國立臺灣大學生物資源暨農學院森林環境暨資源學系 碩士論文
School of Forestry and Resource Conservation College of Bioresources and Agriculture
National Taiwan University Master Thesis
台灣中部柳杉林之長期林分蒸散量推估:熱消散樹液流 法的校正及其於野外資料的應用
Long-term stand transpiration estimates in a Japanese cedar forest, central Taiwan: Calibration of thermal dissipation sap
flow measurements and its application to field data
蘇曼萍 Man-Ping Su
指導教授：久米朋宣 博士 Advisor: Tomonori Kume, Ph.D.
中華民國 106 年 8 月 August 2017
誌謝
碩士班兩年期間受到許多許多的幫忙才得以順利完成論文。感謝家人的支持，使我 可以無後顧之憂的專注於實驗上。感謝久米老師給予的教導與指正，從實驗進行到 論文撰寫，包括兩年期間課業上的幫助，研討會摘要及海報的修改，都非常感謝老 師的細心指教。感謝篠原慶規老師來台灣教導進行樹液流的校正實驗，提供了實驗 上非常大的幫助。感謝梁偉立老師提供許多建議，使能夠發覺更多實驗上的問題。
感謝台大實驗林提供溪頭通量塔及農業氣象站的氣象資料。感謝實驗林 Sophie 學 姊和彭小姐幫我們處理每次出差的住宿。感謝實驗林余瑞珠學姊，每次出差總是麻 煩學姊跟學姊借用儀器。感謝關秉宗老師借用生長錐以及阿坤學長幫忙才能得到 年輪寬度的資料。感謝鄭智馨老師借用電子秤及溫婕妤借用三角瓶使校正實驗能 順利進行。感謝梁小姐幫我們處理出差、儀器購買等的相關事宜。感謝實驗室的大 家，每個月一起開車到溪頭做實驗。感謝 sophie 學姊教導我如何進行樹液流實驗、
樣區維護工作以及解決探針的許多問題。感謝林松駿學長教我如何製作樹液流探 針以及樹液流量測方法。感謝林伯宣學長每次一起去溪頭幫忙做實驗，包括生長錐 的操作、校正實驗的進行等。感謝陳至威及林雋雅學姊幫忙進行溪頭野外的實驗以 及室內的校正實驗。感謝鐘敏華、江明珊、陳奕宏、邱均幫忙進行溪頭野外的實驗。
感謝實驗室的學姊們長久以來的努力，才能讓我有長年的資料可以進行分析觀察。
還要感謝許多人的幫忙與協助，讓我可以順利完成碩士論文!
中文摘要
為了解氣候變遷及森林經營對水循環的影響，長期林分蒸散量的推估是不可或缺 的。樹液流量測方法是推估林分蒸散量的方式之一，且其由 Granier 在 1985 年校 正的熱消散樹液流方法於現今廣為所用。然而，一些前人研究進行校正實驗發現並 非所有樹種都適用於 Granier 的經驗公式，且柳杉為台灣山區森林重要樹種之一，
但還無研究進行校正實驗測定 Granier 樹液流方式對量測柳杉樹液流的精確度。並 且甚少研究檢測推估林分蒸散時，樹木生長的影響。因此本研究目的為 1) 使用樹 幹片段進行校正實驗以確認 Granier 樹液流方法對量測中部台灣柳杉之樹液流的精 確度；2) 利用近 7 年野外量測之樹液流資料檢測真實邊材長度與樹木生長對林分 蒸散量推估的影響；3) 將校正實驗之結果應用於野外資料以改增進的林分蒸散量 推估的精確度；4) 研究長期林分蒸散及氣象資料之年間變異。校正實驗結果顯示 使用 Clearwater 公式修正可以大幅提升推估的精確度，因此真實邊材長度的量測 是很重要的。且校正實驗結果顯示 Granier 經驗公式推估的結果會低估真實流量 約 30%。由於真實邊材長度的量測十分重要，我們在野外樹液流量測樣區進行染 劑注射實驗，且量測樹木生長量，用以檢測邊材長度的改變對蒸散量推估的影響，
亦檢測由樹木生長所造成的邊材面積的改變對長期蒸散量推估的影響。在本研究 中，由樹木生長造成的邊材面積的改變與邊材長度的改變對每年林分蒸散量推估 的影響不大，因為 1) 在本研究期間樹木生長量小，及 2) 因染劑注射實驗修正之 邊材長度造成 2-4 公分之樹液流速變大，雖 2-4 公分樹液流速變大使推估量增加，
但其會與邊材面積減少所造成的推估降低互相抵銷。而為了得到可良好推估蒸散 的方法，我們嘗試將校正實驗之結果使用五種不同方法應用於野外資料。結果顯示 五種方式的差距出乎預期之大。藉由比較其各自的樹液流日變化曲線，以及比較五 種方式計算之年林分蒸散量與潛在蒸發量，得到年林分蒸散量推估的可能範圍大 致為原本推估方式的 1.5 至 3 倍。最後，本實驗推估近 7 年的林分蒸散量。林分蒸
散量的年間變異很小，因為林分蒸散量主要受水蒸氣壓虧缺及太陽輻射影響，而其 年間變異亦不大。溪頭的氣象狀況在過去近 7 年間似乎維持滿穩定的狀態。
關鍵字:校正、樹液流、蒸散
英文摘要
Long-term stand scale transpiration estimation is indispensable for understanding climate change impacts and the effects of forest management on the water cycle. Sap flow measurement is a robust approach for estimating stand transpiration; a thermal dissipation
method calibrated by Granier (1985) has already been widely implemented. However, some experimental sap flow calibration studies have claimed that Granier’s empirical
formula is not universally applicable. Japanese cedar is a dominant tree in the mountain forests of Taiwan; however, no experimental calibration studies have been conducted to determine the accuracy of the Granier sap flow method for this species. Additionally, the potential for incorporating tree growth effects on stand transpiration into the formula has not been examined. Therefore, this study aimed to 1) determine the accuracy of the Granier sap flow method on Japanese cedar in central Taiwan by conducting a calibration experiment using stem segments; 2) examine the effect of actual sapwood depth and tree growth on transpiration estimates based on near 7-year field measurements; 3) apply the results from the calibration experiment to field data to improve the accuracy of transpiration estimates; and 4) investigate inter-annual variation due to long-term transpiration and meteorological factors in a Japanese cedar plantation in Xitou, central Taiwan. The results of the calibration experiment showed that the application of Clearwater formula substantially improved the accuracy of transpiration estimates,
indicating the importance of accurate sapwood depth measurements. The calibration experiments showed that the Granier formula underestimated actual water uptake by approximately 30%. Thus, a dye injection experiment was conducted using samples from the field study site; tree growth was also measured to examine the effects of changes in sapwood depth and area due to tree growth on long-term transpiration estimates. These effects were small, because 1) there was little tree growth during the study period and 2) the 2–4 cm sap flow rate became high due to the correction of sapwood depth for dye injection. However, the increase in the inner sap flow rate was balanced by the decrease in sapwood area due to the sapwood depth correction. To determine the better method to estimate stand transpiration, the calibration results were applied to field data using five different methods. Contrary to expectations, there were distinct differences among the results produced by these methods; stand-scale transpiration, was found to be 1.5–3 times larger than that obtained using the original method. We then conducted stand transpiration estimates for the near 7-year field data. There was little inter-annual variation in stand transpiration, because stand-scale transpiration was mainly affected by vapor pressure deficit and solar radiation, which exhibited little inter-annual variation. Meteorological conditions in Xitou appear to have been stable over the past 7 years.
