Xiang Gao · Zhenyu Du · Qingshan Yang · Jinsong Zhang · Yongtao Li ·Xiaojie Wang · Fengxue Gu · Weiping Hao · Zekun Yang · Dexi Liu ·Jianmin Chu

Abstract Plantations of woody tree species play a crucial role in ecological security along coastal zones.Understanding energy partitioning and evapotranspiration can reveal land–atmosphere interaction processes.We investigated energy fluxes, evapotranspiration, and their related biophysical factors using eddy covariance techniques in a black locust ( Robinia pseudoacacia L.) plantation in 2016,2018, and 2019 on the Yellow River Delta.Downward longwave radiation offsets 84–85% of upward longwave radiation; upward shortwave radiation accounted for 12–13% of downward shortwave radiation.The ratio of net radiation to downward radiation was 18–19% over the three years.During the growing season, latent heat flux was the largest component of net radiation; during the dormant season, the sensible heat flux was the dominant component of net radiation.The seasonal variation in daily evapotranspiration was mainly controlled by net radiation, air temperature, vapor pressure deficit, and leaf area index.Black locust phenology influenced daily evapotranspiration variations, and evapotranspiration was greater under sea winds than under land winds because soil water content at 10-cm depth was greater under sea winds during the day.Seasonal patterns of daily evaporative fraction, Bowen ratio, crop coefficient, Priestley–Taylor coefficient, surface conductance, and decoupling coefficient were mainly controlled by leaf area index.The threshold value of daily surface conductance was approximately 8 mm s -1 over the plantation.

Keywords Black locust plantation·Yellow River Delta ·Eddy covariance·Energy partitioning·Evapotranspiration

Introduction

The quantification of energy exchange between the land surface and the atmosphere is needed for accurate weather prediction and for climate modeling, since land surface thermodynamic and hydrological processes are closely coupled to atmospheric circulation (Ward et al.2014; Odongo et al.2016; Valayamkunnath et al.2018).Studies of surface radiation balance and energy partitioning are promoted by the development of global climate change sciences (Chen et al.2016).The partitioning of net radiation (Rn) into sensible heat flux (H), latent heat flux (Le), soil heat flux (G), and heat storage in vegetation plays a critical role in turbulence flow structure and thermodynamic process of the boundary layer and directly influences the Earth’s climate at local, regional,and global scales (Chen et al.2016; Odongo et al.2016;You et al.2017).Climate change in turn, affects a variety of physical and eco-physiological processes in vegetation and alters energy exchange between the land and the atmosphere(Zhu et al.2014).Previous studies have demonstrated that surface albedo is lower in forests than in croplands, grasslands, and deserts, and therefore, forests absorb more solar radiation (Zhu et al.2014; Gao et al.2018; Yue et al.2019).Forests cover approximately 30% of the Earth’s land surface(Iida et al.2009) and are distributed from tropical to cold temperate zones.Energy partitioning varies with forest type(Betts 2000; Liu et al.2018).For example, the ratio ofLetoRnwas 0.15–0.17 for a young plantation in north China during 2012–2017 (Ma et al.2019), whereas it was 0.570.60 for a sub-alpine forest on the Tibetan Plateau in 2014 and 2015 (Yan et al.2017).Therefore, a better understanding of the energy partitioning in different forest types under various environmental conditions is required to describe the complex interactions between the terrestrial biosphere and the atmosphere as well as water circulation and global climate change (Zhu et al.2014).

Evapotranspiration (Et, i.e.,Lein an energy unit) is the amount of moisture released into the atmosphere from the land surfaces, and is the sum of evaporation from bare soil and wet vegetation canopy and transpiration by plants(Valayamkunnath et al.2018; Iida et al.2020).Etis about 60% of the annual global precipitation from the land surface (Jung et al.2010; Odongo et al.2016), and is the link between the hydrological cycle and energy exchange processes.Leaf stomata regulate transpiration and photosynthesis, andEtis closely coupled to carbon uptake at an ecosystem scale (Aires et al.2008; Jia et al.2016).Variations ofEtare controlled by meteorological factors,vegetation properties and soil water conditions (Gao et al.2018; Valayamkunnath et al.2018; Yue et al.2019).Xiao et al.( 2013) reported annualEtvariations from 150.5 to 957.8 mm across six forest sites in China.In recent years,reforestation rates have been rapidly increasing, and so far plantations, critical for ecological protection, account for about 33% of forestland in China (Chen et al.2014; Ma et al.2019), and has dramatically shown the importance of vegetation in afforested/reforested areas.Therefore, further information of the biophysical controls onEtis required for proper assessment of carbon sequestration and other ecosystem processes in plantations.

