Interannual Variation in Transpiration Peak of a Hill Evergreen Forest in Northern Thailand in the Late Dry Season: Simulation of Evapotranspiration with a Soil-Plant-Air Continuum Model

soil hydraulic properties on forest transpiration using a newly developed soil–plant–air (SPAC) multilayer model. They found that a rooting depth of 4–5 m was needed to effectively simulate heat-pulse velocity variations corresponding to dry-season transpiration and annual discharge or stream flow. Moreover, a penetration test showed that the soil This book represents an overview of the direct measurement techniques of evapotranspiration with related applications to the water use optimization in the agricultural practice and to the ecosystems study. Different measuring techniques at leaf level (porometry), plant-level (sap-flow, lysimetry) and agro-ecosystem level (Surface Renewal, Eddy Covariance, Multi layer BREB), are presented with detailed explanations and examples. For the optimization of the water use in agriculture, detailed measurements on transpiration demands of crops and different cultivars, as well as results of different irrigation schemes and techniques (i.e. subsurface drip) in semi-arid areas for open-field, greenhouse and potted grown plants are presented. Aspects on ET of crops in saline environments, effects of ET on groundwater quality in xeric environments as well as the application of ET to climatic classification are also depicted. The book provides an excellent overview for both, researchers and student,s who intend to address these issues. following:


Introduction
Northern Thailand, which experiences rainy and dry seasons under an Asia monsoon climate, is characterized by hilly and mountainous landscapes. The rainfall tends to increase with altitude (Kuraji et al., 2001;Dairaku et al., 2004). Forests in northern Thailand at 1000 m above sea level (a.s.l.) are classified as lower montane rain forests (Santisuk, 1988). These areas receive high amounts of precipitation and provide a stable supply of high-quality water that is crucial for irrigation and drinking water (Bruijnzeel et al., 2011). Generally, water resources or stream flow are estimated by the difference between precipitation and evapotranspiration (i.e., the sum of canopy interception, soil evaporation, and transpiration). Thus, it is important to examine how forests consume rainwater as evapotranspiration, in conjunction with hydrological and meteorological variables. Such modeling is also essential for water resource management. This study is a continuation of previous studies of transpiration peaks in an evergreen forest in northern Thailand (18 o 48'N, 98 o 54'E, 1265-1450 m a.s.l.) in the late dry season (Tanaka et al., 2003(Tanaka et al., , 2004. Tanaka et al. (2003) concluded that transpiration in evergreen forests peaked in the late dry season. They suggested that reduced canopy wetness lowered evaporation; however, stomatal conductance declined only slightly, even under the driest conditions and highest net radiation. These results counter previous reports of an evapotranspiration decline in Thailand's dry season in evergreen forests (Pinker et al., 1980) and other vegetation (Toda et al., 2002). Tanaka et al. (2004) examined the impact of rooting depth and soil hydraulic properties on forest transpiration using a newly developed soil-plant-air (SPAC) multilayer model. They found that a rooting depth of 4-5 m was needed to effectively simulate heat-pulse velocity variations corresponding to dry-season transpiration and annual discharge or stream flow. Moreover, a penetration test showed that the soil stream flow were used as input and output data, respectively, and the data for 10 (Sep. in 2001-Jun. in 2002, 3 (Oct.-Dec. in 2002), 8 (Jan.-Aug. in 2004, and 12 months (Jan.-Dec. in 2005) around the missing data were used to assimilate the stream flow data. The heat pulse velocity corresponding to water use by an individual tree was monitored in three tree trunks (No. 1: Phoebe paniculata. Nos. 2 and 3: Lithocarpus elegans). The height and diameter of the three trees at 1.2 m were 28.0 m and 0.51 m, 23.0 m and 0. 29 m, and 15.5 m and 0.20 m, respectively, in 199929 m, and 15.5 m and 0.20 m, respectively, in -200529 m, and 15.5 m and 0.20 m, respectively, in (Tanaka et al., 2003. The observation of heat pulse velocity near a ridge, where no water table seemed to form, showed that the water use (or sap flow) of individual trees had a seasonal trend similar to those of the three trees (Tanaka et al., 2004). These trees belonged to the uppermost or second story. Therefore, the water use by these trees should reflect the transpiration over the forest as a whole (e.g., Kelliher et al., 1992;Tanaka et al., 2003) because transpiration from the upper layers is thought to contribute most of the total transpiration. Here, measured seasonal and interannual changes were used to validate the simulated transpiration.

