Introduction

The boreal forest in eastern Siberia, dominated by larch (Larix cajanderii), covers an approximately continuous permafrost area of 1.1 million km2 (Shvidenko and Nilsson 1994). Permafrost regions occupy about 25% of the terrestrial surface in the northern hemisphere of which 60% is in Siberian Russia (Kudrjavtsev et al. 1978). The water use efficiency of larch forest during the short growing season in eastern Siberia (approximately 100 days) is influenced by the soil thawing rate (Lopez et al. 2007a), soil moisture (Sugimoto et al. 2003), and vapor pressure deficit (Dolman et al. 2004). The limited annual rainfall regime of 250 mm is unevenly distributed in summer during the growing season, with over 40 mm of rainfall on some days. Present climate change trends indicate increases in summer precipitation for the period 1986–2004 in this region (Park et al. 2008). Changes in rainfall regime are expected to have a great impact on the water balance of the forest through its influence in the shallow active layer where the tree root system is distributed. However, there is evidence that inter-annual forest evapotranspiration and canopy transpiration does not vary in relation to large annual variation in rainfall (Dolman et al. 2004; Lopez et al. 2007a; Ohta et al. 2008). Furthermore, the modest inter-annual variation in canopy transpiration can be attributed to multiple inter-annual climatic factors that directly affect tree water use and soil thawing rate. In view of the present warming effect on permafrost regions (Fedorov and Konstantinov 2003; Jorgenson et al. 2006), this study explores the forest water dynamics via a controlled experiment where drastic increases in rainfall are simulated by irrigation. The results are compared to a control plot in order to evaluate the effect of increased rainfall under the same environmental conditions. The few studies carried out in boreal regions where irrigation and nutrients were applied have focused on tree growth (Bergh et al. 1999; Albaugh et al. 2004; Waterworth et al. 2007) and forest water relations (Cienciala et al. 1994; Ewers et al. 2001; Waterworth et al. 2007), but to our knowledge there has been no irrigation experiment focused on water dynamics of boreal forests underlain by permafrost.

Therefore, the objectives of this study are (1) to assess the effect of change in soil moisture as a result of irrigation on larch canopy transpiration and canopy conductance, and (2) to determine the water budget of the larch forest ecosystem under control (no irrigation) and increased rainfall (irrigation).

Materials and methods

Study site: control and irrigation plots

The experimental site has been thoroughly described in Ohta et al. (2001) and Lopez et al. (2007a). It is located 35 km NNW from the city of Yakutsk in eastern Siberia (62°15′N, 129°37′E, altitude 220 m) (Fig. 1). Forests in this area are exposed to dry conditions with an annual rainfall of approximately 250 mm (almost half of which occurs during the growing season). The forest is dominated by a 160-year-old Larix cajanderi monoculture with a tree density of 1,000 trees per ha. The sapwood basal area is 4.7 m2 ha−1 and the mean height is 13 m. Larch tree root distribution shows high root concentration in the upper layers, with thicker root diameters (d > 50 mm) in the mineral soil. In the upper 0–10 cm soil organic layer, root diameters ranging between 20 < d < 50 mm are abundant whereas below 40 cm depth, root density is composed of roots with d < 2 mm (Kuwada et al. 2002).

Fig. 1
figure 1

Location of the study site and schematic of the control and irrigation plots and relative height of the study site

Full size image

In this study, two plots of 14 × 14 m (control and irrigation) were set. Irrigation was evenly applied within the plot during 6 consecutive days (20 mm per day), from 17 to 22 July in 2004 with a total of 120 mm applied through sprinklers installed around and within the site at a height of 70 cm. The irrigation occurred between 1600 and 1900 hours. The water used had an electric conductivity of 0.1 mS cm−1 and was carried in tanks from the Lena River. The water temperature at the moment of irrigation was around 2°C lower than air temperature. The understorey vegetation in the plots was cowberry (Vaccinium vitis-idaea). Both plots were at the same relative height in the west–east direction and the slope southward was 1.6%.