The Yellow River Delta on China’s north coastline covers an area of approximately 5.40 × 103km2, and provides habitats on the East Asian-Australasian Flyway (Liu et al.2020).It has a population over two million and supports the largest petroleum industry in China (Zhou et al.2015).The Delta, significantly affected by land–ocean interactions(Kong et al.2015), is a typical ecologically vulnerable area with a low water table (average depth is 1.14 m) and a vast saline-alkali soil area (2.40 × 103km2) (Li et al.2019).Due to human intervention and climate change, the Delta undergoes intensive conversion from natural wetlands to crops and plantations (Chi et al.2018).Studies have shown that plantations lower the water table through large amounts of transpiration and control soil salination by restraining water from deeper layers to the soil surface, and finally promote region sustainable development (Minhas et al.2020).Therefore, plantations improve the ecological environment and play a key role in maintaining ecological security.This needs better understanding through the investigation of land surface processes as impacted by this pivotal ecosystem in the Yellow River Delta.Revealing the maintenance and control mechanisms of energy partitioning andEtby plantations on the Yellow River Delta could improve our understanding of land–atmosphere interactions in the coastal zone.

Black locust (Robinia pseudoacaciaL.) is a source of nectar and a fast-growing timber species native to North America (Jiao et al.2019).More importantly, it is one of the most ecologically beneficial species due to its role in erosion reduction, soil improvement, and carbon sequestration, and is widely used in ecological projects in northern China (Jiao et al.2018).As black locust can be grown on mild saline and alkaline soils (Ma et al.2013), it is a major species for afforestation to improve saline-alkali soils.Black locust plantations cover 8000 ha in the Yellow River Delta (Cao et al.2012).Therefore, it is important to investigate water vapor and energy fluxes and factors that influence them.

Over the past two decades, the eddy covariance (EC)technique has become a standard method for measuring the exchange of water vapor and energy between the land surface and atmosphere (Gao et al.2017).In this study, 3-year energy and water vapor fluxes observed by the EC technique and related biophysical data were used.The objectives were to: (1) characterize energy fluxes andEtvariations on diurnal and seasonal scales; (2) assess the effects of environmental drivers on energy partitioning andEt; and, (3) determine the surface parameters characterizing energy partitioning andEt.

Materials and methods

Study site

Study site is located at the Forest Ecosystem Research Station of the Yellow River Delta in Shandong (37°54′2′′N, 118°49′2′′ E, 3.40 m a.s.l.), Hekou district, Dongying city (Fig.1).It is flat and experiences a warm-temperate,continental monsoon climate with a mean annual temperature of 12.30 °C and a mean annual frost-free period of 210 days.Mean annual precipitation is 574.40 mm with approximately 70% occurring in summer, and the mean annual evaporation is 1962.10 mm (Zhang and Xing 2009).The soil in this site is a Fluvisol (Wang et al.2017a), developed on mixed loess and alluvium.The volumetric water content at field capacity and wilting humidity are 0.28 cm3cm–3and 0.10 cm3cm–3, respectively,with a soil bulk density of 1.45 g cm-3.The soil is a mild saline-alkali with a salt content of 1.53 g kg-1and pH of 8.55.The plantation is 35 years old with density of 581 stems ha–1, average diameter at breast height of 18.6 cm,and canopy height of approximately 14.5 m.The understory covers approximately 80% of the site and is mainly Bermuda grass [Cynodon dactylon(L.) Pers.].

Fig.1 a Location of the study site on the Yellow River Delta and b the observation tower

Measurements

A 25-m tower was constructed in the plantation for mounting turbulent fluxes and meteorological instruments.The details of these instruments are provided in Table S1.All sensors on the tower were connected to a CR1000 datalogger (Campbell Scientific Inc., Logan, UT, USA).Data from 2016, 2018,and 2019 were used.

The distance of the tower to nearest boundary of the plantation was approximately 450 m.Daily precipitation(P) data were acquired from the Yellow River Delta Ecological Research Station of Coastal Wetland, Chinese Academy of Sciences, and groundwater depth (Gd) from a nearby national groundwater depth monitoring station ( http:// xxzx.mwr.gov.cn/).

The period without leaves in a year was the dormant season(DS), and the rest of the year was the growing season (GS).Based on a previous study (Yuan et al.2014), the growing season was divided into three stages, early growing stage (EG);mid growing stage (MG); and, late growing stage (LG).The MG was defined as the period of maximum leaf area to leaves beginning senescence naturally.Twenty black locust trees were randomly selected for measurement of phenological stages and the date recorded at which 50% or more of the selected trees reached each phenological stage.The start and end dates of each phenological stage is shown in Table 1.The plantation was irrigated once annually at the beginning of the GS.

