture, but not temperature. In contrast, the production of forest roots, grassland roots and grassland leaves increased significantly with soil temperature.
  • 5Synthesis. The most commonly measured aspect of phenology, forest leaves, is out of step with the majority of production in forest, as well as phenology in grassland. The invasion of grassland by woody vegetation is characterized by a decoupling of root and shoot phenology, a result that has not been reported previously. Given the global nature of woody plant encroachment, this decoupling may influence our general understanding of productivity and carbon sequestration in response to warming.


The phenology of both leaf (Chmielewski & Rotzer 2001) and flower (Fitter & Fitter 2001; Miller-Rushing et al. 2006) production in temperate regions is advancing in association with climate warming. Warming is expected to accelerate the onset of early growing-season production (Beniston et al. 2003), and leaf phenology plays a large role in models predicting the response of vegetation to warming (Kramer et al. 2000). In contrast, very little is known about root phenology (Jackson et al. 2001), and a recent review of phenology does not mention roots (Cleland et al. 2007). This is surprising given that root production accounts for 50–90% of primary production in temperate vegetation (Ruess et al. 2003; Steinaker & Wilson 2005).

Knowledge about the relationship between shoot and root phenology is needed to understand the overall response of vegetation to warming. If shoot and root phenologies are closely related, above-ground patterns are likely to reflect overall producer phenology. On the other hand, a decoupling of root and shoot phenology would cast into doubt knowledge about total producer phenology based on above-ground patterns. The temporal overlap of root and shoot growth varied among four Mediterranean subshrubs (Palacio & Montserrat-Martí 2007), suggesting that synchrony of above- and below-ground phenology is likely to vary among species and vegetation types.

Contrasting vegetation types may occur along a continuum from synchrony between roots and shoots to complete temporal separation. Root and shoot phenology could be closely related because of physiological coupling, with shoots dependent on roots for soil resources and roots dependent on shoots for photosynthates. This is likely to be the case in short-stature ecosystems, such as grasslands, where most production is below-ground (Steinaker & Wilson 2005), and spring growth utilizes nutrients translocated to tussock crowns the previous winter (Clark 1977). On the other hand, root and shoot phenology could be decoupled in vegetation with a relatively large above-ground component, such as forests, for several reasons: (i) shoots are affected by air temperature (Jackson et al. 2001; Cleland et al. 2007) and roots by soil temperature (Teskey & Hinckley 1981; Tierney et al. 2003), and air warms faster than soil; consequently, the importance of this difference should increase with increasing plant size and physical separation between shoots and roots; (ii) nutrients and photosynthates for new leaf growth are stored in distal twigs and are influenced by atmospheric warming (Landhäusser & Lieffers 2003); and (iii) there is strong light pre-emption in forests, conferring an advantage to rapid leaf emergence (Wilson 1993). Previous studies confirm that a number of results are possible. Leaf expansion in a temperate forest was synchronous with new root growth in the forest floor but preceded root growth in mineral soil by 1–2 weeks (Fahey & Hughes 1994). In a more temperate forest, roots grew throughout the winter in the absence of leaves (Teskey & Hinckley 1981). Shoots preceded roots in arctic tundra (Kummerow & Russel 1980), but roots preceded shoots in cool semi-desert shrubs (Fernandez & Caldwell 1975). Root and shoot phenology are rarely measured simultaneously (cf. Palacio & Montserrat-Martí 2007).

Root phenology studies have yielded conflicting results partly due to the fact that root harvesting collects only about 60% of root mass (Robinson 2004). Images collected from rhizotrons provide a more complete measure of root production with little disturbance (Jackson et al. 2001). Here we examine the correlation between root and leaf phenology in two habitats likely to differ in phenological coupling. These were native grassland, where most production and competition is below-ground, and adjacent invading forest with more above-ground production and shade (Wilson 1993; Steinaker & Wilson 2005). Much of our knowledge about phenology originates from woody species, and grasslands are poorly represented in the literature (Jackson et al. 2001). Both leaf (Jackson et al. 2001; Cleland et al. 2007) and root (Teskey & Hinckley 1981; Tierney et al. 2003) phenology are linked to temperature and soil moisture.