Keywords: calibration, sap flow, transpiration
目錄 Table of contents
口試委員會審定書………i
誌謝………...ii
中文摘要………..iii
英文摘要….………...v
Chapter 1 Introduction………...……6
1-1 Background………..……6
1-2 Meteorological factors for stand-transpiration……….…6
1-3 Transpiration in Taiwan………..………….…8
1-4 Sap flow measurement………..……….…..9
1-5 Calibration of Granier probe……….……….…11
1-6 The goals of this study……….…12
Chapter 2 Materials and methods……….…14
2-1 Sap flow measurement..……….………14
2-2 Calibration experiment………..………18
2-2-1 Samples and tree segment preparation………...……….…18
2-2-2 Process of calibration experiment………..……….…20
2-2-3 Sap flow measurement-sensor arrangement…………..……….…22
2-2-4 Determining of sapwood area and sapwood depth………..………...…23
2-3 Long-term measurement of sap flow and meteorological factors…….……..…24
2-3-1 Experiment site and samples………..….…24
2-3-2 Sapwood depth and sapwood area measurement………..……..…25
2-3-3 Biometric parameters measurement………...……27
2-3-4 Meteorological factors………...…….…27
2-3-5 Sap flow measurement-sensor arrangement………..……….…28
2-4 Data processing for long-term stand scale sap flow………..………29
2-4-1 Estimation of stand scale transpiration………..…………..…29
2-4-2 Effect of sapwood depth and growth on stand transpiration estimate…..30
2-4-3 Application of calibration experiment results to field sap flow data..…..31
Chapter 3 Results and discussions………...….…34
3-1 Calibration experiment………..…34
3-1-1 Sample segments………34
3-1-2 Effect of sensor number………..…36
3-1-3 Effect of segment height……….…36
3-1-4 Accuracy of thermal dissipation probe for Japanese cedar trees……….37
3-1-5 Comparison with other calibration studies………..…38
3-2 Estimation of long-term stand scale transpiration…..………....47
3-2-1 Measurement of sapwood depth………...…..…47
3-2-2 Effect of sapwood depth and radial variation on transpiration estimation……….48
3-2-3 Changed in Biometric parameters……....………...……50
3-2-4 Effect of tree growth……….…..…51
3-3 Application of indoor-calibration experiment results to field sap flow data...58
3-4 Inter-annual variations in stand transpiration……….………66
3-4-1 Meteorological factors and stand scale transpiration from Sep 2010 to Mar 2017...66
3-4-2 Relationship between transpiration and Meteorological factors……...68
3-4-3 Comparison of stand scale transpiration with other cloud forests….…..71
Chapter 4 Conclusions……….…80
Chapter 5 References………...…83
Appendix………..……...…90
圖目錄 Figure list
Fig. 1 Diagram of Granier thermal dissipation method……….……17 Fig. 2 Photo of tree segment preparation………..……19 Fig. 3 Diagram of calibration experiment construction used to test the accuracy of thermal dissipation methods-based sap flow……….………….…21 Fig. 4 Photo of calibration experiment construction used to test the accuracy of thermal dissipation methods-based sap flow….……….…21 Fig. 5 Monthly precipitation and temperature from Sep. 2010 to Mar. 2017………..…..24 Fig. 6 Picture of dye injection experiment…..………..26 Fig. 7 Concept figure of formula used in 5 methods from calibration experiment………33 Fig. 8 Response of sap flow densities to pump pressure changes in segment No.1…...…40 Fig. 9 Sapwood area for every 12 segment at 1 cm (a), 5cm (b) and 10 cm (c), and figures of sapwood area determining through GIMP (d)…….……….……41.42 Fig. 10 Response of K to pump pressure changes in each segment..….………43 Fig. 11 The effect of sensor number to the accuracy of estimation from Granier sensor...44 Fig. 12 Comparison of sap flow densities measured by Granier probes and calculated by real water uptakes in segment No. 4-1, 4-2, 4-3, 5-1, 5-2, 5-3, 6-1, 6-2, 6-3.……44 Fig. 13 Comparison of sap flow densities calculated by real water uptakes and measured by Granier probes without applying corrected formula from Clearwater in all sample segments……….………45 Fig. 14 Comparison of sap flow densities calculated by real water uptakes and measured by Granier probes with applying corrected formula from Clearwater in all sample segments………..………45 Fig. 15 Comparison of sap flow densities calculated by real water uptakes and measured by Granier probes with applying corrected formula from Clearwater in all sample segments (sap flow density ranged from 0 to 50 cm3m-2s-1)..…………..……….46 Fig.16 Difference of sapwood depth between 2010 and 2016 for each tree in east and west sides and averaged sapwood depth……….52 Fig. 17 Regression of two kinds of sapwood depth………..………52 Fig. 18 Comparison of sap flow density between 0-2 cm and 2-4 cm calculated without Clearwater formula in trees No. 5 and No. 7.………..…….…………53 Fig. 19 Comparison of sap flow density between 0-2 cm and 2-4 cm calculated with Clearwater formula in trees No. 5 and No. 7………...……..54 Fig. 20 Accumulation of monthly stand scale transpiration in 2015. Orange curve is the
“original” one; purple dotted line with circle is the one “consider new sapwood depth”; while green dotted line with square is that consider new sapwood depth
but just change 2-4 cm sap flow rate; blue dotted line with triangle is that consider new sapwood depth but just change sapwood area………...…55 Fig. 21 DBH growth for each tree from 2010 to 2016………..……….55 Fig. 22 Accumulation of monthly stand scale transpiration in 2015. Blue curve is the
“consider tree growth” one. While orange curve is the “original” one……...57 Fig. 23 Comparison of sap flow densities calculated by real water uptakes and measured by Granier probes with applying corrected formula from Clearwater in all sample segments (sap flow density ranged from 0 to 50 cm3m-2s-1). The black solid line is y=ax in method 1. The black dotted line is y=ax+b in method 2, 3 and 4. The short black dotted line is y=ax line in method 3……….………..63 Fig. 24 Regression of K and sap flow density calculated by real water uptakes of 12 segments (method 5) (sap flow density ranged from 0 to 50 cm3m-2s1)………....63 Fig. 25 Daily curve of sap flow density of north side in tree No. 9 from 2015/1/11 to 2015/1/20 that applied 5 methods from calibration experiment…………...……64 Fig. 26 Stand scale transpiration applied with 5 methods and potential evaporations in 2015…...…65 Fig. 27 Comparisons of sap flow densities calculated by real water uptakes (Fd_actual), sap flow densities measured by Granier probes with applying corrected formula from Clearwater (Fd_Granier_CW) and Fd_Granier_CW with three formula that derived from calibration experiment (a): y=ax; (b): y=ax+b; (c): y=azb in all sample segments (Fd_Granier_CW ranged from 0 to 50 cm3m-2s-1)………64 Fig. 28 Meteorological factors (daily) such as rainfall (RF), soil water content (SWC), air temperature (T), solar radiation (S) and vapor pressure deficit (VPD) and stand scale transpiration (Et) (daily) from Sep. 2010 to Mar. 2017…………...74 Fig. 29 Relationship between transpiration and VPD for near 7 years………75 Fig. 30 Relationship between transpiration and S for near 7 years……….75 Fig. 31 Relationship between transpiration and VPD for each season in near 7 years…...76 Fig. 32 Relationship between transpiration and S for each season in near 7 years………77 Fig. 33 Relationship between VPD and transpiration under the condition: (a) SWC <