Table 1 Start date and end date of each phenological stage in the growing season in the black locust plantation

According to downward shortwave radiation below/above the canopy (Sdb/Sda), leaf area index (Lai) was calculated by inverting Beer’s law equation:

where κ is the extinction coefficient oflight attenuation(0.54, Li et al.2018).

Data processing

The post-processing of turbulent fluxes contained the procedures for quality control and correction.Quality control of data included basic tests, statistical tests, and tests on fulfillment of theoretical requirements (Foken et al.2004).The correction of turbulent flux data included the traditional coordination rotation (McMillen 1988), sonic temperature correction(Schotanus et al.1983), and density fluctuations correction(Webb et al.1980).Because of unacceptable low quality data,instrument malfunction, or unfavorable weather conditions,18% of all data were rejected.For missing flux data, short gaps (≤ 2 h) were filled using a linear interpolation, and long gaps (＞ 2 h) using the mean diurnal variation (MDV) method described by Falge et al.( 2001).

According to the measurements from net radiation sensors above the canopy, net longwave/shortwave radiation (Ln/Sn),and net radiation (Rn) were expressed as follows:

whereLd/LuandSd/Surepresent downward/upward longwave radiation and downward/upward shortwave radiation,respectively.

Evapotranspiration (Et) was expressed as follow:

where λ denotes the latent heat of the vaporization of water(2.45 kJ g-1, Zhang et al.2016), andLeis the latent heat flux.

The Bowen ratio (β) and evaporative fraction (Ef) were expressed as follows:

The crop coefficient (Kc) was provided by Allen et al.( 1998)as:

whereEt0is reference evapotranspiration,Tasignifies air temperature (°C),Urepresents wind speed (m s-1),Vpdstands for the vapor pressure deficit (kPa),Δis the slope of the water vapor pressure curve (kPa °C-1), andγsignifies the psychrometric constant (kPa °C-1).

The Priestley–Taylor coefficient (α) was given by Priestley and Taylor ( 1972) as:

The aerodynamic conductance (ga) and decoupling coeffi-cient (Ω) were estimated according to Monteith and Unsworth ( 1990) and surface conductance (gs) calculated by inverting the Penman–Monteith equation (Allen et al.1998):

whereU*is friction velocity (m s-1),ρa(bǔ)signifies air density(1.2 kg m-3, Gao et al.2018), and cprepresents the specific heat of dry air (1004.7 J kg-1°C-1, Gao et al.2018).

Statistical analyses

The influence of biophysical factors onEtwith path analysis using SAS for Windows (Version 9, SAS Institute, Cary, NC,USA).The other statistical analyses were performed with SPSS for Windows (Version 18, SPSS Inc., Chicago, IL, USA).Linear regression was used to evaluate the relationships betweenLaiand dailyEf,β,Kc,α,gs, andΩ; exponential regression was used to evaluate the relationships between dailygsandα.

Results

Biophysical factors

Average air temperatures (Ta) ranged from - 12.50 °C in winter to 32.00 °C in summer, and average relative humilities(Ha) fluctuated around 60% (Fig.2 a).Average vapor pressure deficits (Vpd) ranged from ＞ 2.00 kPa in early summer to＜ 0.50 kPa in winter (Fig.2 b).Average groundwater depths(Gd) ranged from 1.38 to 2.16 m between 12 June 2018 and 31 December 2019 (Fig.2 b).Average wind speed (U)usually fluctuated above 2.00 m s–1(Fig.2 c).Annual precipitation (P) was 662.00 mm, 489.30 mm, and 487.50 mm over the three study years.The amount ofirrigation (I) was 80.00 mm each year.Soil water content at 10-cm depth (W10)depended onPandI, and a severe drought occurred in 2019 as soil water levels at 40-cm depth (W40) was lower than 0.17 cm3cm–3(the relative extractable water was equal to 0.40,Ma et al.2019) and lasted 73 days (Fig.2 d).The maximum leaf area index (Lai) was 3.08 m2m–2, 3.14 m2m–2, and 2.20 m2m–2for the three study years.In the mid growing stage(MG) of 2018, the black locust canopy was damaged in a hailstorm (14 August), resulting in a noticeable drop inLai(Fig.2 e).

Fig.2 Seasonal variations in biophysical factors in 2016,2018 and 2019; biophysical factors are a air temperature ( Ta) , and relative humidity ( Ha) ,b vapor pressure deficit ( Vpd) ,c wind speed ( U), groundwater depth ( Gd) , d precipitation ( P),irrigation ( I), soil water content at 10-cm depth ( W10) and 40-cm depth ( W40) , e leaf area index( Lai) ; DS: dormant season, EG:early growing stage, MG: mid growing stage, LG: late growing stage

Energy partitioning

Seasonal variations in daily downward/upward longwave radiation (Ld/Lu) and downward/upward shortwave radiation(Sd/Su) displayed parabolic patterns (Fig.3 a).The maximum value of daily net shortwave radiation (Sn) usually appeared in early June, and daily net longwave radiation (Ln) fluctuated between zero to - 102.73 W m–2(Fig.3 b).Daily albedo increased sharply in winter due to snow cover and peaked in the early growing stage (EG) (Fig.3 c).