Our objectives were to (i) determine whether forest leaf phenology is a good descriptor of total system phenology; (ii) test the hypothesis that the temporal coupling between root and leaf production is higher in grassland than in forest; and (iii) determine how phenology patterns in contrasting cases (leaves and roots, forest and grassland) were related to abiotic factors known to influence phenology.


study area

We worked at the northern edge of the Great Plains of North America, in White Butte Recreation Area (50°28′ N, 104°22′ W, 30 km east of Regina, Saskatchewan, Canada). Our study area is a mosaic of native grassland and aspen forest. In this region, Populus tremuloides Michaux (trembling aspen) has expanded into grassland for at least a century (Steinaker & Wilson 2005) in association with fire suppression and anthropogenic nitrogen deposition (Köchy & Wilson 2001). The forest has a shrubby understorey dominated by Symphoricarpos occidentalis Hook. and Rubus idaeus Regel & Tiling. Grassland is dominated by Stipa comata Trin. & Rupr., Carex spp., Bouteloua gracilis (HBK.) Lag., Agropyron spp., Koeleria gracilis Pers., Poa spp. and Selaginella densa Rydb. Parent soils are regosols on silty sand. The climate is continental with mean daily temperatures of −17 °C in January and 19 °C in July. The mean annual precipitation is 388 mm, falling mainly in May–August. The study year was close to average for both temperature (2004: 8.6 °C; 30-year mean: 10.1 °C) and rainfall (353 and 332 mm; Steinaker 2006).

We measured root and leaf production, as well as soil water and temperature, in both grassland and forest biweekly during the 2004 growing season, from 9 April, when soils were no longer frozen, to 15 October, close to the date of the first autumn snowfall. Measurements were made at ten locations, each separated by > 30 m, in both grassland and forest, for a total of 20 locations. Each location was 15–30 m from the grassland-forest boundary, spread along a 500-m-long section of the boundary. Locations were selected randomly within areas of similar topography and elevation to represent level, undisturbed sites (Steinaker & Wilson 2005). The area has been closed to livestock for the last three decades. Because of similarities in topography, elevation and soil texture between our forest and grassland sites, differences between habitats at this area are associated with aspen invasion. Based on reported aspen invasion rates, our study forest sites were colonized by aspen within the past 3–4 decades (Steinaker & Wilson 2005).

We measured root and leaf production throughout the growing season using sequential imaging. We determined production at different dates during the growing season as relative percentages of the total annual in order to compare seasonal patterns of productivity. Because of the non-destructive nature of these methods, the same plants can be monitored over time, and temporal changes in biomass are not confounded by random spatial variability at each sampling date.

root production

Seasonality of root production was followed using a rhizotron camera (Bartz Technology Corporation, Santa Barbara, CA), a small digital camera that runs through a transparent tube buried in the soil. One transparent rhizotron tube (180 cm long, 6 cm diameter) was installed in each location, for a total of 10 tubes in each habitat. Tubes were installed at a 45° angle to the soil surface, reaching 90 cm deep vertically, a depth which allowed us to record > 90% of root mass (Schenk & Jackson 2002). The end of the tube above-ground was wrapped in black plastic tape to prevent light from entering the tube. The black tape was then wrapped with white tape and the tube was closed with a white plastic cap to minimize solar heating. Warm air was unlikely to travel down the tube because on warm days the soil was cooler than the air, which presumably resulted in cooler, denser air in the tube bottoms that would not mix with warmer, less dense air in the top of the tube. Rhizotron tubes were installed in summer 2000 with measurements beginning in spring 2004 in order to allow fine roots to colonize disturbed soil around the tubes (Johnson et al. 2001). An indexing handle attached to the camera allowed us to follow the same location in the soil through time. Images were taken approximately every 15 days during the growing season.