25%, (b) SWC > 25%. And regression lines of (a) and (b) were shown in (c)…..78 Fig. 34 Relationship between VPD and transpiration under the condition: (a) T < winter average T, (b) T > winter average T. And regression lines of (a) and (b) were shown in (c) ………..……..78
表目錄 Table list
Table 1 Information of all twelve Japanese cedar tree segments………...40 Table 2 Comparisons of calibration experiment with other studies…………...………...46 Table 3 Original sapwood depth and new sapwood depth that calculated from the formula of regression line in Fig. 17.………..53 Table 4 Sapwood area calculated by new sapwood depth (Table 3) and original DBH (obtained in 2010) ………..……..………54 Table 5 DBH for each tree from 2010 to 2016……….………...56 Table 6 Sapwood area calculated by DBH (Table 5) and sapwood depth obtained in 2010 (Table 3)…..……….………..………... 56 Table 7 Comparisons of sap flow measurement based stand transpiration in cloud forest………...78 Table 8 Annual transpiration and meteorological factors such as precipitation (P), solar radiation (S), vapor pressure deficit (VPD), air temperature (T), soil water content (SWC) and ratio between E and P from previous studies and this study………..………….79
Chapter 1 Introduction
1-1 Background
Water is indispensable to creatures also for human beings, hence understanding the mechanism of hydrological cycle is helpful for human life. Evapotranspiration plays a significant part in hydrological cycle and also in climate system, which is a pass that water can move from land or ocean to atmosphere. Evapotranspiration contains two elements, which one is evaporation from land or ocean, and the other one is transpiration from plants.
Transpiration may change according to plant communities (Gerten et al., 2004, Sterling et al., 2013). In recent years, terrestrial evapotranspiration has a decreasing trend probably
due to climate change (Jung et al., 2010). Understanding forest transpiration is helpful for predicting potential changes in water cycling in response to climate changes. Also, the accurate estimation of forest transpiration can help us to understand impacts of deforestation or forest management on water cycle.
1-2 Meteorological factors for stand-transpiration
In hydrological cycle, transpiration is one of the processes that can be influenced by plant (Gerten et al., 2004), which are not only directly regulated by plant physiology, but also indirectly environmental meteorological factors. Therefore, several factors may affect transpiration, such as tree species, tree vigor, soil water content, solar radiation,
precipitation, air temperature, humidity, fog and vapor pressure deficit (VPD) (saturated
vapor pressure – ambient vapor pressure) (Granier, 1987; Granier et al., 1996; Komatsu et al., 2010; Klimešová et al., 2013; Pataki & Oren, 2003; Shen et al., 2015; Wullschleger
et al., 2001).
Generally, under the condition of drought, plants face of drought stress, so
transpiration mechanism from plants are different from that in usual and less affected by other meteorological factors (Granier, 1987; Klimešová et al., 2013; Wullschleger et al.,
2001). While water supply is sufficient, daily and seasonally patterns of plant transpiration are mainly regulated by solar radiation and VPD (Wullschleger et al., 2001).
Although precipitation might increase transpiration, there was no statistically significant relationship between them (Shen et al., 2015). Fog also may affect tree transpiration in three ways. One is that fog lets leaf surface become wet, makes VPD decrease and may cover some stoma so that transpiration decrease (Lin et al., 2015; Misson et al., 2002;
Ritter et al., 2009). Second is that leaf may absorb water from saturated atmosphere or from wet surface of leaf, which probably moderates water stress (Burgess & Dawson, 2004). The third is that fog as a horizontal precipitation which may provide additional input of water so that supplies water source to plants for transpiration (Dawson, 1998).
1-3 Transpiration in Taiwan
In Taiwan, there is about 60% of area covered with forest (林務局第四次全國森林 資 源 調 查 ， 2017), therefore understanding forest transpiration characteristics is
indispensable for realizing forest water cycle. Since forests in Taiwan are mostly located at mountainous area and elevated from 500 to 3,100 m, every place has its own microclimate depending on different meteorological factors. Also, Taiwan is situated between subtropical and tropical region, in which there are widely distributed montane cloud forests which show high biodiversity and contain more water resource (徐嘉君, 2015). In some previous studies in Taiwan, although transpiration was affected by VPD, radiation, and fog (林祐竹, 2011; 陳俐如, 2005; 蔡孜奕, 2013; 羅勻謙, 2004; Chen,
2013; Laplace, 2013), the main factor was different because of different microclimate.
陳俐如 (2005) claimed that transpiration was affected by local circulation induced
specific climate condition such as fog. Other research suggested that fog could affect transpiration (林佑竹, 2011), but fog showed significant daily variation and it may cause different effects to transpiration. All-day fog with rain made transpiration maintain lower, but afternoon fog with rain could not make transpiration become very low. Instead, if fog with rain was sufficient in the afternoon, it could make transpiration become higher because of water supply (林佑竹, 2011).
Xitou located in central Taiwan, fog duration in a year reached up to about 2,000 hours and always appeared in the afternoon and disappeared after sunset (Wey et al., 2011), so plant transpiration might be affected by fog. Also, VPD is related to fog, because VPD may decrease while fog formation (Burgess & Dawson, 2004; Ritter et al., 2009). Consequently, transpiration may be affected by fog through the changes in VPD in Xitou (Chen, 2013; Laplace, 2013). The formation of fog is related to environmental conditions, yet how climate change affects the formation of fog and the structure of cloudy forest is still unknown. Therefore, long-term investigation of transpiration and meteorological factors are needed for solving this problem.
1-4 Sap flow measurement
Various measurement techniques for evapotranspiration have been developed in the last 50 years, such as sap flow measurement, eddy covariance techniques and catchment water balance methods (Wilson et al., 2001). Each method has its own advantages and shortcomings. They are different from the spatial and time scale. Sap flow measurement
is the method that can estimate individual tree transpiration. Therefore, using sap flow measurement can help understanding different tree species’ transpiration characteristics.
Moreover, sap flow measurement is also an ideal method to understand how plants react to environmental conditions, managements (thinning, pruning and harvest), and climate
changes (Smith & Allen, 1996). Many kinds of sap flow measurement methods were developed. Among these methods, thermal dissipation method has been widely used because of its low cost and easily installed (Smith & Allen, 1996; Steppe et al., 2010).