Weather, soil moisture and soil temperature measurements

Weather data were recorded for the period July–August 2004, representing the active phase of the growing season. The variables measured were rainfall (raingauge; Young, USA), short-wave radiation, S r (pyranometer, CPR-PCM-01; Prede, Japan), relative humidity-air temperature, RH and T a (HMP-35D; Vaisala, Finland) and wind speed, u (3-cup anemometer, CYG-3002; Young). These data were collected continuously every 10 min (CR10X datalogger; Campbell Scientific, UK) at the top of a 32-m-high scaffolding tower built in the larch forest by Ohta et al. (2001), located approximately 160 m SW of the control plot.

Soil moisture FDR (frequency domain reflectrometry) sensors were installed at 10, 20, 30, 40, 60, and 80 cm depth (EnvironSMART; Sentek, Australia) at the center of each plot and the results were calibrated with direct weight measurements at all depths. These data were collected continuously every 30 min in separate loggers (CR10X datalogger; Campbell). Daily soil moisture averages were calculated for the same period as above.

The soil thawing depth, represented by the 0°C isotherm was determined by soil temperature data collected from 7 May to 27 September 2004. Soil temperature was measured at 0, 10, 20, 30, 40, 60, 80, 100, 120, 140, 160, 180, and 200 cm depth (three replications) thermistors (104ET; Ishiizuka Denshi, Tokyo, Japan) in the control and irrigation plots in the spring of 2003. Thermistors were calibrated using an ice-water bath with a precision of 0.02°C at 0°C. The overall probe accuracy for the temperature range from −20 to 30°C was ±0.09°C.

Canopy transpiration (E c )

In each plot, ten trees were chosen as a representative sample of each crown class and DBH (diameter at breast height) (Table 1). Sapwood cross-section of the trees studied and total stand sapwood were estimated from cores. The relationship between sapwood and tree diameter was used to estimate sapwood area (Lopez et al. 2007a) for each crown class at each plot. Because sap flow velocity is underestimated when part of the probe is in the non-conducting wood, individual tree sap flow values were corrected for differences in needle length and sapwood thickness (Clearwater et al. 1999).

Table 1 Characteristics of the ten trees selected for sapflow measurements in the control and irrigation plots
Full size table

Sap flow measurements for this study were conducted from 1 July to 31 August at each plot by the thermal dissipation technique (Granier 1987) with 20-mm-long radial sap flow sensors (UP; Umweltanalytische Produkte, Germany) installed at a height of 1.3 m in the stems. Sap flux density (U, m3 m−2 s−1) was estimated by this technique. Sap flow data were collected each 10 s, and 10-min average values were stored on a CR10X datalogger. Total tree sap flow (F) was calculated as the product of U multiplied by the sapwood cross-section. Sap flow measurements carried out in individual trees were scaled up from tree to stand level using sapwood area distribution as described in Granier et al. (1996). Stand sap flow (E c , mm day−1) was calculated independently for each plot as:

$$ E_{c} = S_{T} \sum {p_{i} } U_{i} $$
(1)

where S T is the stand sapwood area per unit of ground area (m2 m−2), p i is the proportion of sapwood in the class i, and U i is the average sap flux density in the class i. Calculations were based on 1-day interval data.

Canopy conductance (g c )

Canopy conductance at the control (g c_cont) and irrigation (g c_irrig) plot was computed by the inverse form of the Penman–Monteith equation (Monteith and Unsworth 1990) assuming that stand sap flow was equal to stand transpiration (E c , mm s−1) as in Eq. 2:

$$ \lambda E_{c} = {\frac{{\Updelta \left( {R_{n} - G} \right) + \rho C_{p} Dg_{a} }}{{\left[ {\Updelta + \gamma \left( {1 + \left( {g_{a} + g_{c} } \right)} \right)} \right]}}} $$
(2)