We recorded twenty-five 18 × 14 mm images equally separated along each rhizotron tube on each date. Thus, we analyzed 250 images for each habitat on each date, for a total 7000 images for the complete season. Image analyses were conducted in the laboratory on a high-resolution colour monitor that magnified the image 10 times. We measured the length of fine roots (< 2 mm diameter) by hand directly on the images. The total length of fine roots produced during each sample interval was calculated for each tube.

leaf production

Images for measuring leaf phenology were taken within 3 days of root images. In grassland, we assessed the timing of annual above-ground production through changes in the number of green pixels in sequential images taken biweekly over a growing season. At each location, we took a downward-looking photograph 1.5 m above the ground over each rhizotron tube, using a digital camera (Olympus Stylus 300) at a resolution of 2048 × 1536 pixels. Markers in the ground allowed us to record images of exactly the same area each time. Pictures were taken between 11 : 00 and 14 : 00 on clear days. The number of green pixels in each image was determined using Adobe Photoshop Elements (Adobe Systems Incorporated, 2001). Because sunlight intensity changes seasonally, we adjusted the luminosity of each image in order to make comparable measurements. The method of counting green pixels is a variation of the point-quadrat method in which plant cover is estimated from the proportion of points intersecting green tissues (Paruelo et al. 2000). The percentage of green pixels in an image was highly correlated with biomass in Colorado (Paruelo et al. 2000) and West Africa (Gerard et al. 2001).

In forest, we took an upward-facing horizontal image of the canopy from 1.75 m above the ground over each rhizotron tube. Thus these images included stems but the greatest within-season changes were expected to occur in leaves and we discuss these changes as leaf phenology. We did not measure understorey phenology because the understorey comprises only about 12% of forest shoot mass (Wilson 1993; Kleb & Wilson 1997). Pixel quantification proved unreliable for forest leaves, probably because of the bright sky background. Therefore we visually quantified leaf cover on each date as a relative percentage of the maximum cover, comparing all the images in a sequence over time. This method was checked using repeated observations by one observer (DFS) and independent observations by two other observers. As with our analysis for grassland, we determined the production at each interval as the difference in leaf cover between sampling dates.

Our measurements of shoot phenology might underestimate shoot production relative to root production because root images included variation in depth whereas shoot images comprised only two-dimensional projections of the canopy. This was addressed in two ways. First, the two-dimensional nature of shoot images did not affect the detection of first production, so the beginning of spring growth was readily comparable using our data. Second, a concurrent study nearby (Steinaker 2006), which measured the length growth of shoots of five woody species, produced temporal patterns very similar to those in this study, indicating that our shoot images produce realistic temporal patterns.

duration of production

We determined the duration (D) of production in order to test whether this varied between roots and leaves, and between grassland and forest. D was calculated as the sum of the differences between the absolute production value on each date (t) and mean annual production:

D= 1/((∑PtPm))

Where Pt is absolute production at time t, and Pm is mean production. In this case, Pm is a constant (100/number of sampling dates) because the production values are relative percentages. Duration could be expressed as the number of sample periods in which production occurred, but D assigns more weight to sample periods with higher production values.

soil moisture and temperature

Soil moisture was measured at four soil depths (10, 30, 45 and 60 cm) on each date using a form of time domain reflectrometry (James et al. 2003). At each location, a tube (PVC, 5 cm diameter, 75 cm deep) allowed access for a soil moisture probe (Sentry 200-AP, Troxler Electronic Laboratories Inc., NC). The probe measures disturbance to an electrical field caused by soil moisture in a disk-shaped soil volume 10 cm deep that extends 10 cm out from the side of the access tube. We transformed each probe reading to relative soil water content (actual value minus the annual minimum value, divided by the annual range) (Dyer & Rice 1997).