Sap flow measurement just measures sap flow of one individual tree, so there are uncertainties using this approach to estimate stand scale transpiration. From sensor to whole tree transpiration, azimuthal variation, radial variation and sapwood area determining variation can affect the estimation of whole tree transpiration (Clearwater et al., 1999; Chiu et al., 2016; Delzon et al., 2004; Kume et al., 2012; Nadezhdina et al.,
2002; Shinohara et al., 2013; Tsuruta et al., 2010). Azimuthal variation may be caused from tree crown fraction or anisotropic radiation. Radial variation results from wood anatomy, that is, new vessel and tracheid have better ability to transport water. From tree to stand scale transpiration, tree to tree variation, sampling strategy and scaling approach may affect the estimation of stand scale transpiration (Köstner et al., 1998; Kume et al., 2010; Lu et al., 2004). According to different plant species and different environment conditions, tree to tree variation is different everywhere. Thus, sampling strategy is different, too. For long-term transpiration estimation, some research have investigated about that. Clausnitzer et al. (2011) and Oishi et al. (2008) published 4-7 years sap flow experiments, they consider tree growth while calculating from sap flow rate to whole tree
transpiration. They found a relationship between sapwood area and DBH, but they did not show how much tree growth affects the amount of transpiration.
Variations mentioned above must be considered to ensure the accurate estimation of stand transpiration. In study site Xitou, Tseng (2011) has measured azimuthal variations of Japanese cedar in four directions and measured radial variations in different depth of xylem. The results implied that the azimuthal variations in this stand less impacted on transpiration estimation than the radial variations. To estimate long-term near 7 years stand transpirations, careful determination of sapwood depth, considering radial variation and considering tree growth are indispensable.
1-5 Calibration of Granier probe
In addition to the variations that mentioned above, an uncertainty in estimating stand transpiration based on the sap flow method still exists. The thermal dissipation probe was previously calibrated by Dr. Granier in 1985, and an empirical formula has been widely used (Granier, 1987). Although some studies using this thermal dissipation probe found that the original empirical formula could produce real sap flow densities (Braun & Schmid, 1999; Clearwater et al., 1999; Do & Rocheteau, 2002; Granier et al. 1996; Lu et al. 2002), some tree species materials in previous studies were not suitable for this empirical formula due to difference of wood characteristics among tree species (Bush et al., 2010;
Niu et al., 2015; Steppe et al., 2010; Sun et al., 2012). Therefore, recently calibration experiment has been recommended for estimation of stand transpiration based on thermal dissipation method for each species. In recent years, calibration experiments were conducted for testing the accuracy of thermal dissipation sensor, and some investigated the effect of different wood characteristic, while some investigated the effect of different structure of thermal dissipation sensor (Ayutthaya et al., 2009; Hultine et al., 2010; Hölttä et al., 2015; Paudel et al., 2013). By absorbing or pushing water through tree trunk or
branch segment, they measured the amount of water that passes through segment, and compared it with sap flow measurements to recalculate their own parameters of the original formula. Generally, sap flow densities calculated by original Granier empirical formula were underestimated, though it were overestimated in some cases (Sun et al., 2012). Therefore, calibration experiment is necessary, which can substantially improve the accuracy of the Granier probe.
1-6 The goals of this study
Japanese cedar is one of the dominant species in mountainous area in Taiwan.
Quantifying the transpiration of Japanese cedar trees is indispensable to understand water cycling there. There are still few studies have done long-term forest transpiration measurements in Taiwan, therefore long-term investigation of transpiration and
meteorological factors are needed for understanding forest transpiration characteristics including inter-annual variation. For the purpose of estimating long-term stand transpiration based on Granier method, the variations derived from scaling, tree growth and the applicability of this method to sample tree species should be examined. No one tested the accuracy of thermal dissipation methods-based sap flow for Japanese cedar trees in Taiwan. Thus, the goals of this study are:
1) To determine the accuracy of thermal dissipation method-based sap flow measurement for Japanese cedar trees in Taiwan using indoor calibration experiments.
2) To provide a method that can estimate long-term stand scale transpiration in terms of potential error from sapwood depth and tree growth to sapwood area.
3) To obtain stand scale transpiration estimation with the consideration of potential error resulting from different application ways of the indoor calibration results.
4) To investigate inter-annual variation of stand scale transpiration of Japanese cedar and meteorological factors such as radiation, soil water content, air temperature, precipitation and VPD.
Chapter 2 Materials and methods
2-1 Sap flow measurement
Thermal dissipation method was adopted in this study for field long-term sap flow measurement and for indoor calibration experiment on Japanese cedar (Cryptomeria japonica). The thermal dissipation sensors used in this study were handmade which has
the same specification with the sensor in Granier (1987). One thermal dissipation sensor set consists of two probes which have copper-constantan thermocouple junction each other to sensor temperature; one is named heater probe that contains a heating and a temperature-sensing device, and the other one is called referenced probe that contains only a temperature-sensing device. These two probes are about 2 cm long. Before inserting sensors, the bark of the position which sensor will be inserted should be removed to reveal sapwood, and to ensure that the temperature which measured by sensor represented the average temperature of the 2cm long probe in wood. These two probes are inserted into xylem of trees, which one in upper side is heater probe and that in lower side is referenced probe (Fig. 1). Heater probes are provided 0.2W constant heating energy from electricity. Through the temperature difference between the two probes, sap
flow density can be calculated by the empirical formula from Granier (1987). The Granier’s empirical formula was used to calculate the sap flow density (cm3cm-2s-1):
Sap flow density (cm3m−2s−1)= 119 × K1.231 (1)
K =∆TM− ∆T
∆T (2)
where K is the dimensionless flow index; ∆𝑇𝑀 is the value ∆𝑇 obtained under zero sap flow condition; ∆𝑇 is the difference of temperature between heater and reference probe.
Small temperature difference represents that sap flows quickly in xylem, and vice versa.
Thermal dissipation sensors were connected to data logger which recorded every 30 seconds and calculated mean temperature difference every 30 minutes from the field data.
For calibration experiment, data logger recorded every 1 second and calculated mean temperature difference every 10 seconds. Because the sap flow density calculation formula of the thermal dissipation method required the temperature difference under zero sap flow condition, for field data, ∆𝑇𝑀 was different every day depending on the highest
∆𝑇 in each day. For calibration experiment, after installing sensors, started providing
heat and recorded temperature difference without water uptake till the difference of temperature became stable, and took the highest ∆𝑇 as the ∆𝑇𝑀 for each segment.
If the sapwood depth was less than 2 cm, a correction formula from Clearwater would be applied to recalculate sap flow density to reduce the effective of inactive xylem.