where Δ is the rate of change of saturation vapour pressure (Pa C−1), R n is the net radiation (W m−2), ρ is the density of dry air at constant pressure (J kg−1 C−1), D is the vapour pressure deficit (Pa), g a is the aerodynamic conductance (m s−1), g c is the canopy conductance (m s−1), λ is the latent heat of vaporization of water (J kg−1), and γ is the psychrometric constant (Pa C−1). In this study, the soil heat flux (G) is taken as 10% of the R n based on previous results (Iwahana et al. 2005). Aerodynamic conductance (g a , m s−1) was calculated as

$$ g_{a} = {\frac{{k^{2} u}}{{\left\{ {\ln \left[ {{\raise0.7ex\hbox{${(z - d)}$} \!\mathord{\left/ {\vphantom {{(z - d)} {z_{0} }}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{${z_{0} }$}}} \right]} \right\}^{2} }}} $$
(3)

The displacement height (d) was set as 0.67 h and the roughness length (z 0 ) as 0.1 h, where h is stand height, k is von Karman’s constant (0.40), and u is wind speed at height z above the canopy (Brutsaert 1982). In order to account for the time lag between sapflux and canopy conductance, daily values were used.

Water budget

The water budget of the forest ecosystem was estimated using the soil equivalent depth of water from July to August. Daily forest evapotranspiration (E a ) calculated from latent heat measured by the eddy covariance method with the open-path gas analyzer for this year and location were obtained from Ohta et al. (2008). E a in the irrigation plot was estimated based on the ratio between E c_irrig and E c_cont. E a includes canopy transpiration (E c ) and evaporation from the understory vegetation (E u ), allowing us to equate the water lost as evapotranspiration with the water changes in the forest soil at each plot. In order to determine the movement of water in the upper soil profile, this was divided into the organic layer (0–10 cm), the mineral soil where the bulk of the tree roots are distributed (10–40 cm), and deeper soil layers (40–60 and 60–80 cm).

Results

Meteorological conditions, soil moisture and soil thawing depth

The year 2004 was relatively cool and dry compared to mean values for Yakutsk (Rivas-Martinez 2007). Total precipitation during the growing season from May to September was 110 mm (Lopez et al. 2007a). Air temperature in 2004 was high for a short period, from 4 to 10 July, with a maximum daily value of 30.6°C that was followed by daily T a below 10.0°C (Fig. 2). Rainfall from 1 July until the irrigation day (17 July) was 2.9 mm, while rainfall from the day irrigation started to 31 August amounted to 60.5 mm.

Fig. 2
figure 2

Daily courses of meteorological parameters in 2004: a solar radiation (S r ) and air temperature (T a ); b wind speed (u) and vapour pressure deficit (D); c rain and irrigation (upside down bars). Dashed lines indicate the period when irrigation took place (17–22 July)

Full size image

Before irrigation was applied, soil moisture at the irrigation plot was higher than at the control plot (Fig. 3), which represents the micro-topographic differences that exist within the forest. After 6 days of irrigation (120 mm), soil moisture on the upper 10 cm layer increased 1.7 m3 m−3, while at the 20, 30, and 40 cm depth, the increase was 2.3, 3.2, and 4.6 m3 m−3, respectively. However, soil moisture at 60 and 80 cm showed the largest changes with 5.9 and 14.3 m3 m−3, respectively. Meanwhile, in the control plot during this period, especially in the upper soil layers, moisture continued decreasing while in the lower layers it remained unchanged. From 26 July to 4 August, after 26.1 mm of rainfall, soil moisture in the control plot increased in the upper soil layers (0–20 cm) while no variation was observed at lower layers (30–80 cm depth). Rainfall during this period does not appear to contribute to increase soil moisture in the irrigation plot since no variation was observed at any of the measured depths, and it can be inferred that this water was lost as runoff.