Three temperature readings from each location on each date were also obtained at the same four soil depths (10, 30, 45 and 60 cm) using mercury thermometers. The replicates from each location, date and depth were averaged prior to analysis.

statistical analysis

We tested for differences between habitats in soil moisture and temperature using repeated-measures anova, with depth nested within habitat. Multiple comparisons among soil depth classes within each habitat were performed using Tukey-Kramer HSD tests. Differences in production duration between habitats for both shoots and roots were tested using one-way anova. We checked for normality of data and homogeneity of variances using Shapiro-Wilks and Levene tests, respectively. Response variables that departed from normality were [ln(x + 1)]-transformed. All analyses were performed using JMP software (JMP, SAS Institute, 1994).



In grassland, maximum leaf production occurred just one sample interval (15 days) before maximum root production (Fig. 1a). As a result, leaf and root production were significantly positively correlated over time (Fig. 2a).

Figure 1.

Mean (+ 1 SE) relative production of shoots (black) and roots (white) in adjacent grassland (a) and forest (b). Values are the percentage of total annual production.

Figure 2.

The relationship between leaf and fine root production in grassland (a) and forest (b). Each point is relative production from a 15-day period (Fig. 1) in each of the study sites. Points with no production for either component were excluded from the analysis.

In forest, maximum leaf production occurred three sample intervals (45 days) before maximum root production (Fig. 1b), and leaf and root production were significantly negatively correlated over time (Fig. 2b). Thus root and leaf phenology were characterized by synchrony in grassland but not in forest.

The timing of root production was similar in both habitats, with first, maximum, and last production occurring on the same sample dates (Fig. 1).

Leaf production started on the same dates in each habitat, but the forest maximum occurred 30 days before the grassland maximum, and last production was 15 days earlier in forest (Fig. 1). Overall, forest leaf phenology was accelerated relative to grassland.

Root production occurred throughout the spring and summer compared to leaf production, which was restricted to the early part of the growing season (Fig. 1). Thus, production duration (D) was significantly greater for roots than leaves in both habitats (roots: 0.0104 ± 0.0003 [mean ± SE, averaged across habitats]; leaves: 0.0073 ± 0.0001; P < 0.001). Thus the duration of root production was about 40% greater than shoot production. Production duration, however, was not significantly different between habitats, whether calculated for leaves or roots (grassland: 0.0086 (averaged across roots and leaves); forest: 0.0085; P > 0.05).

A significant interaction occurred between time and depth for root production in both habitats (P < 0.001). In grassland, this resulted from root production varying over time at shallow depths but not at greater depths (Fig. 3a). In forest, the interaction reflected a wave of maximum root production from the shallowest depth at the start of the growing season to maximum depths at the end of the growing season (Fig. 3b).

Figure 3.

Root production over time at five depths in grassland (a) and forest (b). Significant interactions between time and depth occurred because root production in grassland showed a seasonal response only near the soil surface. In forest, maximum root production occurred at successively greater depths during the growing season.

soil moisture and temperature

Soil moisture varied significantly with time (F12,977 = 24.38, P < 0.001) but did not differ significantly between habitats (F1,977 = 0.23, P > 0.05, Fig. 4a,b). Seasonal patterns of soil moisture were similar between habitats. Soil water content was relatively high in spring and decreased in early-mid summer. Late summer precipitation (34 mm during 21–26 August) restored moisture to both grassland and forest soils.

Figure 4.

Soil moisture (top) and temperature (bottom) in grassland (left) and forest (right) at four depths. Letters beside the legends compare depths within each habitat (all dates averaged). Depth labels with identical letters did not differ significantly (P > 0.05, Tukey–Kramer HSD test).

Soil temperature varied significantly with time (F12,2016 = 2316.08, P < 0.001) and soils were significantly warmer in grassland than forest (F1,2016 = 7.10, P < 0.05; Fig. 4c,d). An interaction between habitat and sample date (F12,2016 = 31.16, P < 0.001) occurred because grassland soils were warmer than forest for most of the growing season, but similar between habitats at the end of the growing season (Fig. 4c,d).

correlations with production

Grassland leaf production increased significantly with soil temperature (Table 1), because peak grassland leaf production (Fig. 1a) occurred during the high temperatures of midsummer (Fig. 4). In contrast, forest leaf production increased significantly with soil moisture (Table 1) because forest leaf production was greatest earlier (Fig. 1b) when soil moisture was higher (Fig. 4). In both habitats, fine root production increased over the growing season as soil temperature increased and soil water decreased. Between habitats, root production had similar temporal relationships to abiotic factors, but leaf production did not.