Fig.3 Seasonal variations in a downward/upward shortwave radiation ( Sd/ Su) and downward/upward longwave radiation( Ld/ Lu) , b net shortwave/longwave radiation ( Sn/ Ln) , and c albedo in 2016, 2018 and 2019

Annual energy balance ratios were 0.68–0.69 with a slope of 0.57, goodness-of-fit (R2) values of 0.85–0.86, and the intercepts between 12.05 and 12.50 W m-2(Table 2 ).

Table 2 Characteristics of the energy balance and annual ratios of Σ( Le + H)/Σ ( Rn - G) in 2016, 2018, and 2019; Le latent heat flux; H sensible heat flux; Rn net radiation; G soil heat flux

Daily net radiation (Rn) ranged from - 18.49 W m-2in winter to 265.41 W m-2in summer (Fig.S1a).There was one peak in the seasonal course of daily sensible heat flux(H), appearing in mid-April (Fig.S1b).Daily latent heat flux (Le) was usually smaller than 10.00 W m-2in winter and greater than 120.00 W m-2after rainy days in summer(Fig.S1c).Daily soil heat flux (G) ranged from - 19.71 to 13.73 W m-2(Fig.S1d).The monthly averageRnandHshowed clear diurnal variations and peaked at noon.The largest diurnal peaks ofRnandHwere 690.76 (May 2019)and 332.21 (April 2016) W m-2, respectively.The monthly averageLeandGshowed clear diurnal variations from March to October in both years becauseLeandGin other months were small.The largest diurnal peaks ofLeandGwere 239.23 (September 2018) and 57.05 (April 2019) W m-2, respectively (Fig.4).

Fig.4 Diurnal cycle of monthly average net radiation( Rn ) and its components in 2016, 2018 and 2019

Albedo in the growing season (GS) was similar to that in the dormant season (DS), and the annual albedo was either 0.12 or 0.13.BothLd/LuandRn/(Sd+Ld) were smaller in the DS than in the GS.AnnualLd/Luwas either 0.85 or 0.84 and annualRn/(Sd+Ld) was 0.18–0.19.In 2016, 2018, and 2019,H/Rnin the DS was 0.69, 0.66, and 0.61, evaporative fraction(Ef) in the GS was 0.44, 0.42, and 0.40, respectively.AnnualH/Rn(0.31–0.34) was comparable to annualEf(0.34–0.38).The ratio ofGtoRnwas relatively lower in different periods(Table S2).

AverageSd,Sn,Ld,Lu,RnandLewere larger in the GS.AverageHwas smaller in the GS and averageGwas positive and negative in the GS and in the DS, respectively (Fig.S2).

Evapotranspiration

Daily evapotranspiration (Et) was usually less than 0.50 mm d-1in the DS and peak values of dailyEtappeared afterPevents (Fig.5).In particular, the reference evapotranspiration (Et0) was relatively high andEtrelatively weak before the EG in the DS.In the EG, theEtincreased rapidly and theEt0relatively slowly.In the MG,Et0decreased slowly and the day-to-day variation was not synchronous with theEtdue to drought andP.In the late growing stage (LG), theEtdecreased with decreasingEt0.AnnualEtwas 497.16 mm,503.19 mm, and 479.66 mm, and annualEt/(P+I) was 0.67,0.86, and 0.85 for the three study years.Water balance was- 24.6 mm, - 11.4 mm, and - 18.9 mm in the DS of 2016,2018, and 2019, respectively (Table S3).

Fig.5 Seasonal variations in a reference evapotranspiration( Et0) , and b evapotranspiration( Et) in 2016, 2018 and 2019

The major factors controlling seasonal variation in dailyEtwereRn,Ta,Vpd, andLai(Table 3).The effect ofTaonEtwas indirect and viaLaiandVpd; the effect ofLaionEtwas direct.ForRn, the direct effect was smaller than the indirect effect, and mainly viaLaiandVpd, in 2018 and 2019.ForVpd, the direct effect was greater than the indirect effect in 2018 and 2019.In addition,Uin 2016 andW10in 2018 had significant negative and positive correlations with dailyEt(p＜ 0.01), respectively.