A one-dimensional SPAC multilayer model for evapotranspiration
We used a one-dimensional SPAC multilayer model (Tanaka & Hashimoto, 2006) consisting of a soil multilayer model (Kondo & Xu, 1997) and a canopy multilayer model (Tanaka et al., 2003;Fig. 1). The soil multilayer model considers the variation in albedo and evaporation efficiency with changes in soil moisture at the top of the soil column (Kondo & Xu, 1997). The canopy multilayer model (Tanaka et al., 2003) for sensible and latent heat and CO 2 gas exchange consists of a second-order closure model for atmospheric diffusion coupled with a radiation transfer model (Tanaka, 2002), a rainfall interception model (Tanaka, 2002), a Farquhar-type photosynthesis model (Farquhar et al., 1980), and a stomatal conductance model (Ball, 1988). The rainfall interception model assumes that rainfall does not wet the lower sides of leaves with stomata, only the upper sides without stomata, while condensation wets both sides. In the photosynthesis model, the maximum potential rate of electron transport and dark respiration at 25°C (J MAX25 and R d25 ) were scaled to V cMAX at 25°C (V cMAX25 ); that is, J MAX25 = 2.14 V cMAX25 after Tanaka et al. (2002) and R d25 = 0.015 V cMAX25 after Collatz et al (1991). The assumptions, functions, and procedures in the calculations of the gross CO 2 assimilation rate A, dark respiration R d , and stomatal conductance g s were described by Tanaka et al. (2002). Combined, the two multilayer models by Kondo and Xu (1997) and Tanaka et al. (2003) consider the loss of soil moisture by water uptake (or transpiration) and the effect of soil water content on stomatal closure (Tanaka et al., 2004). The water uptake at depth z was assumed to be proportional to the ratio of the extractable to the entire extractable soilwater content (W e ; Tanaka et al., 2006). The sum of the water uptakes corresponds to the temporal canopy transpiration. When canopy transpiration can be supplied by the entire extractable soilwater content at 0-1m depths (see Case A in Fig. 1), where the major plant nutrients C, N, P, and K appear to be concentrated (Jackson et al., 2000), W e is calculated between the depths of 0-1m from which the water uptake is supplied. In the other case, W e is calculated as the extractable soilwater content between 1m and the maximum rooting depth Z ROOT (here, = 4 m) (see Case B in Fig. 1), and the water uptake is supplied from soil layers at 1m to Z ROOT . The water uptake at depth z, regardless of the vertical root distribution, is expressed as Fig. 1. A one-dimensional soil-plant-air continuum multilayer model for evapotranspiration (i.e., the sum of canopy transpiration E t , canopy interception E i , and soil evaporation E s ). Discharge D was calculated as the downward water flux passing through the rooting depth (4 m), and volumetric soil moisture at the bottom of the soil layer (9 m) was set to the saturated volumetric soil moisture θ s . E u and θ MIN are the water uptake by roots and soil moisture, respectively, at the lower limit of soil water potential ψ LL (-100 m), where trees cannot take up water. Case A shows water uptake when canopy transpiration can be www.intechopen.com supplied by the entire extractable soilwater content at 0-1m depths, while Case B shows water uptake in the other case. E u at depth z was assumed to be proportional to the ratio of the extractable soil water content (i.e., θ(z) -θ MIN ) to the entire extractable soil water content W e at 0-1m depths (Case A) or from 1m to the maximum rooting depth Z Root (= 4 m) (Case B) (Tanaka et al., 2006). The canopy height was set to 30 m (Tanaka et al., 2003).