Table 1.  Seasonal correlations (Pearson product-moment) between production and soil moisture and temperature in grassland and forest. Asterisks show significance probabilities (*P < 0.05; **P < 0.001; NS: no significant difference)
 Soil waterSoil temperature


The phenology of forest leaf production was distinct from forest root phenology, and both above- and below-ground components of grassland phenology (Fig. 1). Thus forest leaf phenology is not a reliable index of overall vegetation phenology, or of phenology in other vegetation types.

Forest leaf production typically precedes root production (Hendrick & Pregitzer 1996; Joslin et al. 2001) because the atmosphere warms faster than the soil. This occurred in our study: early forest leaf production was associated with warmer air (3.9 °C; 15-day average) than soil (0.9 °C, Fig. 4) at the start of the growing season. Root production might also follow leaf production because fine root growth depends heavily on newly fixed carbon from the canopy (Fitter et al. 1999; Joslin et al. 2001).

Forest leaf production may precede grassland leaf production for two reasons. First, expanding forest leaves receive carbohydrates and nutrients translocated from supporting twigs (Landhäusser & Lieffers 2003) which, in the spring, are affected by the same warm temperatures as leaves. In contrast, grasses translocate nutrients from below-ground crowns (Clark 1977) in cooler soil. Second, woody vegetation is characterized by intense light competition relative to grasslands (Wilson 1993) and woody species presumably have been selected for early leaf production to allow light pre-emption. Leaf phenology also varied among growth forms in Patagonian steppe (Jobbágy & Sala 2000).

Grassland leaf and root production may be more synchronous than that in forest because, as noted above, nutrients for new leaf production are translocated from below-ground storage. In general, root and shoot phenology may be most closely coupled in vegetation with large below-ground components, such as grasslands and tundra, and less coupled in habitats with larger above-ground components, such as shrubland and forest. The generality of this pattern needs to be tested.

Plots of leaf vs. root production displayed considerable variation (Fig. 2). This is not surprising because there is no expectation that leaf and root production are linked exclusively to one another. The salient point in Fig. 2 is that the significant relationships had opposite signs in the two habitats, indicating greater synchronicity in grassland than forest.

Relationships between leaf and root production need to be interpreted in light of the fact that different methods were used to measure each (rhizotron vs. ordinary camera images). Rhizotrons have measured leaf production in the greenhouse (Pärtel & Wilson 2001), producing wholly comparable above and below-ground data, and in principle this could also be done in the field.

Root production showed similar temporal patterns in the two habitats, probably because the roots experience similar temporal patterns in temperature and moisture (Fig. 4). The presence of a tree canopy did not appear to affect forest root phenology relative to that in the grassland. Root production in both habitats was positively correlated with temperature (Table 1). Maximum root length occurred in midsummer when potential evapotranspiration was high and soil water was low. Other studies also showed negative temporal (Teskey & Hinckley 1981; Hendrick & Pregitzer 1993) and spatial (Davis et al. 2004; Dukes et al. 2005) correlations between soil water and fine root production, as well as positive temporal (Teskey & Hinckley 1981; Steele et al. 1997; Tierney et al. 2003) and spatial (Tryon & Chapin 1983; King et al. 1999) correlations between soil temperature and fine root production.