Table 3 Path analysis between daily Et and net radiation ( Rn ),wind speed ( U), air temperature( Ta ), vapor pressure deficit( Vpd ), soil water content at 10-cm depth ( W10 ), and leaf area index ( Lai ) in the black locust plantation

The range in wind direction (Wd) from zero to 90° and 180° to 270° was used as typical sea wind and land wind,respectively.Because the canopy cover had an important effect onEt, and dramatically decreased by severe drought,onlyEtand biophysical data in the MG of 2016 were used to analyze the effect of wind direction onEt.Average λEtunder sea winds (80.01 W m-2) was greater than under land winds (67.59 W m-2); average sea and land wind directions were 57.76° and 205.57°, respectively (Fig.S3a and b).AverageRnunder sea and land winds were 139.99 W m-2and 155.37 W m-2, respectively (Fig.S3c).After 8:00 h,VpdandTaunder land winds were relatively greater (Fig.S3d and e).Between 8:00 h and 18:00 h,W10under sea winds was relatively greater, and around noon,Uunder sea winds was relatively greater (Fig.S3f and g).

Surface parameters

Daily evaporative fraction (Ef), crop coefficient (Kc), Priestley–Taylor coefficient (α), surface conductance (gs), and decoupling coefficient (Ω) had similar seasonal patterns, all relatively smaller in the DS, increasing with canopy growth and development, and decreasing with leaf senescence(Fig.6 a, c–f).Daily Bowen ratio (β) was opposite compared with the other surface parameters (Fig.6 b).Average dailyEf,Kc,α,gs, andΩreached maximum values of 0.42, 0.49,0.48, 4.43 mm s-1, and 0.14 in the MG of 2016, respectively.Average dailyβwas 5.94, 5.38, and 5.84 in the DS of 2016,2018, and 2019, respectively (Table S4).

Fig.6 Seasonal variations in daily a evaporative fraction ( Ef) ,b Bowen ratio ( β), c crop coefficient ( Kc) , d Priestley–Taylor coefficient ( α), e surface conductance ( gs) , and decoupling coefficient ( Ω) in 2016, 2018 and 2019.Data for rainy days are not shown

The linear relationships between dailyEf,Kc,α,gs, andΩandLaihad goodness-of-fit values of 0.44–0.78 (Fig.S4a and c–f).Dailyβdecreased exponentially with increasingLai, and the goodness-of-fit value was 0.96 (Fig.S4b).An exponential relationship existed between dailyαandgs;αwas insensitive whengswas greater than approximately 8 mm s-1, and the asymptotic value of dailyαwas 0.66(Fig.7).

Fig.7 Relationship between daily α and gs in the growing season;when gs ＜ 7 mm s -1 , the dependent variables were bin-averaged into 0.5 mm s -1 gs increments; when 7 mm s -1 ＜ gs ＜ 11 mm s -1 , the dependent variables were bin-averaged into 2 mm s -1 gs increments;when gs ＞ 11 mm s -1 , the dependent variables were bin-averaged into 10 mm s -1 gs increments.*Represents a significance level of p ＜ 0.01

Discussion

Energy balance

The energy balance closure is used to determine the performance of the eddy covariance (EC) measurements.A comprehensive evaluation of energy balance closure in FLUXNET showed that energy balance ratios ranged from 0.34 to 1.69 (Wilson et al.2002).Another evaluation of energy balance closure in ChinaFLUX showed that energy balance ratio values ranged from 0.58 to 1.00 (Li et al.2005).Our results in Table 2 are within the ranges in FLUXNET and ChinaFLUX, comparable to the results in a hilly tea plantation, east China (Geng et al.2020) and in a subalpine forest, southwest China (Yan et al.2017),and better than the results in a young plantation, north China (Ma et al.2019), indicating that the eddy covariance measurements were reliable in estimating surface energy balance components in the study plantation.Reasons for the energy unbalance may be related to sampling error, instrument bias, neglected energy sinks, high/low frequency loss, advection, and other complications(Wilson et al.2002).At this study site, the open canopy with understory vegetation resulted in a patchy underlying surface, which might lead to the observed inconsistencies.Energy for photosynthesis into biomass and for heat storage in air below the EC measurement level were excluded in our analysis, which might have led to an overestimation of available energy.Moreover, the long- and short-wave radiation transmission at the soil-vegetation interface might be altered by surface cover patchiness, and this phenomenon might alter the allocation of available energy (Ma et al.2019).