Here θ MIN is the volumetric soil water content at which trees cannot take up water. This corresponds to the value at the upper limit of the soil water potential (ψ UL = -100 m). W e in Case A or B is expressed as This assumption of water uptake is simple compared to another frequently used weighting scheme (e.g., Dickinson et al., 1993;Desborough, 1997) based on the assumption that the root length density distribution is proportional to water extraction throughout the profile. Radersma and Ong (2004) did not find a clear relationship between root length density and water extraction. Other researchers have questioned the various proposed relationships between root length density and water uptake (Dardanelli et al., 2004). These findings suggest that the process of water uptake by roots is not entirely clear. Therefore, we used a simpler assumption. Stomatal conductance was assumed to decrease with the ratio R We of integrated extractable water content W e from the surface to the rooting depth (i.e., 0 (( where g sW is the stomatal conductance in well-watered soil and f(R We ) is a function of the ratio R We = W e /W es ranging from 0 to 1. The function f(R We ) was calculated as Equation (4), including the values of the slope and intercept, is close to the relationship between the extractable water content and stomatal conductance shown by Gollan et al. (1985).
In the canopy multilayer model, the evapotranspiration depends on the LAI, the slope m in Ball's stomatal conductance model (Ball, 1988), and V cmax at 25 o C (V cmax25 ) in a Farquhar-type photosynthesis model (Farquhar et al., 1980). These parameters are based on the estimated LAI and determined by referring to the measured net photosynthesis rate and stomatal conductance for single leaves (Tanaka et al., 2003). The values were set at 4.5, 10, and 25 μmol m -2 s -1 , respectively. The vertical profile of the LAI is also a required parameter. It was assumed to obey a beta distribution, with the greatest leaf area density at 0.7 times the canopy height (B-type canopy; see Figure 6 of Tanaka et al. 2003). Tanaka et al. (2003) investigated the impact of each parameter on evapotranspiration. Kondo and Xu (1997) verified the method by comparing observed and calculated results for four soil textures (i.e., volcanic ash, clay loam, silty sand, and sand). Silty sand was selected as the sub-watershed soil texture, whose observed relationship between volumetric soil water content θ and soil water potential ψ was close to that in the model (Tanaka et al., 2004). The soil and rooting depth were set at 9 and 4 m, respectively (Tanaka et al., 2004). Kondo and Xu (1997) and Tanaka et al. (2003) detailed the other parameter values used in the simulation. The canopy (height = 30 m) was divided into 50 layers. Each soil layer was 0.1 m thick. The time interval was set at 3 min in the soil multilayer model because of the thin soil layers (0.1 m), but it was set at 15 min in the canopy multilayer model. The model simulated soil evaporation E s , canopy interception (i.e., evaporation from a wet canopy) E i , transpiration E t , discharge D, and soil moisture. The profiles of all the meteorological elements were calculated repeatedly among all sub-models until the differences between the previous and new values of leaf temperature, air temperature, humidity, ambient CO 2 concentration, downward and upward longwave radiation, and water storage on both upper and lower sides of the leaves were within 1% (Tanaka, 2002). The maximum number of repetitions was set at 100 (Tanaka, 2002). Here, D was calculated as the downward water flux passing through the rooting depth (Fig. 1). The initial soil moisture condition at the beginning of 1998 calculated by Tanaka et al. (2004) was used here. The initial soil moisture condition was calculated repeatedly until it corresponded to the value at the end of 2000 in the study by Tanaka et al. (2004). This implies that the total rainfall was used as E t , E i , and E s , and discharged without changing into stored soil water between the beginning and end of the 3year period (1998)(1999)(2000). Soil moisture at the bottom of the soil layer (= 9 m; Fig. 1) was set to the saturated soil moisture θs. This initial condition did not consider the impact of the decrease in rainfall in the rainy season in 1997 caused by the 1997-1998 El Niño (Wang & Weisberg, 2000). The initial soil moisture appeared greater because of the impact of more rainfall in 2001. Heat pulse velocity was not monitored in the late dry season in 1998. Therefore, simulation results for 1999-2005 are shown here. The initial soil moisture at the beginning of 1999 was calculated using hydrometeorological variables in 1998.