Fig. 3
figure 3

Soil moisture and soil thawing depth in the a control and b irrigation plots at 10, 20, 30, 40, 60, and 80 cm. The range of soil moisture fluctuation in each of the soil layers on the left vertical axis and the STD on the right one. Soil thawing depth on 31 August was 119 cm in the control plot and 125 cm in the irrigation plot

Full size image

On 1 July, the soil thawing depth (STD) was 55.0 and 57.0 cm in the control and irrigation plots, respectively. The first half of July was the warmest period of the growing season 2004, and accordingly the soil thawing rate was the highest (1.4 cm day−1). On the day irrigation began, the STD was 79.0 and 81.0 cm in the control and irrigation plots, respectively. After 6 days of irrigation, the STD increased by 5 cm (0.8 cm day−1) and 17 cm (2.8 cm day−1) in the control and irrigation plots, respectively, but from the time irrigation ceased and until the end of August, the soil thawing rate was 0.9 and 0.7 cm day−1 in the control and irrigation plots, respectively.

Scaled up transpiration (E c ) and canopy conductance (g c )

Before irrigation was applied (1–16 July), E c_cont and E c_irrig were on average 1.0 mm day−1 in each plot and the cumulative E c during this period was 15.9 and 15.4 mm in the control and irrigation plot, respectively. Maximum E c (1.2 mm day−1) was reached on 7 July during the warmer period of the growing season. After irrigation, maximum E c_irrig was 1.1 mm day−1 on 22 July while for the same day E c_cont was 0.7 mm day−1 (Fig. 4a). On the days following irrigation, the difference between E c_irrig and E c_cont oscillated between 0.1 and 0.4 mm day−1 (18–69% higher). In the control plot, soil moisture appeared to recover after a period of continuous rain from 26 July to 4 August (25.1 mm). This was further accentuated after rainfall events on 13–14 August (19.1 mm) that brought E c_irrig and E c_cont to a similar average value of approximately 0.7 mm day−1. In total, from 1 July to 31 August, E c_cont was 42.5 mm and E c_irrig was 46.4 mm, that is, a 9% difference, which did not correspond with the large differences in water input in the irrigation plot compared to the control plot.

Fig. 4
figure 4

a Daily courses of canopy transpiration (E c ) in the control and irrigation plots from 1 July to 31 August 2004. b Daily canopy conductance, calculated from sapflow measurements, in the control and irrigation plots. The dashed lines indicate the period when irrigation took place

Full size image

Maximum g c values in the morning were on average 2.3 and 2.4 mm s−1 in the irrigation and control plots, respectively before the irrigation started (Fig. 4b). From 19 to 22 July, maximum g c , 3.9 mm s−1 and 3.2 mm in the irrigation and control plots, respectively. According to Cermak et al. (2007), water crown storage capacity increased considerably during periods of high soil water content and low evaporative demand, which describe the conditions in the irrigation plot in our experiment. Thus, morning g c was driven by tree crown water storage that refilled in the previous evening (Schulze et al. 1985) under moisture saturated conditions. During irrigation and 1 week after it, g c_irrig remained higher than g c_cont; however, after a more modest rainfall event, by the end of July g c reached the same values and remained stable for the rest of the growing season. This also coincided with decreasing temperatures and lower VPD. The period (during irrigation and the short time that followed) when g c was higher, reflects the control exerted by moisture in the upper soil layer (Lopez et al. 2007a).

Water budget in the control and irrigation plot

In the first half of July, before irrigation was applied, rainfall was 2.9 mm. During this period, in the control plot, E a was 22.2 mm and the soil water stored in the 0–80 cm soil layer was 11.9 mm. Thus, during this period, 31.2 mm of newly available water in the forest ecosystem did not come from rainfall but from soil thawing layers. Similarly, in the irrigation plot, the contribution from thawed soils in the first half of July was 34.2 mm.

Rainfall during July–August was 63.4 mm. During this period in the control plot, rainfall contribution to soil water storage in the 0–80 cm layer was 34.5 mm while the contribution from soil thawing layers was 46.5 mm, reaching a total of 81.0 mm, of which 36.9 mm was used for E a_cont (61.5 mm). Rainfall contribution to E a_cont was 24.6 mm. By the end of August, the water stored in the 0–80 cm soil layer was +44.1 mm (Table 2) that remained in the soil for the next growing season. The amount of water that moved out of the 0–80 cm soil layer as runoff or to deeper layers was 4.3 mm (23–31 July).