Because thermal dissipation method assumed that probes integrate temperature and sap
flow density along the probe length (2cm) (Lu et al., 2004), the ideal case is when
sapwood depth equal to the length of the probes. But, when sapwood depth was smaller
than 2 cm, the sap flow density which sensor measured always underestimated true mean
sap flow density. It is because that when some part of the probe is in inactive sapwood,
the temperature integrated in the probe will be higher, and then the difference of
temperature becomes higher to make the sap flow density underestimated. Therefore,
Clearwater et al. (1999) provided a method to calculate sap flow density that just in active
xylem to deal with this problem of underestimation when active sapwood depth is below
2 cm:
∆T = a∆TSW + b∆TM ,
∆TSW = ∆T − b∆TM
a (3)
This formula considered that the temperature which probe measured contains two parts,
one is the temperature measured in sapwood called ∆TSW, and the other is the
temperature measured in inactive xylem called ∆TM. While a is the proportion of probe
in sapwood and b is the proportion of probe in the inactive xylem. Under the condition
that sapwood depth is known and lower than 2 cm, corrected sap flow density for the
active xylem can be calculated by replacing ∆𝑇 in formula 2 with ∆𝑇𝑆𝑊 in formula 3.
However, because 𝑎 and 𝑏 are the proportions of sapwood and inactive xylem
respectively, under the condition that the value of 𝑏 higher than 0.5 (i.e. the width of
inactive xylem is higher than 1.0 cm), the value of ∆𝑇 may be lower than the value of 𝑏∆𝑇𝑀 or just a little bit higher than it. Therefore, the value of ∆𝑇𝑆𝑊 may be lower than 0 or may be higher than 0 but close to 0. If the value of ∆𝑇𝑆𝑊 is lower than 0, the value
of 𝐾 will also be lower than 0, so that sap flow density cannot be calculated. On the
other hand, if the value of ∆𝑇𝑆𝑊 is higher than 0 and close to 0, the value of 𝐾 will be
very high, so that sap flow density will be extremely high which is distinct from that in
usual. Consequently, to prevent the abnormal phenomenon, since the value of 𝑏 is higher
than 0.5, the correct formula 3 from Clearwater et al. (1999) will not be adopted in this
study.
Fig. 1 Diagram of Granier thermal dissipation method.
2-2 Calibration experiment
2-2-1 Samples and tree segment preparation
Six sample trees of Japanese cedar (Cryptomeria japonica) were harvested from Xitou tract, experimental forest of National Taiwan university. The stem diameter at breast high of these six samples were ranged from 15 to 20 cm. Twelve stem segments were cut from these six sample trees. Three sample trees were cut on 6 March 2016, and the other three trees were cut on 27 August 2016. In the first three sample trees, one segment was cut out from each tree (i.e., total three segments). In the last three sample trees, three segments were cut out from each tree (i.e., total nine segments). The height of segments should be lower than under branch height, and the north side and upside of the segment were recorded on each segment in the field. The tree height and diameter at breast height of each tree were measured and recorded. Also, the height of segments was recorded for the last three trees (9 segments). Segments were about 50 to 60 cm long and were covered with wet towels in two sides to avoid stem dehydration. Finally, segments were brought to laboratory to do experiments.
Before doing calibration experiment, following sample preparation must be carefully done as it had greatly impact on the success of the experiment. Each stem segment was recut in both sides in laboratory before doing experiments to assure conductivity of the segments. The bark on the up side of segments was removed about 3-5 cm strip to make
sure that water passed segments only through xylem; also, removing bark can make attachment much easier to stick on the segment because of its smooth surface. Then, adhesive and silicone were applied to the inside of the attachment (a plastic cylinder with a cover) and the position which bark was removed from the segment. The attachment was putted on the tree segment and adjusted to fit the tree segment to tightly bounded avoiding leaking water or air. After the adhesive and silicone were dry, the segment was prepared to do a calibration experiment (Fig. 2).
Fig. 2 Photo of tree segment preparation.
2-2-2 Process of calibration experiment
The calibration experiment (Fig. 3, Fig. 4) using twelve Japanese cedar stem segments were conducted as follow:
1) Scaffold (75cm high) were set up for hanging the tree segments.
2) A flask was equipped with two plastic tubes, which one was connected to pump and the other one was connected to attachment that stick on tree segment.
3) Tree segment was hung on the scaffold; a bucket with water was placed under a tree segment and the bottom side of the segment (about 4-5 cm) was under water.
4) Thermal dissipation probes were inserted into the tree segments to collected data.
5) By adjusting the pump pressure, water was pumped from bottom to top; also, constant water flow rates at different level were generated to test the accuracy of thermal dissipation methods-based sap flow under different sap flow rates.
6) A volumetric cylinder which was equipped with one plastic tube that connected to the bucket under tree segment was set up. Based on the principle of Pascal, the volume of water which took out from tree segment could be measured from the cylinder.
7) After the sap flow rate became stable under each pump pressure, the volume of water took out from tree segment was recorded every 1 minute in a 10 minutes period or every 30 seconds in a 5 minute period. Then the measurements with the values derived from Granier probe were compared with it during the period.
8) Safranin stain solution was used to dye the tree segment to get the active sapwood area.
After dying, tree segment about 1cm, 5cm and 10cm far from the bottom side were cut and photos were taken. Sapwood area and sapwood depth data were got from image processing using the photo which was about 1cm far from bottom side of each segment.
Fig. 3 Diagram of calibration experiment construction used to test the accuracy of thermal dissipation methods-based sap flow.
Fig. 4 Photo of calibration experiment construction used to test the accuracy of thermal dissipation methods-based sap flow.
2-2-3 Sap flow measurement-sensor arrangement
In the calibration experiment, four sets of thermal dissipation sensors were inserted into four directions (north, east, south and west) for each stem segment. This was in order to prevent that the position where sensor inserted was a nod or without sap flowing, and
to get more representative data of thermal dissipation sap flow for each segment. Finally, the sap flow rate calculated by Granier’s empirical formula (formula 1 and 2) in four
directions were averaged, and this averaged sap flow density represented the value that measured by Granier sensor for each segment, which was compared with sap flow density calculated by real water uptake.
As the calibration experiment compared sap flow densities that calculated from thermal dissipation sensor and from real water uptake, there were two kinds of sap flow densities were calculated. One was measured from thermal dissipation sensor and the
other one was measured from the volumetric cylinder measurement system. For thermal dissipation sensor, the Granier’s empirical formula (formula 1 and 2) was used to
calculate sap flow density (cm3cm-2s-1). If the sapwood depth was less than 2 cm, a correction formula from Clearwater (formula 3) would be applied to recalculate. For volumetric cylinder measurement, the sap flow density (cm3m-2s-1) was calculated by the formula:
Sap flow density (cm3m−2s−1) = V
A × T (4)
where V is the volume (cm3) of water uptake in the measurement period; A is the sapwood area (m2) of the segment; and T is the time (s) of the measurement period.
2-2-4 Determining of sapwood area and sapwood depth
In order to determine active xylem area of each segment, 0.1 % safranin stain solution was used to dye each segment after calibration experiment for about 1 hour. After dying tree segments, about 1 cm long at the bottom side of segment was cut as a disk, in which appearance of active xylem can be identified. Then a photo was taken for image processing to get sapwood area and sapwood depth in four azimuths. Sapwood area and depth were calculated by image analysis software Image J and GNU Image Manipulation Program (GIMP). Sapwood depth for each azimuth was averaged from five point measurements. One point was at the azimuth, two points were 0.5 cm apart from the azimuth, and two points were 1.0 cm apart from the azimuth.