Results
Figure 2 shows seasonal and interannual temporal variations in hydrometeorological variables in 1999-2005. The study area has three seasons in terms of air temperature and rainfall changes: a rainy season and early and late dry seasons (Tanaka et al., 2003). The light gray, gray, and black bars in Fig. 2a indicate the point 30 days before the day when the rainfall amount exceeded 150 mm (i.e., the wet period; WP) ( Fig. 2b), the days excluding those in the WP whose following 5 days had mean air temperatures below 21ºC (i.e., the cool dry period; CDP), and the days excluding those in the WP whose following 5 days had average values of air temperature over 21ºC (i.e., hot dry period; HDP), respectively. The horizontal bars in Fig. 2b show the points at which the 90 previous days had less than 50 mm of total rainfall (i.e., a drought condition; DC). The CDP was concentrated in the early dry season, while the HDP was concentrated in the late dry season. The HDP occasionally appeared in the early dry seasons, in much shorter periods than in the late dry seasons. The www.intechopen.com show the days for which the previous 90 days had less than 50 mm of total rainfall (i.e., drought conditions; DC). The horizontal bars in (e) show the days without measurements of stream flow. During the days with missing data, stream flows were estimated by data assimilation using a river flow model (Fukushima, 1988). The shaded areas in (e) correspond to the data assimilation periods. The solar elevation at noon and the day length were higher and longer, respectively, in the rainy season, but the frequent appearance of clouds modified the less intense solar radiation. The volumetric soil moisture values at 0.1 and 0.5 m were also lowest in the late dry season (Fig. 2d). The duration of the driest conditions increased with the DC period, particularly in 2004. The stream flow was never interrupted, even in the late dry season (Fig.  2e). This implies that the deeper soil portion was still moist even though the soil surface layer at depths of 0.1-0.5 m was dry. The peak stream flow appeared in the late rainy season or at the end of WP (e.g., September-November). The peak value was the largest in 2002 due to the rainfall amount. The horizontal bars in Fig. 2e show the days without stream flow measurements. For days with missing data, stream flows were complemented by data assimilation with a model for river flow (Fukushima, 1988). The shaded areas in Fig. 2e correspond to the data assimilation periods. The total measured and modeled stream flows were 3025 mm and 3040 mm, respectively, for 835 days during the periods with measured stream flow. The total amount of complemented stream flow was 498 mm for the 164 days with missing data, and the estimated error was extremely small (several millimeters).  were smaller in the CDP due to weaker atmospheric conditions (lower temperature, lower VPD, and less intensive solar radiation due to the decline in solar elevation), even though the soil was wetter. The simulated E s was small under humid conditions within a canopy due to both the lower VPD and the wetter soil, while E s increased in the HDP and/or the www.intechopen.com DC. Figure 3b shows the cumulative results from Fig. 3a. The annual amounts of simulated ET, E s , E i , and E t were 707, 49, 151, and 507 mm yr -1 , respectively, for the 7-year period. The closed canopy reduced E s , and almost all E i disappeared outside the rainy season. Thus, the percentages of E s , E i , and E t within ET were 7, 21, and 72% in the 7-year period, respectively. Assuming a negligible difference in the storage of soil moisture between the beginning of 1999 and the end of 2005, the annual ET was 694 mm yr -1 , the difference between rainfall (1881 mm yr -1 ) and stream flow (1187 mm yr -1 ). The ET value from the water budget was very close to the simulated ET. The error in stream flow estimated by data assimilation appeared to be negligible (several millimeters over 7 years), as the above-mentioned.  