Environmental controls on seasonal ecosystem evapotranspiration/potential evapotranspiration ratio as determined by the global eddy flux measurements

1. Jiangsu Provincial Key Laboratory of Agricultural Meteorology, College of Applied Meteorology, 5 Nanjing University of Information Science and Technology, Nanjing 210044, China; 2. Eastern Forest Environmental Threat Assessment Center, Southern Research Station, USDA Forest Service, Raleigh, NC 27606, USA; 3. Department of Forestry and Environmental Resources, North Carolina State University, Raleigh, NC 27695, USA. 10


Introduction
Evapotranspiration (ET) is one of the major hydrological processes that link energy, water, and carbon cycles in terrestrial ecosystems (Fang et al., 2015;Sun et al., 2011a;Sun et al., 2011b;Sun et al., 2010). In contrast to potential ET (PET) that depends only on atmospheric water demand (Lu et al., 2005), actual evapotranspiration (AET) is arguably the most 45 uncertain ecohydrologic variable for quantifying watershed water budgets (Baldocchi and Ryu, 2011;Fang et al., 2015;Hao et al., 2015a) and for understanding the ecological impacts of climate and land use change (Hao et al., 2015b), and climate variability (Hao et al., 2014). In recent years, one of the most important research questions of ecohydrology focused on how ecosystem dynamics, precipitation, AET, and PET interact in different 50 ecosystems at seasonal and long term scales under a changing environment .
The ratio of AET to PET is traditionally termed as crop coefficient (Kc), and has been widely used to as a parameter to estimate crop water demand by water managers (Allen and Pereira, 2009;Irmak et al., 2013a).However, this parameter has not been well examined 55 for other ecosystems (Zhang et al., 2012;Zhou et al., 2010). The ratio of AET to PET has also been used as an indicator of regional terrestrial water availability, wetness or drought index, and plant water stress (Anderson et al., 2012;Mu et al., 2012).When the AET/PET ratio is close to 1.0, the soil water meets ecosystem water use demand. The ratio of AET/PET or water stress level can be drastically different among different ecosystems in 60 different environmental conditions, because AET is mainly controlled by climate (precipitation and PET) (Zhang et al., 2001) and ecosystem species composition and Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-237, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci. Published: 24 May 2016 c Author(s) 2016. CC-BY 3.0 License. structure (i.e., leaf area index, rooting depth, stomata conductance) (Sun et al., 2011a). The seasonal PET values for a particular region are generally stable (Lu et al., 2005;Rao et al., 2011), and deviation of AET/PET from the norm indicates variability in AET, which 65 responds to precipitation and water availability when PET is stable (Rao et al., 2011).
However, under a changing climate, the AET/PET patterns can be rather complex since both AET and PET are affected by air temperature and precipitation (Sun et al., 2015a;Sun et al., 2015b) and corresponding changes in ecosystem characteristics (e.g., plant species shift) (Sun et al., 2014;Vose et al., 2011). 70 In the agricultural water management community, the crop coefficient method remains a popular one for approximating crop water use, despite recent advances in direct ET measurement methods (Allen and Pereira, 2009;Allen et al., 1998;Baldocchi et al., 2001;Fang et al., 2015). The Kc is termed as single crop coefficient (Allen et al., 1998;Allen et al., 2006;Tabari et al., 2013) which is affected by growing periods, crop species, canopy 75 conductance, and soil evaporation in the field scale (Allen et al., 1998;Ding et al., 2015;Shukla et al., 2014b). Moreover, Kc can be influenced by soil characteristics, vegetative soil cover, height, plant species distribution, and leaf area index in a larger spatial scale (Anda et al., 2014;Consoli and Vanella, 2014;Descheemaeker et al., 2011).
Although the Food and Agriculture Organization of the United Nations provides various 80 guidelines for several crops (Allen et al., 1998), local measurements are still required to estimate Kc to account for local crop varieties and for year-to-year variation in weather conditions (Pereira et al., 2015).
Although the Kc method has been widely used for estimating AET for crops, it has not been widely used for natural ecosystems for the purpose of estimating AET due to limited 85 Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-237, 2016  continuous measurements in these systems. However, as discussed earlier, ecologists and hydrologist have started to use Kc to quantify ecosystem stress levels, and consider Kc as a variable rather than a constant. Past studies found that Kc was influenced by the growing stages and leaf area index for maize (Ding et al., 2015;Kang et al., 2003), winter wheat (Allen et al., 1998;Kang et al., 2003), watermelon (Shukla et al., 2014b), and fruit 90 trees (Marsal et al., 2014b;Taylor et al., 2015 (Allen and Pereira, 2009;Allen et al., 2011). Therefore, the goal of this study was to explore how Kc varies among multiple 100 ecosystems with various vegetation types over multiple seasons. Another goal was to determine the key biophysical and environmental factors such as latitude, precipitation, and leaf area index that could be used to estimate Kc, and if Kc can be modeled with a reasonable accuracy in a larger spatial scale. We examined the Kc variations for seven land cover types by analyzing the FLUXNET eddy flux data (Baldocchi et al., 2001;Fang et al., 105 2015). Specifically, our objectives were to 1) understand the variation of monthly Kc for seven distinct land covers by analyzing the influences of environmental factors (e.g., precipitation, site latitude) on Kc; and 2) to develop simple land cover-specific regression Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-237, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci.