2-3 Long-term measurement of sap flow and meteorological factors
2-3-1 Experiment site and samples
A long-term sap flow measurements plot is located at Xitou, which is situated in Nantou in central Taiwan. The area of our plot is 20*20 m (400 m2). In Xitou, the average annual temperature is about 16.6 ⁰C, and the average annual rainfall is about 2,600 mm (Wey et al., 2011). The monthly precipitation and average temperature from Sep. 2010 to Mar. 2017 were shown in Figure 5, and air temperature and rainfall were high in summer and low in winter.
In our plot, sap flow measurements for Japanese cedar (Cryptomeria japonica) trees have been conducted since August 2010. In this study, the period of sap flow data used to estimate stand transpiration was from 1 September 2010 to 31 March 2017, totally 6
Fig. 5 Monthly precipitation and temperature from Sep. 2010 to Mar. 2017.
years plus 7 months. In this long-term period, the number of sample trees with sensors and the number of trees in the plot were not constant. In the period from August 2010 to 23 March 2012, sample size for sap flow measurements was 19 trees. While in the period 24 March 2012 to 31 March 2017, the sample size was 15 trees. On the other hand, for the number of trees in the plot, from August 2010 to 31 December 2014, there were total 25 trees; from 1 January 2015 to 31 June 2016, there were 24 trees; and from 1 July 2016 to 31 March 2017, there were 23 trees. The more details was shown in Tseng (2011).
2-3-2 Sapwood depth and sapwood area measurement
The sapwood depth of trees at breast height on east and west sides were measured in June 2010 by using increment borer and then determined sapwood depth visually. In August 2016, the sapwood depth of trees at breast height on east and west sides were measured again.
However, to improve the accuracy of stand transpiration estimation, the sapwood depth for each tree should be measured more carefully. To obtain sapwood depth accurately, dye injection method could be used (Gebauer et al., 2008; Meinzer et al., 2001). Therefore, the dye injection method was adopted in this study to measure sapwood depth in 17 trees covering sap flow measurement samples. The dye injection experiment in this study was conducted as follow (Fig. 6). First, a cup was fastened on tree stem and
filled with water. Second, a hole was drilled which diameter was 8 mm into tree xylem under water. Third, water was replaced with 0.1 % safranin stain solution. Finally, about 3-5 hours later, increment borer was used to get increment core above the hole about 1 to 2 cm, and then sapwood depth which the portion had been dyed was measured; also, the sapwood depth visually determined was recorded. Sapwood depth measured by dye experiment was compared with that determined by eye, and the relation between these two was established in this study. This relationship was applied to sapwood depth which was determined by eye in June 2010, to convert to the more accurate sapwood depth based on dye experiment. Sapwood area estimation was performed based on the accurate sapwood depth:
Sapwood area (m2) = π[(DBH 2 )
2
− (DBH
2 − sapwood depth)
2
] (5)
Fig. 6 Picture of dye injection experiment.
2-3-3 Biometric parameters measurement
In order to understand the basic information of all trees in the plot, tree height and diameter in the breast height were measured in June 2010. For long-term tree growth measurements, in August 2016, the increment cores of trees at breast height on east and west sides were obtained by using an increment borer. The width of growth rings of each increment core in each year for 7 years was measured under a magnifier. Then the width of growth rings in east and west side were averaged, and the double of the values represented the diameters of tree growth in each year.
2-3-4 Meteorological factors
Several meteorological factors were used in this study, which were provided from the Experimental Forest of National Taiwan University. Environment data contain solar radiation, relative humidity, air temperature, precipitation and soil water content (weighted averaged from 5 cm, 20cm and 50 cm depth under soil), which have been continuously measured by Xitou flux tower and Xitou agricultural weather station. Lack of air temperature data were filled with other data which use the relationship between these two places data. The period of data was from 1 September 2010 to 31 March 2017.
Vapor pressure deficit was calculated from air temperature and relative humidity.
2-3-5 Sap flow measurement-sensor arrangement
From 10 August 2010 to 23 March 2012, there were 19 trees that inserted sensors.
In these 19 trees, there were three kinds of sensor arrangement. One is that 5 sensors were inserted in each tree, another one is that 4 sensors were inserted in each tree, and the other one is that 2 sensors were inserted in each tree. For the first kind, there were totally 6 trees. Among these 5 sets of sensors, 4 sets were inserted into four directions (north, east, south and west) and in the deep of 0-2 cm, while 1 set was inserted into the deep of 2-4 cm. For the second kind, there were 2 trees only. Four sensors were inserted in four directions and were in the deep of 0-2 cm. For the third kind, there were 11 trees. Two sensors were inserted in the directions of east and west with the deep of 0-2 cm. The more detail information was shown in Tseng (2011).
From 24 March 2012 to 31 March 2017, there were 15 trees for sap flow measurements. Two sets of sensors were inserted into tree xylem in the directions of north and south at the depth of 0-2cm in the 15 trees.
For stand scale transpiration estimation in this study, data of 15 trees with two sets of sensors from Sep 2010 to Mar 2017 were used. For radial variation examination, data of 2 trees with five sets of sensors from 10 Aug 2010 to 10 Feb 2011 were used.
2-4 Data processing for long-term stand scale sap flow
2-4-1 Estimation of stand scale transpiration
Transpiration from one tree can be scaled up to stand scale (Clausnitzer et al., 2011;
Chiu et al., 2016; Kume et al., 2010; Oishi et al., 2008; Shinohara et al., 2013). The formula used in this study was:
Et (mm/day) = JsAs_stand AG (6) Js(cm3m−2s−1) = ∑ni=1Qi
∑ni=1As_treei (7) Asstand(m2) = ∑ As_treei
x
i=1
(8)
Qi(cm3/s) = (Fd0−2× As0−2) + (Fd2−4× As2−4) (9)
where Js(cm3m−2s−1) is the mean stand sap flow rate; As_stand(m2)is the sum of
sapwood areas of all trees in sample plot; AG(m2) is the area of the plot, n is the
number of sample trees with sap flow measurements; As_tree is the sapwood area of each
tree; x is the total number of trees in the plot. Fd0−2 is sap flow density in 0-2cm;
Fd2−4 is sap flow density in 2-4cm; As0−2 is sapwood area in 0-2cm; As2−4 is sapwood area in 2-4cm.
The data (0-2cm and 2-4cm sap flow density) in the period (from 8/10/2010 to
2/10/2012) were used to calculate the ratio between 0-2cm and 2-4cm, and then this ratio
was used to estimate 2-4cm sap flow density from 0-2cm sap flow density for all data.
2-4-2 Effect of sapwood depth and growth on stand transpiration estimates
In order to know the change of sapwood depth during about 6 years, in August 2016, the sapwood depth of trees at breast height on east and west sides were measured again, and sapwood depths were determined by eye. Besides, after conducting dye injection experiment, sapwood depth determined by dye were compared with that determined by eye, and the regression line between them was used to correct all sapwood depth determined by eye in 2010. Also, the radial variation was examined by comparing sap flow rate in 0-2 cm and 2-4 cm with the corrected sapwood depth and Clearwater formula (formula 3); the ratio of sap flow rate between 0-2 cm and 2-4 cm was calculated.