The horizontal bars in Fig. 4a indicate the duration of the measurement of heat pulse velocity for each tree with the same sensor and position. The values above the bars correspond to the heat pulse velocity (m s -1 ) at 18 mm per 5 days of transpiration. These changes depended on where and how deep the probe and sensor were placed in the trunks (Phillips et al., 1996;Takizawa et al., 1996;James et al., 2002). Probes and sensors were inserted in the trunks. The simulated E t captured the variation in heat pulse velocity. Both E t and the heat pulse velocities exhibited late-dry-season transpiration peaks in 1999, 2000, and particularly 2002. However, the peaks were smaller in 2003 and 2004 and almost the same as E t in the rainy season. Figure 5 shows the simulated mean±1 standard deviation values of daily transpiration during the HDP (see Fig. 4b

Discussion
The SPAC multilayer model (Tanaka & Hashimoto, 2006) was used to simulate the interannual variation in ET at a hill evergreen forest in northern Thailand between 1999 and 2005. The simulated annual ET was close to the difference between rainfall and stream flow (i.e., the ET from the water budget) during the 7-year period. The simulated transpiration E t captured the measured heat pulse velocity corresponding to water use by an individual tree, particularly the decrease in the late-dry-season E t in 2004 and 2005. The assumption that the decrease in extractable soil moisture had reduced impact on stomatal closure (i.e., equation 4) thus appeared to be appropriate for the estimation of transpiration at the forest canopy www.intechopen.com level. To confirm this assumption, the model should be applied to data for different vegetation under drought conditions. Tanaka et al. (2009) found that this assumption was reasonable for modeling deciduous teak plantations with shallower rooting depth under drought conditions in northern Thailand. Although the simulated canopy interception E i was not validated here, Tanaka et al. (2003) simulated the E i of the forest in 1998-1999 using a canopy multilayer model (Tanaka, 2002) but no soil multilayer model (Kondo & Xu, 1997). They compared the simulated E i with that estimated from the difference between rainfall and the sum of throughfall and stemflow and found that the annual values were close. Although the simulated soil evaporation E s was also not validated, E s likely occupied a small portion of the ET due to the decline in downward solar radiation and wind velocity because of the closed canopy, as in the simulation (Fig. 3; the ratio of E i to ET is annually 7%). Thus, this study is the first to reveal features of the interannual variation in ET (i.e., the sum of E s , E i , and E t ).
The simulated annual ET (= 707 mm yr -1 ) was smaller than that estimated from the water budget (= 694 mm yr -1 ) by 13 mm. The smaller LAI decreased E t and E i and increased E s (Tanaka et al., 2009). Smaller values of V cMAX25 or shallower rooting decreased E t and changed little both E i and E s within the model (Tanaka et al., 2004(Tanaka et al., , 2009). These impacts on ET were larger in the late dry season and the HDP, when atmospheric evaporation demand was stronger than in the rainy season because of the higher temperatures and VPD and the intensive solar radiation (Tanaka et al., 2003(Tanaka et al., , 2004. Therefore smaller values of V cMAX25 or shallower rooting depths are required for simulated ET to become closer to that estimated from measurements. Then the seasonal and interannual variation in ET would show slightly smaller E t values in the HDP, and neither E s nor E i would change significantly. In tuning with smaller LAI values, the decrease in E i and E t and increase in E s must be considered.