Methods
This synthesis study used the LaThuile eddy flux dataset that was developed by FLUXNET (http://fluxnet.ornl.gov/; Fig. 1), a global network that measures the exchanges of carbon 120 dioxide, water vapor, and energy between the biosphere and atmosphere (Baldocchi et al., 2001). The FLUXNET data (Baldocchi et al., 2001;Baldocchi and Ryu, 2011)  We used an existing database that was developed from the eddy flux measurements from 81 sites (Fang et al., 2015). According to the International Geosphere-Biosphere Program (IGBP) land cover classification system, these eddy flux sites represent nine land forest (DB), evergreen needle leaf forest (ENF) and evergreen broad leaf forest (EBF), and mixed forest (MF). For each eddy flux tower site (Figure 1), we acquired AET and associated micro-meteorological data, such as vapor pressure deficit (VPD), precipitation (P), winds speed (WS), net radiation (Rn). Reference evapotranspiration(ET0) was calculated by the FAO Penman-Monteith equation as follows (Allen et al., 1998): where Rn is net radiation at the cover surface (MJ m -2 d -1 ), G is soil heat flux (MJ m -2 d -1 ), T is mean air temperature at 2 m height (°C), u2 is wind speed at 2 m height (m s -1 ), es is saturation vapour pressure (kPa), ea is actual vapour pressure (kPa), es-ea is the saturation vapour pressure deficit (kPa), Δ is slope vapour pressure curve (kPa °C -1 ), and γ is the 140 psychrometric constant (kPa °C -1 ).
The crop coefficient (Kc) is defined as the ratio of the measured AET and the ET0 calculated by equation (1)

Seasonal variations and long term means of Kc by land cover
The average monthly Kc based on eddy flux data from 2000 to 2007 increased gradually from January to July and then decreased (Fig. 2). EBF had the highest mean monthly Kc (1.01±0.17) (mean ± standard error) in August. Kc for both EBF and ENF varied less (376-425 mm). In contrast, CRO had relatively low precipitation with a high ET0.

Environmental controls on Kc
At the annual temporal scale, annual Kc was negatively (p<0.05) correlated with the latitude of the sites (Fig.5) (Fig. 2). Therefore, Kc was closely related to the 185 monthly precipitation.
Growing season, site latitude and monthly precipitation affected the monthly Kc, in addition to leaf area index (Fig. 7). Kc was obviously influenced by the leaf area index (LAI) for all land covers except EBF. The determination coefficients for different land covers were OS> MF>GRA> ENF>DB>CRO. The LAI could reach 6 m 2 m -2 in most land 190 covers, while in OS and CRO the LAI were only 3-4 m 2 m -2 .

3.3.Kc models
A series empirical Kc model were developed using a multiple linear regression approach with precipitation, leaf area index (LAI), and site latitude as independent variables ( are for a specific ecosystem are available.

Discussion
Our study estimated annual and seasonal crop coefficient (Kc) for seven land cover types using measured global eddy flux data. We comprehensively evaluated environmental controls (i.e., precipitation, LAI, and site latitude) on annual and growing seasons Kc and 210 developed a series of multiple linear regression models that can be used for estimating monthly AET over time and space.

Crop coefficient variation in different seasons
Several recent studies had shown that Kc reached the maximum value in middle of the growing season in many ecosystems, such as a P. euphratica forest in the riparian area  the summer. It is likely that the orchards had higher evapotranspiration rates than natural forests due to irrigation in orchards.

Environmental control factors for Kc
The ecosystem covers and the distributions of the vegetation classes were determined by the latitude (Potter et al., 1993). Crop coefficient varied predominately by ecosystems, Kc 230 increased as the site latitude decreased for the same land cover (Fig. 5). As the latitude decreased, the temperature and the solar radiation increased and the vegetation characteristics would be different for the same land cover type. Models developed from the FLUXNET data may be best used on flat areas for a given latitude given that eddy covariance towers were generally installed on flat lands (Baldocchi et al., 2001). For areas 235 with complex topography, the relationship between Kc and site latitude may be more complicated.
Spatial variations of Kc are characteristic of ecosystems, but Kc is also affected by climate factors such as rainfall and temperature. For example, Kc was highly correlated with precipitation for most land covers (Fig. 6).The rainfall is the major source of soil water 240 and AET in natural ecosystems (Parent and Anctil, 2012). During dry years or periods, a

Modeling the dynamics of Kc
Our study results are consistent with previous studies that show that the growing stage   (Taylor et al., 2015). The LAI and total monthly precipitation varied in both time and space while the site latitude only represents spatial influences on Kc. Thus, the multiple linear regression equations developed from this study take account of both spatial and temporal changes in land surface characteristics and offer a powerful 270 tool to estimate of seasonal dynamic Kc for different ecosystems (Table 1).

Conclusions
To seek a convenient method to calculate monthly AET in large spatial scale, we   Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2016-237, 2016 Manuscript under review for journal Hydrol. Earth Syst. Sci.