In order to estimate long-term stand transpiration, this study examined effect of tree growth on stand transpiration. To investigate tree growth, the measurement of growth ring width was conducted in August 2016 using increment cores, which were derived from east and west side of the individuals. The DBH change for each year was double of growth ring width (i.e. average of east and west), so DBH in each year can be calculated by accumulating DBH change to DBH which was measured in 2010. Then, sapwood area for each year was estimated according to the formula 5.
To identify the effect of sapwood depth and tree growth to stand transpiration estimation, sap flow data in 2015 was used. Here, three kinds of estimations were tested, one was “original”, another was “consider new sapwood depth” and the other was
“consider tree growth”. The “original” one did not consider corrected sapwood depth and
ratio of sap flow rate between 0-2 cm and 2-4 cm obtained in this study. The sapwood depth determined by eye in 2010 was used to calculate sapwood area and to determine the ratio of sap flow rate between 0-2 cm and 2-4 cm without Clearwater formula (formula
3). Also, the DBH measured in 2010 was used to calculated sapwood area. On the other hand, the “consider new sapwood depth” one used corrected sapwood depth and ratio of
sap flow rate between 0-2 cm and 2-4 cm based on the corrected sapwood depth with Clearwater formula (formula 3) to stand transpiration estimation, but still used DBH measured from 2010. Last, the “consider tree growth” one did not consider corrected sapwood depth and ratio of sap flow rate between 0-2 cm and 2-4 cm obtained based on it; DBH in 2015, which was calculated by accumulating DBH change to DBH that was measured in 2010, was used to calculate sapwood area. By comparing yearly stand scale transpiration estimated from “original” and the other two methods, the effect of sapwood depth and tree growth on stand transpiration can be identified.
2-4-3 Application of calibration experiment results to field sap flow data
To improve the accuracy of stand scale transpiration estimation, the results of calibration experiment were applied to the long-term field data to correct the sap flow
density measured by thermal dissipation sensor. Five methods were used in this study.
Here, sap flow data in 2015 was used to examine applicability of the five methods.
The concept figure of five methods was shown in figure 7. Y axis was considered as the sap flow density calculated by real water uptake, and X axis was considered as the sap flow density calculated by original Granier’s empirical formula (formula 1). First method, y=ax (blue line in Fig. 7) was used for all field data. Second method, y=ax+b (orange line in Fig. 7) was used for all field data. Third method, both y=ax+b (orange line in Fig. 7) and y=ax (grey line in Fig. 7) were used. y=ax+b (orange line in Fig. 7) corrected field data which were higher than the lowest value of y that calibration experiment had tested. While, y=ax (grey line in Fig. 7) was used when field data was lower than the lowest value of y that calibration experiment had tested. Forth method, y=ax+b (orange line in Fig. 7) corrected field data which were higher than the lowest value of y that calibration experiment was conducted. On the other hand, for field data which were lower than the lowest value of y that calibration experiment was conducted, there was no correction. Fifth method, based on calibration experiment, new parameters were calculated to correct the formula 1, so I used a new formula to calculate sap flow.
After correcting field data, this study examined which one had better correction for sap flow density estimation. According to the diurnal variations in sap flow and comparisons with potential evaporation (Thornthwaite and Hamon, which calculation
was same with Alkaeed et al., 2006). Also, these five methods also applied to data from calibration experiment to identify which one had better correction for sap flow density estimation. Through above, the better methods for estimating the most likely stand transpiration was examined. Here, the estimation of transpiration in 2015 considered new sapwood depth and tree growth.
Fig. 7 Concept figure of formula used in 5 methods from calibration experiment.
Fd_actual represents the sap flow rate which was calculated by real water uptake.
Fd_Granier_CW represents the sap flow rate which was calculated by Granier empirical formula and Clearwater formula.
Fd_actual (cm3m-2s-1) Fd_Granier_CW (cm3 m-2 s-1 )
Chapter 3 Results and discussions
3-1 Calibration experiment
3-1-1 Sample segments
The information of all twelve segment samples was shown in Table 1. The age of trees was 37 years old for tree No.1, 2 and 3; and was 35 years old for tree No. 4, 5 and 6. Diameter at breast height of all segments ranged from 15.5 to 19.8 cm, and tree height ranged from 26.5 to 19.3 m. Height of segments were not recorded for No.1, No.2 and No.3 segments, while it were recorded in another nine segments. The diameter of segment for the first three segments were smaller than other nine segments, it was about 12 cm.
For another nine segments, diameter of segment ranged from 12 to 15 cm. For 12 segments, sapwood depths in 7 segments were less than 2 cm. Sapwood depth in No. 3 was the smallest, and sapwood depths in three azimuths (east, south and west) were smaller than 1 cm. For those sapwood depth less than 2 cm, Clearwater formula (Eq. 3) was applied to avoid underestimating. However, if sapwood depth was less than 1 cm (No. 3 E, S, W; No. 5-1 S; No. 5-2 W; No. 5-3 S; No. 6-1 S), the sap flow density corrected by Clearwater formula will be extremely high, so this data would not be used in this study.
Response of sap flow density to pump pressure changes in segment No. 1 was shown in Fig. 8. When pump pressure changed, sap flow density changed obviously, but it took about 1 hour to reach a stable state. The thermal dissipation method underestimated the
corresponding real water uptake, while after applying corrected formula from Clearwater, the estimation of thermal dissipation probe was closer to real water uptake.
The photos of sapwood area after dye experiment for each segment and response of K (Eq. 2) to pump pressure changes in each segment were shown in Figure 9 and Figure 10. We can see that in most cases, the position of dyed area in the tree segments in 1 cm, 5cm and 10cm did not change a lot except for segment No. 1. So that we used the averaged sapwood depths from each azimuth, which was calculated from five points measurements.
One point was at azimuth, two points were 0.5 cm apart from azimuth and the other two points were 1.0 cm apart from azimuth. For the response of K to pump pressure changes in all segments (Fig. 10), it showed that there was azimuth variation in these 6 trees, but sap flow rates did not be highest in only one azimuth. From the dyed sapwood area in 1 cm, 5cm and 10cm, we could identify the azimuth with higher sap flow density, and it was mostly agree with that determining from the figure of response of K to pump pressure changes. We also could find that sapwood depth was not the reason that cause underestimated under the low sap flow rate condition. Because although sapwood depth was about 2 cm, the accuracy of the Granier sensor was low under the low sap flow rate condition.
3-1-2 Effect of sensor number
In calibration experiment, four sets of Granier sensors were used for each tree segment in order to improve the accuracy of real sap flow density estimation. To confirm potential errors due to the number of azimuthal direction, we calculated the ratio of sap flow density calculated by Granier and Clearwater formula (Fd_Granier_CW) and sap flow density calculated by real water uptake (Fd_actual) for the four combinations (1, 2, 3 and 4 sensors) (Fig. 11). The results showed that the estimations from just one set of sensor and from two sets of sensors might overestimate or underestimate real sap flow density significantly. While sensor number increased, the ratio of sap flow density calculated by Granier and Clearwater formula to real water uptake got closer to 1. That is to say, more sensor number could lead to more accurate estimation.