Forest root production was probably also related to photosynthate allocation, first to shallow roots and later to deeper roots (Fig. 3b). The cause of this temporal partitioning was not temperature because a similar pattern was not shown by this factor (Fig. 4). The last peak in forest root production did correspond with a late-season peak in soil moisture (Fig. 4), which was also reflected by root production in shallow grassland soils (Fig. 3a). In the case of forest roots, however, there was an intermediate peak that was not associated with a rainfall event (Fig. 3b), and, in contrast to grasslands, shallow forest roots did not respond to the late season pulse in water availability. Taken together, the evidence discounts water as a mechanism behind this pattern and suggests that the photosynthate allocation progressed from shallow to deeper roots. Clearly root production depends on both exogenous and endogenous influences (Tierney et al. 2003).

Grassland shoot production also increased significantly with temperature (Table 1), probably due to the close physiological links between roots and shoots. Forest leaf production in contrast, increased significantly with soil moisture, because leaf production was higher earlier in the year (Fig. 1) when soils were relatively wet (Fig. 4).

The invasion of grasslands by woody plants is a global phenomenon (Archer et al. 2002). Our data show that this invasion greatly advances leaf phenology, although not the timing of root production (Fig. 1). Leaf and root phenology, which are temporally coupled in grassland (Fig. 2a), are decoupled by tree invasion (Fig. 2b). This consequence of woody invasion has not been reported previously.

In the context of climate warming, an increment of temperature increase may have a greater impact on grasslands than forests because of the dampening effect of forest trees on temperature (Fig. 4). Thus for each unit of atmospheric temperature increase, soil temperature should increase more in grassland than in forest. Warmer soils under grasslands may favour fine root production and contribute to the exclusion of trees in competition for below-ground resources. In support of this, dendrochronological studies show that tree recruitment into grasslands tends to occur during cool periods (Kitzberger et al. 2000; Chhin & Wang 2002). Increased grassland fine root production due to increasing soil temperature may also restrict tree establishment by limiting soil nitrogen availability both via uptake and microbial immobilization, because fine roots are an important source of carbon for soil microbes (Knops et al. 2002). This explains why inorganic nitrogen availability in semi-arid grassland soils is lowest in the summer (Li & Wilson 1998; Hook & Burke 2000), when soil temperature and root production rates are highest. Aspen forest expansion in our region increases with soil N availability (Köchy & Wilson 2001), so a microbial reduction of the nitrogen availability in grassland soils might retard aspen invasion.

The difference between root and shoot phenology, and the preponderance of root production in temperate vegetation, may have implications for climate change models which consider moisture and temperature effects on total plant production without distinguishing shoots from roots (Smith et al. 1997; Adams et al. 2004). Simulation models indicate that a 5–10-day lengthening of the growing season can increase the net productivity of temperate forests by as much as one-third because early spring photosynthetic tissue disproportionately increases total productivity (Jackson et al. 2001). Leaf production, however, is a relatively small portion of the total (8–20% in our systems), so accelerated leaf phenology may have limited ecological consequences, particularly in habitats where system productivity is limited by below-ground resources, as in the case of natural grasslands (Sala et al. 1988). Finally, the duration of root production in our forest was about 40% greater than that of shoot production, and root production in some temperate forests occurs even during winter in the absence of deciduous leaves (Teskey & Hinckley 1981). This suggests that models based on shoot production alone are using an underestimate of the length of the growing season of the bulk of the vegetation.

In conclusion, leaf phenology was positively correlated with root phenology in grassland, suggesting that grassland leaf phenology reflects overall system seasonality. In forest, however, leaf phenology was much earlier than root phenology, and forest leaf phenology was not a reliable indicator of phenology of the 80% of the total system production accounted for by fine roots. The duration of root production was 40% greater than that of shoots. Taken together, these previously undescribed aspects of root phenology are likely to have implications for predicting vegetation responses to warming.


Authors thank D. Donald and F. Gendron for field assistance, R. Bardgett, B. Schamp, B. Vaness and referees for improving the manuscript, Saskatchewan Environment for access to White Butte Recreational Area, and Instituto Nacional de Tecnología Agropecuaria (INTA) of Argentina and the Natural Sciences and Engineering Research Council (NSERC) of Canada for research support.