Table 2 Water budget in the control plot
Full size table

In the irrigation plot, in addition to rainfall, 120 mm were irrigated, making a total of 183.4 mm for the period July–August. The total E a_irrig, during the period July–August was 71.8 mm. During this period, rainfall and irrigation contribution to soil water storage in the 0–80 cm layer was 52.2 mm while the contribution from soil thawing layers was 35.1 mm, reaching a total of 87.3 mm of which 37.3 mm was used for E a_irrig. Rainfall and irrigation contribution to E a_irrig was 34.5 mm. By the end of August, the water stored in the 0–80 cm soil layer was +41.1 mm (Table 3). The amount of water that moved out of the 0–80 cm soil layer as runoff or to deeper layers was 105.7 mm. Based on the timing of irrigation and precipitation as well as on the corresponding amount of water that left the 0–80 cm soil layer, calculations show that 60.4 mm of runoff came from irrigation and 45.2 mm from rain.

Table 3 Water budget in the irrigation plot
Full size table

In the control plot, the total amount of water available for the period July–August was 109.9 mm (rain + thawing soil layers) of which 56% was used for E a_cont, 40% remained in the 0-80 cm soil layer, and 4% moved downward to deeper layers or left as runoff. Meanwhile in the irrigation plot, the total amount of water input for the same period was 218.5 mm (rain + irrigation + thawing soil layers) of which 33% was used for E a_irrig, 19% remained in the soil layer, and 48% moved downward to deeper layers or left as runoff. However, if the runoff or the water that moved to deeper layers is subtracted from the total water input, the available water for the forest ecosystem was 112.9 mm, of which 64% was used for E a_irrig and 36% was stored in the 0–80 cm soil layer.

Discussion

Soil moisture and soil thawing depth

Fine-weather conditions at the beginning of July favored higher rates of transpiration from larch trees that depleted soil moisture from the upper soil layers. During this period, it was the 0–20 cm layer of soil that lost more water because of higher root density in this layer (Kuwada et al. 2002). Sudden soil moisture increases were observed at 60 and 80 cm depth in the control and irrigation plots in the first half of July, associated with the thawing of soil layers, which is a common characteristic in permafrost regions (Sugimoto et al. 2003; Iwahana et al. 2005; Lopez et al. 2007b). The moisture made available by soil thawing moves upward and becomes the source for tree water uptake, although this mechanism depends strongly on the evaporative demand (Lopez et al. 2008) and the timing of the rainless span within the growing season (Hamada et al. 2004; Lopez et al. 2007a). When irrigation took place, soil moisture at the 10 cm depth did not change sharply because of the predominantly organic matter composition of this layer and the coarse texture of the mineral soil below. Irrigation replenished soil moisture in the already thawed soil profile but after saturation was reached the water is assumed to move horizontally in the form of runoff as has been previously reported in the Siberian Taiga (Ohta et al. 2008). Soil saturation is inferred by the lack of change in moisture after continuous rainfall at the end of July.

When irrigation was applied, the heat inherent to water accelerated the soil thawing rate. However, it also caused earlier soil upward freezing in the irrigation plot because of the heat accumulation properties of saturated soil water conditions at the thawing front (Romanovsky et al. 1997; Hinkel et al. 2001). The latent heat of this saturated layer was responsible for keeping the freezing front at 0°C for a considerably longer period (Outcalt et al. 1990). A moisture-saturated layer at the boundary between the active layer and permafrost can have a positive feedback effect on permafrost stability in this region (Shur Yu 1998), since higher energy (higher air temperatures) is required to thaw this layer. The response of silty-clay soils, largely present in this region (Fedorov and Konstantinov 2003), to increases in rainfall will differ (Bernier et al. 2006; Lopez et al. 2007b), and this needs to be considered for modeling.