Energy partitioning

The mean annual averages ofLd(downward long radiation),Lu(upward long radiation),Sd(downward short radiation),andSu(upward short radiation) over the three study years was 327.49 W m-2, 383.82 W m-2, 184.31 W m-2, and 21.90 W m-2(Fig.S2).These were comparable to a single crop farmland at the same latitude on the Loess Plateau (Gao et al.2018).Compared to those of an alpine meadow on the Tibetan Plateau (You et al.2017),LdandLuwere relatively greater andSdandSurelatively smaller.Solar altitude mainly controls the diel and seasonal variations inSd, which is also influenced by aerosols and clouds.Suis mainly controlled bySdand influenced by land cover, in particular snow.Annual average albedo on our study site was smaller than that of an alpine meadow (0.31, You et al.2017) and a single crop farmland (0.18, Gao et al.2018), being consistent with the widely accepted recognition that albedo is often lower in forests than in croplands and grasslands (Zhu et al.2014).Daily albedo peaked in the early growing stage (EG) at our site (Fig.3 c), suggesting that the peak might be caused by the reflection oflight on young leaves with pubescence.The peak becomes less obvious with the young leaf growing,during which the pubescence gradually fall off.Surface temperatures control the daily and seasonal variations inLu, andLdis mainly influenced by air moisture and temperature.In the growing season (GS), humid and warm air resulted in greaterLd/Lu(Table S2).Annual averageLd/Luin the plantation was comparable to that of a sub-alpine spruce forest on the Tibetan Plateau (0.76, Zhu et al.2014), and of a mixed cropping system on the Loess Plateau (0.84, Chen et al.

2016).Similar to previous studies (Zhu et al.2014; Chen et al.2016), the averageRn/(Sd+Ld) was greater in the GS because of greaterLd/Lu.Annual averageRn/(Sd+Ld) was comparable to those of croplands on the Loess Plateau (Chen et al.2016; Gao et al.2018).

As vegetation growth dominates energy partitioning,H/Rnand evaporative fraction (Ef) in the GS clearly differed from those in the dormant season (DS) (Table S2).In the DS, 3-year averageH/Rnwas 0.65, indicating that net radiation (Rn) was mainly portioned to sensible heat f lux(H) at our site as found in forests, shrublands, grasslands and croplands in previous studies (Zhu et al.2014; Jia et al.2016; You et al.2017; Gao et al.2018).In the GS,Efwas greater thanH/Rn(Table S2), indicating that latent heat f lux(Le) was the largest component ofRnin the black locust plantation.Similar to previous studies (Zhu et al.2014; Ma et al.2019),Efwas ＜ 0.5 in the GS, and the limited water and open canopy might have resulted in this portioning.On an annual scale,Efwas 0.38, 0.37, and 0.34 in 2016, 2018,and 2019, respectively (Table S2).These values were higher than those of a sub-alpine forest (0.28, Zhu et al.2014) but smaller for a hilly tea plantation (Geng et al.2020).Because daily soil heat f lux (G) was usually positive and negative during periods ofincreasing and decreasing temperature,respectively.AnnualG/Rnwas small at our site (Table S2),consistent with previous studies of various ecosystems (Zhu et al.2014; Jia et al.2016; You et al.2017; Gao et al.2018).

In the diurnal cycle of average monthly energy f luxes,Lewas a major component ofRnin May–September in 2016 and in August and September in 2018 and 2019.Hwas a major component ofRnin other months (Fig.4).The different vegetation growth and seasonal distribution of precipitation (P) may be responsible for this observation.At our site,Lewas slightly above zero at night, indicating that evapotranspiration (Et) occurred at night;Hwas a little below zero at night, suggesting that the heat from the atmosphere might be used forEtand offset the heat loss from the soil and vegetative due to long-wave radiation.These observations are similar to previous studies in various ecosystems (Wang et al.2010; Zhu et al.2014; Chen et al.2016; Jia et al.2016;Hayat et al.2020).At our site,Gwas positive in the daytime and negative at night (Fig.4) due to energy dissipation into the soil in the daytime and out at night, resulting in relatively small dailyGvalues (Fig.S4d).AsGwas influenced by the shade of vegetation and litter and seasonal variation inRn, the amplitude of the diurnal cycle ofGat our site was between 6.88 (January 2018) and 72.84 (June 2016) W m-2,comparable to that of a sub-alpine spruce forest (Zhu et al.2014), and usually smaller than that of croplands, grasslands and shrublands (Gu et al.2005; Jia et al.2016; Gao et al.2018), indicating the measurement ofGin high vegetation may be less important than in low vegetation on a 30-min time scale.

Evapotranspiration

Annual evapotranspiration (Et) (479.66–503.19 mm) at our site was comparable to that of a sub-alpine forest (Zhu et al.2014) and an urban forest ecosystem in north China (Xie et al.2016), smaller than that of an evergreen broad-leaved forest, south China, and a poplar plantation, north China(Xiao et al.2013), and greater than that of a temperate mixed forest, northeast China (Xiao et al.2013) and a young plantation (Ma et al.2019).AnnualEtvariations in different sites may be related to meteorological factors, vegetation properties and soil water conditions.The ratio of annualEttoPand irrigation (I) is a key parameter quantifying the effect of land cover change on regional hydrology (Ma et al.2019).AnnualEt/(P+I) was 0.67–0.86 at our site, comparable to that of other forest ecosystems (Xiao et al.2013).AnnualEt/(P+I) was ＜ 1, indicating that the plantation was going through the process of salinity eluviation or leaching.However, cumulativeEtwas higher than cumulativePin the dormant season (DS).This indicates that the DS was a process of salt accumulation.This also suggests thatIwas necessary to mitigate the effects of the drought and salt stress for black locust growth in the early growing stage (EG).