The simulated E t and monitored heat pulse velocity showed that the late-dry season E t peak was lower in 2004 and 2005 (Fig. 4). The rooting depth was set at 4 m. This is deeper than the 1 m depth often used in land surface models within general circulation models (Kleidon & Heimann, 1998). Tanaka et al. (2004) noted that, during dry periods, rainwater remaining from the previous rainy season may be sufficient for trees with greater water use capacities. The longest DC periods, however, as well as the longest dry season lengths in 2003-2004 and 2004-2005, largely limited E t in the late dry season beyond this larger water use capacity. Kume et al. (2007) compared heat pulse velocities (or sap flow) among two large trees corresponding to Nos. 1 and 2 in this study (Fig. 4b) and two smaller trees (4.8 m and 1.4 m height) in the study forest in 2003 and 2004. They found that the reduced impact of soil drought on sap flow was clearer in the smaller trees in the late dry season in 2004, likely due to their shallower rooting depths. They suggested that the larger trees might avoid water uptake limitations with their deeper roots. Furthermore, transpiration over the whole forest canopy could also have been limited in the late dry season in 2005 (Figs. 4b and5). The annual ET (= 694 mm) was small compared with values reported for other tropical and sub-tropical forests (e.g., Doley 1981); in the latter, values often exceed 1000 mm, with maxima of 1750 mm, and the ratio of annual discharge to annual rainfall exceeds that of the annual ET. Air temperature decreases with altitude, while rainfall tends to increase with altitude (Kuraji, 2001;Dairaku et al., 2004) in northern Thailand, and the downward solar www.intechopen.com radiation decreases due to the frequent appearance of clouds during the rainy season (Fig.  2c). Under the weaker evaporative demand in the rainy season, most rainwater probably infiltrates the soil layers with relatively smaller ET values. This rainwater is likely used by evergreen trees with deeper roots, even in the following dry season. Such trees can continuously transpire, control their leaf temperature, and assimilate carbon, although latedry-season transpiration peaks likely decreased in 2004 and 2005. Hence the trees can maintain leaves all year round. The population of deciduous trees increases as altitude decreases below 1000 m a.s.l. in northern Thailand (Santisuk, 1998). Tanaka et al. (2009) numerically simulated canopy net assimilation A n , ET, and soil moisture in a deciduous teak plantation with shallow rooting depth (< 1 m) in a dry tropical climate in northern Thailand (18˚25'N, 99˚43'; 380 m a.s.l.) using the SPAC multilayer model. That site had annual rainfall of 1361 mm for 2001-2008 and higher annual temperature of 25.4°C (K. . The first experiment in that study involved seasonally varying LAI estimates based on timeseries of radiative transmittance through the canopy, and the second experiment applied an annually constant LAI. The first simulation captured the measured seasonal changes in soil surface moisture; the simulated transpiration agreed with seasonal changes in heat pulse velocity, corresponding to the water use of individual trees. The simulated A n almost always became positive during leafy seasons. The simulated annual ET was ~900-1200 mm. However, in the second simulation in the dry season, A n and E t became negative and small, respectively, because the decline in stomatal conductance due to severe soil drought limited the assimilation. The simultaneous increase in leaf temperature increased dark respiration. These experiments revealed that leaflessness in the dry season is reasonable for carbon gain, and trees cannot maintain leaves year round at the site. Therefore, it may be more difficult (easier) for trees to maintain leaves in the late dry season as altitudes decrease (increase) in northern Thailand.

Conclusion
The E t simulated with the SPAC multilayer model and heat pulse velocities indicated that the late-dry-season transpiration peak weakened in 2004 and 2005, even with an assumed rooting depth of 4 m. The 2003The -2005 dry seasons were relatively longer, and they had the second longest (= 67 days) and longest (= 108 days) DC days. The soil moisture likely became insufficient beyond the rooting depth limitations on soil water use because of the duration of drought conditions along with the stronger atmospheric evaporative demand.

Acknowledgments
We are grateful to Nipon Tangtham  Aspects on ET of crops in saline environments, effects of ET on groundwater quality in xeric environments as well as the application of ET to climatic classification are also depicted. The book provides an excellent overview for both, researchers and student,s who intend to address these issues.

How to reference
In order to correctly reference this scholarly work, feel free to copy and paste the following: Tanaka K., Wakahara T., Shiraki K., Yoshifuji N. and Suzuki M. (2011). Interannual Variation in Transpiration Peak of a Hill Evergreen Forest in Northern Thailand in the Late Dry Season: Simulation of Evapotranspiration with a Soil-Plant-Air Continuum Model, Evapotranspiration -From Measurements to Agricultural and Environmental Applications, Dr. Giacomo Gerosa (Ed.), ISBN: 978-953-307-512-9, InTech, Available from: http://www.intechopen.com/books/evapotranspiration-from-measurements-to-agricultural-and-environmentalapplications/interannual-variation-in-transpiration-peak-of-a-hill-evergreen-forest-in-northern-thailand-in-the-l