Canopy transpiration and canopy conductance

During the period of irrigation, E c_irrig increased between 18 and 69% depending on weather conditions. After rainfall restored soil moisture availability, E c_irrig remained higher than E c_cont, but gradually decreased until the difference was 16% on average at the beginning of August. By the end of August, the difference was 9%, suggesting that rainfall had replenished soil moisture enough to meet forest water demand, regardless of irrigation (Bergh et al. 1999; Albaugh et al. 2004). The relatively cool conditions of the growing season in 2004 probably reduced the evaporative demand effect that could have otherwise accentuated the difference in canopy transpiration between the two plots. Higher E c_irrig under non-limiting soil moisture conditions was driven by stomatal opening, which is enhanced when the upper soil layers are well supplied with moisture (Lopez et al. 2007a). Thus, the difference between E c_irrig and E c_cont represents the water that is not transpired by the forest ecosystem because of soil moisture deficit. The lower E c_cont values observed are probably because of the presence of ABA (abscisic acid) signaling from the roots (Roberts and Dumbroff 1986) or leaf water potential, although pre-dawn water potential is a poor indicator of water deficit (Cienciala et al. 1994).

Canopy conductance increased after irrigation reaching a morning maximum that was between 15 and 30% higher than g c_cont. Factors other than S r and D have been reported to affect morning g c in the Siberian larch forest but soil moisture was disregarded (Arneth et al. 1996), although this latter study was short-term. Ewers et al. (2001) reported increases in g c after fertilization and irrigation accompanied by increases in LAI (leaf area index). The relationship between g c and soil moisture was not clear when g c was estimated from E a obtained by the eddy correlation technique (Ohta et al. 2001), but as reported in Lopez et al. (2007a), there was a good response to soil moisture when g c was estimated from sap flow measurements since it better represented the local conditions. During irrigation, the upper soil layers remained well supplied with moisture which, combined with a high VPD, enhanced the stomatal opening. However, g c at each plot reached similar values with more modest precipitation and when the evaporative demand decreased.

Water budget

The differences in water input in the control and irrigation plots were not reflected in water loss as canopy transpiration (E c ) or in forest evapotranspiration (E a ). Under normal and (simulated) extreme rainfall regimes, the contribution from thawing soil layers to E a and the water stored in the soil at the end of the growing season in both plots was similar. Ohta et al. (2001) estimated that E a in 1998 was 73% of the total water availability (rainfall and snow water equivalent); however, water supplied from soil thawing layers was not counted and, as seen in this study, this accounted for almost 60% of E a in 2004. Even though irrigation water was responsible for the increase in E c (8.4%) as well as E a (14.3%) in the irrigation plot, the largest part of it moved to deeper soil layers or moved out of the plot soil as runoff. The lack of a strong dependency of forest E a on rainfall (and irrigation) in this experiment confirmed the low response in inter-annual variability in canopy transpiration and forest evapotranspiration to large variations in inter-annual rainfall (Yamazaki et al. 2004; Ohta et al. 2008; Park et al. 2008). Based on the differences in soil water uptake in each plot after irrigation, it is possible to assume that the distribution of rainfall within the growing season determines the rate at which forests uptake water from the soil. The increase in extreme rainfall events and lengthening of rainless periods as a result of global warming (Karl et al. 1995) can intensify the use of water stored in the soil. Another consequence of extreme rain events is that if they occur early in the growing season, runoff will be intensified because of shallow frozen soil layers (Lopez et al. 2007b). In order to obtain more reliable results from models of the Siberian Taiga, the total contribution of water from thawing soil layers must be included rather than considering only annual precipitation.

Conclusions

 

  1. 1.

    Despite large differences in water input into the control and irrigation plot, E c showed similar values (42.5 and 46.4 mm, respectively), during July–August 2004. This similarity can be explained by the contribution of water from stored soil moisture.

  2. 2.

    The effect of extreme rain on g c was limited to a short time within the growing season, reflecting the strong upper soil moisture and evaporative demand control.

  3. 3.

    The water budget showed that the actual water input in the control and irrigation plot was 109.9 and 218.5 mm, respectively. Regardless of this difference, water dynamics were similar in both plots. In the irrigation plot, 105.6 mm of water moved out of the plot as runoff or to deeper layers, leaving the forest ecosystem a water availability of 112.9 mm, roughly that in the control plot.