3-1-3 Effect of segment height
To understand whether different height of segments in the same tree may have similar results or not, tree No.4, 5 and 6 were cut into three segments for each tree. If different height of segment had dissimilar results, all sample segments should cut at sensor installed height. While if different height of segment had similar results, sample segments can be taken from anywhere under the first living branch height.
Therefore, sap flow densities measured by Granier probes (Fd_Granier) and calculated by water uptakes (Fd_actual) in tree No.4, 5 and 6 were compared (Fig. 12).
In figure 12, three segments of each tree in No.4, 5 and 6 showed similar trends. Height of segments has little impact on relationship between Fd_actual and Fd_Granier.
However, it was obvious that different trees had their own trend. So, the effect of different trees was larger than different height of segments. To sum up, different height of segments might have similar results, so all twelve sample segments in this study could have high representative in each whole sample tree.
3-1-4 Accuracy of thermal dissipation probe for Japanese cedar trees
Sap flow densities calculated by real water uptakes (Fd_actual) and measured by thermal dissipation probe (Fd_Granier) were compared in all sample segments (Fig. 13).
In Figure 13, the value of Fd_Granier was calculated only by Granier empirical formula without applying the corrected formula from Clearwater. So, thermal dissipation probes in most of sample segments were underestimated because the sapwood depths were shorter than 2 cm. Under this condition, Granier probes underestimated about 40% of the real water uptakes.
Fd_actual and measured by thermal dissipation probe with applying the corrected formula from Clearwater (Fd_Granier_CW) were compared in all sample segments (Fig.
14). With the corrected formula from Clearwater, Fd_Granier_CW in most of sample segments corresponded to Fd_actual accurately. Under this condition, Granier probes could estimate the real water uptakes accurately less than 10% error. This result implied that the original Granier formula may be suitable for Japanese cedar, and that sapwood depth significantly impacted on the accuracy of thermal dissipation method; hence, careful determination of sapwood depth was the key for the transpiration estimates.
On the other hand, in long term field sap flow density data, we found that the sap flow density of Japanese cedar trees in Taiwan ranged from 0 to 50 (cm3m-2s-1). Under the condition, Fd_Granier_CW may underestimate about 30% of Fd_actual (Fig. 15). This result suggested that the original Granier formula may not be suitable for Japanese cedar trees with the sap flow rate ranging from 0 to 50 (cm3m-2s-1) in central Taiwan.
3-1-5. Comparison with other calibration studies
Some studies showed that thermal dissipation original formula was not suitable for some tree species, and thermal dissipation methods in most cases were underestimated real sap flow rate though some cases were overestimated (Table 2). Previous studies that shown in Table 2 all used Granier 2 cm probe to conduct calibration experiment.
The accuracy of Granier estimation was low under low sap flow rate, while the accuracy of Granier estimation was relatively high when considerating both high and low
sap flow rate (Gutierrez & Santiago, 2006; Sun et al., 2012; Wiedemann et al., 2016), which was same with this study. Although the measurement errors was high when sap flow rate was high in some cases (Bush et al., 2010, Steppe et al., 2010; Sun et al., 2012).
After applying Clearwater formula (formula 3), the substantial improvement in accuracy of estimation could be found in Bush et al. (2010) and Paudel et al. (2013), which was the same with this study. However, some results showed that the application of Clearwater formula did not approve the accuracy (Sun et al., 2012).
In these previous studies, sample materials were derived from branches or stem segments, the range of sap flow rate may be affected by the origin of the materials. For branches, the data in low sap flow rate could be obtained, but for stem segments the data in low sap flow rate were rarely found (Table 2).
This study showed that the Fd_Granier_CW underestimated Fd_actual, probably due to xylem anatomy but not due to sapwood depth, pump pressure (see appendix 1), and azimuthal variations in Fd. Consequently, this study suggested that simple calibration experiment can approve the accuracy of Granier probe, and it is recommended to conduct calibration experiment for each species. The Granier probes and original formula were mostly suitable for softwood species, although it were not suitable for some softwood species (Table 2; Bush et al., 2010).
Fig. 8 Response of sap flow densities to pump pressure changes in segment No.1. The red line represents the period that recorded real water uptake through volumetric cylinder and also represents sap flow density which calculated from real water uptake.
Table 1 Information of all twelve Japanese cedar tree segments.
Segment DBH Diameter of segment
Tree height
sapwood area
Height of segment
No. (cm) (cm) (m) N E S W (cm2) (cm)
1 15.5 12.0 17.8 1.3 1.7 2.1 1.9 1.8 ± 0.4 43.9 - 2016/3/11
2 15.8 12.0 19.0 1.5 1.2 1.3 1.9 1.5 ± 0.3 41.0 - 2016/3/8
3 16.8 12.0 16.6 1.3 0.3 0.5 0.9 0.8 ± 0.4 23.6 - 2016/3/13
4-1 14.6 1.8 1.5 1.8 1.9 1.8 ± 0.2 69.9 3.6 2016/9/1
4-2 16.8 13.5 16.8 2.0 1.9 1.9 2.1 2.0 ± 0.1 76.6 5.1 2016/9/2
4-3 12.3 2.0 1.8 2.4 2.0 2.0 ± 0.2 67.6 6.2 2016/9/6
5-1 14.7 1.6 1.8 0.1 1.6 1.3 ± 0.8 50.2 2.0 2016/9/4
5-2 16.1 14.0 16.5 2.3 2.4 1.4 0.8 1.7 ± 0.7 56.9 2.7 2016/9/5
5-3 13.7 2.4 1.8 1.0 2.2 1.9 ± 0.6 55.7 3.5 2016/9/7
6-1 14.5 2.7 2.0 1.0 2.8 2.1 ± 0.8 87.9 5.2 2016/9/4
6-2 19.8 14.2 19.3 2.3 2.2 1.9 2.1 2.1 ± 0.2 79.6 5.9 2016/9/6
6-3 13.9 2.3 2.2 2.0 1.5 2.0 ± 0.3 75.8 8.5 2016/9/8
Sapwood depth (cm)
average
Experiment Date
No. 1
No. 2
No. 3
No. 4-1
No. 4-2
No. 4-3
Fig. 9 Sapwood area for every 12 segment at 1 cm (a), 5cm (b) and 10 cm (c), and figures of sapwood area determined through GIMP (d).
No. 5-1
No. 5-2
No. 5-3
No. 6-1
No. 6-2
No. 6-3
No. 1
No. 2
No. 3
No. 4-1
No. 4-2
No. 4-3
No. 5-1
No. 5-2
No. 5-3
No. 6-1
No. 6-2
No. 6-3
Fig. 10 Response of K to pump pressure changes in each segment. Measured period represents the measured time of volumetric cylinder.
Fig. 11 The effect of sensor number to the accuracy of estimation from Granier sensor.
Fig. 12 Comparison of sap flow densities measured by Granier probes and calculated by real water uptakes in segment No. 4-1, 4-2, 4-3, 5-1, 5-2, 5-3, 6-1, 6-2, 6-3. Every dot is averaged from four azimuths in each segment, and the black solid line is one by one line.