As similar to many terrestrial ecosystems (Xiao et al.2013; Gao et al.2018), net radiation (Rn) andTa(air temperature) were major meteorological factors controlling dailyEtat our site, suggesting the critical role of energy supply in water vapor loss in the plantation (Sun et al.2019).As with a coastal salt marsh ecosystem in east China (Huang et al.2019), vapor pressure deficit (Vpd)was also a major controlling dailyEt, because of relatively higher relative humilities (Ha) in the coastal zone, which influences water vapor transport (Fig.2 a).As leaves are the effective area for transpiration, leaf area index (Lai)had a significantly positive correlation with dailyEtat our site, and this was comparable with a single crop cropland (Gao et al.2018).Seasonal variation in daily reference evapotranspiration (Et0) was not consistent with dailyEt(Fig.5), showing that the seasonal pattern of dailyEtin the black locust plantation was related to vegetation phenology, indicating that seasonal patterns of dailyEtwas controlled by vegetation growth.Seasonal variation inLaishould be is the critical factor of seasonal variation in dailyEtduring the different phenological stages of black locust.The effect of vegetation phenology onEtwas also assessed in a riparianTamarixspp.stand (Yuan et al.2014), and the results were similar to our study.However, dailyEtdid not dramatically decrease after a noticeable drop ofLaiin the mid growing stage (MG) of 2018 (Figs.2 e, 6 b).This is mainly because understory vegetation and the soil surface received more radiation,resulting in higher understory transpiration and soil evaporation after the canopy was damaged in a hailstorm.In the same year, soil water at 10-cm depth (W10) had a significant positive correlation with dailyEt, because the ratio of understory vegetation transpiration and soil evaporation to totalEtmay have been greater, as the upper soil layer is the water source for understory transpiration and soil evaporation.Therefore, more focus should be given to the interactions between understory and overstory vegetation.In 2016, wind speed (U) had negative effect on dailyEtviaLaiat our site (Table 3), mainly because the general trend ofUis opposite that ofLaiin the growing season (GS).

Although annualEt/(P+I) was ＜ 1, drought still occurred in the plantation and seasonal precipitation (P) was not uniform (Fig.2 d).Salinity stress increased with drought.Similar to previous studies (Ma et al.2019; Yue et al.2019),drought inhibited dailyEtby approximately 40% in 2019,reducing black locust growth and resulting in smaller peak value ofLai.With the background of climate warming and periods of drought extending on the North China Plain(Wang et al.2017b), the black locust plantation in the Yellow River Delta may be facing severe pressure to survive.According to Ma et al.( 2013), while black locust roots grow into groundwater, the species would die because of higher groundwater salinity.As the 1.38 to 2.16 m water table on this site, root systems can only survive to a 1.38 m depth and this may weaken drought resistance of black locust.In addition, the Yellow River Delta is subsiding due to a combination of factors (Liu et al.2019) which may lead to lower groundwater depth (Gd), and needs future attention.

Wind direction influenced evapotranspiration (Et) in the plantation (i.e.,Etunder sea winds were greater than that under land winds, Fig.S3a), and consistent with the results from a coastal salt marsh ecosystem (Huang et al.2019).However, the cause of the occurrence in the previous studies was different from that in this study.In the previous study,Taunder sea winds was almost equal to that under land winds,Vpdwere relatively smaller under sea winds, and solar radiation along withUwere relatively larger under sea winds in the daytime.And solar radiation was the dominant control onEtresulting in largerEtunder sea winds (Huang et al.2019).In this study,Rn,Ta, andVpdwere larger under land winds in the daytime, onlyUandW10was larger under sea winds at mid-day, respectively (Fig.S3b–g).We suspects thatEtunder sea winds was greater at our site because of higher soil water levels under sea winds in the daytime, indicating that soil water was not sufficient all the time and also that sea winds might occur afterP.

Surface parameters

Seasonal variations in daily evaporative fraction (Ef), Bowen ratio (β), crop coefficient (Kc), Priestley–Taylor coefficient(α), surface conductance (gs), and decoupling coefficient (Ω)were similar to those of previous studies (Zhu et al.2014; Jia et al.2016; Gao et al.2018).DailyEfandβwere the important surface parameters of energy partitioning, and dailyEfwas opposite from dailyβin the mid growing stage (MG)(Fig.6 a, b).Average dailyβin the MG was higher than that in a single crop cropland and a hill tea plantation (Gao et al.2018; Geng et al.2020), and smaller than that in an alpine meadow (You et al.2017) and a semi-arid shrubland (Jia et al.2016).DailyKcis an important surface parameter for planning irrigation systems and for estimating evapotranspiration (Et) in croplands (Guo et al.2020).This has attracted considerable attention in natural ecosystems in recent years(Yang and Zhou 2011; Yuan et al.2014).Average dailyKcin the MG at our site was smaller than that in a desert oasis in northwest China (Zhang et al.2016), and larger than that in a temperate desert steppe in Inner Mongolia, China (Yang and Zhou 2011).Dailyα,gs, andΩare the bulk surface parameters used to assess the effect of biophysical factors onEt(Jiao et al.2018).Average dailyαin the MG was approximately 0.50 (Table S4), which indicates that available soil water in the root zone was insufficient.Average dailygsin the MG in the black locust plantation was smaller than that in a hill tea plantation and a subalpine forest with greaterEt(Yan et al.2017; Geng et al.2020), and previous studies had shown that higherEtis associated with largergsin various vegetation types (Zhang et al.2016).Average dailyΩin the MG in this study was relatively smaller (0.12–0.21,Table S4) comparable to that of a young plantation (Ma et al.2019), and larger than that of a sub-alpine spruce forest (Zhu et al.2014).The smaller dailyΩindicates that the atmosphere and canopy were coupled, and vapor pressure deficit (Vpd) was an important factor controllingEtin the black locust plantation (Table 3; Jiao et al.2018).

DailyEf,α,gs, andΩincreased with increasing leaf area index (Lai) and dailyβdecreased (Fig.S4), which is consistent with other studies (Jia et al.2016; Ma et al.2019).DailyKcincreased linearly withLaiat our site, which is in agreement with a riparianTamarixspp.stand (Yuan et al.2014),indicating thatEtwas mainly contributed by leaf transpiration as quantified byLaiwithout a threshold, and implying that smallerLaimight be a reason for a relatively smaller evapotranspiration (Et) in the black locust plantation.DailyKcincreased exponentially with increasingLaiin cropland with largerLaiassociated with higherEt(Zhang et al.2016).Abiotic factors (e.g., soil moisture andVpd) also influenced surface parameters (Jia et al.2016).This implies that day-today f luctuations and trends of seasonal variations in surface parameters were mainly controlled by abiotic factors andLai, respectively.The effects ofLaion surface parameters illustrates that vegetation growth played a key role in energy partitioning,Etestimating, and land surface development.

The logarithmic curve between surface conductance (gs)and Priestley–Taylor coefficient (α) is consistent with results from various ecosystems (Jia et al.2016; Gao et al.2018;Ma et al.201), indicating strong physiological and phenological regulation of energy partitioning and evapotranspiration (Et).Dailygsincreased with increasinggsuntil the threshold (ca.8 mm s-1; Fig.7), indicatingEtwas strongly influenced bygswhengs＜ 8 mm s-1.A theoretical study on a fully developed canopy indicated thatαwas insensitive togswhengs＞ 16 mm s-1(McNaughton and Spriggs 1986; Gao et al.2018).The asymptotic value of dailyαat our site was 0.66 (Fig.7), smaller than the results from many ecosystems with higherLaiandEt(Yan et al.2017; Gao et al.2018;Geng et al.2020).The smaller asymptotic value of daily α andgsthreshold in this study were possibly due to the open canopy coupled with drought and salinity stress.

Conclusion

Downward longwave radiation offset 84%–85% of upward longwave radiation, upward shortwave radiation accounted for 12%–13% of downward shortwave radiation, and the ratio of net radiation to downward radiation was 18%–19%.In the growing season, latent heat f lux was the largest component of net radiation; in the dormant season, sensible heat f lux was the dominant component of net radiation.Seasonal variations of daily evapotranspiration were controlled by net radiation, air temperature, vapor pressure deficit, and leaf area index.Black locust phenology influenced seasonal variation in daily evapotranspiration, and evapotranspiration was greater under sea winds than under land winds because soil water content at 10-cm depth was larger under daytime sea winds.Seasonal patterns of daily evaporative fraction,Bowen ratio, crop coefficient, Priestley–Taylor coefficient,surface conductance, and decoupling coefficient were regulated by leaf area index and the threshold value of daily surface conductance was approximately 8 mm s-1.The open canopy, drought and salinity stress played an important role in energy partitioning, evapotranspiration and surface development at our site.These results should be a valuable reference for sustainable management of black locust plantations in the Yellow River Delta.